From 46b96f24cda66f039527ca44e786a9c6c06693ca Mon Sep 17 00:00:00 2001 From: Keyi Zhong Date: Tue, 6 Dec 2022 20:08:41 +0800 Subject: [PATCH] kae driver base line --- kae_driver/.vscode/sftp.json | 10 + kae_driver/Makefile | 67 + kae_driver/hisilicon/Kconfig | 91 + kae_driver/hisilicon/Makefile | 9 + kae_driver/hisilicon/hisi_acc_qm.h | 516 ++ kae_driver/hisilicon/hisi_qm.h | 39 + kae_driver/hisilicon/hpre/Makefile | 2 + kae_driver/hisilicon/hpre/hpre.h | 107 + kae_driver/hisilicon/hpre/hpre_crypto.c | 2217 ++++++ kae_driver/hisilicon/hpre/hpre_main.c | 1525 ++++ kae_driver/hisilicon/migration/Makefile | 2 + .../hisilicon/migration/acc_vf_migration.c | 1726 +++++ .../hisilicon/migration/acc_vf_migration.h | 242 + kae_driver/hisilicon/qm.c | 6630 +++++++++++++++++ kae_driver/hisilicon/sec/Makefile | 3 + kae_driver/hisilicon/sec/sec_algs.c | 1122 +++ kae_driver/hisilicon/sec/sec_drv.c | 1321 ++++ kae_driver/hisilicon/sec/sec_drv.h | 428 ++ kae_driver/hisilicon/sec2/Makefile | 2 + kae_driver/hisilicon/sec2/sec.h | 228 + kae_driver/hisilicon/sec2/sec_crypto.c | 2569 +++++++ kae_driver/hisilicon/sec2/sec_crypto.h | 410 + kae_driver/hisilicon/sec2/sec_main.c | 1370 ++++ kae_driver/hisilicon/sgl.c | 290 + kae_driver/hisilicon/trng/Makefile | 2 + kae_driver/hisilicon/trng/trng.c | 341 + kae_driver/hisilicon/zip/Makefile | 2 + kae_driver/hisilicon/zip/zip.h | 88 + kae_driver/hisilicon/zip/zip_crypto.c | 839 +++ kae_driver/hisilicon/zip/zip_main.c | 1354 ++++ kae_driver/include_linux/uacce.h | 149 + kae_driver/include_uapi_linux/uacce.h | 38 + kae_driver/uacce/Kconfig | 13 + kae_driver/uacce/Makefile | 2 + kae_driver/uacce/uacce.c | 595 ++ 35 files changed, 24349 insertions(+) create mode 100644 kae_driver/.vscode/sftp.json create mode 100644 kae_driver/Makefile create mode 100755 kae_driver/hisilicon/Kconfig create mode 100755 kae_driver/hisilicon/Makefile create mode 100755 kae_driver/hisilicon/hisi_acc_qm.h create mode 100755 kae_driver/hisilicon/hisi_qm.h create mode 100755 kae_driver/hisilicon/hpre/Makefile create mode 100755 kae_driver/hisilicon/hpre/hpre.h create mode 100755 kae_driver/hisilicon/hpre/hpre_crypto.c create mode 100755 kae_driver/hisilicon/hpre/hpre_main.c create mode 100755 kae_driver/hisilicon/migration/Makefile create mode 100755 kae_driver/hisilicon/migration/acc_vf_migration.c create mode 100755 kae_driver/hisilicon/migration/acc_vf_migration.h create mode 100755 kae_driver/hisilicon/qm.c create mode 100755 kae_driver/hisilicon/sec/Makefile create mode 100755 kae_driver/hisilicon/sec/sec_algs.c create mode 100755 kae_driver/hisilicon/sec/sec_drv.c create mode 100755 kae_driver/hisilicon/sec/sec_drv.h create mode 100755 kae_driver/hisilicon/sec2/Makefile create mode 100755 kae_driver/hisilicon/sec2/sec.h create mode 100755 kae_driver/hisilicon/sec2/sec_crypto.c create mode 100755 kae_driver/hisilicon/sec2/sec_crypto.h create mode 100755 kae_driver/hisilicon/sec2/sec_main.c create mode 100755 kae_driver/hisilicon/sgl.c create mode 100755 kae_driver/hisilicon/trng/Makefile create mode 100755 kae_driver/hisilicon/trng/trng.c create mode 100755 kae_driver/hisilicon/zip/Makefile create mode 100755 kae_driver/hisilicon/zip/zip.h create mode 100755 kae_driver/hisilicon/zip/zip_crypto.c create mode 100755 kae_driver/hisilicon/zip/zip_main.c create mode 100755 kae_driver/include_linux/uacce.h create mode 100755 kae_driver/include_uapi_linux/uacce.h create mode 100755 kae_driver/uacce/Kconfig create mode 100755 kae_driver/uacce/Makefile create mode 100755 kae_driver/uacce/uacce.c diff --git a/kae_driver/.vscode/sftp.json b/kae_driver/.vscode/sftp.json new file mode 100644 index 0000000..827e980 --- /dev/null +++ b/kae_driver/.vscode/sftp.json @@ -0,0 +1,10 @@ +{ + "name": "My Server", + "host": "175.33.44.16", + "protocol": "sftp", + "port": 22, + "username": "root", + "password": "Acc@2022", + "remotePath": "/home/zky/KAE_working", + "uploadOnSave": true +} diff --git a/kae_driver/Makefile b/kae_driver/Makefile new file mode 100644 index 0000000..c4fce47 --- /dev/null +++ b/kae_driver/Makefile @@ -0,0 +1,67 @@ +KERNEL_PATH := /lib/modules/5.10*/build +KSP := $(shell if test -d /lib/modules/5.10*/source; then \ + echo /lib/modules/5.10*/source; \ + else \ + echo /lib/modules/5.10*/build; \ + fi) + +obj-m += uacce/ +obj-m += hisilicon/ + +DIRS := $(shell find . -maxdepth 3 -type d) +TARGET = $(foreach dir,$(DIRS),$(wildcard \ + $(dir)/*.o) $(dir)/*.ko $(dir)/*.tmp_versions $(dir)/*.depend $(dir)/*.mod.c $(dir)/*.order $(dir)/*.symvers) + +defaul: + $(MAKE) -C $(KERNEL_PATH) M=$(shell pwd) modules \ + CONFIG_CC_STACKPROTECTOR_STRONG=y \ + CONFIG_UACCE=m \ + CONFIG_CRYPTO_QM_UACCE=m \ + CONFIG_CRYPTO_DEV_HISI_SGL=m \ + CONFIG_CRYPTO_DEV_HISI_QM=m \ + CONFIG_CRYPTO_DEV_HISI_ZIP=m \ + CONFIG_CRYPTO_DEV_HISI_HPRE=m \ + CONFIG_CRYPTO_DEV_HISI_SEC2=m \ + CONFIG_CRYPTO_DEV_HISI_TRNG=m +#copy: +# cp -f $(shell pwd)/include_linux/uacce.h $(KSP)/include/linux +# cp -f $(shell pwd)/include_uapi_linux/uacce.h $(KSP)/include/uapi/linux + +install: + $(shell mkdir -p /lib/modules/5.10*/extra) + $(shell find . -name "*.ko" -exec cp -f {} /lib/modules/5.10*/extra \;) + depmod -a + $(shell if test -e /etc/modprobe.d/10-unsupported-modules.conf; then \ + sed -i "s/^allow_unsupported_modules.*/allow_unsupported_modules 1/" /etc/modprobe.d/10-unsupported-modules.conf; \ + fi) + -modprobe uacce + -modprobe hisi_qm + -modprobe hisi_sec2 uacce_mode=2 enable_sm4_ctr=1 pf_q_num=256 + -modprobe hisi_hpre uacce_mode=2 pf_q_num=256 + -modprobe hisi_zip uacce_mode=2 pf_q_num=256 + -modprobe hisi_trng_v2 uacce_mode=2 pf_q_num=256 + -echo "options hisi_sec2 uacce_mode=2 enable_sm4_ctr=1 pf_q_num=256" > /etc/modprobe.d/hisi_sec2.conf + -echo "options hisi_hpre uacce_mode=2 pf_q_num=256" > /etc/modprobe.d/hisi_hpre.conf + -echo "options hisi_zip uacce_mode=2 pf_q_num=256" > /etc/modprobe.d/hisi_zip.conf + -echo "options hisi_trng_v2 uacce_mode=2 pf_q_num=256" > /etc/modprobe.d/hisi_trng_v2.conf +uninstall: + modprobe -r hisi_zip + modprobe -r hisi_hpre + modprobe -r hisi_sec2 + modprobe -r hisi_trng_v2 + modprobe -r hisi_qm + modprobe -r uacce + rm -rf /lib/modules/5.10*/extra/uacce.ko + rm -rf /lib/modules/5.10*/extra/hisi_qm.ko + rm -rf /lib/modules/5.10*/extra/hisi_sec2.ko + rm -rf /lib/modules/5.10*/extra/hisi_hpre.ko + rm -rf /lib/modules/5.10*/extra/hisi_trng_v2.ko + rm -rf /lib/modules/5.10*/extra/hisi_zip.ko + rm -rf /etc/modprobe.d/hisi_sec2.conf + rm -rf /etc/modprobe.d/hisi_hpre.conf + rm -rf /etc/modprobe.d/hisi_zip.conf + rm -rf /etc/modprobe.d/hisi_trng_v2.conf + depmod -a + +clean: + rm -rf $(TARGET) diff --git a/kae_driver/hisilicon/Kconfig b/kae_driver/hisilicon/Kconfig new file mode 100755 index 0000000..f9f749e --- /dev/null +++ b/kae_driver/hisilicon/Kconfig @@ -0,0 +1,91 @@ +# SPDX-License-Identifier: GPL-2.0 + +config CRYPTO_DEV_HISI_SEC + tristate "Support for Hisilicon SEC crypto block cipher accelerator" + select CRYPTO_SKCIPHER + select CRYPTO_ALGAPI + select CRYPTO_LIB_DES + select SG_SPLIT + depends on ARM64 || COMPILE_TEST + depends on HAS_IOMEM + help + Support for Hisilicon SEC Engine in Hip06 and Hip07 + + To compile this as a module, choose M here: the module + will be called hisi_sec. + +config CRYPTO_DEV_HISI_SEC2 + tristate "Support for HiSilicon SEC2 crypto block cipher accelerator" + select CRYPTO_SKCIPHER + select CRYPTO_ALGAPI + select CRYPTO_LIB_DES + select CRYPTO_DEV_HISI_QM + select CRYPTO_AEAD + select CRYPTO_AUTHENC + select CRYPTO_HMAC + select CRYPTO_SHA1 + select CRYPTO_SHA256 + select CRYPTO_SHA512 + select CRYPTO_SM4 + depends on PCI && PCI_MSI + depends on UACCE || UACCE=n + depends on ARM64 || (COMPILE_TEST && 64BIT) + depends on ACPI + help + Support for HiSilicon SEC Engine of version 2 in crypto subsystem. + It provides AES, SM4, and 3DES algorithms with ECB + CBC, and XTS cipher mode, and AEAD algorithms. + + To compile this as a module, choose M here: the module + will be called hisi_sec2. + +config CRYPTO_DEV_HISI_QM + tristate + depends on ARM64 || COMPILE_TEST + depends on PCI && PCI_MSI + depends on UACCE || UACCE=n + depends on ACPI + help + HiSilicon accelerator engines use a common queue management + interface. Specific engine driver may use this module. + +config CRYPTO_DEV_HISI_ZIP + tristate "Support for HiSilicon ZIP accelerator" + depends on PCI && PCI_MSI + depends on ARM64 || (COMPILE_TEST && 64BIT) + depends on !CPU_BIG_ENDIAN || COMPILE_TEST + depends on UACCE || UACCE=n + depends on ACPI + select CRYPTO_DEV_HISI_QM + help + Support for HiSilicon ZIP Driver + +config CRYPTO_DEV_HISI_HPRE + tristate "Support for HISI HPRE accelerator" + depends on PCI && PCI_MSI + depends on UACCE || UACCE=n + depends on ARM64 || (COMPILE_TEST && 64BIT) + depends on ACPI + select CRYPTO_DEV_HISI_QM + select CRYPTO_DH + select CRYPTO_RSA + select CRYPTO_CURVE25519 + select CRYPTO_ECDH + help + Support for HiSilicon HPRE(High Performance RSA Engine) + accelerator, which can accelerate RSA and DH algorithms. + +config CRYPTO_DEV_HISI_TRNG + tristate "Support for HISI TRNG Driver" + depends on ARM64 && ACPI + select HW_RANDOM + select CRYPTO_RNG + help + Support for HiSilicon TRNG Driver. + +config CRYPTO_DEV_HISI_MIGRATION + tristate "Support for HISI MIGRATION Driver" + depends on ARM64 && ACPI + select PCI_IOV + help + Support for HiSilicon vfio pci to support VF live migration. diff --git a/kae_driver/hisilicon/Makefile b/kae_driver/hisilicon/Makefile new file mode 100755 index 0000000..da68eab --- /dev/null +++ b/kae_driver/hisilicon/Makefile @@ -0,0 +1,9 @@ +# SPDX-License-Identifier: GPL-2.0 +obj-$(CONFIG_CRYPTO_DEV_HISI_HPRE) += hpre/ +obj-$(CONFIG_CRYPTO_DEV_HISI_SEC) += sec/ +obj-$(CONFIG_CRYPTO_DEV_HISI_SEC2) += sec2/ +obj-$(CONFIG_CRYPTO_DEV_HISI_QM) += hisi_qm.o +hisi_qm-objs = qm.o sgl.o +obj-$(CONFIG_CRYPTO_DEV_HISI_ZIP) += zip/ +obj-$(CONFIG_CRYPTO_DEV_HISI_TRNG) += trng/ +obj-$(CONFIG_CRYPTO_DEV_HISI_MIGRATION) += migration/ diff --git a/kae_driver/hisilicon/hisi_acc_qm.h b/kae_driver/hisilicon/hisi_acc_qm.h new file mode 100755 index 0000000..4ea4c23 --- /dev/null +++ b/kae_driver/hisilicon/hisi_acc_qm.h @@ -0,0 +1,516 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ +#ifndef HISI_ACC_QM_H +#define HISI_ACC_QM_H + +#include +#include +#include +#include +#include + +#define QM_QNUM_V1 4096 +#define QM_QNUM_V2 1024 +#define QM_MAX_VFS_NUM_V2 63 + +/* qm user domain */ +#define QM_ARUSER_M_CFG_1 0x100088 +#define AXUSER_SNOOP_ENABLE BIT(30) +#define AXUSER_CMD_TYPE GENMASK(14, 12) +#define AXUSER_CMD_SMMU_NORMAL 1 +#define AXUSER_NS BIT(6) +#define AXUSER_NO BIT(5) +#define AXUSER_FP BIT(4) +#define AXUSER_SSV BIT(0) +#define AXUSER_BASE (AXUSER_SNOOP_ENABLE | \ + FIELD_PREP(AXUSER_CMD_TYPE, \ + AXUSER_CMD_SMMU_NORMAL) | \ + AXUSER_NS | AXUSER_NO | AXUSER_FP) +#define QM_ARUSER_M_CFG_ENABLE 0x100090 +#define ARUSER_M_CFG_ENABLE 0xfffffffe +#define QM_AWUSER_M_CFG_1 0x100098 +#define QM_AWUSER_M_CFG_ENABLE 0x1000a0 +#define AWUSER_M_CFG_ENABLE 0xfffffffe +#define QM_WUSER_M_CFG_ENABLE 0x1000a8 +#define WUSER_M_CFG_ENABLE 0xffffffff + +/* mailbox */ +#define QM_MB_CMD_SQC 0x0 +#define QM_MB_CMD_CQC 0x1 +#define QM_MB_CMD_EQC 0x2 +#define QM_MB_CMD_AEQC 0x3 +#define QM_MB_CMD_SQC_BT 0x4 +#define QM_MB_CMD_CQC_BT 0x5 +#define QM_MB_CMD_SQC_VFT_V2 0x6 +#define QM_MB_CMD_STOP_QP 0x8 +#define QM_MB_CMD_SRC 0xc +#define QM_MB_CMD_DST 0xd + +#define QM_MB_CMD_SEND_BASE 0x300 +#define QM_MB_EVENT_SHIFT 8 +#define QM_MB_BUSY_SHIFT 13 +#define QM_MB_OP_SHIFT 14 +#define QM_MB_CMD_DATA_ADDR_L 0x304 +#define QM_MB_CMD_DATA_ADDR_H 0x308 +#define QM_MB_MAX_WAIT_CNT 6000 + +/* doorbell */ +#define QM_DOORBELL_CMD_SQ 0 +#define QM_DOORBELL_CMD_CQ 1 +#define QM_DOORBELL_CMD_EQ 2 +#define QM_DOORBELL_CMD_AEQ 3 + +#define QM_DOORBELL_SQ_CQ_BASE_V2 0x1000 +#define QM_DOORBELL_EQ_AEQ_BASE_V2 0x2000 +#define QM_QP_MAX_NUM_SHIFT 11 +#define QM_DB_CMD_SHIFT_V2 12 +#define QM_DB_RAND_SHIFT_V2 16 +#define QM_DB_INDEX_SHIFT_V2 32 +#define QM_DB_PRIORITY_SHIFT_V2 48 + +/* qm cache */ +#define QM_CACHE_CTL 0x100050 +#define SQC_CACHE_ENABLE BIT(0) +#define CQC_CACHE_ENABLE BIT(1) +#define SQC_CACHE_WB_ENABLE BIT(4) +#define SQC_CACHE_WB_THRD GENMASK(10, 5) +#define CQC_CACHE_WB_ENABLE BIT(11) +#define CQC_CACHE_WB_THRD GENMASK(17, 12) +#define QM_AXI_M_CFG 0x1000ac +#define AXI_M_CFG 0xffff +#define QM_AXI_M_CFG_ENABLE 0x1000b0 +#define AM_CFG_SINGLE_PORT_MAX_TRANS 0x300014 +#define AXI_M_CFG_ENABLE 0xffffffff +#define QM_PEH_AXUSER_CFG 0x1000cc +#define QM_PEH_AXUSER_CFG_ENABLE 0x1000d0 +#define PEH_AXUSER_CFG 0x401001 +#define PEH_AXUSER_CFG_ENABLE 0xffffffff + +#define QM_MIN_QNUM 2 +#define HISI_ACC_SGL_SGE_NR_MAX 255 +#define QM_SHAPER_CFG 0x100164 +#define QM_SHAPER_ENABLE BIT(30) +#define QM_SHAPER_TYPE1_OFFSET 10 +#define QM_VF_STATE 0x0060 + +/* page number for queue file region */ +#define QM_DOORBELL_PAGE_NR 1 + +/* uacce mode of the driver */ +#define UACCE_MODE_NOUACCE 0 /* don't use uacce */ +#define UACCE_MODE_SVA 1 /* use uacce sva mode */ +#define UACCE_MODE_DESC "0(default) means only register to crypto, 1 means both register to crypto and uacce" + +enum qm_stop_reason { + QM_NORMAL, + QM_SOFT_RESET, + QM_FLR, +}; + +enum qm_state { + QM_INIT = 0, + QM_START, + QM_CLOSE, + QM_STOP, +}; + +enum qp_state { + QP_INIT = 1, + QP_START, + QP_STOP, + QP_CLOSE, +}; + +enum qm_hw_ver { + QM_HW_UNKNOWN = -1, + QM_HW_V1 = 0x20, + QM_HW_V2 = 0x21, + QM_HW_V3 = 0x30, +}; + +enum qm_fun_type { + QM_HW_PF, + QM_HW_VF, +}; + +enum qm_debug_file { + CURRENT_QM, + CURRENT_Q, + CLEAR_ENABLE, + DEBUG_FILE_NUM, +}; + +enum vf_state { + VF_READY = 0x0, + VF_NOT_READY, + VF_PREPARE, +}; + +enum qm_cap_bits { + QM_SUPPORT_DB_ISOLATION = 0x0, + QM_SUPPORT_FUNC_QOS, + QM_SUPPORT_STOP_QP, + QM_SUPPORT_MB_COMMAND, + QM_SUPPORT_SVA_PREFETCH, + QM_SUPPORT_RPM, +}; + +struct dfx_diff_registers { + u32 *regs; + u32 reg_offset; + u32 reg_len; +}; + +struct qm_dfx { + atomic64_t err_irq_cnt; + atomic64_t aeq_irq_cnt; + atomic64_t abnormal_irq_cnt; + atomic64_t create_qp_err_cnt; + atomic64_t mb_err_cnt; +}; + +struct debugfs_file { + enum qm_debug_file index; + struct mutex lock; + struct qm_debug *debug; +}; + +struct qm_debug { + u32 curr_qm_qp_num; + u32 sqe_mask_offset; + u32 sqe_mask_len; + struct qm_dfx dfx; + struct dentry *debug_root; + struct dentry *qm_d; + struct debugfs_file files[DEBUG_FILE_NUM]; + unsigned int *qm_last_words; + /* ACC engines recoreding last regs */ + unsigned int *last_words; + struct dfx_diff_registers *qm_diff_regs; + struct dfx_diff_registers *acc_diff_regs; +}; + +struct qm_shaper_factor { + u32 func_qos; + u64 cir_b; + u64 cir_u; + u64 cir_s; + u64 cbs_s; +}; + +struct qm_dma { + void *va; + dma_addr_t dma; + size_t size; +}; + +struct hisi_qm_status { + u32 eq_head; + bool eqc_phase; + u32 aeq_head; + bool aeqc_phase; + atomic_t flags; + int stop_reason; +}; + +struct hisi_qm; + +struct hisi_qm_err_info { + char *acpi_rst; + u32 msi_wr_port; + u32 ecc_2bits_mask; + u32 qm_shutdown_mask; + u32 dev_shutdown_mask; + u32 qm_reset_mask; + u32 dev_reset_mask; + u32 ce; + u32 nfe; + u32 fe; +}; + +struct hisi_qm_err_status { + u32 is_qm_ecc_mbit; + u32 is_dev_ecc_mbit; +}; + +struct hisi_qm_err_ini { + int (*hw_init)(struct hisi_qm *qm); + void (*hw_err_enable)(struct hisi_qm *qm); + void (*hw_err_disable)(struct hisi_qm *qm); + u32 (*get_dev_hw_err_status)(struct hisi_qm *qm); + void (*clear_dev_hw_err_status)(struct hisi_qm *qm, u32 err_sts); + void (*open_axi_master_ooo)(struct hisi_qm *qm); + void (*close_axi_master_ooo)(struct hisi_qm *qm); + void (*open_sva_prefetch)(struct hisi_qm *qm); + void (*close_sva_prefetch)(struct hisi_qm *qm); + void (*log_dev_hw_err)(struct hisi_qm *qm, u32 err_sts); + void (*show_last_dfx_regs)(struct hisi_qm *qm); + void (*err_info_init)(struct hisi_qm *qm); +}; + +struct hisi_qm_cap_info { + u32 type; + /* Register offset */ + u32 offset; + /* Bit offset in register */ + u32 shift; + u32 mask; + u32 v1_val; + u32 v2_val; + u32 v3_val; +}; + +struct hisi_qm_list { + struct mutex lock; + struct list_head list; + int (*register_to_crypto)(struct hisi_qm *qm); + void (*unregister_from_crypto)(struct hisi_qm *qm); +}; + +struct hisi_qm_poll_data { + struct hisi_qm *qm; + struct work_struct work; + u16 *qp_finish_id; +}; + +struct hisi_qm { + enum qm_hw_ver ver; + enum qm_fun_type fun_type; + const char *dev_name; + struct pci_dev *pdev; + void __iomem *io_base; + void __iomem *db_io_base; + + /* Capbility version, 0: not supports */ + u32 cap_ver; + u32 sqe_size; + u32 qp_base; + u32 qp_num; + u32 qp_in_used; + u32 ctrl_qp_num; + u32 max_qp_num; + u32 vfs_num; + u32 db_interval; + u16 eq_depth; + u16 aeq_depth; + struct list_head list; + struct hisi_qm_list *qm_list; + + struct qm_dma qdma; + struct qm_sqc *sqc; + struct qm_cqc *cqc; + struct qm_eqe *eqe; + struct qm_aeqe *aeqe; + dma_addr_t sqc_dma; + dma_addr_t cqc_dma; + dma_addr_t eqe_dma; + dma_addr_t aeqe_dma; + + struct hisi_qm_status status; + const struct hisi_qm_err_ini *err_ini; + struct hisi_qm_err_info err_info; + struct hisi_qm_err_status err_status; + unsigned long misc_ctl; /* driver removing and reset sched */ + /* Device capability bit */ + unsigned long caps; + + struct rw_semaphore qps_lock; + struct idr qp_idr; + struct hisi_qp *qp_array; + struct hisi_qm_poll_data *poll_data; + + struct mutex mailbox_lock; + + const struct hisi_qm_hw_ops *ops; + + struct qm_debug debug; + + u32 error_mask; + + struct workqueue_struct *wq; + struct work_struct rst_work; + struct work_struct cmd_process; + + const char *algs; + bool use_sva; + bool is_frozen; + + resource_size_t phys_base; + resource_size_t db_phys_base; + struct uacce_device *uacce; + int mode; + struct qm_shaper_factor *factor; + u32 mb_qos; + u32 type_rate; +}; + +struct hisi_qp_status { + atomic_t used; + u16 sq_tail; + u16 cq_head; + bool cqc_phase; + atomic_t flags; +}; + +struct hisi_qp_ops { + int (*fill_sqe)(void *sqe, void *q_parm, void *d_parm); +}; + +struct hisi_qp { + u32 qp_id; + u16 sq_depth; + u16 cq_depth; + u8 alg_type; + u8 req_type; + + struct qm_dma qdma; + void *sqe; + struct qm_cqe *cqe; + dma_addr_t sqe_dma; + dma_addr_t cqe_dma; + + struct hisi_qp_status qp_status; + struct hisi_qp_ops *hw_ops; + void *qp_ctx; + void (*req_cb)(struct hisi_qp *qp, void *data); + void (*event_cb)(struct hisi_qp *qp); + + struct hisi_qm *qm; + bool is_resetting; + bool is_in_kernel; + u16 pasid; + struct uacce_queue *uacce_q; +}; + +static inline int q_num_set(const char *val, const struct kernel_param *kp, + unsigned int device) +{ + struct pci_dev *pdev = pci_get_device(PCI_VENDOR_ID_HUAWEI, + device, NULL); + u32 n, q_num; + int ret; + + if (!val) + return -EINVAL; + + if (!pdev) { + q_num = min_t(u32, QM_QNUM_V1, QM_QNUM_V2); + pr_info("No device found currently, suppose queue number is %u\n", + q_num); + } else { + if (pdev->revision == QM_HW_V1) + q_num = QM_QNUM_V1; + else + q_num = QM_QNUM_V2; + } + + ret = kstrtou32(val, 10, &n); + if (ret || n < QM_MIN_QNUM || n > q_num) + return -EINVAL; + + return param_set_int(val, kp); +} + +static inline int vfs_num_set(const char *val, const struct kernel_param *kp) +{ + u32 n; + int ret; + + if (!val) + return -EINVAL; + + ret = kstrtou32(val, 10, &n); + if (ret < 0) + return ret; + + if (n > QM_MAX_VFS_NUM_V2) + return -EINVAL; + + return param_set_int(val, kp); +} + +static inline int mode_set(const char *val, const struct kernel_param *kp) +{ + u32 n; + int ret; + + if (!val) + return -EINVAL; + + ret = kstrtou32(val, 10, &n); + if (ret != 0 || (n != UACCE_MODE_SVA && + n != UACCE_MODE_NOUACCE)) + return -EINVAL; + + return param_set_int(val, kp); +} + +static inline int uacce_mode_set(const char *val, const struct kernel_param *kp) +{ + return mode_set(val, kp); +} + +static inline void hisi_qm_init_list(struct hisi_qm_list *qm_list) +{ + INIT_LIST_HEAD(&qm_list->list); + mutex_init(&qm_list->lock); +} + +int hisi_qm_init(struct hisi_qm *qm); +void hisi_qm_uninit(struct hisi_qm *qm); +int hisi_qm_start(struct hisi_qm *qm); +int hisi_qm_stop(struct hisi_qm *qm, enum qm_stop_reason r); +int hisi_qm_start_qp(struct hisi_qp *qp, unsigned long arg); +int hisi_qm_stop_qp(struct hisi_qp *qp); +int hisi_qp_send(struct hisi_qp *qp, const void *msg); +void hisi_qm_debug_init(struct hisi_qm *qm); +void hisi_qm_debug_regs_clear(struct hisi_qm *qm); +int hisi_qm_sriov_enable(struct pci_dev *pdev, int max_vfs); +int hisi_qm_sriov_disable(struct pci_dev *pdev, bool is_frozen); +int hisi_qm_sriov_configure(struct pci_dev *pdev, int num_vfs); +void hisi_qm_dev_err_init(struct hisi_qm *qm); +void hisi_qm_dev_err_uninit(struct hisi_qm *qm); +int hisi_qm_diff_regs_init(struct hisi_qm *qm, + struct dfx_diff_registers *dregs, int reg_len); +void hisi_qm_diff_regs_uninit(struct hisi_qm *qm, int reg_len); +void hisi_qm_acc_diff_regs_dump(struct hisi_qm *qm, struct seq_file *s, + struct dfx_diff_registers *dregs, int regs_len); + +pci_ers_result_t hisi_qm_dev_err_detected(struct pci_dev *pdev, + pci_channel_state_t state); +pci_ers_result_t hisi_qm_dev_slot_reset(struct pci_dev *pdev); +void hisi_qm_reset_prepare(struct pci_dev *pdev); +void hisi_qm_reset_done(struct pci_dev *pdev); + +int hisi_qm_wait_mb_ready(struct hisi_qm *qm); +int hisi_qm_mb(struct hisi_qm *qm, u8 cmd, dma_addr_t dma_addr, u16 queue, + bool op); + +struct hisi_acc_sgl_pool; +struct hisi_acc_hw_sgl *hisi_acc_sg_buf_map_to_hw_sgl(struct device *dev, + struct scatterlist *sgl, struct hisi_acc_sgl_pool *pool, + u32 index, dma_addr_t *hw_sgl_dma); +void hisi_acc_sg_buf_unmap(struct device *dev, struct scatterlist *sgl, + struct hisi_acc_hw_sgl *hw_sgl); +struct hisi_acc_sgl_pool *hisi_acc_create_sgl_pool(struct device *dev, + u32 count, u32 sge_nr); +void hisi_acc_free_sgl_pool(struct device *dev, + struct hisi_acc_sgl_pool *pool); +int hisi_qm_alloc_qps_node(struct hisi_qm_list *qm_list, int qp_num, + u8 alg_type, int node, struct hisi_qp **qps); +void hisi_qm_free_qps(struct hisi_qp **qps, int qp_num); +void hisi_qm_dev_shutdown(struct pci_dev *pdev); +void hisi_qm_wait_task_finish(struct hisi_qm *qm, struct hisi_qm_list *qm_list); +int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list); +void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list); +int hisi_qm_resume(struct device *dev); +int hisi_qm_suspend(struct device *dev); +void hisi_qm_pm_uninit(struct hisi_qm *qm); +void hisi_qm_pm_init(struct hisi_qm *qm); +int hisi_qm_get_dfx_access(struct hisi_qm *qm); +void hisi_qm_put_dfx_access(struct hisi_qm *qm); +void hisi_qm_regs_dump(struct seq_file *s, struct debugfs_regset32 *regset); +u32 hisi_qm_get_hw_info(struct hisi_qm *qm, + const struct hisi_qm_cap_info *info_table, + u32 index, bool is_read); +#endif diff --git a/kae_driver/hisilicon/hisi_qm.h b/kae_driver/hisilicon/hisi_qm.h new file mode 100755 index 0000000..3e66dbc --- /dev/null +++ b/kae_driver/hisilicon/hisi_qm.h @@ -0,0 +1,39 @@ +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */ +#ifndef _UAPI_HISI_QM_H +#define _UAPI_HISI_QM_H + +#include + +/** + * struct hisi_qp_ctx - User data for hisi qp. + * @id: qp_index return to user space + * @qc_type: Accelerator algorithm type + */ +struct hisi_qp_ctx { + __u16 id; + __u16 qc_type; +}; + +/** + * struct hisi_qp_info - User data for hisi qp. + * @sqe_size: Submission queue element size + * @sq_depth: The number of sqe + * @cq_depth: The number of cqe + * @reserved: Reserved data + */ +struct hisi_qp_info { + __u32 sqe_size; + __u16 sq_depth; + __u16 cq_depth; + __u64 reserved; +}; + +#define HISI_QM_API_VER_BASE "hisi_qm_v1" +#define HISI_QM_API_VER2_BASE "hisi_qm_v2" +#define HISI_QM_API_VER3_BASE "hisi_qm_v3" + +/* UACCE_CMD_QM_SET_QP_CTX: Set qp algorithm type */ +#define UACCE_CMD_QM_SET_QP_CTX _IOWR('H', 10, struct hisi_qp_ctx) +/* UACCE_CMD_QM_SET_QP_INFO: Set qp depth and BD size */ +#define UACCE_CMD_QM_SET_QP_INFO _IOWR('H', 11, struct hisi_qp_info) +#endif diff --git a/kae_driver/hisilicon/hpre/Makefile b/kae_driver/hisilicon/hpre/Makefile new file mode 100755 index 0000000..4fd32b7 --- /dev/null +++ b/kae_driver/hisilicon/hpre/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_HPRE) += hisi_hpre.o +hisi_hpre-objs = hpre_main.o hpre_crypto.o diff --git a/kae_driver/hisilicon/hpre/hpre.h b/kae_driver/hisilicon/hpre/hpre.h new file mode 100755 index 0000000..21e780e --- /dev/null +++ b/kae_driver/hisilicon/hpre/hpre.h @@ -0,0 +1,107 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ +#ifndef __HISI_HPRE_H +#define __HISI_HPRE_H + +#include +#include "../hisi_acc_qm.h" + +#define HPRE_SQE_SIZE sizeof(struct hpre_sqe) +#define HPRE_PF_DEF_Q_NUM 64 +#define HPRE_PF_DEF_Q_BASE 0 + +/* + * type used in qm sqc DW6. + * 0 - Algorithm which has been supported in V2, like RSA, DH and so on; + * 1 - ECC algorithm in V3. + */ +#define HPRE_V2_ALG_TYPE 0 +#define HPRE_V3_ECC_ALG_TYPE 1 + +enum { + HPRE_CLUSTER0, + HPRE_CLUSTER1, + HPRE_CLUSTER2, + HPRE_CLUSTER3, + HPRE_CLUSTERS_NUM_MAX +}; + +enum hpre_ctrl_dbgfs_file { + HPRE_CLEAR_ENABLE, + HPRE_CLUSTER_CTRL, + HPRE_DEBUG_FILE_NUM, +}; + +enum hpre_dfx_dbgfs_file { + HPRE_SEND_CNT, + HPRE_RECV_CNT, + HPRE_SEND_FAIL_CNT, + HPRE_SEND_BUSY_CNT, + HPRE_OVER_THRHLD_CNT, + HPRE_OVERTIME_THRHLD, + HPRE_INVALID_REQ_CNT, + HPRE_DFX_FILE_NUM +}; + +#define HPRE_DEBUGFS_FILE_NUM (HPRE_DEBUG_FILE_NUM + HPRE_CLUSTERS_NUM_MAX - 1) + +struct hpre_debugfs_file { + int index; + enum hpre_ctrl_dbgfs_file type; + spinlock_t lock; + struct hpre_debug *debug; +}; + +struct hpre_dfx { + atomic64_t value; + enum hpre_dfx_dbgfs_file type; +}; + +/* + * One HPRE controller has one PF and multiple VFs, some global configurations + * which PF has need this structure. + * Just relevant for PF. + */ +struct hpre_debug { + struct hpre_dfx dfx[HPRE_DFX_FILE_NUM]; + struct hpre_debugfs_file files[HPRE_DEBUGFS_FILE_NUM]; +}; + +struct hpre { + struct hisi_qm qm; + struct hpre_debug debug; + unsigned long status; +}; + +enum hpre_alg_type { + HPRE_ALG_NC_NCRT = 0x0, + HPRE_ALG_NC_CRT = 0x1, + HPRE_ALG_KG_STD = 0x2, + HPRE_ALG_KG_CRT = 0x3, + HPRE_ALG_DH_G2 = 0x4, + HPRE_ALG_DH = 0x5, + HPRE_ALG_ECC_MUL = 0xD, + /* shared by x25519 and x448, but x448 is not supported now */ + HPRE_ALG_CURVE25519_MUL = 0x10, +}; + +struct hpre_sqe { + __le32 dw0; + __u8 task_len1; + __u8 task_len2; + __u8 mrttest_num; + __u8 resv1; + __le64 key; + __le64 in; + __le64 out; + __le16 tag; + __le16 resv2; +#define _HPRE_SQE_ALIGN_EXT 7 + __le32 rsvd1[_HPRE_SQE_ALIGN_EXT]; +}; + +struct hisi_qp *hpre_create_qp(u8 type); +int hpre_algs_register(struct hisi_qm *qm); +void hpre_algs_unregister(struct hisi_qm *qm); +bool hpre_check_alg_support(struct hisi_qm *qm, u32 alg); +#endif diff --git a/kae_driver/hisilicon/hpre/hpre_crypto.c b/kae_driver/hisilicon/hpre/hpre_crypto.c new file mode 100755 index 0000000..ef02dad --- /dev/null +++ b/kae_driver/hisilicon/hpre/hpre_crypto.c @@ -0,0 +1,2217 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "hpre.h" + +struct hpre_ctx; + +#define HPRE_CRYPTO_ALG_PRI 1000 +#define HPRE_ALIGN_SZ 64 +#define HPRE_BITS_2_BYTES_SHIFT 3 +#define HPRE_RSA_512BITS_KSZ 64 +#define HPRE_RSA_1536BITS_KSZ 192 +#define HPRE_CRT_PRMS 5 +#define HPRE_CRT_Q 2 +#define HPRE_CRT_P 3 +#define HPRE_CRT_INV 4 +#define HPRE_DH_G_FLAG 0x02 +#define HPRE_TRY_SEND_TIMES 100 +#define HPRE_INVLD_REQ_ID (-1) + +#define HPRE_SQE_ALG_BITS 5 +#define HPRE_SQE_DONE_SHIFT 30 +#define HPRE_DH_MAX_P_SZ 512 + +#define HPRE_DFX_SEC_TO_US 1000000 +#define HPRE_DFX_US_TO_NS 1000 + +/* due to nist p521 */ +#define HPRE_ECC_MAX_KSZ 66 + +/* size in bytes of the n prime */ +#define HPRE_ECC_NIST_P192_N_SIZE 24 +#define HPRE_ECC_NIST_P256_N_SIZE 32 +#define HPRE_ECC_NIST_P384_N_SIZE 48 + +/* size in bytes */ +#define HPRE_ECC_HW256_KSZ_B 32 +#define HPRE_ECC_HW384_KSZ_B 48 + +/* capability register mask of driver */ +#define HPRE_DRV_RSA_MASK_CAP BIT(0) +#define HPRE_DRV_DH_MASK_CAP BIT(1) +#define HPRE_DRV_ECDH_MASK_CAP BIT(2) +#define HPRE_DRV_X25519_MASK_CAP BIT(5) + +typedef void (*hpre_cb)(struct hpre_ctx *ctx, void *sqe); + +struct hpre_rsa_ctx { + /* low address: e--->n */ + char *pubkey; + dma_addr_t dma_pubkey; + + /* low address: d--->n */ + char *prikey; + dma_addr_t dma_prikey; + + /* low address: dq->dp->q->p->qinv */ + char *crt_prikey; + dma_addr_t dma_crt_prikey; + + struct crypto_akcipher *soft_tfm; +}; + +struct hpre_dh_ctx { + /* + * If base is g we compute the public key + * ya = g^xa mod p; [RFC2631 sec 2.1.1] + * else if base if the counterpart public key we + * compute the shared secret + * ZZ = yb^xa mod p; [RFC2631 sec 2.1.1] + * low address: d--->n, please refer to Hisilicon HPRE UM + */ + char *xa_p; + dma_addr_t dma_xa_p; + + char *g; /* m */ + dma_addr_t dma_g; +}; + +struct hpre_ecdh_ctx { + /* low address: p->a->k->b */ + unsigned char *p; + dma_addr_t dma_p; + + /* low address: x->y */ + unsigned char *g; + dma_addr_t dma_g; +}; + +struct hpre_curve25519_ctx { + /* low address: p->a->k */ + unsigned char *p; + dma_addr_t dma_p; + + /* gx coordinate */ + unsigned char *g; + dma_addr_t dma_g; +}; + +struct hpre_ctx { + struct hisi_qp *qp; + struct device *dev; + struct hpre_asym_request **req_list; + struct hpre *hpre; + spinlock_t req_lock; + unsigned int key_sz; + bool crt_g2_mode; + struct idr req_idr; + union { + struct hpre_rsa_ctx rsa; + struct hpre_dh_ctx dh; + struct hpre_ecdh_ctx ecdh; + struct hpre_curve25519_ctx curve25519; + }; + /* for ecc algorithms */ + unsigned int curve_id; +}; + +struct hpre_asym_request { + char *src; + char *dst; + struct hpre_sqe req; + struct hpre_ctx *ctx; + union { + struct akcipher_request *rsa; + struct kpp_request *dh; + struct kpp_request *ecdh; + struct kpp_request *curve25519; + } areq; + int err; + int req_id; + hpre_cb cb; + struct timespec64 req_time; +}; + +static int hpre_alloc_req_id(struct hpre_ctx *ctx) +{ + unsigned long flags; + int id; + + spin_lock_irqsave(&ctx->req_lock, flags); + id = idr_alloc(&ctx->req_idr, NULL, 0, ctx->qp->sq_depth, GFP_ATOMIC); + spin_unlock_irqrestore(&ctx->req_lock, flags); + + return id; +} + +static void hpre_free_req_id(struct hpre_ctx *ctx, int req_id) +{ + unsigned long flags; + + spin_lock_irqsave(&ctx->req_lock, flags); + idr_remove(&ctx->req_idr, req_id); + spin_unlock_irqrestore(&ctx->req_lock, flags); +} + +static int hpre_add_req_to_ctx(struct hpre_asym_request *hpre_req) +{ + struct hpre_ctx *ctx; + struct hpre_dfx *dfx; + int id; + + ctx = hpre_req->ctx; + id = hpre_alloc_req_id(ctx); + if (unlikely(id < 0)) + return -EINVAL; + + ctx->req_list[id] = hpre_req; + hpre_req->req_id = id; + + dfx = ctx->hpre->debug.dfx; + if (atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value)) + ktime_get_ts64(&hpre_req->req_time); + + return id; +} + +static void hpre_rm_req_from_ctx(struct hpre_asym_request *hpre_req) +{ + struct hpre_ctx *ctx = hpre_req->ctx; + int id = hpre_req->req_id; + + if (hpre_req->req_id >= 0) { + hpre_req->req_id = HPRE_INVLD_REQ_ID; + ctx->req_list[id] = NULL; + hpre_free_req_id(ctx, id); + } +} + +static struct hisi_qp *hpre_get_qp_and_start(u8 type) +{ + struct hisi_qp *qp; + int ret; + + qp = hpre_create_qp(type); + if (!qp) { + pr_err("Can not create hpre qp!\n"); + return ERR_PTR(-ENODEV); + } + + ret = hisi_qm_start_qp(qp, 0); + if (ret < 0) { + hisi_qm_free_qps(&qp, 1); + pci_err(qp->qm->pdev, "Can not start qp!\n"); + return ERR_PTR(-EINVAL); + } + + return qp; +} + +static int hpre_get_data_dma_addr(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, dma_addr_t *tmp) +{ + struct device *dev = hpre_req->ctx->dev; + enum dma_data_direction dma_dir; + + if (is_src) { + hpre_req->src = NULL; + dma_dir = DMA_TO_DEVICE; + } else { + hpre_req->dst = NULL; + dma_dir = DMA_FROM_DEVICE; + } + *tmp = dma_map_single(dev, sg_virt(data), len, dma_dir); + if (unlikely(dma_mapping_error(dev, *tmp))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + return 0; +} + +static int hpre_prepare_dma_buf(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, dma_addr_t *tmp) +{ + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + void *ptr; + int shift; + + shift = ctx->key_sz - len; + if (unlikely(shift < 0)) + return -EINVAL; + + ptr = dma_alloc_coherent(dev, ctx->key_sz, tmp, GFP_ATOMIC); + if (unlikely(!ptr)) + return -ENOMEM; + + if (is_src) { + scatterwalk_map_and_copy(ptr + shift, data, 0, len, 0); + hpre_req->src = ptr; + } else { + hpre_req->dst = ptr; + } + + return 0; +} + +static int hpre_hw_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, int is_dh) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + dma_addr_t tmp = 0; + int ret; + + /* when the data is dh's source, we should format it */ + if ((sg_is_last(data) && len == ctx->key_sz) && + ((is_dh && !is_src) || !is_dh)) + ret = hpre_get_data_dma_addr(hpre_req, data, len, is_src, &tmp); + else + ret = hpre_prepare_dma_buf(hpre_req, data, len, is_src, &tmp); + + if (unlikely(ret)) + return ret; + + if (is_src) + msg->in = cpu_to_le64(tmp); + else + msg->out = cpu_to_le64(tmp); + + return 0; +} + +static void hpre_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t tmp; + + tmp = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, tmp))) + return; + + if (src) { + if (req->src) + dma_free_coherent(dev, ctx->key_sz, req->src, tmp); + else + dma_unmap_single(dev, tmp, ctx->key_sz, DMA_TO_DEVICE); + } + + tmp = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, tmp))) + return; + + if (req->dst) { + if (dst) + scatterwalk_map_and_copy(req->dst, dst, 0, + ctx->key_sz, 1); + dma_free_coherent(dev, ctx->key_sz, req->dst, tmp); + } else { + dma_unmap_single(dev, tmp, ctx->key_sz, DMA_FROM_DEVICE); + } +} + +static int hpre_alg_res_post_hf(struct hpre_ctx *ctx, struct hpre_sqe *sqe, + void **kreq) +{ + struct hpre_asym_request *req; + unsigned int err, done, alg; + int id; + +#define HPRE_NO_HW_ERR 0 +#define HPRE_HW_TASK_DONE 3 +#define HREE_HW_ERR_MASK GENMASK(10, 0) +#define HREE_SQE_DONE_MASK GENMASK(1, 0) +#define HREE_ALG_TYPE_MASK GENMASK(4, 0) + id = (int)le16_to_cpu(sqe->tag); + req = ctx->req_list[id]; + hpre_rm_req_from_ctx(req); + *kreq = req; + + err = (le32_to_cpu(sqe->dw0) >> HPRE_SQE_ALG_BITS) & + HREE_HW_ERR_MASK; + + done = (le32_to_cpu(sqe->dw0) >> HPRE_SQE_DONE_SHIFT) & + HREE_SQE_DONE_MASK; + + if (likely(err == HPRE_NO_HW_ERR && done == HPRE_HW_TASK_DONE)) + return 0; + + alg = le32_to_cpu(sqe->dw0) & HREE_ALG_TYPE_MASK; + dev_err_ratelimited(ctx->dev, "alg[0x%x] error: done[0x%x], etype[0x%x]\n", + alg, done, err); + + return -EINVAL; +} + +static int hpre_ctx_set(struct hpre_ctx *ctx, struct hisi_qp *qp, int qlen) +{ + struct hpre *hpre; + + if (!ctx || !qp || qlen < 0) + return -EINVAL; + + spin_lock_init(&ctx->req_lock); + ctx->qp = qp; + ctx->dev = &qp->qm->pdev->dev; + + hpre = container_of(ctx->qp->qm, struct hpre, qm); + ctx->hpre = hpre; + ctx->req_list = kcalloc(qlen, sizeof(void *), GFP_KERNEL); + if (!ctx->req_list) + return -ENOMEM; + ctx->key_sz = 0; + ctx->crt_g2_mode = false; + idr_init(&ctx->req_idr); + + return 0; +} + +static void hpre_ctx_clear(struct hpre_ctx *ctx, bool is_clear_all) +{ + if (is_clear_all) { + idr_destroy(&ctx->req_idr); + kfree(ctx->req_list); + hisi_qm_free_qps(&ctx->qp, 1); + } + + ctx->crt_g2_mode = false; + ctx->key_sz = 0; +} + +static bool hpre_is_bd_timeout(struct hpre_asym_request *req, + u64 overtime_thrhld) +{ + struct timespec64 reply_time; + u64 time_use_us; + + ktime_get_ts64(&reply_time); + time_use_us = (reply_time.tv_sec - req->req_time.tv_sec) * + HPRE_DFX_SEC_TO_US + + (reply_time.tv_nsec - req->req_time.tv_nsec) / + HPRE_DFX_US_TO_NS; + + if (time_use_us <= overtime_thrhld) + return false; + + return true; +} + +static void hpre_dh_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req; + struct kpp_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.dh; + areq->dst_len = ctx->key_sz; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + hpre_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static void hpre_rsa_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req; + struct akcipher_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + areq = req->areq.rsa; + areq->dst_len = ctx->key_sz; + hpre_hw_data_clr_all(ctx, req, areq->dst, areq->src); + akcipher_request_complete(areq, ret); + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static void hpre_alg_cb(struct hisi_qp *qp, void *resp) +{ + struct hpre_ctx *ctx = qp->qp_ctx; + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_sqe *sqe = resp; + struct hpre_asym_request *req = ctx->req_list[le16_to_cpu(sqe->tag)]; + + if (unlikely(!req)) { + atomic64_inc(&dfx[HPRE_INVALID_REQ_CNT].value); + return; + } + + req->cb(ctx, resp); +} + +static void hpre_stop_qp_and_put(struct hisi_qp *qp) +{ + hisi_qm_stop_qp(qp); + hisi_qm_free_qps(&qp, 1); +} + +static int hpre_ctx_init(struct hpre_ctx *ctx, u8 type) +{ + struct hisi_qp *qp; + int ret; + + qp = hpre_get_qp_and_start(type); + if (IS_ERR(qp)) + return PTR_ERR(qp); + + qp->qp_ctx = ctx; + qp->req_cb = hpre_alg_cb; + + ret = hpre_ctx_set(ctx, qp, qp->sq_depth); + if (ret) + hpre_stop_qp_and_put(qp); + + return ret; +} + +static int hpre_msg_request_set(struct hpre_ctx *ctx, void *req, bool is_rsa) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (is_rsa) { + struct akcipher_request *akreq = req; + + if (akreq->dst_len < ctx->key_sz) { + akreq->dst_len = ctx->key_sz; + return -EOVERFLOW; + } + + tmp = akcipher_request_ctx(akreq); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_rsa_cb; + h_req->areq.rsa = akreq; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + } else { + struct kpp_request *kreq = req; + + if (kreq->dst_len < ctx->key_sz) { + kreq->dst_len = ctx->key_sz; + return -EOVERFLOW; + } + + tmp = kpp_request_ctx(kreq); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_dh_cb; + h_req->areq.dh = kreq; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->key = cpu_to_le64(ctx->dh.dma_xa_p); + } + + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->dw0 |= cpu_to_le32(0x1 << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + + return 0; +} + +static int hpre_send(struct hpre_ctx *ctx, struct hpre_sqe *msg) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + int ctr = 0; + int ret; + + do { + atomic64_inc(&dfx[HPRE_SEND_CNT].value); + ret = hisi_qp_send(ctx->qp, msg); + if (ret != -EBUSY) + break; + atomic64_inc(&dfx[HPRE_SEND_BUSY_CNT].value); + } while (ctr++ < HPRE_TRY_SEND_TIMES); + + if (likely(!ret)) + return ret; + + if (ret != -EBUSY) + atomic64_inc(&dfx[HPRE_SEND_FAIL_CNT].value); + + return ret; +} + +static int hpre_dh_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_msg_request_set(ctx, req, false); + if (unlikely(ret)) + return ret; + + if (req->src) { + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 1); + if (unlikely(ret)) + goto clear_all; + } else { + msg->in = cpu_to_le64(ctx->dh.dma_g); + } + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 1); + if (unlikely(ret)) + goto clear_all; + + if (ctx->crt_g2_mode && !req->src) + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH_G2); + else + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH); + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_is_dh_params_length_valid(unsigned int key_sz) +{ +#define _HPRE_DH_GRP1 768 +#define _HPRE_DH_GRP2 1024 +#define _HPRE_DH_GRP5 1536 +#define _HPRE_DH_GRP14 2048 +#define _HPRE_DH_GRP15 3072 +#define _HPRE_DH_GRP16 4096 + switch (key_sz) { + case _HPRE_DH_GRP1: + case _HPRE_DH_GRP2: + case _HPRE_DH_GRP5: + case _HPRE_DH_GRP14: + case _HPRE_DH_GRP15: + case _HPRE_DH_GRP16: + return 0; + default: + return -EINVAL; + } +} + +static int hpre_dh_set_params(struct hpre_ctx *ctx, struct dh *params) +{ + struct device *dev = ctx->dev; + unsigned int sz; + + if (params->p_size > HPRE_DH_MAX_P_SZ) + return -EINVAL; + + if (hpre_is_dh_params_length_valid(params->p_size << + HPRE_BITS_2_BYTES_SHIFT)) + return -EINVAL; + + sz = ctx->key_sz = params->p_size; + ctx->dh.xa_p = dma_alloc_coherent(dev, sz << 1, + &ctx->dh.dma_xa_p, GFP_KERNEL); + if (!ctx->dh.xa_p) + return -ENOMEM; + + memcpy(ctx->dh.xa_p + sz, params->p, sz); + + /* If g equals 2 don't copy it */ + if (params->g_size == 1 && *(char *)params->g == HPRE_DH_G_FLAG) { + ctx->crt_g2_mode = true; + return 0; + } + + ctx->dh.g = dma_alloc_coherent(dev, sz, &ctx->dh.dma_g, GFP_KERNEL); + if (!ctx->dh.g) { + dma_free_coherent(dev, sz << 1, ctx->dh.xa_p, + ctx->dh.dma_xa_p); + ctx->dh.xa_p = NULL; + return -ENOMEM; + } + + memcpy(ctx->dh.g + (sz - params->g_size), params->g, params->g_size); + + return 0; +} + +static void hpre_dh_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (ctx->dh.g) { + dma_free_coherent(dev, sz, ctx->dh.g, ctx->dh.dma_g); + ctx->dh.g = NULL; + } + + if (ctx->dh.xa_p) { + memzero_explicit(ctx->dh.xa_p, sz); + dma_free_coherent(dev, sz << 1, ctx->dh.xa_p, + ctx->dh.dma_xa_p); + ctx->dh.xa_p = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +static int hpre_dh_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct dh params; + int ret; + + if (crypto_dh_decode_key(buf, len, ¶ms) < 0) + return -EINVAL; + + /* Free old secret if any */ + hpre_dh_clear_ctx(ctx, false); + + ret = hpre_dh_set_params(ctx, ¶ms); + if (ret < 0) + goto err_clear_ctx; + + memcpy(ctx->dh.xa_p + (ctx->key_sz - params.key_size), params.key, + params.key_size); + + return 0; + +err_clear_ctx: + hpre_dh_clear_ctx(ctx, false); + return ret; +} + +static unsigned int hpre_dh_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return ctx->key_sz; +} + +static int hpre_dh_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return hpre_ctx_init(ctx, HPRE_V2_ALG_TYPE); +} + +static void hpre_dh_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_dh_clear_ctx(ctx, true); +} + +static void hpre_rsa_drop_leading_zeros(const char **ptr, size_t *len) +{ + while (!**ptr && *len) { + (*ptr)++; + (*len)--; + } +} + +static bool hpre_rsa_key_size_is_support(unsigned int len) +{ + unsigned int bits = len << HPRE_BITS_2_BYTES_SHIFT; + +#define _RSA_1024BITS_KEY_WDTH 1024 +#define _RSA_2048BITS_KEY_WDTH 2048 +#define _RSA_3072BITS_KEY_WDTH 3072 +#define _RSA_4096BITS_KEY_WDTH 4096 + + switch (bits) { + case _RSA_1024BITS_KEY_WDTH: + case _RSA_2048BITS_KEY_WDTH: + case _RSA_3072BITS_KEY_WDTH: + case _RSA_4096BITS_KEY_WDTH: + return true; + default: + return false; + } +} + +static int hpre_rsa_enc(struct akcipher_request *req) +{ + struct crypto_akcipher *tfm = crypto_akcipher_reqtfm(req); + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + void *tmp = akcipher_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) { + akcipher_request_set_tfm(req, ctx->rsa.soft_tfm); + ret = crypto_akcipher_encrypt(req); + akcipher_request_set_tfm(req, tfm); + return ret; + } + + if (unlikely(!ctx->rsa.pubkey)) + return -EINVAL; + + ret = hpre_msg_request_set(ctx, req, true); + if (unlikely(ret)) + return ret; + + msg->dw0 |= cpu_to_le32(HPRE_ALG_NC_NCRT); + msg->key = cpu_to_le64(ctx->rsa.dma_pubkey); + + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0); + if (unlikely(ret)) + goto clear_all; + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0); + if (unlikely(ret)) + goto clear_all; + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_rsa_dec(struct akcipher_request *req) +{ + struct crypto_akcipher *tfm = crypto_akcipher_reqtfm(req); + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + void *tmp = akcipher_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) { + akcipher_request_set_tfm(req, ctx->rsa.soft_tfm); + ret = crypto_akcipher_decrypt(req); + akcipher_request_set_tfm(req, tfm); + return ret; + } + + if (unlikely(!ctx->rsa.prikey)) + return -EINVAL; + + ret = hpre_msg_request_set(ctx, req, true); + if (unlikely(ret)) + return ret; + + if (ctx->crt_g2_mode) { + msg->key = cpu_to_le64(ctx->rsa.dma_crt_prikey); + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | + HPRE_ALG_NC_CRT); + } else { + msg->key = cpu_to_le64(ctx->rsa.dma_prikey); + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | + HPRE_ALG_NC_NCRT); + } + + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0); + if (unlikely(ret)) + goto clear_all; + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0); + if (unlikely(ret)) + goto clear_all; + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_rsa_set_n(struct hpre_ctx *ctx, const char *value, + size_t vlen, bool private) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + ctx->key_sz = vlen; + + /* if invalid key size provided, we use software tfm */ + if (!hpre_rsa_key_size_is_support(ctx->key_sz)) + return 0; + + ctx->rsa.pubkey = dma_alloc_coherent(ctx->dev, vlen << 1, + &ctx->rsa.dma_pubkey, + GFP_KERNEL); + if (!ctx->rsa.pubkey) + return -ENOMEM; + + if (private) { + ctx->rsa.prikey = dma_alloc_coherent(ctx->dev, vlen << 1, + &ctx->rsa.dma_prikey, + GFP_KERNEL); + if (!ctx->rsa.prikey) { + dma_free_coherent(ctx->dev, vlen << 1, + ctx->rsa.pubkey, + ctx->rsa.dma_pubkey); + ctx->rsa.pubkey = NULL; + return -ENOMEM; + } + memcpy(ctx->rsa.prikey + vlen, ptr, vlen); + } + memcpy(ctx->rsa.pubkey + vlen, ptr, vlen); + + /* Using hardware HPRE to do RSA */ + return 1; +} + +static int hpre_rsa_set_e(struct hpre_ctx *ctx, const char *value, + size_t vlen) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + if (!ctx->key_sz || !vlen || vlen > ctx->key_sz) + return -EINVAL; + + memcpy(ctx->rsa.pubkey + ctx->key_sz - vlen, ptr, vlen); + + return 0; +} + +static int hpre_rsa_set_d(struct hpre_ctx *ctx, const char *value, + size_t vlen) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + if (!ctx->key_sz || !vlen || vlen > ctx->key_sz) + return -EINVAL; + + memcpy(ctx->rsa.prikey + ctx->key_sz - vlen, ptr, vlen); + + return 0; +} + +static int hpre_crt_para_get(char *para, size_t para_sz, + const char *raw, size_t raw_sz) +{ + const char *ptr = raw; + size_t len = raw_sz; + + hpre_rsa_drop_leading_zeros(&ptr, &len); + if (!len || len > para_sz) + return -EINVAL; + + memcpy(para + para_sz - len, ptr, len); + + return 0; +} + +static int hpre_rsa_setkey_crt(struct hpre_ctx *ctx, struct rsa_key *rsa_key) +{ + unsigned int hlf_ksz = ctx->key_sz >> 1; + struct device *dev = ctx->dev; + u64 offset; + int ret; + + ctx->rsa.crt_prikey = dma_alloc_coherent(dev, hlf_ksz * HPRE_CRT_PRMS, + &ctx->rsa.dma_crt_prikey, + GFP_KERNEL); + if (!ctx->rsa.crt_prikey) + return -ENOMEM; + + ret = hpre_crt_para_get(ctx->rsa.crt_prikey, hlf_ksz, + rsa_key->dq, rsa_key->dq_sz); + if (ret) + goto free_key; + + offset = hlf_ksz; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->dp, rsa_key->dp_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_Q; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->q, rsa_key->q_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_P; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->p, rsa_key->p_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_INV; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->qinv, rsa_key->qinv_sz); + if (ret) + goto free_key; + + ctx->crt_g2_mode = true; + + return 0; + +free_key: + offset = hlf_ksz * HPRE_CRT_PRMS; + memzero_explicit(ctx->rsa.crt_prikey, offset); + dma_free_coherent(dev, hlf_ksz * HPRE_CRT_PRMS, ctx->rsa.crt_prikey, + ctx->rsa.dma_crt_prikey); + ctx->rsa.crt_prikey = NULL; + ctx->crt_g2_mode = false; + + return ret; +} + +/* If it is clear all, all the resources of the QP will be cleaned. */ +static void hpre_rsa_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all) +{ + unsigned int half_key_sz = ctx->key_sz >> 1; + struct device *dev = ctx->dev; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (ctx->rsa.pubkey) { + dma_free_coherent(dev, ctx->key_sz << 1, + ctx->rsa.pubkey, ctx->rsa.dma_pubkey); + ctx->rsa.pubkey = NULL; + } + + if (ctx->rsa.crt_prikey) { + memzero_explicit(ctx->rsa.crt_prikey, + half_key_sz * HPRE_CRT_PRMS); + dma_free_coherent(dev, half_key_sz * HPRE_CRT_PRMS, + ctx->rsa.crt_prikey, ctx->rsa.dma_crt_prikey); + ctx->rsa.crt_prikey = NULL; + } + + if (ctx->rsa.prikey) { + memzero_explicit(ctx->rsa.prikey, ctx->key_sz); + dma_free_coherent(dev, ctx->key_sz << 1, ctx->rsa.prikey, + ctx->rsa.dma_prikey); + ctx->rsa.prikey = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +/* + * we should judge if it is CRT or not, + * CRT: return true, N-CRT: return false . + */ +static bool hpre_is_crt_key(struct rsa_key *key) +{ + u16 len = key->p_sz + key->q_sz + key->dp_sz + key->dq_sz + + key->qinv_sz; + +#define LEN_OF_NCRT_PARA 5 + + /* N-CRT less than 5 parameters */ + return len > LEN_OF_NCRT_PARA; +} + +static int hpre_rsa_setkey(struct hpre_ctx *ctx, const void *key, + unsigned int keylen, bool private) +{ + struct rsa_key rsa_key; + int ret; + + hpre_rsa_clear_ctx(ctx, false); + + if (private) + ret = rsa_parse_priv_key(&rsa_key, key, keylen); + else + ret = rsa_parse_pub_key(&rsa_key, key, keylen); + if (ret < 0) + return ret; + + ret = hpre_rsa_set_n(ctx, rsa_key.n, rsa_key.n_sz, private); + if (ret <= 0) + return ret; + + if (private) { + ret = hpre_rsa_set_d(ctx, rsa_key.d, rsa_key.d_sz); + if (ret < 0) + goto free; + + if (hpre_is_crt_key(&rsa_key)) { + ret = hpre_rsa_setkey_crt(ctx, &rsa_key); + if (ret < 0) + goto free; + } + } + + ret = hpre_rsa_set_e(ctx, rsa_key.e, rsa_key.e_sz); + if (ret < 0) + goto free; + + if ((private && !ctx->rsa.prikey) || !ctx->rsa.pubkey) { + ret = -EINVAL; + goto free; + } + + return 0; + +free: + hpre_rsa_clear_ctx(ctx, false); + return ret; +} + +static int hpre_rsa_setpubkey(struct crypto_akcipher *tfm, const void *key, + unsigned int keylen) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ret = crypto_akcipher_set_pub_key(ctx->rsa.soft_tfm, key, keylen); + if (ret) + return ret; + + return hpre_rsa_setkey(ctx, key, keylen, false); +} + +static int hpre_rsa_setprivkey(struct crypto_akcipher *tfm, const void *key, + unsigned int keylen) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ret = crypto_akcipher_set_priv_key(ctx->rsa.soft_tfm, key, keylen); + if (ret) + return ret; + + return hpre_rsa_setkey(ctx, key, keylen, true); +} + +static unsigned int hpre_rsa_max_size(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) + return crypto_akcipher_maxsize(ctx->rsa.soft_tfm); + + return ctx->key_sz; +} + +static int hpre_rsa_init_tfm(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ctx->rsa.soft_tfm = crypto_alloc_akcipher("rsa-generic", 0, 0); + if (IS_ERR(ctx->rsa.soft_tfm)) { + pr_err("Can not alloc_akcipher!\n"); + return PTR_ERR(ctx->rsa.soft_tfm); + } + + ret = hpre_ctx_init(ctx, HPRE_V2_ALG_TYPE); + if (ret) + crypto_free_akcipher(ctx->rsa.soft_tfm); + + return ret; +} + +static void hpre_rsa_exit_tfm(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + + hpre_rsa_clear_ctx(ctx, true); + crypto_free_akcipher(ctx->rsa.soft_tfm); +} + +static void hpre_key_to_big_end(u8 *data, int len) +{ + int i, j; + + for (i = 0; i < len / 2; i++) { + j = len - i - 1; + swap(data[j], data[i]); + } +} + +static void hpre_ecc_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all, + bool is_ecdh) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + unsigned int shift = sz << 1; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (is_ecdh && ctx->ecdh.p) { + /* ecdh: p->a->k->b */ + memzero_explicit(ctx->ecdh.p + shift, sz); + dma_free_coherent(dev, sz << 3, ctx->ecdh.p, ctx->ecdh.dma_p); + ctx->ecdh.p = NULL; + } else if (!is_ecdh && ctx->curve25519.p) { + /* curve25519: p->a->k */ + memzero_explicit(ctx->curve25519.p + shift, sz); + dma_free_coherent(dev, sz << 2, ctx->curve25519.p, + ctx->curve25519.dma_p); + ctx->curve25519.p = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +/* + * The bits of 192/224/256/384/521 are supported by HPRE, + * and convert the bits like: + * bits<=256, bits=256; 256key_sz << 1; + unsigned int shiftb = ctx->key_sz << 2; + void *p = ctx->ecdh.p + ctx->key_sz - cur_sz; + void *a = ctx->ecdh.p + shifta - cur_sz; + void *b = ctx->ecdh.p + shiftb - cur_sz; + void *x = ctx->ecdh.g + ctx->key_sz - cur_sz; + void *y = ctx->ecdh.g + shifta - cur_sz; + const struct ecc_curve *curve = ecc_get_curve(ctx->curve_id); + char *n; + + if (unlikely(!curve)) + return -EINVAL; + + n = kzalloc(ctx->key_sz, GFP_KERNEL); + if (!n) + return -ENOMEM; + + fill_curve_param(p, curve->p, cur_sz, curve->g.ndigits); + fill_curve_param(a, curve->a, cur_sz, curve->g.ndigits); + fill_curve_param(b, curve->b, cur_sz, curve->g.ndigits); + fill_curve_param(x, curve->g.x, cur_sz, curve->g.ndigits); + fill_curve_param(y, curve->g.y, cur_sz, curve->g.ndigits); + fill_curve_param(n, curve->n, cur_sz, curve->g.ndigits); + + if (params->key_size == cur_sz && memcmp(params->key, n, cur_sz) >= 0) { + kfree(n); + return -EINVAL; + } + + kfree(n); + return 0; +} + +static unsigned int hpre_ecdh_get_curvesz(unsigned short id) +{ + switch (id) { + case ECC_CURVE_NIST_P192: + return HPRE_ECC_NIST_P192_N_SIZE; + case ECC_CURVE_NIST_P256: + return HPRE_ECC_NIST_P256_N_SIZE; + case ECC_CURVE_NIST_P384: + return HPRE_ECC_NIST_P384_N_SIZE; + default: + break; + } + + return 0; +} + +static int hpre_ecdh_set_param(struct hpre_ctx *ctx, struct ecdh *params) +{ + struct device *dev = ctx->dev; + unsigned int sz, shift, curve_sz; + int ret; + + ctx->key_sz = hpre_ecdh_supported_curve(ctx->curve_id); + if (!ctx->key_sz) + return -EINVAL; + + curve_sz = hpre_ecdh_get_curvesz(ctx->curve_id); + if (!curve_sz || params->key_size > curve_sz) + return -EINVAL; + + sz = ctx->key_sz; + + if (!ctx->ecdh.p) { + ctx->ecdh.p = dma_alloc_coherent(dev, sz << 3, &ctx->ecdh.dma_p, + GFP_KERNEL); + if (!ctx->ecdh.p) + return -ENOMEM; + } + + shift = sz << 2; + ctx->ecdh.g = ctx->ecdh.p + shift; + ctx->ecdh.dma_g = ctx->ecdh.dma_p + shift; + + ret = hpre_ecdh_fill_curve(ctx, params, curve_sz); + if (ret) { + dev_err(dev, "failed to fill curve_param, ret = %d!\n", ret); + dma_free_coherent(dev, sz << 3, ctx->ecdh.p, ctx->ecdh.dma_p); + ctx->ecdh.p = NULL; + return ret; + } + + return 0; +} + +static bool hpre_key_is_zero(char *key, unsigned short key_sz) +{ + int i; + + for (i = 0; i < key_sz; i++) + if (key[i]) + return false; + + return true; +} + +static int ecdh_gen_privkey(struct hpre_ctx *ctx, struct ecdh *params) +{ + struct device *dev = ctx->dev; + int ret; + + ret = crypto_get_default_rng(); + if (ret) { + dev_err(dev, "failed to get default rng, ret = %d!\n", ret); + return ret; + } + + ret = crypto_rng_get_bytes(crypto_default_rng, (u8 *)params->key, + params->key_size); + crypto_put_default_rng(); + if (ret) + dev_err(dev, "failed to get rng, ret = %d!\n", ret); + + return ret; +} + +static int hpre_ecdh_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + char key[HPRE_ECC_MAX_KSZ]; + unsigned int sz, sz_shift; + struct ecdh params; + int ret; + + if (crypto_ecdh_decode_key(buf, len, ¶ms) < 0) { + dev_err(dev, "failed to decode ecdh key!\n"); + return -EINVAL; + } + + /* Use stdrng to generate private key */ + if (!params.key || !params.key_size) { + params.key = key; + params.key_size = hpre_ecdh_get_curvesz(ctx->curve_id); + ret = ecdh_gen_privkey(ctx, ¶ms); + if (ret) + return ret; + } + + if (hpre_key_is_zero(params.key, params.key_size)) { + dev_err(dev, "Invalid hpre key!\n"); + return -EINVAL; + } + + hpre_ecc_clear_ctx(ctx, false, true); + + ret = hpre_ecdh_set_param(ctx, ¶ms); + if (ret < 0) { + dev_err(dev, "failed to set hpre param, ret = %d!\n", ret); + return ret; + } + + sz = ctx->key_sz; + sz_shift = (sz << 1) + sz - params.key_size; + memcpy(ctx->ecdh.p + sz_shift, params.key, params.key_size); + + return 0; +} + +static void hpre_ecdh_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t dma; + + dma = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (src && req->src) + dma_free_coherent(dev, ctx->key_sz << 2, req->src, dma); + + dma = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (req->dst) + dma_free_coherent(dev, ctx->key_sz << 1, req->dst, dma); + if (dst) + dma_unmap_single(dev, dma, ctx->key_sz << 1, DMA_FROM_DEVICE); +} + +static void hpre_ecdh_cb(struct hpre_ctx *ctx, void *resp) +{ + unsigned int curve_sz = hpre_ecdh_get_curvesz(ctx->curve_id); + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req = NULL; + struct kpp_request *areq; + u64 overtime_thrhld; + char *p; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.ecdh; + areq->dst_len = ctx->key_sz << 1; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + p = sg_virt(areq->dst); + memmove(p, p + ctx->key_sz - curve_sz, curve_sz); + memmove(p + curve_sz, p + areq->dst_len - curve_sz, curve_sz); + + hpre_ecdh_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static int hpre_ecdh_msg_request_set(struct hpre_ctx *ctx, + struct kpp_request *req) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (req->dst_len < ctx->key_sz << 1) { + req->dst_len = ctx->key_sz << 1; + return -EINVAL; + } + + tmp = kpp_request_ctx(req); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_ecdh_cb; + h_req->areq.ecdh = req; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->key = cpu_to_le64(ctx->ecdh.dma_p); + + msg->dw0 |= cpu_to_le32(0x1U << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + return 0; +} + +static int hpre_ecdh_src_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + unsigned int tmpshift; + dma_addr_t dma = 0; + void *ptr; + int shift; + + /* Src_data include gx and gy. */ + shift = ctx->key_sz - (len >> 1); + if (unlikely(shift < 0)) + return -EINVAL; + + ptr = dma_alloc_coherent(dev, ctx->key_sz << 2, &dma, GFP_KERNEL); + if (unlikely(!ptr)) + return -ENOMEM; + + tmpshift = ctx->key_sz << 1; + scatterwalk_map_and_copy(ptr + tmpshift, data, 0, len, 0); + memcpy(ptr + shift, ptr + tmpshift, len >> 1); + memcpy(ptr + ctx->key_sz + shift, ptr + tmpshift + (len >> 1), len >> 1); + + hpre_req->src = ptr; + msg->in = cpu_to_le64(dma); + return 0; +} + +static int hpre_ecdh_dst_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + dma_addr_t dma; + + if (unlikely(!data || !sg_is_last(data) || len != ctx->key_sz << 1)) { + dev_err(dev, "data or data length is illegal!\n"); + return -EINVAL; + } + + hpre_req->dst = NULL; + dma = dma_map_single(dev, sg_virt(data), len, DMA_FROM_DEVICE); + if (unlikely(dma_mapping_error(dev, dma))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + msg->out = cpu_to_le64(dma); + return 0; +} + +static int hpre_ecdh_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_ecdh_msg_request_set(ctx, req); + if (unlikely(ret)) { + dev_err(dev, "failed to set ecdh request, ret = %d!\n", ret); + return ret; + } + + if (req->src) { + ret = hpre_ecdh_src_data_init(hpre_req, req->src, req->src_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init src data, ret = %d!\n", ret); + goto clear_all; + } + } else { + msg->in = cpu_to_le64(ctx->ecdh.dma_g); + } + + ret = hpre_ecdh_dst_data_init(hpre_req, req->dst, req->dst_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init dst data, ret = %d!\n", ret); + goto clear_all; + } + + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_ECC_MUL); + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_ecdh_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + return ret; +} + +static unsigned int hpre_ecdh_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + /* max size is the pub_key_size, include x and y */ + return ctx->key_sz << 1; +} + +static int hpre_ecdh_nist_p192_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P192; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static int hpre_ecdh_nist_p256_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P256; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static int hpre_ecdh_nist_p384_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P384; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static void hpre_ecdh_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_ecc_clear_ctx(ctx, true, true); +} + +static void hpre_curve25519_fill_curve(struct hpre_ctx *ctx, const void *buf, + unsigned int len) +{ + u8 secret[CURVE25519_KEY_SIZE] = { 0 }; + unsigned int sz = ctx->key_sz; + const struct ecc_curve *curve; + unsigned int shift = sz << 1; + void *p; + + /* + * The key from 'buf' is in little-endian, we should preprocess it as + * the description in rfc7748: "k[0] &= 248, k[31] &= 127, k[31] |= 64", + * then convert it to big endian. Only in this way, the result can be + * the same as the software curve-25519 that exists in crypto. + */ + memcpy(secret, buf, len); + curve25519_clamp_secret(secret); + hpre_key_to_big_end(secret, CURVE25519_KEY_SIZE); + + p = ctx->curve25519.p + sz - len; + + curve = ecc_get_curve25519(); + + /* fill curve parameters */ + fill_curve_param(p, curve->p, len, curve->g.ndigits); + fill_curve_param(p + sz, curve->a, len, curve->g.ndigits); + memcpy(p + shift, secret, len); + fill_curve_param(p + shift + sz, curve->g.x, len, curve->g.ndigits); + memzero_explicit(secret, CURVE25519_KEY_SIZE); +} + +static int hpre_curve25519_set_param(struct hpre_ctx *ctx, const void *buf, + unsigned int len) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + unsigned int shift = sz << 1; + + /* p->a->k->gx */ + if (!ctx->curve25519.p) { + ctx->curve25519.p = dma_alloc_coherent(dev, sz << 2, + &ctx->curve25519.dma_p, + GFP_KERNEL); + if (!ctx->curve25519.p) + return -ENOMEM; + } + + ctx->curve25519.g = ctx->curve25519.p + shift + sz; + ctx->curve25519.dma_g = ctx->curve25519.dma_p + shift + sz; + + hpre_curve25519_fill_curve(ctx, buf, len); + + return 0; +} + +static int hpre_curve25519_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + int ret = -EINVAL; + + if (len != CURVE25519_KEY_SIZE || + !crypto_memneq(buf, curve25519_null_point, CURVE25519_KEY_SIZE)) { + dev_err(dev, "key is null or key len is not 32bytes!\n"); + return ret; + } + + /* Free old secret if any */ + hpre_ecc_clear_ctx(ctx, false, false); + + ctx->key_sz = CURVE25519_KEY_SIZE; + ret = hpre_curve25519_set_param(ctx, buf, CURVE25519_KEY_SIZE); + if (ret) { + dev_err(dev, "failed to set curve25519 param, ret = %d!\n", ret); + hpre_ecc_clear_ctx(ctx, false, false); + return ret; + } + + return 0; +} + +static void hpre_curve25519_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t dma; + + dma = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (src && req->src) + dma_free_coherent(dev, ctx->key_sz, req->src, dma); + + dma = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (req->dst) + dma_free_coherent(dev, ctx->key_sz, req->dst, dma); + if (dst) + dma_unmap_single(dev, dma, ctx->key_sz, DMA_FROM_DEVICE); +} + +static void hpre_curve25519_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req = NULL; + struct kpp_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.curve25519; + areq->dst_len = ctx->key_sz; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + hpre_key_to_big_end(sg_virt(areq->dst), CURVE25519_KEY_SIZE); + + hpre_curve25519_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static int hpre_curve25519_msg_request_set(struct hpre_ctx *ctx, + struct kpp_request *req) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (unlikely(req->dst_len < ctx->key_sz)) { + req->dst_len = ctx->key_sz; + return -EINVAL; + } + + tmp = kpp_request_ctx(req); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_curve25519_cb; + h_req->areq.curve25519 = req; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->key = cpu_to_le64(ctx->curve25519.dma_p); + + msg->dw0 |= cpu_to_le32(0x1U << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + return 0; +} + +static void hpre_curve25519_src_modulo_p(u8 *ptr) +{ + int i; + + for (i = 0; i < CURVE25519_KEY_SIZE - 1; i++) + ptr[i] = 0; + + /* The modulus is ptr's last byte minus '0xed'(last byte of p) */ + ptr[i] -= 0xed; +} + +static int hpre_curve25519_src_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + u8 p[CURVE25519_KEY_SIZE] = { 0 }; + const struct ecc_curve *curve; + dma_addr_t dma = 0; + u8 *ptr; + + if (len != CURVE25519_KEY_SIZE) { + dev_err(dev, "sourc_data len is not 32bytes, len = %u!\n", len); + return -EINVAL; + } + + ptr = dma_alloc_coherent(dev, ctx->key_sz, &dma, GFP_KERNEL); + if (unlikely(!ptr)) + return -ENOMEM; + + scatterwalk_map_and_copy(ptr, data, 0, len, 0); + + if (!crypto_memneq(ptr, curve25519_null_point, CURVE25519_KEY_SIZE)) { + dev_err(dev, "gx is null!\n"); + goto err; + } + + /* + * Src_data(gx) is in little-endian order, MSB in the final byte should + * be masked as described in RFC7748, then transform it to big-endian + * form, then hisi_hpre can use the data. + */ + ptr[31] &= 0x7f; + hpre_key_to_big_end(ptr, CURVE25519_KEY_SIZE); + + curve = ecc_get_curve25519(); + + fill_curve_param(p, curve->p, CURVE25519_KEY_SIZE, curve->g.ndigits); + + /* + * When src_data equals (2^255 - 19) ~ (2^255 - 1), it is out of p, + * we get its modulus to p, and then use it. + */ + if (memcmp(ptr, p, ctx->key_sz) == 0) { + dev_err(dev, "gx is p!\n"); + goto err; + } else if (memcmp(ptr, p, ctx->key_sz) > 0) { + hpre_curve25519_src_modulo_p(ptr); + } + + hpre_req->src = ptr; + msg->in = cpu_to_le64(dma); + return 0; + +err: + dma_free_coherent(dev, ctx->key_sz, ptr, dma); + return -EINVAL; +} + +static int hpre_curve25519_dst_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + dma_addr_t dma; + + if (!data || !sg_is_last(data) || len != ctx->key_sz) { + dev_err(dev, "data or data length is illegal!\n"); + return -EINVAL; + } + + hpre_req->dst = NULL; + dma = dma_map_single(dev, sg_virt(data), len, DMA_FROM_DEVICE); + if (unlikely(dma_mapping_error(dev, dma))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + msg->out = cpu_to_le64(dma); + return 0; +} + +static int hpre_curve25519_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_curve25519_msg_request_set(ctx, req); + if (unlikely(ret)) { + dev_err(dev, "failed to set curve25519 request, ret = %d!\n", ret); + return ret; + } + + if (req->src) { + ret = hpre_curve25519_src_init(hpre_req, req->src, req->src_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init src data, ret = %d!\n", + ret); + goto clear_all; + } + } else { + msg->in = cpu_to_le64(ctx->curve25519.dma_g); + } + + ret = hpre_curve25519_dst_init(hpre_req, req->dst, req->dst_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init dst data, ret = %d!\n", ret); + goto clear_all; + } + + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_CURVE25519_MUL); + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_curve25519_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + return ret; +} + +static unsigned int hpre_curve25519_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return ctx->key_sz; +} + +static int hpre_curve25519_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static void hpre_curve25519_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_ecc_clear_ctx(ctx, true, false); +} + +static struct akcipher_alg rsa = { + .sign = hpre_rsa_dec, + .verify = hpre_rsa_enc, + .encrypt = hpre_rsa_enc, + .decrypt = hpre_rsa_dec, + .set_pub_key = hpre_rsa_setpubkey, + .set_priv_key = hpre_rsa_setprivkey, + .max_size = hpre_rsa_max_size, + .init = hpre_rsa_init_tfm, + .exit = hpre_rsa_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "rsa", + .cra_driver_name = "hpre-rsa", + .cra_module = THIS_MODULE, + }, +}; + +static struct kpp_alg dh = { + .set_secret = hpre_dh_set_secret, + .generate_public_key = hpre_dh_compute_value, + .compute_shared_secret = hpre_dh_compute_value, + .max_size = hpre_dh_max_size, + .init = hpre_dh_init_tfm, + .exit = hpre_dh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "dh", + .cra_driver_name = "hpre-dh", + .cra_module = THIS_MODULE, + }, +}; + +static struct kpp_alg ecdh_curves[] = { + { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p192_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p192", + .cra_driver_name = "hpre-ecdh-nist-p192", + .cra_module = THIS_MODULE, + }, + }, { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p256_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p256", + .cra_driver_name = "hpre-ecdh-nist-p256", + .cra_module = THIS_MODULE, + }, + }, { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p384_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p384", + .cra_driver_name = "hpre-ecdh-nist-p384", + .cra_module = THIS_MODULE, + }, + } +}; + +static struct kpp_alg curve25519_alg = { + .set_secret = hpre_curve25519_set_secret, + .generate_public_key = hpre_curve25519_compute_value, + .compute_shared_secret = hpre_curve25519_compute_value, + .max_size = hpre_curve25519_max_size, + .init = hpre_curve25519_init_tfm, + .exit = hpre_curve25519_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "curve25519", + .cra_driver_name = "hpre-curve25519", + .cra_module = THIS_MODULE, + }, +}; + +static int hpre_register_rsa(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_RSA_MASK_CAP)) + return 0; + + rsa.base.cra_flags = 0; + ret = crypto_register_akcipher(&rsa); + if (ret) + dev_err(&qm->pdev->dev, "failed to register rsa (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_rsa(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_RSA_MASK_CAP)) + return; + + crypto_unregister_akcipher(&rsa); +} + +static int hpre_register_dh(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_DH_MASK_CAP)) + return 0; + + ret = crypto_register_kpp(&dh); + if (ret) + dev_err(&qm->pdev->dev, "failed to register dh (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_dh(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_DH_MASK_CAP)) + return; + + crypto_unregister_kpp(&dh); +} + +static int hpre_register_ecdh(struct hisi_qm *qm) +{ + int ret, i; + + if (!hpre_check_alg_support(qm, HPRE_DRV_ECDH_MASK_CAP)) + return 0; + + for (i = 0; i < ARRAY_SIZE(ecdh_curves); i++) { + ret = crypto_register_kpp(&ecdh_curves[i]); + if (ret) { + dev_err(&qm->pdev->dev, "failed to register %s (%d)!\n", + ecdh_curves[i].base.cra_name, ret); + goto unreg_kpp; + } + } + + return 0; + +unreg_kpp: + for (--i; i >= 0; --i) + crypto_unregister_kpp(&ecdh_curves[i]); + + return ret; +} + +static void hpre_unregister_ecdh(struct hisi_qm *qm) +{ + int i; + + if (!hpre_check_alg_support(qm, HPRE_DRV_ECDH_MASK_CAP)) + return; + + for (i = ARRAY_SIZE(ecdh_curves) - 1; i >= 0; --i) + crypto_unregister_kpp(&ecdh_curves[i]); +} + +static int hpre_register_x25519(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_X25519_MASK_CAP)) + return 0; + + ret = crypto_register_kpp(&curve25519_alg); + if (ret) + dev_err(&qm->pdev->dev, "failed to register x25519 (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_x25519(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_X25519_MASK_CAP)) + return; + + crypto_unregister_kpp(&curve25519_alg); +} + +int hpre_algs_register(struct hisi_qm *qm) +{ + int ret; + + ret = hpre_register_rsa(qm); + if (ret) + return ret; + + ret = hpre_register_dh(qm); + if (ret) + goto unreg_rsa; + + ret = hpre_register_ecdh(qm); + if (ret) + goto unreg_dh; + + ret = hpre_register_x25519(qm); + if (ret) + goto unreg_ecdh; + + return ret; + +unreg_ecdh: + hpre_unregister_ecdh(qm); +unreg_dh: + hpre_unregister_dh(qm); +unreg_rsa: + hpre_unregister_rsa(qm); + return ret; +} + +void hpre_algs_unregister(struct hisi_qm *qm) +{ + hpre_unregister_x25519(qm); + hpre_unregister_ecdh(qm); + hpre_unregister_dh(qm); + hpre_unregister_rsa(qm); +} diff --git a/kae_driver/hisilicon/hpre/hpre_main.c b/kae_driver/hisilicon/hpre/hpre_main.c new file mode 100755 index 0000000..732b4c3 --- /dev/null +++ b/kae_driver/hisilicon/hpre/hpre_main.c @@ -0,0 +1,1525 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2018-2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "../../include_linux/uacce.h" +#include "hpre.h" + +#define HPRE_QM_ABNML_INT_MASK 0x100004 +#define HPRE_CTRL_CNT_CLR_CE_BIT BIT(0) +#define HPRE_COMM_CNT_CLR_CE 0x0 +#define HPRE_CTRL_CNT_CLR_CE 0x301000 +#define HPRE_FSM_MAX_CNT 0x301008 +#define HPRE_VFG_AXQOS 0x30100c +#define HPRE_VFG_AXCACHE 0x301010 +#define HPRE_RDCHN_INI_CFG 0x301014 +#define HPRE_AWUSR_FP_CFG 0x301018 +#define HPRE_BD_ENDIAN 0x301020 +#define HPRE_ECC_BYPASS 0x301024 +#define HPRE_RAS_WIDTH_CFG 0x301028 +#define HPRE_POISON_BYPASS 0x30102c +#define HPRE_BD_ARUSR_CFG 0x301030 +#define HPRE_BD_AWUSR_CFG 0x301034 +#define HPRE_TYPES_ENB 0x301038 +#define HPRE_RSA_ENB BIT(0) +#define HPRE_ECC_ENB BIT(1) +#define HPRE_DATA_RUSER_CFG 0x30103c +#define HPRE_DATA_WUSER_CFG 0x301040 +#define HPRE_INT_MASK 0x301400 +#define HPRE_INT_STATUS 0x301800 +#define HPRE_HAC_INT_MSK 0x301400 +#define HPRE_HAC_RAS_CE_ENB 0x301410 +#define HPRE_HAC_RAS_NFE_ENB 0x301414 +#define HPRE_HAC_RAS_FE_ENB 0x301418 +#define HPRE_HAC_INT_SET 0x301500 +#define HPRE_RNG_TIMEOUT_NUM 0x301A34 +#define HPRE_CORE_INT_ENABLE 0 +#define HPRE_CORE_INT_DISABLE GENMASK(21, 0) +#define HPRE_RDCHN_INI_ST 0x301a00 +#define HPRE_CLSTR_BASE 0x302000 +#define HPRE_CORE_EN_OFFSET 0x04 +#define HPRE_CORE_INI_CFG_OFFSET 0x20 +#define HPRE_CORE_INI_STATUS_OFFSET 0x80 +#define HPRE_CORE_HTBT_WARN_OFFSET 0x8c +#define HPRE_CORE_IS_SCHD_OFFSET 0x90 + +#define HPRE_RAS_CE_ENB 0x301410 +#define HPRE_RAS_NFE_ENB 0x301414 +#define HPRE_RAS_FE_ENB 0x301418 +#define HPRE_OOO_SHUTDOWN_SEL 0x301a3c +#define HPRE_HAC_RAS_FE_ENABLE 0 + +#define HPRE_CORE_ENB (HPRE_CLSTR_BASE + HPRE_CORE_EN_OFFSET) +#define HPRE_CORE_INI_CFG (HPRE_CLSTR_BASE + HPRE_CORE_INI_CFG_OFFSET) +#define HPRE_CORE_INI_STATUS (HPRE_CLSTR_BASE + HPRE_CORE_INI_STATUS_OFFSET) +#define HPRE_HAC_ECC1_CNT 0x301a04 +#define HPRE_HAC_ECC2_CNT 0x301a08 +#define HPRE_HAC_SOURCE_INT 0x301600 +#define HPRE_CLSTR_ADDR_INTRVL 0x1000 +#define HPRE_CLUSTER_INQURY 0x100 +#define HPRE_CLSTR_ADDR_INQRY_RSLT 0x104 +#define HPRE_TIMEOUT_ABNML_BIT 6 +#define HPRE_PASID_EN_BIT 9 +#define HPRE_REG_RD_INTVRL_US 10 +#define HPRE_REG_RD_TMOUT_US 1000 +#define HPRE_DBGFS_VAL_MAX_LEN 20 +#define PCI_DEVICE_ID_HUAWEI_HPRE_PF 0xa258 +#define PCI_DEVICE_ID_HUAWEI_HPRE_VF 0xa259 +#define HPRE_QM_USR_CFG_MASK GENMASK(31, 1) +#define HPRE_QM_AXI_CFG_MASK GENMASK(15, 0) +#define HPRE_QM_VFG_AX_MASK GENMASK(7, 0) +#define HPRE_BD_USR_MASK GENMASK(1, 0) +#define HPRE_PREFETCH_CFG 0x301130 +#define HPRE_SVA_PREFTCH_DFX 0x30115C +#define HPRE_PREFETCH_ENABLE (~(BIT(0) | BIT(30))) +#define HPRE_PREFETCH_DISABLE BIT(30) +#define HPRE_SVA_DISABLE_READY (BIT(4) | BIT(8)) + +/* clock gate */ +#define HPRE_CLKGATE_CTL 0x301a10 +#define HPRE_PEH_CFG_AUTO_GATE 0x301a2c +#define HPRE_CLUSTER_DYN_CTL 0x302010 +#define HPRE_CORE_SHB_CFG 0x302088 +#define HPRE_CLKGATE_CTL_EN BIT(0) +#define HPRE_PEH_CFG_AUTO_GATE_EN BIT(0) +#define HPRE_CLUSTER_DYN_CTL_EN BIT(0) +#define HPRE_CORE_GATE_EN (BIT(30) | BIT(31)) + +#define HPRE_AM_OOO_SHUTDOWN_ENB 0x301044 +#define HPRE_AM_OOO_SHUTDOWN_ENABLE BIT(0) +#define HPRE_WR_MSI_PORT BIT(2) + +#define HPRE_CORE_ECC_2BIT_ERR BIT(1) +#define HPRE_OOO_ECC_2BIT_ERR BIT(5) + +#define HPRE_QM_BME_FLR BIT(7) +#define HPRE_QM_PM_FLR BIT(11) +#define HPRE_QM_SRIOV_FLR BIT(12) + +#define HPRE_SHAPER_TYPE_RATE 640 +#define HPRE_VIA_MSI_DSM 1 +#define HPRE_SQE_MASK_OFFSET 8 +#define HPRE_SQE_MASK_LEN 24 + +#define HPRE_DFX_BASE 0x301000 +#define HPRE_DFX_COMMON1 0x301400 +#define HPRE_DFX_COMMON2 0x301A00 +#define HPRE_DFX_CORE 0x302000 +#define HPRE_DFX_BASE_LEN 0x55 +#define HPRE_DFX_COMMON1_LEN 0x41 +#define HPRE_DFX_COMMON2_LEN 0xE +#define HPRE_DFX_CORE_LEN 0x43 + +#define HPRE_DEV_ALG_MAX_LEN 256 + +static const char hpre_name[] = "hisi_hpre"; +static struct dentry *hpre_debugfs_root; +static const struct pci_device_id hpre_dev_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_HPRE_PF) }, + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_HPRE_VF) }, + { 0, } +}; + +MODULE_DEVICE_TABLE(pci, hpre_dev_ids); + +struct hpre_hw_error { + u32 int_msk; + const char *msg; +}; + +struct hpre_dev_alg { + u32 alg_msk; + const char *alg; +}; + +static const struct hpre_dev_alg hpre_dev_algs[] = { + { + .alg_msk = BIT(0), + .alg = "rsa\n" + }, { + .alg_msk = BIT(1), + .alg = "dh\n" + }, { + .alg_msk = BIT(2), + .alg = "ecdh\n" + }, { + .alg_msk = BIT(3), + .alg = "ecdsa\n" + }, { + .alg_msk = BIT(4), + .alg = "sm2\n" + }, { + .alg_msk = BIT(5), + .alg = "x25519\n" + }, { + .alg_msk = BIT(6), + .alg = "x448\n" + }, { + /* sentinel */ + } +}; + +static struct hisi_qm_list hpre_devices = { + .register_to_crypto = hpre_algs_register, + .unregister_from_crypto = hpre_algs_unregister, +}; + +static const char * const hpre_debug_file_name[] = { + [HPRE_CLEAR_ENABLE] = "rdclr_en", + [HPRE_CLUSTER_CTRL] = "cluster_ctrl", +}; + +enum hpre_cap_type { + HPRE_QM_NFE_MASK_CAP, + HPRE_QM_RESET_MASK_CAP, + HPRE_QM_OOO_SHUTDOWN_MASK_CAP, + HPRE_QM_CE_MASK_CAP, + HPRE_NFE_MASK_CAP, + HPRE_RESET_MASK_CAP, + HPRE_OOO_SHUTDOWN_MASK_CAP, + HPRE_CE_MASK_CAP, + HPRE_CLUSTER_NUM_CAP, + HPRE_CORE_TYPE_NUM_CAP, + HPRE_CORE_NUM_CAP, + HPRE_CLUSTER_CORE_NUM_CAP, + HPRE_CORE_ENABLE_BITMAP_CAP, + HPRE_DRV_ALG_BITMAP_CAP, + HPRE_DEV_ALG_BITMAP_CAP, + HPRE_CORE1_ALG_BITMAP_CAP, + HPRE_CORE2_ALG_BITMAP_CAP, + HPRE_CORE3_ALG_BITMAP_CAP, + HPRE_CORE4_ALG_BITMAP_CAP, + HPRE_CORE5_ALG_BITMAP_CAP, + HPRE_CORE6_ALG_BITMAP_CAP, + HPRE_CORE7_ALG_BITMAP_CAP, + HPRE_CORE8_ALG_BITMAP_CAP, + HPRE_CORE9_ALG_BITMAP_CAP, + HPRE_CORE10_ALG_BITMAP_CAP +}; + +static const struct hisi_qm_cap_info hpre_basic_info[] = { + {HPRE_QM_NFE_MASK_CAP, 0x3124, 0, GENMASK(31, 0), 0x0, 0x1C37, 0x7C37}, + {HPRE_QM_RESET_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0xC37, 0x6C37}, + {HPRE_QM_OOO_SHUTDOWN_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0x4, 0x6C37}, + {HPRE_QM_CE_MASK_CAP, 0x312C, 0, GENMASK(31, 0), 0x0, 0x8, 0x8}, + {HPRE_NFE_MASK_CAP, 0x3130, 0, GENMASK(31, 0), 0x0, 0x3FFFFE, 0xFFFFFE}, + {HPRE_RESET_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x3FFFFE, 0xBFFFFE}, + {HPRE_OOO_SHUTDOWN_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x22, 0xBFFFFE}, + {HPRE_CE_MASK_CAP, 0x3138, 0, GENMASK(31, 0), 0x0, 0x1, 0x1}, + {HPRE_CLUSTER_NUM_CAP, 0x313c, 20, GENMASK(3, 0), 0x0, 0x4, 0x1}, + {HPRE_CORE_TYPE_NUM_CAP, 0x313c, 16, GENMASK(3, 0), 0x0, 0x2, 0x2}, + {HPRE_CORE_NUM_CAP, 0x313c, 8, GENMASK(7, 0), 0x0, 0x8, 0xA}, + {HPRE_CLUSTER_CORE_NUM_CAP, 0x313c, 0, GENMASK(7, 0), 0x0, 0x2, 0xA}, + {HPRE_CORE_ENABLE_BITMAP_CAP, 0x3140, 0, GENMASK(31, 0), 0x0, 0xF, 0x3FF}, + {HPRE_DRV_ALG_BITMAP_CAP, 0x3144, 0, GENMASK(31, 0), 0x0, 0x03, 0x27}, + {HPRE_DEV_ALG_BITMAP_CAP, 0x3148, 0, GENMASK(31, 0), 0x0, 0x03, 0x7F}, + {HPRE_CORE1_ALG_BITMAP_CAP, 0x314c, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE2_ALG_BITMAP_CAP, 0x3150, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE3_ALG_BITMAP_CAP, 0x3154, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE4_ALG_BITMAP_CAP, 0x3158, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE5_ALG_BITMAP_CAP, 0x315c, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE6_ALG_BITMAP_CAP, 0x3160, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE7_ALG_BITMAP_CAP, 0x3164, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE8_ALG_BITMAP_CAP, 0x3168, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE9_ALG_BITMAP_CAP, 0x316c, 0, GENMASK(31, 0), 0x0, 0x10, 0x10}, + {HPRE_CORE10_ALG_BITMAP_CAP, 0x3170, 0, GENMASK(31, 0), 0x0, 0x10, 0x10} +}; + +static const struct hpre_hw_error hpre_hw_errors[] = { + { + .int_msk = BIT(0), + .msg = "core_ecc_1bit_err_int_set" + }, { + .int_msk = BIT(1), + .msg = "core_ecc_2bit_err_int_set" + }, { + .int_msk = BIT(2), + .msg = "dat_wb_poison_int_set" + }, { + .int_msk = BIT(3), + .msg = "dat_rd_poison_int_set" + }, { + .int_msk = BIT(4), + .msg = "bd_rd_poison_int_set" + }, { + .int_msk = BIT(5), + .msg = "ooo_ecc_2bit_err_int_set" + }, { + .int_msk = BIT(6), + .msg = "cluster1_shb_timeout_int_set" + }, { + .int_msk = BIT(7), + .msg = "cluster2_shb_timeout_int_set" + }, { + .int_msk = BIT(8), + .msg = "cluster3_shb_timeout_int_set" + }, { + .int_msk = BIT(9), + .msg = "cluster4_shb_timeout_int_set" + }, { + .int_msk = GENMASK(15, 10), + .msg = "ooo_rdrsp_err_int_set" + }, { + .int_msk = GENMASK(21, 16), + .msg = "ooo_wrrsp_err_int_set" + }, { + .int_msk = BIT(22), + .msg = "pt_rng_timeout_int_set" + }, { + .int_msk = BIT(23), + .msg = "sva_fsm_timeout_int_set" + }, { + /* sentinel */ + } +}; + +static const u64 hpre_cluster_offsets[] = { + [HPRE_CLUSTER0] = + HPRE_CLSTR_BASE + HPRE_CLUSTER0 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER1] = + HPRE_CLSTR_BASE + HPRE_CLUSTER1 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER2] = + HPRE_CLSTR_BASE + HPRE_CLUSTER2 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER3] = + HPRE_CLSTR_BASE + HPRE_CLUSTER3 * HPRE_CLSTR_ADDR_INTRVL, +}; + +static const struct debugfs_reg32 hpre_cluster_dfx_regs[] = { + {"CORES_EN_STATUS ", HPRE_CORE_EN_OFFSET}, + {"CORES_INI_CFG ", HPRE_CORE_INI_CFG_OFFSET}, + {"CORES_INI_STATUS ", HPRE_CORE_INI_STATUS_OFFSET}, + {"CORES_HTBT_WARN ", HPRE_CORE_HTBT_WARN_OFFSET}, + {"CORES_IS_SCHD ", HPRE_CORE_IS_SCHD_OFFSET}, +}; + +static const struct debugfs_reg32 hpre_com_dfx_regs[] = { + {"READ_CLR_EN ", HPRE_CTRL_CNT_CLR_CE}, + {"AXQOS ", HPRE_VFG_AXQOS}, + {"AWUSR_CFG ", HPRE_AWUSR_FP_CFG}, + {"BD_ENDIAN ", HPRE_BD_ENDIAN}, + {"ECC_CHECK_CTRL ", HPRE_ECC_BYPASS}, + {"RAS_INT_WIDTH ", HPRE_RAS_WIDTH_CFG}, + {"POISON_BYPASS ", HPRE_POISON_BYPASS}, + {"BD_ARUSER ", HPRE_BD_ARUSR_CFG}, + {"BD_AWUSER ", HPRE_BD_AWUSR_CFG}, + {"DATA_ARUSER ", HPRE_DATA_RUSER_CFG}, + {"DATA_AWUSER ", HPRE_DATA_WUSER_CFG}, + {"INT_STATUS ", HPRE_INT_STATUS}, + {"INT_MASK ", HPRE_HAC_INT_MSK}, + {"RAS_CE_ENB ", HPRE_HAC_RAS_CE_ENB}, + {"RAS_NFE_ENB ", HPRE_HAC_RAS_NFE_ENB}, + {"RAS_FE_ENB ", HPRE_HAC_RAS_FE_ENB}, + {"INT_SET ", HPRE_HAC_INT_SET}, + {"RNG_TIMEOUT_NUM ", HPRE_RNG_TIMEOUT_NUM}, +}; + +static const char *hpre_dfx_files[HPRE_DFX_FILE_NUM] = { + "send_cnt", + "recv_cnt", + "send_fail_cnt", + "send_busy_cnt", + "over_thrhld_cnt", + "overtime_thrhld", + "invalid_req_cnt" +}; + +/* define the HPRE's dfx regs region and region length */ +static struct dfx_diff_registers hpre_diff_regs[] = { + { + .reg_offset = HPRE_DFX_BASE, + .reg_len = HPRE_DFX_BASE_LEN, + }, { + .reg_offset = HPRE_DFX_COMMON1, + .reg_len = HPRE_DFX_COMMON1_LEN, + }, { + .reg_offset = HPRE_DFX_COMMON2, + .reg_len = HPRE_DFX_COMMON2_LEN, + }, { + .reg_offset = HPRE_DFX_CORE, + .reg_len = HPRE_DFX_CORE_LEN, + }, +}; + +bool hpre_check_alg_support(struct hisi_qm *qm, u32 alg) +{ + u32 cap_val; + + cap_val = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_DRV_ALG_BITMAP_CAP, qm->cap_ver); + if (alg & cap_val) + return true; + + return false; +} + +static int hpre_set_qm_algs(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + char *algs, *ptr; + u32 alg_msk; + int i; + + if (!qm->use_sva) + return 0; + + algs = devm_kzalloc(dev, HPRE_DEV_ALG_MAX_LEN * sizeof(char), GFP_KERNEL); + if (!algs) + return -ENOMEM; + + alg_msk = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_DEV_ALG_BITMAP_CAP, qm->cap_ver); + + for (i = 0; i < ARRAY_SIZE(hpre_dev_algs); i++) + if (alg_msk & hpre_dev_algs[i].alg_msk) + strcat(algs, hpre_dev_algs[i].alg); + + ptr = strrchr(algs, '\n'); + if (ptr) + *ptr = '\0'; + + qm->uacce->algs = algs; + + return 0; +} + +static int hpre_diff_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + + hisi_qm_acc_diff_regs_dump(qm, s, qm->debug.acc_diff_regs, + ARRAY_SIZE(hpre_diff_regs)); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_diff_regs); + +static int hpre_com_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_com_regs); + +static int hpre_cluster_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_cluster_regs); + +static const struct kernel_param_ops hpre_uacce_mode_ops = { + .set = uacce_mode_set, + .get = param_get_int, +}; + +/* + * uacce_mode = 0 means hpre only register to crypto, + * uacce_mode = 1 means hpre both register to crypto and uacce. + */ +static u32 uacce_mode = UACCE_MODE_NOUACCE; +module_param_cb(uacce_mode, &hpre_uacce_mode_ops, &uacce_mode, 0444); +MODULE_PARM_DESC(uacce_mode, UACCE_MODE_DESC); + +static int pf_q_num_set(const char *val, const struct kernel_param *kp) +{ + return q_num_set(val, kp, PCI_DEVICE_ID_HUAWEI_HPRE_PF); +} + +static const struct kernel_param_ops hpre_pf_q_num_ops = { + .set = pf_q_num_set, + .get = param_get_int, +}; + +static u32 pf_q_num = HPRE_PF_DEF_Q_NUM; +module_param_cb(pf_q_num, &hpre_pf_q_num_ops, &pf_q_num, 0444); +MODULE_PARM_DESC(pf_q_num, "Number of queues in PF of CS(2-1024)"); + +static const struct kernel_param_ops vfs_num_ops = { + .set = vfs_num_set, + .get = param_get_int, +}; + +static u32 vfs_num; +module_param_cb(vfs_num, &vfs_num_ops, &vfs_num, 0444); +MODULE_PARM_DESC(vfs_num, "Number of VFs to enable(1-63), 0(default)"); + +static inline int hpre_cluster_num(struct hisi_qm *qm) +{ + return hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CLUSTER_NUM_CAP, qm->cap_ver); +} + +static inline int hpre_cluster_core_mask(struct hisi_qm *qm) +{ + return hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CORE_ENABLE_BITMAP_CAP, qm->cap_ver); +} + +struct hisi_qp *hpre_create_qp(u8 type) +{ + int node = cpu_to_node(smp_processor_id()); + struct hisi_qp *qp = NULL; + int ret; + + if (type != HPRE_V2_ALG_TYPE && type != HPRE_V3_ECC_ALG_TYPE) + return NULL; + + /* + * type: 0 - RSA/DH. algorithm supported in V2, + * 1 - ECC algorithm in V3. + */ + ret = hisi_qm_alloc_qps_node(&hpre_devices, 1, type, node, &qp); + if (!ret) + return qp; + + return NULL; +} + +static void hpre_config_pasid(struct hisi_qm *qm) +{ + u32 val1, val2; + + if (qm->ver >= QM_HW_V3) + return; + + val1 = readl_relaxed(qm->io_base + HPRE_DATA_RUSER_CFG); + val2 = readl_relaxed(qm->io_base + HPRE_DATA_WUSER_CFG); + if (qm->use_sva) { + val1 |= BIT(HPRE_PASID_EN_BIT); + val2 |= BIT(HPRE_PASID_EN_BIT); + } else { + val1 &= ~BIT(HPRE_PASID_EN_BIT); + val2 &= ~BIT(HPRE_PASID_EN_BIT); + } + writel_relaxed(val1, qm->io_base + HPRE_DATA_RUSER_CFG); + writel_relaxed(val2, qm->io_base + HPRE_DATA_WUSER_CFG); +} + +static int hpre_cfg_by_dsm(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + union acpi_object *obj; + guid_t guid; + + if (guid_parse("b06b81ab-0134-4a45-9b0c-483447b95fa7", &guid)) { + dev_err(dev, "Hpre GUID failed\n"); + return -EINVAL; + } + + /* Switch over to MSI handling due to non-standard PCI implementation */ + obj = acpi_evaluate_dsm(ACPI_HANDLE(dev), &guid, + 0, HPRE_VIA_MSI_DSM, NULL); + if (!obj) { + dev_err(dev, "ACPI handle failed!\n"); + return -EIO; + } + + ACPI_FREE(obj); + + return 0; +} + +static int hpre_set_cluster(struct hisi_qm *qm) +{ + u32 cluster_core_mask = hpre_cluster_core_mask(qm); + u8 clusters_num = hpre_cluster_num(qm); + struct device *dev = &qm->pdev->dev; + unsigned long offset; + u32 val = 0; + int ret, i; + + for (i = 0; i < clusters_num; i++) { + offset = i * HPRE_CLSTR_ADDR_INTRVL; + + /* clusters initiating */ + writel(cluster_core_mask, + qm->io_base + offset + HPRE_CORE_ENB); + writel(0x1, qm->io_base + offset + HPRE_CORE_INI_CFG); + ret = readl_relaxed_poll_timeout(qm->io_base + offset + + HPRE_CORE_INI_STATUS, val, + ((val & cluster_core_mask) == + cluster_core_mask), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) { + dev_err(dev, + "cluster %d int st status timeout!\n", i); + return -ETIMEDOUT; + } + } + + return 0; +} + +/* + * For Kunpeng 920, we should disable FLR triggered by hardware (BME/PM/SRIOV). + * Or it may stay in D3 state when we bind and unbind hpre quickly, + * as it does FLR triggered by hardware. + */ +static void disable_flr_of_bme(struct hisi_qm *qm) +{ + u32 val; + + val = readl(qm->io_base + QM_PEH_AXUSER_CFG); + val &= ~(HPRE_QM_BME_FLR | HPRE_QM_SRIOV_FLR); + val |= HPRE_QM_PM_FLR; + writel(val, qm->io_base + QM_PEH_AXUSER_CFG); + writel(PEH_AXUSER_CFG_ENABLE, qm->io_base + QM_PEH_AXUSER_CFG_ENABLE); +} + +static void hpre_open_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + /* Enable prefetch */ + val = readl_relaxed(qm->io_base + HPRE_PREFETCH_CFG); + val &= HPRE_PREFETCH_ENABLE; + writel(val, qm->io_base + HPRE_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_PREFETCH_CFG, + val, !(val & HPRE_PREFETCH_DISABLE), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) + pci_err(qm->pdev, "failed to open sva prefetch\n"); +} + +static void hpre_close_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + val = readl_relaxed(qm->io_base + HPRE_PREFETCH_CFG); + val |= HPRE_PREFETCH_DISABLE; + writel(val, qm->io_base + HPRE_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_SVA_PREFTCH_DFX, + val, !(val & HPRE_SVA_DISABLE_READY), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) + pci_err(qm->pdev, "failed to close sva prefetch\n"); +} + +static void hpre_enable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + HPRE_CLKGATE_CTL); + val |= HPRE_CLKGATE_CTL_EN; + writel(val, qm->io_base + HPRE_CLKGATE_CTL); + + val = readl(qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + val |= HPRE_PEH_CFG_AUTO_GATE_EN; + writel(val, qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + + val = readl(qm->io_base + HPRE_CLUSTER_DYN_CTL); + val |= HPRE_CLUSTER_DYN_CTL_EN; + writel(val, qm->io_base + HPRE_CLUSTER_DYN_CTL); + + val = readl_relaxed(qm->io_base + HPRE_CORE_SHB_CFG); + val |= HPRE_CORE_GATE_EN; + writel(val, qm->io_base + HPRE_CORE_SHB_CFG); +} + +static void hpre_disable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + HPRE_CLKGATE_CTL); + val &= ~HPRE_CLKGATE_CTL_EN; + writel(val, qm->io_base + HPRE_CLKGATE_CTL); + + val = readl(qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + val &= ~HPRE_PEH_CFG_AUTO_GATE_EN; + writel(val, qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + + val = readl(qm->io_base + HPRE_CLUSTER_DYN_CTL); + val &= ~HPRE_CLUSTER_DYN_CTL_EN; + writel(val, qm->io_base + HPRE_CLUSTER_DYN_CTL); + + val = readl_relaxed(qm->io_base + HPRE_CORE_SHB_CFG); + val &= ~HPRE_CORE_GATE_EN; + writel(val, qm->io_base + HPRE_CORE_SHB_CFG); +} + +static int hpre_set_user_domain_and_cache(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u32 val; + int ret; + + /* disabel dynamic clock gate before sram init */ + hpre_disable_clock_gate(qm); + + writel(HPRE_QM_USR_CFG_MASK, qm->io_base + QM_ARUSER_M_CFG_ENABLE); + writel(HPRE_QM_USR_CFG_MASK, qm->io_base + QM_AWUSER_M_CFG_ENABLE); + writel_relaxed(HPRE_QM_AXI_CFG_MASK, qm->io_base + QM_AXI_M_CFG); + + /* HPRE need more time, we close this interrupt */ + val = readl_relaxed(qm->io_base + HPRE_QM_ABNML_INT_MASK); + val |= BIT(HPRE_TIMEOUT_ABNML_BIT); + writel_relaxed(val, qm->io_base + HPRE_QM_ABNML_INT_MASK); + + if (qm->ver >= QM_HW_V3) + writel(HPRE_RSA_ENB | HPRE_ECC_ENB, + qm->io_base + HPRE_TYPES_ENB); + else + writel(HPRE_RSA_ENB, qm->io_base + HPRE_TYPES_ENB); + + writel(HPRE_QM_VFG_AX_MASK, qm->io_base + HPRE_VFG_AXCACHE); + writel(0x0, qm->io_base + HPRE_BD_ENDIAN); + writel(0x0, qm->io_base + HPRE_INT_MASK); + writel(0x0, qm->io_base + HPRE_POISON_BYPASS); + writel(0x0, qm->io_base + HPRE_COMM_CNT_CLR_CE); + writel(0x0, qm->io_base + HPRE_ECC_BYPASS); + + writel(HPRE_BD_USR_MASK, qm->io_base + HPRE_BD_ARUSR_CFG); + writel(HPRE_BD_USR_MASK, qm->io_base + HPRE_BD_AWUSR_CFG); + writel(0x1, qm->io_base + HPRE_RDCHN_INI_CFG); + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_RDCHN_INI_ST, val, + val & BIT(0), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) { + dev_err(dev, "read rd channel timeout fail!\n"); + return -ETIMEDOUT; + } + + ret = hpre_set_cluster(qm); + if (ret) + return -ETIMEDOUT; + + /* This setting is only needed by Kunpeng 920. */ + if (qm->ver == QM_HW_V2) { + ret = hpre_cfg_by_dsm(qm); + if (ret) + return ret; + + disable_flr_of_bme(qm); + } + + /* Config data buffer pasid needed by Kunpeng 920 */ + hpre_config_pasid(qm); + + hpre_enable_clock_gate(qm); + + return ret; +} + +static void hpre_cnt_regs_clear(struct hisi_qm *qm) +{ + u8 clusters_num = hpre_cluster_num(qm); + unsigned long offset; + int i; + + /* clear clusterX/cluster_ctrl */ + for (i = 0; i < clusters_num; i++) { + offset = HPRE_CLSTR_BASE + i * HPRE_CLSTR_ADDR_INTRVL; + writel(0x0, qm->io_base + offset + HPRE_CLUSTER_INQURY); + } + + /* clear rdclr_en */ + writel(0x0, qm->io_base + HPRE_CTRL_CNT_CLR_CE); + + hisi_qm_debug_regs_clear(qm); +} + +static void hpre_master_ooo_ctrl(struct hisi_qm *qm, bool enable) +{ + u32 val1, val2; + + val1 = readl(qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + if (enable) { + val1 |= HPRE_AM_OOO_SHUTDOWN_ENABLE; + val2 = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + } else { + val1 &= ~HPRE_AM_OOO_SHUTDOWN_ENABLE; + val2 = 0x0; + } + + if (qm->ver > QM_HW_V2) + writel(val2, qm->io_base + HPRE_OOO_SHUTDOWN_SEL); + + writel(val1, qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); +} + +static void hpre_hw_error_disable(struct hisi_qm *qm) +{ + u32 ce, nfe; + + ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CE_MASK_CAP, qm->cap_ver); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + + /* disable hpre hw error interrupts */ + writel(ce | nfe | HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_INT_MASK); + /* disable HPRE block master OOO when nfe occurs on Kunpeng930 */ + hpre_master_ooo_ctrl(qm, false); +} + +static void hpre_hw_error_enable(struct hisi_qm *qm) +{ + u32 ce, nfe; + + ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CE_MASK_CAP, qm->cap_ver); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + + /* clear HPRE hw error source if having */ + writel(ce | nfe | HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_HAC_SOURCE_INT); + + /* configure error type */ + writel(ce, qm->io_base + HPRE_RAS_CE_ENB); + writel(nfe, qm->io_base + HPRE_RAS_NFE_ENB); + writel(HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_RAS_FE_ENB); + + /* enable HPRE block master OOO when nfe occurs on Kunpeng930 */ + hpre_master_ooo_ctrl(qm, true); + + /* enable hpre hw error interrupts */ + writel(HPRE_CORE_INT_ENABLE, qm->io_base + HPRE_INT_MASK); +} + +static inline struct hisi_qm *hpre_file_to_qm(struct hpre_debugfs_file *file) +{ + struct hpre *hpre = container_of(file->debug, struct hpre, debug); + + return &hpre->qm; +} + +static u32 hpre_clear_enable_read(struct hpre_debugfs_file *file) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + + return readl(qm->io_base + HPRE_CTRL_CNT_CLR_CE) & + HPRE_CTRL_CNT_CLR_CE_BIT; +} + +static int hpre_clear_enable_write(struct hpre_debugfs_file *file, u32 val) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + u32 tmp; + + if (val != 1 && val != 0) + return -EINVAL; + + tmp = (readl(qm->io_base + HPRE_CTRL_CNT_CLR_CE) & + ~HPRE_CTRL_CNT_CLR_CE_BIT) | val; + writel(tmp, qm->io_base + HPRE_CTRL_CNT_CLR_CE); + + return 0; +} + +static u32 hpre_cluster_inqry_read(struct hpre_debugfs_file *file) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + int cluster_index = file->index - HPRE_CLUSTER_CTRL; + unsigned long offset = HPRE_CLSTR_BASE + + cluster_index * HPRE_CLSTR_ADDR_INTRVL; + + return readl(qm->io_base + offset + HPRE_CLSTR_ADDR_INQRY_RSLT); +} + +static void hpre_cluster_inqry_write(struct hpre_debugfs_file *file, u32 val) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + int cluster_index = file->index - HPRE_CLUSTER_CTRL; + unsigned long offset = HPRE_CLSTR_BASE + cluster_index * + HPRE_CLSTR_ADDR_INTRVL; + + writel(val, qm->io_base + offset + HPRE_CLUSTER_INQURY); +} + +static ssize_t hpre_ctrl_debug_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct hpre_debugfs_file *file = filp->private_data; + struct hisi_qm *qm = hpre_file_to_qm(file); + char tbuf[HPRE_DBGFS_VAL_MAX_LEN]; + u32 val; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->type) { + case HPRE_CLEAR_ENABLE: + val = hpre_clear_enable_read(file); + break; + case HPRE_CLUSTER_CTRL: + val = hpre_cluster_inqry_read(file); + break; + default: + goto err_input; + } + spin_unlock_irq(&file->lock); + + hisi_qm_put_dfx_access(qm); + ret = snprintf(tbuf, HPRE_DBGFS_VAL_MAX_LEN, "%u\n", val); + return simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return -EINVAL; +} + +static ssize_t hpre_ctrl_debug_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct hpre_debugfs_file *file = filp->private_data; + struct hisi_qm *qm = hpre_file_to_qm(file); + char tbuf[HPRE_DBGFS_VAL_MAX_LEN]; + unsigned long val; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= HPRE_DBGFS_VAL_MAX_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, HPRE_DBGFS_VAL_MAX_LEN - 1, + pos, buf, count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + if (kstrtoul(tbuf, 0, &val)) + return -EFAULT; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->type) { + case HPRE_CLEAR_ENABLE: + ret = hpre_clear_enable_write(file, val); + if (ret) + goto err_input; + break; + case HPRE_CLUSTER_CTRL: + hpre_cluster_inqry_write(file, val); + break; + default: + ret = -EINVAL; + goto err_input; + } + + ret = count; + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return ret; +} + +static const struct file_operations hpre_ctrl_debug_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = hpre_ctrl_debug_read, + .write = hpre_ctrl_debug_write, +}; + +static int hpre_debugfs_atomic64_get(void *data, u64 *val) +{ + struct hpre_dfx *dfx_item = data; + + *val = atomic64_read(&dfx_item->value); + + return 0; +} + +static int hpre_debugfs_atomic64_set(void *data, u64 val) +{ + struct hpre_dfx *dfx_item = data; + struct hpre_dfx *hpre_dfx = NULL; + + if (dfx_item->type == HPRE_OVERTIME_THRHLD) { + hpre_dfx = dfx_item - HPRE_OVERTIME_THRHLD; + atomic64_set(&hpre_dfx[HPRE_OVER_THRHLD_CNT].value, 0); + } else if (val) { + return -EINVAL; + } + + atomic64_set(&dfx_item->value, val); + + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(hpre_atomic64_ops, hpre_debugfs_atomic64_get, + hpre_debugfs_atomic64_set, "%llu\n"); + +static int hpre_create_debugfs_file(struct hisi_qm *qm, struct dentry *dir, + enum hpre_ctrl_dbgfs_file type, int indx) +{ + struct hpre *hpre = container_of(qm, struct hpre, qm); + struct hpre_debug *dbg = &hpre->debug; + struct dentry *file_dir; + + if (dir) + file_dir = dir; + else + file_dir = qm->debug.debug_root; + + if (type >= HPRE_DEBUG_FILE_NUM) + return -EINVAL; + + spin_lock_init(&dbg->files[indx].lock); + dbg->files[indx].debug = dbg; + dbg->files[indx].type = type; + dbg->files[indx].index = indx; + debugfs_create_file(hpre_debug_file_name[type], 0600, file_dir, + dbg->files + indx, &hpre_ctrl_debug_fops); + + return 0; +} + +static int hpre_pf_comm_regs_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct debugfs_regset32 *regset; + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOMEM; + + regset->regs = hpre_com_dfx_regs; + regset->nregs = ARRAY_SIZE(hpre_com_dfx_regs); + regset->base = qm->io_base; + regset->dev = dev; + + debugfs_create_file("regs", 0444, qm->debug.debug_root, + regset, &hpre_com_regs_fops); + + return 0; +} + +static int hpre_cluster_debugfs_init(struct hisi_qm *qm) +{ + u8 clusters_num = hpre_cluster_num(qm); + struct device *dev = &qm->pdev->dev; + char buf[HPRE_DBGFS_VAL_MAX_LEN]; + struct debugfs_regset32 *regset; + struct dentry *tmp_d; + int i, ret; + + for (i = 0; i < clusters_num; i++) { + ret = snprintf(buf, HPRE_DBGFS_VAL_MAX_LEN, "cluster%d", i); + if (ret < 0) + return -EINVAL; + tmp_d = debugfs_create_dir(buf, qm->debug.debug_root); + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOMEM; + + regset->regs = hpre_cluster_dfx_regs; + regset->nregs = ARRAY_SIZE(hpre_cluster_dfx_regs); + regset->base = qm->io_base + hpre_cluster_offsets[i]; + regset->dev = dev; + + debugfs_create_file("regs", 0444, tmp_d, regset, + &hpre_cluster_regs_fops); + ret = hpre_create_debugfs_file(qm, tmp_d, HPRE_CLUSTER_CTRL, + i + HPRE_CLUSTER_CTRL); + if (ret) + return ret; + } + + return 0; +} + +static int hpre_ctrl_debug_init(struct hisi_qm *qm) +{ + int ret; + + ret = hpre_create_debugfs_file(qm, NULL, HPRE_CLEAR_ENABLE, + HPRE_CLEAR_ENABLE); + if (ret) + return ret; + + ret = hpre_pf_comm_regs_debugfs_init(qm); + if (ret) + return ret; + + return hpre_cluster_debugfs_init(qm); +} + +static void hpre_dfx_debug_init(struct hisi_qm *qm) +{ + struct dfx_diff_registers *hpre_regs = qm->debug.acc_diff_regs; + struct hpre *hpre = container_of(qm, struct hpre, qm); + struct hpre_dfx *dfx = hpre->debug.dfx; + struct dentry *parent; + int i; + + parent = debugfs_create_dir("hpre_dfx", qm->debug.debug_root); + for (i = 0; i < HPRE_DFX_FILE_NUM; i++) { + dfx[i].type = i; + debugfs_create_file(hpre_dfx_files[i], 0644, parent, &dfx[i], + &hpre_atomic64_ops); + } + + if (qm->fun_type == QM_HW_PF && hpre_regs) + debugfs_create_file("diff_regs", 0444, parent, + qm, &hpre_diff_regs_fops); +} + +static int hpre_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + qm->debug.debug_root = debugfs_create_dir(dev_name(dev), + hpre_debugfs_root); + + qm->debug.sqe_mask_offset = HPRE_SQE_MASK_OFFSET; + qm->debug.sqe_mask_len = HPRE_SQE_MASK_LEN; + ret = hisi_qm_diff_regs_init(qm, hpre_diff_regs, + ARRAY_SIZE(hpre_diff_regs)); + if (ret) { + dev_warn(dev, "Failed to init HPRE diff regs!\n"); + goto debugfs_remove; + } + + hisi_qm_debug_init(qm); + + if (qm->pdev->device == PCI_DEVICE_ID_HUAWEI_HPRE_PF) { + ret = hpre_ctrl_debug_init(qm); + if (ret) + goto failed_to_create; + } + + hpre_dfx_debug_init(qm); + + return 0; + +failed_to_create: + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(hpre_diff_regs)); +debugfs_remove: + debugfs_remove_recursive(qm->debug.debug_root); + return ret; +} + +static void hpre_debugfs_exit(struct hisi_qm *qm) +{ + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(hpre_diff_regs)); + + debugfs_remove_recursive(qm->debug.debug_root); +} + +static int hpre_qm_init(struct hisi_qm *qm, struct pci_dev *pdev) +{ + int ret; + + if (pdev->revision == QM_HW_V1) { + pci_warn(pdev, "HPRE version 1 is not supported!\n"); + return -EINVAL; + } + + qm->mode = uacce_mode; + qm->pdev = pdev; + qm->ver = pdev->revision; + qm->sqe_size = HPRE_SQE_SIZE; + qm->dev_name = hpre_name; + + qm->fun_type = (pdev->device == PCI_DEVICE_ID_HUAWEI_HPRE_PF) ? + QM_HW_PF : QM_HW_VF; + if (qm->fun_type == QM_HW_PF) { + qm->qp_base = HPRE_PF_DEF_Q_BASE; + qm->qp_num = pf_q_num; + qm->debug.curr_qm_qp_num = pf_q_num; + qm->qm_list = &hpre_devices; + } + + ret = hisi_qm_init(qm); + if (ret) { + pci_err(pdev, "Failed to init hpre qm configures!\n"); + return ret; + } + + ret = hpre_set_qm_algs(qm); + if (ret) { + pci_err(pdev, "Failed to set hpre algs!\n"); + hisi_qm_uninit(qm); + } + + return ret; +} + +static int hpre_show_last_regs_init(struct hisi_qm *qm) +{ + int cluster_dfx_regs_num = ARRAY_SIZE(hpre_cluster_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hpre_com_dfx_regs); + u8 clusters_num = hpre_cluster_num(qm); + struct qm_debug *debug = &qm->debug; + void __iomem *io_base; + int i, j, idx; + + debug->last_words = kcalloc(cluster_dfx_regs_num * clusters_num + + com_dfx_regs_num, sizeof(unsigned int), GFP_KERNEL); + if (!debug->last_words) + return -ENOMEM; + + for (i = 0; i < com_dfx_regs_num; i++) + debug->last_words[i] = readl_relaxed(qm->io_base + + hpre_com_dfx_regs[i].offset); + + for (i = 0; i < clusters_num; i++) { + io_base = qm->io_base + hpre_cluster_offsets[i]; + for (j = 0; j < cluster_dfx_regs_num; j++) { + idx = com_dfx_regs_num + i * cluster_dfx_regs_num + j; + debug->last_words[idx] = readl_relaxed( + io_base + hpre_cluster_dfx_regs[j].offset); + } + } + + return 0; +} + +static void hpre_show_last_regs_uninit(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + kfree(debug->last_words); + debug->last_words = NULL; +} + +static void hpre_show_last_dfx_regs(struct hisi_qm *qm) +{ + int cluster_dfx_regs_num = ARRAY_SIZE(hpre_cluster_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hpre_com_dfx_regs); + u8 clusters_num = hpre_cluster_num(qm); + struct qm_debug *debug = &qm->debug; + struct pci_dev *pdev = qm->pdev; + void __iomem *io_base; + int i, j, idx; + u32 val; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + /* dumps last word of the debugging registers during controller reset */ + for (i = 0; i < com_dfx_regs_num; i++) { + val = readl_relaxed(qm->io_base + hpre_com_dfx_regs[i].offset); + if (debug->last_words[i] != val) + pci_info(pdev, "Common_core:%s \t= 0x%08x => 0x%08x\n", + hpre_com_dfx_regs[i].name, debug->last_words[i], val); + } + + for (i = 0; i < clusters_num; i++) { + io_base = qm->io_base + hpre_cluster_offsets[i]; + for (j = 0; j < cluster_dfx_regs_num; j++) { + val = readl_relaxed(io_base + + hpre_cluster_dfx_regs[j].offset); + idx = com_dfx_regs_num + i * cluster_dfx_regs_num + j; + if (debug->last_words[idx] != val) + pci_info(pdev, "cluster-%d:%s \t= 0x%08x => 0x%08x\n", + i, hpre_cluster_dfx_regs[j].name, debug->last_words[idx], val); + } + } +} + +static void hpre_log_hw_error(struct hisi_qm *qm, u32 err_sts) +{ + const struct hpre_hw_error *err = hpre_hw_errors; + struct device *dev = &qm->pdev->dev; + + while (err->msg) { + if (err->int_msk & err_sts) + dev_warn(dev, "%s [error status=0x%x] found\n", + err->msg, err->int_msk); + err++; + } +} + +static u32 hpre_get_hw_err_status(struct hisi_qm *qm) +{ + return readl(qm->io_base + HPRE_INT_STATUS); +} + +static void hpre_clear_hw_err_status(struct hisi_qm *qm, u32 err_sts) +{ + u32 nfe; + + writel(err_sts, qm->io_base + HPRE_HAC_SOURCE_INT); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + writel(nfe, qm->io_base + HPRE_RAS_NFE_ENB); +} + +static void hpre_open_axi_master_ooo(struct hisi_qm *qm) +{ + u32 value; + + value = readl(qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + writel(value & ~HPRE_AM_OOO_SHUTDOWN_ENABLE, + qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + writel(value | HPRE_AM_OOO_SHUTDOWN_ENABLE, + qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); +} + +static void hpre_err_info_init(struct hisi_qm *qm) +{ + struct hisi_qm_err_info *err_info = &qm->err_info; + + err_info->fe = HPRE_HAC_RAS_FE_ENABLE; + err_info->ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_QM_CE_MASK_CAP, qm->cap_ver); + err_info->nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_QM_NFE_MASK_CAP, qm->cap_ver); + err_info->ecc_2bits_mask = HPRE_CORE_ECC_2BIT_ERR | HPRE_OOO_ECC_2BIT_ERR; + err_info->dev_shutdown_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_shutdown_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_QM_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_reset_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_QM_RESET_MASK_CAP, qm->cap_ver); + err_info->dev_reset_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_RESET_MASK_CAP, qm->cap_ver); + err_info->msi_wr_port = HPRE_WR_MSI_PORT; + err_info->acpi_rst = "HRST"; +} + +static const struct hisi_qm_err_ini hpre_err_ini = { + .hw_init = hpre_set_user_domain_and_cache, + .hw_err_enable = hpre_hw_error_enable, + .hw_err_disable = hpre_hw_error_disable, + .get_dev_hw_err_status = hpre_get_hw_err_status, + .clear_dev_hw_err_status = hpre_clear_hw_err_status, + .log_dev_hw_err = hpre_log_hw_error, + .open_axi_master_ooo = hpre_open_axi_master_ooo, + .open_sva_prefetch = hpre_open_sva_prefetch, + .close_sva_prefetch = hpre_close_sva_prefetch, + .show_last_dfx_regs = hpre_show_last_dfx_regs, + .err_info_init = hpre_err_info_init, +}; + +static int hpre_pf_probe_init(struct hpre *hpre) +{ + struct hisi_qm *qm = &hpre->qm; + int ret; + + ret = hpre_set_user_domain_and_cache(qm); + if (ret) + return ret; + + hpre_open_sva_prefetch(qm); + + qm->err_ini = &hpre_err_ini; + qm->err_ini->err_info_init(qm); + hisi_qm_dev_err_init(qm); + ret = hpre_show_last_regs_init(qm); + if (ret) + pci_err(qm->pdev, "Failed to init last word regs!\n"); + + return ret; +} + +static int hpre_probe_init(struct hpre *hpre) +{ + u32 type_rate = HPRE_SHAPER_TYPE_RATE; + struct hisi_qm *qm = &hpre->qm; + int ret; + + if (qm->fun_type == QM_HW_PF) { + ret = hpre_pf_probe_init(hpre); + if (ret) + return ret; + /* Enable shaper type 0 */ + if (qm->ver >= QM_HW_V3) { + type_rate |= QM_SHAPER_ENABLE; + qm->type_rate = type_rate; + } + } + + return 0; +} + +static int hpre_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct hisi_qm *qm; + struct hpre *hpre; + int ret; + + hpre = devm_kzalloc(&pdev->dev, sizeof(*hpre), GFP_KERNEL); + if (!hpre) + return -ENOMEM; + + qm = &hpre->qm; + ret = hpre_qm_init(qm, pdev); + if (ret) { + pci_err(pdev, "Failed to init HPRE QM (%d)!\n", ret); + return ret; + } + + ret = hpre_probe_init(hpre); + if (ret) { + pci_err(pdev, "Failed to probe (%d)!\n", ret); + goto err_with_qm_init; + } + + ret = hisi_qm_start(qm); + if (ret) + goto err_with_err_init; + + ret = hpre_debugfs_init(qm); + if (ret) + dev_warn(&pdev->dev, "init debugfs fail!\n"); + + ret = hisi_qm_alg_register(qm, &hpre_devices); + if (ret < 0) { + pci_err(pdev, "fail to register algs to crypto!\n"); + goto err_with_qm_start; + } + + if (qm->uacce) { + ret = uacce_register(qm->uacce); + if (ret) { + pci_err(pdev, "failed to register uacce (%d)!\n", ret); + goto err_with_alg_register; + } + } + + if (qm->fun_type == QM_HW_PF && vfs_num) { + ret = hisi_qm_sriov_enable(pdev, vfs_num); + if (ret < 0) + goto err_with_alg_register; + } + + hisi_qm_pm_init(qm); + + return 0; + +err_with_alg_register: + hisi_qm_alg_unregister(qm, &hpre_devices); + +err_with_qm_start: + hpre_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + +err_with_err_init: + hpre_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + +err_with_qm_init: + hisi_qm_uninit(qm); + + return ret; +} + +static void hpre_remove(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + + hisi_qm_pm_uninit(qm); + hisi_qm_wait_task_finish(qm, &hpre_devices); + hisi_qm_alg_unregister(qm, &hpre_devices); + if (qm->fun_type == QM_HW_PF && qm->vfs_num) + hisi_qm_sriov_disable(pdev, true); + + hpre_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + + if (qm->fun_type == QM_HW_PF) { + hpre_cnt_regs_clear(qm); + qm->debug.curr_qm_qp_num = 0; + hpre_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + } + + hisi_qm_uninit(qm); +} + +static const struct dev_pm_ops hpre_pm_ops = { + SET_RUNTIME_PM_OPS(hisi_qm_suspend, hisi_qm_resume, NULL) +}; + +static const struct pci_error_handlers hpre_err_handler = { + .error_detected = hisi_qm_dev_err_detected, + .slot_reset = hisi_qm_dev_slot_reset, + .reset_prepare = hisi_qm_reset_prepare, + .reset_done = hisi_qm_reset_done, +}; + +static struct pci_driver hpre_pci_driver = { + .name = hpre_name, + .id_table = hpre_dev_ids, + .probe = hpre_probe, + .remove = hpre_remove, + .sriov_configure = IS_ENABLED(CONFIG_PCI_IOV) ? + hisi_qm_sriov_configure : NULL, + .err_handler = &hpre_err_handler, + .shutdown = hisi_qm_dev_shutdown, + .driver.pm = &hpre_pm_ops, +}; + +static void hpre_register_debugfs(void) +{ + if (!debugfs_initialized()) + return; + + hpre_debugfs_root = debugfs_create_dir(hpre_name, NULL); +} + +static void hpre_unregister_debugfs(void) +{ + debugfs_remove_recursive(hpre_debugfs_root); +} + +static int __init hpre_init(void) +{ + int ret; + + hisi_qm_init_list(&hpre_devices); + hpre_register_debugfs(); + + ret = pci_register_driver(&hpre_pci_driver); + if (ret) { + hpre_unregister_debugfs(); + pr_err("hpre: can't register hisi hpre driver.\n"); + } + + return ret; +} + +static void __exit hpre_exit(void) +{ + pci_unregister_driver(&hpre_pci_driver); + hpre_unregister_debugfs(); +} + +module_init(hpre_init); +module_exit(hpre_exit); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Zaibo Xu "); +MODULE_AUTHOR("Meng Yu "); +MODULE_DESCRIPTION("Driver for HiSilicon HPRE accelerator"); diff --git a/kae_driver/hisilicon/migration/Makefile b/kae_driver/hisilicon/migration/Makefile new file mode 100755 index 0000000..6493f52 --- /dev/null +++ b/kae_driver/hisilicon/migration/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_MIGRATION) += hisi_migration.o +hisi_migration-objs = acc_vf_migration.o diff --git a/kae_driver/hisilicon/migration/acc_vf_migration.c b/kae_driver/hisilicon/migration/acc_vf_migration.c new file mode 100755 index 0000000..920f199 --- /dev/null +++ b/kae_driver/hisilicon/migration/acc_vf_migration.c @@ -0,0 +1,1726 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2021 HiSilicon Limited. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "acc_vf_migration.h" + +#define VDM_OFFSET(x) offsetof(struct vfio_device_migration_info, x) +static struct dentry *mig_debugfs_root; +static int mig_root_ref; + +/* return 0 mailbox ready, -ETIMEDOUT hardware timeout */ +static int qm_wait_mb_ready(struct hisi_qm *qm) +{ + u32 val; + + return readl_relaxed_poll_timeout(qm->io_base + QM_MB_CMD_SEND_BASE, + val, !((val >> QM_MB_BUSY_SHIFT) & + 0x1), POLL_PERIOD, POLL_TIMEOUT); +} + +/* return 0 VM acc device ready, -ETIMEDOUT hardware timeout */ +static int qm_wait_dev_ready(struct hisi_qm *qm) +{ + u32 val; + + return readl_relaxed_poll_timeout(qm->io_base + QM_VF_STATE, + val, !(val & 0x1), POLL_PERIOD, POLL_TIMEOUT); +} + + +/* 128 bit should be written to hardware at one time to trigger a mailbox */ +static void qm_mb_write(struct hisi_qm *qm, const void *src) +{ + void __iomem *fun_base = qm->io_base + QM_MB_CMD_SEND_BASE; + unsigned long tmp0 = 0; + unsigned long tmp1 = 0; + + if (!IS_ENABLED(CONFIG_ARM64)) { + memcpy_toio(fun_base, src, 16); + wmb(); + return; + } + + asm volatile("ldp %0, %1, %3\n" + "stp %0, %1, %2\n" + "dsb sy\n" + : "=&r" (tmp0), + "=&r" (tmp1), + "+Q" (*((char __iomem *)fun_base)) + : "Q" (*((char *)src)) + : "memory"); +} + +static void qm_mb_pre_init(struct qm_mailbox *mailbox, u8 cmd, + u16 queue, bool op) +{ + mailbox->w0 = cpu_to_le16(cmd | + (op ? 0x1 << QM_MB_OP_SHIFT : 0) | + (0x1 << QM_MB_BUSY_SHIFT)); + mailbox->queue_num = cpu_to_le16(queue); + mailbox->rsvd = 0; +} + +static int qm_mb_nolock(struct hisi_qm *qm, struct qm_mailbox *mailbox) +{ + int cnt = 0; + + if (unlikely(qm_wait_mb_ready(qm))) { + dev_err(&qm->pdev->dev, "QM mailbox is busy to start!\n"); + return -EBUSY; + } + + qm_mb_write(qm, mailbox); + while (true) { + if (!qm_wait_mb_ready(qm)) + break; + if (++cnt > QM_MB_MAX_WAIT_CNT) { + dev_err(&qm->pdev->dev, "QM mailbox operation timeout!\n"); + return -EBUSY; + } + } + return 0; +} + +static int qm_mb(struct hisi_qm *qm, u8 cmd, dma_addr_t dma_addr, u16 queue, + bool op) +{ + struct qm_mailbox mailbox; + int ret; + + dev_dbg(&qm->pdev->dev, "QM mailbox request to q%u: %u-0x%llx\n", + queue, cmd, (unsigned long long)dma_addr); + + qm_mb_pre_init(&mailbox, cmd, queue, op); + mailbox.base_l = cpu_to_le32(lower_32_bits(dma_addr)); + mailbox.base_h = cpu_to_le32(upper_32_bits(dma_addr)); + + mutex_lock(&qm->mailbox_lock); + ret = qm_mb_nolock(qm, &mailbox); + mutex_unlock(&qm->mailbox_lock); + + return ret; +} + +/* + * Each state Reg is checked 100 times, + * with a delay of 100 microseconds after each check + */ +static u32 acc_check_reg_state(struct hisi_qm *qm, u32 regs) +{ + int check_times = 0; + u32 state; + + state = readl(qm->io_base + regs); + while (state && check_times < ERROR_CHECK_TIMEOUT) { + udelay(CHECK_DELAY_TIME); + state = readl(qm->io_base + regs); + check_times++; + } + + return state; +} + +/* Check the PF's RAS state and Function INT state */ +static int qm_check_int_state(struct acc_vf_migration *acc_vf_dev) +{ + struct hisi_qm *vfqm = acc_vf_dev->vf_qm; + struct hisi_qm *qm = acc_vf_dev->pf_qm; + struct device *dev = &qm->pdev->dev; + u32 state; + + /* Check RAS state */ + state = acc_check_reg_state(qm, QM_ABNORMAL_INT_STATUS); + if (state) { + dev_err(dev, "failed to check QM RAS state!\n"); + return -EBUSY; + } + + /* Check Function Communication state between PF and VF */ + state = acc_check_reg_state(vfqm, QM_IFC_INT_STATUS); + if (state) { + dev_err(dev, "failed to check QM IFC INT state!\n"); + return -EBUSY; + } + state = acc_check_reg_state(vfqm, QM_IFC_INT_SET_V); + if (state) { + dev_err(dev, "failed to check QM IFC INT SET state!\n"); + return -EBUSY; + } + + /* Check submodule task state */ + switch (acc_vf_dev->acc_type) { + case HISI_SEC: + state = acc_check_reg_state(qm, SEC_CORE_INT_STATUS); + if (state) { + dev_err(dev, "failed to check QM SEC Core INT state!\n"); + return -EBUSY; + } + break; + case HISI_HPRE: + state = acc_check_reg_state(qm, HPRE_HAC_INT_STATUS); + if (state) { + dev_err(dev, "failed to check QM HPRE HAC INT state!\n"); + return -EBUSY; + } + break; + case HISI_ZIP: + state = acc_check_reg_state(qm, HZIP_CORE_INT_STATUS); + if (state) { + dev_err(dev, "failed to check QM ZIP Core INT state!\n"); + return -EBUSY; + } + break; + default: + dev_err(dev, "failed to detect acc module type!\n"); + return -EINVAL; + } + + return 0; +} + +static int qm_read_reg(struct hisi_qm *qm, u32 reg_addr, + u32 *data, u8 nums) +{ + int i; + + if (nums < 1 || nums > QM_REGS_MAX_LEN) { + dev_err(&qm->pdev->dev, "QM read input parameter is error!\n"); + return -EINVAL; + } + + for (i = 0; i < nums; i++) { + data[i] = readl(qm->io_base + reg_addr); + reg_addr += QM_REG_ADDR_OFFSET; + } + + return 0; +} + +static int qm_write_reg(struct hisi_qm *qm, u32 reg_addr, + u32 *data, u8 nums) +{ + int i; + + if (nums < 1 || nums > QM_REGS_MAX_LEN) { + dev_err(&qm->pdev->dev, "QM write input parameter is error!\n"); + return -EINVAL; + } + + for (i = 0; i < nums; i++) { + writel(data[i], qm->io_base + reg_addr); + reg_addr += QM_REG_ADDR_OFFSET; + } + + return 0; +} + +static int qm_get_vft(struct hisi_qm *qm, u32 *base, u32 *number) +{ + u64 sqc_vft; + int ret; + + ret = qm_mb(qm, QM_MB_CMD_SQC_VFT_V2, 0, 0, 1); + if (ret) + return ret; + + sqc_vft = readl(qm->io_base + QM_MB_CMD_DATA_ADDR_L) | + ((u64)readl(qm->io_base + QM_MB_CMD_DATA_ADDR_H) << + QM_XQC_ADDR_OFFSET); + *base = QM_SQC_VFT_BASE_MASK_V2 & (sqc_vft >> QM_SQC_VFT_BASE_SHIFT_V2); + *number = (QM_SQC_VFT_NUM_MASK_V2 & + (sqc_vft >> QM_SQC_VFT_NUM_SHIFT_V2)) + 1; + + return 0; +} + +static int qm_get_sqc(struct hisi_qm *qm, u64 *addr) +{ + int ret; + + ret = qm_mb(qm, QM_MB_CMD_SQC_BT, 0, 0, 1); + if (ret) + return ret; + + *addr = readl(qm->io_base + QM_MB_CMD_DATA_ADDR_L) | + ((u64)readl(qm->io_base + QM_MB_CMD_DATA_ADDR_H) << + QM_XQC_ADDR_OFFSET); + + return 0; +} + +static int qm_get_cqc(struct hisi_qm *qm, u64 *addr) +{ + int ret; + + ret = qm_mb(qm, QM_MB_CMD_CQC_BT, 0, 0, 1); + if (ret) + return ret; + + *addr = readl(qm->io_base + QM_MB_CMD_DATA_ADDR_L) | + ((u64)readl(qm->io_base + QM_MB_CMD_DATA_ADDR_H) << + QM_XQC_ADDR_OFFSET); + + return 0; +} + +static int qm_rw_regs_read(struct hisi_qm *qm, struct acc_vf_data *vf_data) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + ret = qm_read_reg(qm, QM_VF_AEQ_INT_MASK, &vf_data->aeq_int_mask, 1); + if (ret) { + dev_err(dev, "failed to read QM_VF_AEQ_INT_MASK!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_VF_EQ_INT_MASK, &vf_data->eq_int_mask, 1); + if (ret) { + dev_err(dev, "failed to read QM_VF_EQ_INT_MASK!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_IFC_INT_SOURCE_V, + &vf_data->ifc_int_source, 1); + if (ret) { + dev_err(dev, "failed to read QM_IFC_INT_SOURCE_V!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_IFC_INT_MASK, &vf_data->ifc_int_mask, 1); + if (ret) { + dev_err(dev, "failed to read QM_IFC_INT_MASK!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_IFC_INT_SET_V, &vf_data->ifc_int_set, 1); + if (ret) { + dev_err(dev, "failed to read QM_IFC_INT_SET_V!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_PAGE_SIZE, &vf_data->page_size, 1); + if (ret) { + dev_err(dev, "failed to read QM_PAGE_SIZE!\n"); + return ret; + } + + ret = qm_read_reg(qm, QM_VF_STATE, &vf_data->vf_state, 1); + if (ret) { + dev_err(dev, "failed to read QM_VF_STATE!\n"); + return ret; + } + + /* QM_EQC_DW has 7 regs */ + ret = qm_read_reg(qm, QM_EQC_DW0, vf_data->qm_eqc_dw, 7); + if (ret) { + dev_err(dev, "failed to read QM_EQC_DW!\n"); + return ret; + } + + /* QM_AEQC_DW has 7 regs */ + ret = qm_read_reg(qm, QM_AEQC_DW0, vf_data->qm_aeqc_dw, 7); + if (ret) { + dev_err(dev, "failed to read QM_AEQC_DW!\n"); + return ret; + } + + return 0; +} + +static int qm_rw_regs_write(struct hisi_qm *qm, struct acc_vf_data *vf_data) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + /* check VF state */ + if (unlikely(qm_wait_mb_ready(qm))) { + dev_err(&qm->pdev->dev, "QM device is not ready to write!\n"); + return -EBUSY; + } + + ret = qm_write_reg(qm, QM_VF_AEQ_INT_MASK, &vf_data->aeq_int_mask, 1); + if (ret) { + dev_err(dev, "failed to write QM_VF_AEQ_INT_MASK!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_VF_EQ_INT_MASK, &vf_data->eq_int_mask, 1); + if (ret) { + dev_err(dev, "failed to write QM_VF_EQ_INT_MASK!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_IFC_INT_SOURCE_V, + &vf_data->ifc_int_source, 1); + if (ret) { + dev_err(dev, "failed to write QM_IFC_INT_SOURCE_V!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_IFC_INT_MASK, &vf_data->ifc_int_mask, 1); + if (ret) { + dev_err(dev, "failed to write QM_IFC_INT_MASK!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_IFC_INT_SET_V, &vf_data->ifc_int_set, 1); + if (ret) { + dev_err(dev, "failed to write QM_IFC_INT_SET_V!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_PAGE_SIZE, &vf_data->page_size, 1); + if (ret) { + dev_err(dev, "failed to write QM_PAGE_SIZE!\n"); + return ret; + } + + ret = qm_write_reg(qm, QM_VF_STATE, &vf_data->vf_state, 1); + if (ret) { + dev_err(dev, "failed to write QM_VF_STATE!\n"); + return ret; + } + + /* QM_EQC_DW has 7 regs */ + ret = qm_write_reg(qm, QM_EQC_DW0, vf_data->qm_eqc_dw, 7); + if (ret) { + dev_err(dev, "failed to write QM_EQC_DW!\n"); + return ret; + } + + /* QM_AEQC_DW has 7 regs */ + ret = qm_write_reg(qm, QM_AEQC_DW0, vf_data->qm_aeqc_dw, 7); + if (ret) { + dev_err(dev, "failed to write QM_AEQC_DW!\n"); + return ret; + } + + return 0; +} + +/* + * the vf QM have unbind from host, insmod in the VM + * so, qm just have the addr from pci dev + * others is null. + * so we need read from the SEC hardware REGs. + */ +static int vf_migration_data_store(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct acc_vf_data *vf_data = acc_vf_dev->vf_data; + struct device *dev = &qm->pdev->dev; + int ret; + + ret = qm_rw_regs_read(qm, vf_data); + if (ret) { + dev_err(dev, "failed to read QM regs!\n"); + return -EINVAL; + } + + /* + * every Reg is 32 bit, the dma address is 64 bit + * so, the dma address is store in the Reg2 and Reg1 + */ + vf_data->eqe_dma = vf_data->qm_eqc_dw[2]; + vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET; + vf_data->eqe_dma |= vf_data->qm_eqc_dw[1]; + vf_data->aeqe_dma = vf_data->qm_aeqc_dw[2]; + vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET; + vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[1]; + + /* Through SQC_BT/CQC_BT to get sqc and cqc address */ + ret = qm_get_sqc(qm, &vf_data->sqc_dma); + if (ret) { + dev_err(dev, "failed to read SQC addr!\n"); + return -EINVAL; + } + + ret = qm_get_cqc(qm, &vf_data->cqc_dma); + if (ret) { + dev_err(dev, "failed to read CQC addr!\n"); + return -EINVAL; + } + + return 0; +} + +static void qm_dev_cmd_init(struct hisi_qm *qm) +{ + /* clear VF communication status registers. */ + writel(0x1, qm->io_base + QM_IFC_INT_SOURCE_V); + + /* enable pf and vf communication. */ + writel(0x0, qm->io_base + QM_IFC_INT_MASK); +} + +static void qm_db(struct hisi_qm *qm, u16 qn, u8 cmd, + u16 index, u8 priority) +{ + void __iomem *io_base = qm->io_base; + u16 randata = 0; + u64 doorbell; + + if (cmd == QM_DOORBELL_CMD_SQ || cmd == QM_DOORBELL_CMD_CQ) + io_base = qm->db_io_base + (u64)qn * qm->db_interval + + QM_DOORBELL_SQ_CQ_BASE_V2; + else + io_base += QM_DOORBELL_EQ_AEQ_BASE_V2; + + doorbell = qn | ((u64)cmd << QM_DB_CMD_SHIFT_V2) | + ((u64)randata << QM_DB_RAND_SHIFT_V2) | + ((u64)index << QM_DB_INDEX_SHIFT_V2) | + ((u64)priority << QM_DB_PRIORITY_SHIFT_V2); + + writeq(doorbell, io_base); +} + +static void vf_qm_fun_restart(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct acc_vf_data *vf_data = acc_vf_dev->vf_data; + struct device *dev = &qm->pdev->dev; + int i; + + /* + * When the Guest is rebooted or reseted, the SMMU page table + * will be destroyed, and the QP queue cannot be returned + * normally at this time. so if Guest acc driver have removed, + * don't need to restart QP. + */ + if (vf_data->vf_state != VF_READY) { + dev_err(dev, "failed to restart VF!\n"); + return; + } + + for (i = 0; i < qm->qp_num; i++) + qm_db(qm, i, QM_DOORBELL_CMD_SQ, 0, 1); +} + +static int vf_match_info_check(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct acc_vf_data *vf_data = acc_vf_dev->vf_data; + struct hisi_qm *pf_qm = acc_vf_dev->pf_qm; + struct device *dev = &qm->pdev->dev; + u32 que_iso_state; + int ret; + + /* vf acc type check */ + if (vf_data->acc_type != acc_vf_dev->acc_type) { + dev_err(dev, "failed to match VF acc type!\n"); + return -EINVAL; + } + + /* vf qp num check */ + ret = qm_get_vft(qm, &qm->qp_base, &qm->qp_num); + if (ret || qm->qp_num <= 1) { + dev_err(dev, "failed to get vft qp nums!\n"); + return ret; + } + + if (vf_data->qp_num != qm->qp_num) { + dev_err(dev, "failed to match VF qp num!\n"); + return -EINVAL; + } + + /* vf isolation state check */ + ret = qm_read_reg(pf_qm, QM_QUE_ISO_CFG_V, &que_iso_state, 1); + if (ret) { + dev_err(dev, "failed to read QM_QUE_ISO_CFG_V!\n"); + return ret; + } + if (vf_data->que_iso_cfg != que_iso_state) { + dev_err(dev, "failed to match isolation state!\n"); + return -EINVAL; + } + + return 0; +} + +static int vf_migration_data_recover(struct hisi_qm *qm, + struct acc_vf_data *vf_data) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + qm->eqe_dma = vf_data->eqe_dma; + qm->aeqe_dma = vf_data->aeqe_dma; + qm->sqc_dma = vf_data->sqc_dma; + qm->cqc_dma = vf_data->cqc_dma; + + qm->qp_base = vf_data->qp_base; + qm->qp_num = vf_data->qp_num; + + ret = qm_rw_regs_write(qm, vf_data); + if (ret) { + dev_err(dev, "Set VF regs failed!\n"); + return ret; + } + + ret = qm_mb(qm, QM_MB_CMD_SQC_BT, qm->sqc_dma, 0, 0); + if (ret) { + dev_err(dev, "Set sqc failed!\n"); + return ret; + } + + ret = qm_mb(qm, QM_MB_CMD_CQC_BT, qm->cqc_dma, 0, 0); + if (ret) { + dev_err(dev, "Set cqc failed!\n"); + return ret; + } + + /* which ACC module need to reinit? */ + qm_dev_cmd_init(qm); + + return 0; +} + +static int vf_qm_cache_wb(struct hisi_qm *qm) +{ + unsigned int val; + + writel(0x1, qm->io_base + QM_CACHE_WB_START); + if (readl_relaxed_poll_timeout(qm->io_base + QM_CACHE_WB_DONE, + val, val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT)) { + dev_err(&qm->pdev->dev, "vf QM writeback sqc cache fail!\n"); + return -EINVAL; + } + + return 0; +} + +static int vf_qm_func_stop(struct hisi_qm *qm) +{ + return qm_mb(qm, QM_MB_CMD_PAUSE_QM, 0, 0, 0); +} + +static int pf_qm_get_qp_num(struct hisi_qm *qm, int vf_id, + u32 *rbase, u32 *rnumber) +{ + unsigned int val; + u64 sqc_vft; + int ret; + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); + if (ret) + return ret; + + writel(0x1, qm->io_base + QM_VFT_CFG_OP_WR); + /* 0 mean SQC VFT */ + writel(0x0, qm->io_base + QM_VFT_CFG_TYPE); + writel(vf_id, qm->io_base + QM_VFT_CFG); + + writel(0x0, qm->io_base + QM_VFT_CFG_RDY); + writel(0x1, qm->io_base + QM_VFT_CFG_OP_ENABLE); + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); + if (ret) + return ret; + + sqc_vft = readl(qm->io_base + QM_VFT_CFG_DATA_L) | + ((u64)readl(qm->io_base + QM_VFT_CFG_DATA_H) << + QM_XQC_ADDR_OFFSET); + *rbase = QM_SQC_VFT_BASE_MASK_V2 & + (sqc_vft >> QM_SQC_VFT_BASE_SHIFT_V2); + *rnumber = (QM_SQC_VFT_NUM_MASK_V2 & + (sqc_vft >> QM_SQC_VFT_NUM_SHIFT_V2)) + 1; + + return 0; +} + +static int pf_qm_state_pre_save(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct acc_vf_data *vf_data = acc_vf_dev->vf_data; + struct device *dev = &qm->pdev->dev; + int vf_id = acc_vf_dev->vf_id; + int ret; + + /* Vf acc type save */ + vf_data->acc_type = acc_vf_dev->acc_type; + + /* Vf qp num save from PF */ + ret = pf_qm_get_qp_num(qm, vf_id, &vf_data->qp_base, &vf_data->qp_num); + if (ret) { + dev_err(dev, "failed to get vft qp nums!\n"); + return -EINVAL; + } + + /* Vf isolation state save from PF */ + ret = qm_read_reg(qm, QM_QUE_ISO_CFG_V, &vf_data->que_iso_cfg, 1); + if (ret) { + dev_err(dev, "failed to read QM_QUE_ISO_CFG_V!\n"); + return ret; + } + + return 0; +} + +static int vf_qm_state_save(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct device *dev = &acc_vf_dev->vf_dev->dev; + int ret; + + /* + * check VM task driver state + * if vf_ready == 0x1, skip migrate. + */ + if (unlikely(qm_wait_dev_ready(qm))) { + acc_vf_dev->mig_ignore = true; + dev_err(&qm->pdev->dev, "QM device is not ready to read!\n"); + return 0; + } + + /* First stop the ACC vf function */ + ret = vf_qm_func_stop(qm); + if (ret) { + dev_err(dev, "failed to stop QM VF function!\n"); + return ret; + } + + /* Check the VF's RAS and Interrution state */ + ret = qm_check_int_state(acc_vf_dev); + if (ret) { + dev_err(dev, "failed to check QM INT state!\n"); + goto state_error; + } + + /* hisi_qm_cache_wb store cache data to DDR */ + ret = vf_qm_cache_wb(qm); + if (ret) { + dev_err(dev, "failed to writeback QM Cache!\n"); + goto state_error; + } + + ret = vf_migration_data_store(qm, acc_vf_dev); + if (ret) { + dev_err(dev, "failed to get and store migration data!\n"); + goto state_error; + } + + return 0; + +state_error: + vf_qm_fun_restart(qm, acc_vf_dev); + return ret; +} + +static int vf_qm_state_resume(struct hisi_qm *qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct device *dev = &acc_vf_dev->vf_dev->dev; + int ret; + + /* recover data to VF */ + ret = vf_migration_data_recover(qm, acc_vf_dev->vf_data); + if (ret) { + dev_err(dev, "failed to recover the VF!\n"); + return ret; + } + + /* restart all destination VF's QP */ + vf_qm_fun_restart(qm, acc_vf_dev); + + return 0; +} + +static int acc_vf_set_device_state(struct acc_vf_migration *acc_vf_dev, + u32 state) +{ + struct vfio_device_migration_info *mig_ctl = acc_vf_dev->mig_ctl; + struct device *dev = &acc_vf_dev->vf_dev->dev; + struct hisi_qm *pfqm = acc_vf_dev->pf_qm; + struct hisi_qm *qm = acc_vf_dev->vf_qm; + int ret = 0; + + if (state == mig_ctl->device_state) + return 0; + + switch (state) { + case VFIO_DEVICE_STATE_RUNNING: + if (!mig_ctl->data_size) + break; + + if (mig_ctl->device_state == VFIO_DEVICE_STATE_RESUMING) { + ret = vf_qm_state_resume(qm, acc_vf_dev); + if (ret) { + dev_err(dev, "failed to resume device!\n"); + return -EFAULT; + } + } + + break; + case VFIO_DEVICE_STATE_SAVING | VFIO_DEVICE_STATE_RUNNING: + /* ACC should in the pre cycle to read match information data */ + ret = pf_qm_state_pre_save(pfqm, acc_vf_dev); + if (ret) { + dev_err(dev, "failed to pre save device state!\n"); + return -EFAULT; + } + + /* set the pending_byte and match data size */ + mig_ctl->data_size = QM_MATCH_SIZE; + mig_ctl->pending_bytes = mig_ctl->data_size; + + break; + case VFIO_DEVICE_STATE_SAVING: + /* stop the vf function */ + ret = vf_qm_state_save(qm, acc_vf_dev); + if (ret) { + dev_err(dev, "failed to save device state!\n"); + return -EFAULT; + } + + if (acc_vf_dev->mig_ignore) { + mig_ctl->data_size = 0; + mig_ctl->pending_bytes = 0; + break; + } + + /* set the pending_byte and data_size */ + mig_ctl->data_size = sizeof(struct acc_vf_data); + mig_ctl->pending_bytes = mig_ctl->data_size; + + break; + case VFIO_DEVICE_STATE_STOP: + case VFIO_DEVICE_STATE_RESUMING: + break; + default: + ret = -EFAULT; + } + + if (!ret) { + dev_info(dev, "migration state: %s ----------> %s!\n", + vf_dev_state[mig_ctl->device_state], + vf_dev_state[state]); + mig_ctl->device_state = state; + } + + return ret; +} + +static int acc_vf_data_transfer(struct acc_vf_migration *acc_vf_dev, + char __user *buf, size_t count, u64 pos, bool iswrite) +{ + struct vfio_device_migration_info *mig_ctl = acc_vf_dev->mig_ctl; + void *data_addr = acc_vf_dev->vf_data; + int ret = 0; + + if (!count) { + dev_err(&acc_vf_dev->vf_dev->dev, + "Qemu operation data size error!\n"); + return -EINVAL; + } + + data_addr += pos - mig_ctl->data_offset; + if (iswrite) { + ret = copy_from_user(data_addr, buf, count) ? + -EFAULT : count; + if (ret == count) + mig_ctl->pending_bytes += count; + } else { + ret = copy_to_user(buf, data_addr, count) ? + -EFAULT : count; + if (ret == count) + mig_ctl->pending_bytes -= count; + } + + return ret; +} + +static int acc_vf_region_migration_rw(struct acc_vf_migration *acc_vf_dev, + char __user *buf, size_t count, loff_t *ppos, bool iswrite) +{ + struct vfio_device_migration_info *mig_ctl = acc_vf_dev->mig_ctl; + struct device *dev = &acc_vf_dev->vf_dev->dev; + struct hisi_qm *qm = acc_vf_dev->vf_qm; + u64 pos = *ppos & VFIO_PCI_OFFSET_MASK; + u32 device_state; + int ret = 0; + + switch (pos) { + case VDM_OFFSET(device_state): + if (count != sizeof(mig_ctl->device_state)) { + ret = -EINVAL; + break; + } + + if (iswrite) { + if (copy_from_user(&device_state, buf, count)) { + ret = -EFAULT; + break; + } + + ret = acc_vf_set_device_state(acc_vf_dev, + device_state) ? ret : count; + } else { + ret = copy_to_user(buf, &mig_ctl->device_state, + count) ? -EFAULT : count; + } + break; + case VDM_OFFSET(reserved): + ret = -EFAULT; + break; + case VDM_OFFSET(pending_bytes): + if (count != sizeof(mig_ctl->pending_bytes)) { + ret = -EINVAL; + break; + } + + if (iswrite) + ret = -EFAULT; + else + ret = copy_to_user(buf, &mig_ctl->pending_bytes, + count) ? -EFAULT : count; + break; + case VDM_OFFSET(data_offset): + if (count != sizeof(mig_ctl->data_offset)) { + ret = -EINVAL; + break; + } + if (iswrite) + ret = copy_from_user(&mig_ctl->data_offset, buf, count) ? + -EFAULT : count; + else + ret = copy_to_user(buf, &mig_ctl->data_offset, count) ? + -EFAULT : count; + break; + case VDM_OFFSET(data_size): + if (count != sizeof(mig_ctl->data_size)) { + ret = -EINVAL; + break; + } + + if (iswrite) + ret = copy_from_user(&mig_ctl->data_size, buf, count) ? + -EFAULT : count; + else + ret = copy_to_user(buf, &mig_ctl->data_size, count) ? + -EFAULT : count; + break; + default: + ret = -EFAULT; + break; + } + + /* Transfer data section */ + if (pos >= mig_ctl->data_offset && + pos < MIGRATION_REGION_SZ) { + ret = acc_vf_data_transfer(acc_vf_dev, buf, + count, pos, iswrite); + if (ret != count) + return ret; + } + + if (mig_ctl->device_state == VFIO_DEVICE_STATE_RESUMING && + mig_ctl->pending_bytes == QM_MATCH_SIZE && + mig_ctl->data_size == QM_MATCH_SIZE) { + /* check the VF match information */ + ret = vf_match_info_check(qm, acc_vf_dev); + if (ret) { + dev_err(dev, "failed to check match information!\n"); + return -EFAULT; + } + ret = count; + + /* clear the VF match data size */ + mig_ctl->pending_bytes = 0; + mig_ctl->data_size = 0; + } + return ret; +} + +static int acc_vf_region_migration_mmap(struct acc_vf_migration *acc_vf_dev, + struct acc_vf_region *region, + struct vm_area_struct *vma) +{ + return -EFAULT; +} + +static void acc_vf_region_migration_release(struct acc_vf_migration *acc_vf_dev, + struct acc_vf_region *region) +{ + kfree(acc_vf_dev->mig_ctl); + acc_vf_dev->mig_ctl = NULL; +} + +static const struct acc_vf_region_ops acc_vf_region_ops_migration = { + .rw = acc_vf_region_migration_rw, + .release = acc_vf_region_migration_release, + .mmap = acc_vf_region_migration_mmap, +}; + +static int acc_vf_register_region(struct acc_vf_migration *acc_vf_dev, + const struct acc_vf_region_ops *ops, + void *data) +{ + struct acc_vf_region *regions; + + regions = krealloc(acc_vf_dev->regions, + (acc_vf_dev->num_regions + 1) * sizeof(*regions), + GFP_KERNEL); + if (!regions) + return -ENOMEM; + + acc_vf_dev->regions = regions; + regions[acc_vf_dev->num_regions].type = + VFIO_REGION_TYPE_MIGRATION; + regions[acc_vf_dev->num_regions].subtype = + VFIO_REGION_SUBTYPE_MIGRATION; + regions[acc_vf_dev->num_regions].ops = ops; + regions[acc_vf_dev->num_regions].size = + MIGRATION_REGION_SZ; + regions[acc_vf_dev->num_regions].flags = + VFIO_REGION_INFO_FLAG_READ | VFIO_REGION_INFO_FLAG_WRITE; + regions[acc_vf_dev->num_regions].data = data; + acc_vf_dev->num_regions++; + + return 0; +} + +static long acc_vf_get_region_info(void *device_data, + unsigned int cmd, unsigned long arg) +{ + int num_vdev_regions = vfio_pci_num_regions(device_data); + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + struct vfio_region_info_cap_type cap_type; + struct acc_vf_region *regions; + struct vfio_region_info info; + struct vfio_info_cap caps; + unsigned long minsz; + int index, ret; + + minsz = offsetofend(struct vfio_region_info, offset); + + if (cmd != VFIO_DEVICE_GET_REGION_INFO) + return -EINVAL; + + if (copy_from_user(&info, (void __user *)arg, minsz)) + return -EFAULT; + + if (info.argsz < minsz) + return -EINVAL; + + if (info.index < VFIO_PCI_NUM_REGIONS + num_vdev_regions) + goto default_handle; + + index = info.index - VFIO_PCI_NUM_REGIONS - num_vdev_regions; + if (index > acc_vf_dev->num_regions) { + dev_err(&acc_vf_dev->vf_dev->dev, + "failed to check region numbers!\n"); + return -EINVAL; + } + + info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index); + regions = acc_vf_dev->regions; + info.size = regions[index].size; + info.flags = regions[index].flags; + caps.buf = NULL; + caps.size = 0; + cap_type.header.id = VFIO_REGION_INFO_CAP_TYPE; + cap_type.header.version = 1; + cap_type.type = regions[index].type; + cap_type.subtype = regions[index].subtype; + + ret = vfio_info_add_capability(&caps, &cap_type.header, + sizeof(cap_type)); + if (ret) + return ret; + + if (regions[index].ops->add_cap) { + ret = regions[index].ops->add_cap(acc_vf_dev, + ®ions[index], &caps); + if (ret) { + kfree(caps.buf); + return ret; + } + } + + if (caps.size) { + info.flags |= VFIO_REGION_INFO_FLAG_CAPS; + if (info.argsz < sizeof(info) + caps.size) { + info.argsz = sizeof(info) + caps.size; + info.cap_offset = 0; + } else { + vfio_info_cap_shift(&caps, sizeof(info)); + if (copy_to_user((void __user *)arg + sizeof(info), + caps.buf, caps.size)) { + kfree(caps.buf); + return -EFAULT; + } + info.cap_offset = sizeof(info); + } + kfree(caps.buf); + } + + return copy_to_user((void __user *)arg, &info, minsz) ? + -EFAULT : 0; + +default_handle: + ret = vfio_pci_ioctl(device_data, cmd, arg); + if (ret) + return ret; + + if (info.index == VFIO_PCI_BAR0_REGION_INDEX) { + if (!acc_vf_dev->in_dirty_track) + return ret; + + /* read default handler's data back */ + if (copy_from_user(&info, (void __user *)arg, minsz)) + return -EFAULT; + + info.flags = VFIO_REGION_INFO_FLAG_READ | + VFIO_REGION_INFO_FLAG_WRITE; + /* update customized region info */ + if (copy_to_user((void __user *)arg, &info, minsz)) + return -EFAULT; + } + + if (info.index == VFIO_PCI_BAR2_REGION_INDEX) { + info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index); + /* + * ACC VF dev BAR2 region(64K) consists of both functional + * register space and migration control register space. + * Report only the first 32K(functional region) to Guest. + */ + info.size = pci_resource_len(acc_vf_dev->vf_dev, info.index) >> 1; + info.flags = VFIO_REGION_INFO_FLAG_READ | + VFIO_REGION_INFO_FLAG_WRITE | + VFIO_REGION_INFO_FLAG_MMAP; + if (copy_to_user((void __user *)arg, &info, minsz)) + return -EFAULT; + } + + return ret; +} + +static int acc_vf_open(void *device_data) +{ + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + struct vfio_device_migration_info *mig_ctl; + __u64 mig_offset; + void *vf_data; + int ret; + + if (!try_module_get(THIS_MODULE)) + return -ENODEV; + + mutex_lock(&acc_vf_dev->reflock); + if (!acc_vf_dev->refcnt) { + ret = acc_vf_register_region(acc_vf_dev, + &acc_vf_region_ops_migration, + NULL); + if (ret) + goto region_error; + vfio_pci_set_vendor_regions(device_data, + acc_vf_dev->num_regions); + + /* the data region must follow migration info */ + mig_offset = sizeof(struct vfio_device_migration_info); + mig_ctl = kzalloc(MIGRATION_REGION_SZ, GFP_KERNEL); + if (!mig_ctl) { + ret = -ENOMEM; + goto mig_error; + } + acc_vf_dev->mig_ctl = mig_ctl; + + vf_data = (void *)mig_ctl + mig_offset; + acc_vf_dev->vf_data = vf_data; + + mig_ctl->device_state = VFIO_DEVICE_STATE_RUNNING; + mig_ctl->data_offset = mig_offset; + mig_ctl->data_size = 0; + } + + ret = vfio_pci_open(device_data); + if (ret) + goto open_error; + + acc_vf_dev->refcnt++; + mutex_unlock(&acc_vf_dev->reflock); + + return 0; + +open_error: + if (!acc_vf_dev->refcnt) { + kfree(acc_vf_dev->mig_ctl); + acc_vf_dev->mig_ctl = NULL; + } +mig_error: + vfio_pci_set_vendor_regions(device_data, 0); +region_error: + mutex_unlock(&acc_vf_dev->reflock); + module_put(THIS_MODULE); + return ret; +} + +static void acc_vf_release(void *device_data) +{ + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + int i; + + mutex_lock(&acc_vf_dev->reflock); + if (!--acc_vf_dev->refcnt) { + for (i = 0; i < acc_vf_dev->num_regions; i++) { + if (!acc_vf_dev->regions[i].ops) + continue; + acc_vf_dev->regions[i].ops->release(acc_vf_dev, + &acc_vf_dev->regions[i]); + } + kfree(acc_vf_dev->regions); + acc_vf_dev->regions = NULL; + acc_vf_dev->num_regions = 0; + vfio_pci_set_vendor_regions(device_data, 0); + + kfree(acc_vf_dev->mig_ctl); + acc_vf_dev->mig_ctl = NULL; + } + vfio_pci_release(device_data); + mutex_unlock(&acc_vf_dev->reflock); + module_put(THIS_MODULE); +} + +static void acc_vf_reset(void *device_data) +{ + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + struct hisi_qm *qm = acc_vf_dev->vf_qm; + struct device *dev = &qm->pdev->dev; + u32 vf_state = VF_NOT_READY; + int ret; + + dev_info(dev, "QEMU prepare to Reset Guest!\n"); + ret = qm_write_reg(qm, QM_VF_STATE, &vf_state, 1); + if (ret) + dev_err(dev, "failed to write QM_VF_STATE\n"); +} + +static long acc_vf_ioctl(void *device_data, + unsigned int cmd, unsigned long arg) +{ + switch (cmd) { + case VFIO_DEVICE_GET_REGION_INFO: + return acc_vf_get_region_info(device_data, cmd, arg); + case VFIO_DEVICE_RESET: + acc_vf_reset(device_data); + return vfio_pci_ioctl(device_data, cmd, arg); + default: + return vfio_pci_ioctl(device_data, cmd, arg); + } +} + +static ssize_t acc_vf_read(void *device_data, char __user *buf, + size_t count, loff_t *ppos) +{ + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + int num_vdev_regions = vfio_pci_num_regions(device_data); + unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos); + int num_vendor_region = acc_vf_dev->num_regions; + struct acc_vf_region *region; + + if (index >= VFIO_PCI_NUM_REGIONS + num_vdev_regions + + num_vendor_region) { + dev_err(&acc_vf_dev->vf_dev->dev, + "failed to check read regions index!\n"); + return -EINVAL; + } + + if (index < VFIO_PCI_NUM_REGIONS + num_vdev_regions) + return vfio_pci_read(device_data, buf, count, ppos); + + index -= VFIO_PCI_NUM_REGIONS + num_vdev_regions; + + region = &acc_vf_dev->regions[index]; + if (!region->ops->rw) { + dev_err(&acc_vf_dev->vf_dev->dev, + "failed to check regions read ops!\n"); + return -EINVAL; + } + + return region->ops->rw(acc_vf_dev, buf, count, ppos, false); +} + +static ssize_t acc_vf_write(void *device_data, const char __user *buf, + size_t count, loff_t *ppos) +{ + struct acc_vf_migration *acc_vf_dev = + vfio_pci_vendor_data(device_data); + int num_vdev_regions = vfio_pci_num_regions(device_data); + unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos); + int num_vendor_region = acc_vf_dev->num_regions; + struct acc_vf_region *region; + + if (index == VFIO_PCI_BAR0_REGION_INDEX) + pr_debug("vfio bar 0 write\n"); + + if (index >= VFIO_PCI_NUM_REGIONS + num_vdev_regions + + num_vendor_region) { + dev_err(&acc_vf_dev->vf_dev->dev, + "failed to check write regions index!\n"); + return -EINVAL; + } + + if (index < VFIO_PCI_NUM_REGIONS + num_vdev_regions) + return vfio_pci_write(device_data, buf, count, ppos); + + index -= VFIO_PCI_NUM_REGIONS + num_vdev_regions; + + region = &acc_vf_dev->regions[index]; + + if (!region->ops->rw) { + dev_err(&acc_vf_dev->vf_dev->dev, + "failed to check regions write ops!\n"); + return -EINVAL; + } + + return region->ops->rw(acc_vf_dev, (char __user *)buf, + count, ppos, true); +} + +static int acc_vf_mmap(void *device_data, struct vm_area_struct *vma) +{ + return vfio_pci_mmap(device_data, vma); +} + +static void acc_vf_request(void *device_data, unsigned int count) +{ + vfio_pci_request(device_data, count); +} + +static struct vfio_device_ops acc_vf_device_ops_node = { + .name = "acc_vf", + .open = acc_vf_open, + .release = acc_vf_release, + .ioctl = acc_vf_ioctl, + .read = acc_vf_read, + .write = acc_vf_write, + .mmap = acc_vf_mmap, + .request = acc_vf_request, +}; + +static ssize_t acc_vf_debug_read(struct file *filp, char __user *buffer, + size_t count, loff_t *pos) +{ + char buf[VFIO_DEV_DBG_LEN]; + int len; + + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "echo 0: test vf data store\n" + "echo 1: test vf data writeback\n" + "echo 2: test vf send mailbox\n" + "echo 3: dump vf dev data\n" + "echo 4: dump migration state\n"); + + return simple_read_from_buffer(buffer, count, pos, buf, len); +} + +static ssize_t acc_vf_debug_write(struct file *filp, const char __user *buffer, + size_t count, loff_t *pos) +{ + struct acc_vf_migration *acc_vf_dev = filp->private_data; + struct device *dev = &acc_vf_dev->vf_dev->dev; + struct hisi_qm *qm = acc_vf_dev->vf_qm; + char tbuf[VFIO_DEV_DBG_LEN]; + unsigned long val; + u64 data; + int len, ret; + + if (*pos) + return 0; + + if (count >= VFIO_DEV_DBG_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, VFIO_DEV_DBG_LEN - 1, + pos, buffer, count); + if (len < 0) + return len; + tbuf[len] = '\0'; + if (kstrtoul(tbuf, 0, &val)) + return -EFAULT; + + switch (val) { + case STATE_SAVE: + ret = vf_qm_state_save(qm, acc_vf_dev); + if (ret) + return -EINVAL; + break; + case STATE_RESUME: + ret = vf_qm_state_resume(qm, acc_vf_dev); + if (ret) + return -EINVAL; + break; + case MB_TEST: + data = readl(qm->io_base + QM_MB_CMD_SEND_BASE); + dev_info(dev, "debug mailbox addr: 0x%lx, mailbox val: 0x%llx\n", + (uintptr_t)qm->phys_base, data); + break; + case MIG_DATA_DUMP: + dev_info(dev, "dumped vf migration data:\n"); + print_hex_dump(KERN_INFO, "Mig Data:", DUMP_PREFIX_OFFSET, + VFIO_DBG_LOG_LEN, 1, + (unsigned char *)acc_vf_dev->vf_data, + sizeof(struct acc_vf_data), false); + break; + case MIG_DEV_SHOW: + if (!acc_vf_dev->mig_ctl) + dev_info(dev, "migration region have release!\n"); + else + dev_info(dev, + "device state: %u\n" + "data offset: %llu\n" + "data size: %llu\n" + "pending bytes: %llu\n" + "data addr: 0x%lx\n", + acc_vf_dev->mig_ctl->device_state, + acc_vf_dev->mig_ctl->data_offset, + acc_vf_dev->mig_ctl->data_size, + acc_vf_dev->mig_ctl->pending_bytes, + (uintptr_t)acc_vf_dev->vf_data); + break; + default: + return -EINVAL; + } + + return count; +} + +static const struct file_operations acc_vf_debug_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = acc_vf_debug_read, + .write = acc_vf_debug_write, +}; + +static ssize_t acc_vf_state_read(struct file *filp, char __user *buffer, + size_t count, loff_t *pos) +{ + struct acc_vf_migration *acc_vf_dev = filp->private_data; + char buf[VFIO_DEV_DBG_LEN]; + u32 state; + int len; + + if (!acc_vf_dev->mig_ctl) { + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", "Invalid\n"); + } else { + state = acc_vf_dev->mig_ctl->device_state; + switch (state) { + case VFIO_DEVICE_STATE_RUNNING: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "RUNNING\n"); + break; + case VFIO_DEVICE_STATE_SAVING | VFIO_DEVICE_STATE_RUNNING: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "SAVING and RUNNING\n"); + break; + case VFIO_DEVICE_STATE_SAVING: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "SAVING\n"); + break; + case VFIO_DEVICE_STATE_STOP: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "STOP\n"); + break; + case VFIO_DEVICE_STATE_RESUMING: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "RESUMING\n"); + break; + default: + len = scnprintf(buf, VFIO_DEV_DBG_LEN, "%s\n", + "Error\n"); + } + } + + return simple_read_from_buffer(buffer, count, pos, buf, len); +} + +static const struct file_operations acc_vf_state_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = acc_vf_state_read, +}; + +static void vf_debugfs_init(struct acc_vf_migration *acc_vf_dev) +{ + char name[VFIO_DEV_DBG_LEN]; + int node_id; + + if (!mig_root_ref) + mig_debugfs_root = debugfs_create_dir("vfio_acc", NULL); + mutex_lock(&acc_vf_dev->reflock); + mig_root_ref++; + mutex_unlock(&acc_vf_dev->reflock); + + node_id = dev_to_node(&acc_vf_dev->vf_dev->dev); + if (node_id < 0) + node_id = 0; + + if (acc_vf_dev->acc_type == HISI_SEC) + scnprintf(name, VFIO_DEV_DBG_LEN, "sec_vf%d-%d", + node_id, acc_vf_dev->vf_id); + else if (acc_vf_dev->acc_type == HISI_HPRE) + scnprintf(name, VFIO_DEV_DBG_LEN, "hpre_vf%d-%d", + node_id, acc_vf_dev->vf_id); + else + scnprintf(name, VFIO_DEV_DBG_LEN, "zip_vf%d-%d", + node_id, acc_vf_dev->vf_id); + + acc_vf_dev->debug_root = debugfs_create_dir(name, mig_debugfs_root); + + debugfs_create_file("debug", 0644, acc_vf_dev->debug_root, + acc_vf_dev, &acc_vf_debug_fops); + debugfs_create_file("state", 0444, acc_vf_dev->debug_root, + acc_vf_dev, &acc_vf_state_fops); +} + +static void vf_debugfs_exit(struct acc_vf_migration *acc_vf_dev) +{ + debugfs_remove_recursive(acc_vf_dev->debug_root); + + mutex_lock(&acc_vf_dev->reflock); + mig_root_ref--; + mutex_unlock(&acc_vf_dev->reflock); + + if (!mig_root_ref) + debugfs_remove_recursive(mig_debugfs_root); +} + +static int qm_acc_type_init(struct acc_vf_migration *acc_vf_dev) +{ + struct hisi_qm *qm = acc_vf_dev->vf_qm; + int i; + + acc_vf_dev->acc_type = 0; + for (i = 0; i < ARRAY_SIZE(vf_acc_types); i++) { + if (!strncmp(qm->dev_name, vf_acc_types[i].name, + strlen(vf_acc_types[i].name))) + acc_vf_dev->acc_type = vf_acc_types[i].type; + } + if (!acc_vf_dev->acc_type) { + dev_err(&acc_vf_dev->vf_dev->dev, "failed to check acc type!\n"); + return -EINVAL; + } + + return 0; +} + +static int vf_qm_pci_init(struct pci_dev *pdev, struct hisi_qm *vfqm) +{ + struct device *dev = &pdev->dev; + u32 val; + int ret; + + ret = pci_request_mem_regions(pdev, vfqm->dev_name); + if (ret < 0) { + dev_err(dev, "failed to request mem regions!\n"); + return ret; + } + + vfqm->phys_base = pci_resource_start(pdev, PCI_BAR_2); + vfqm->io_base = devm_ioremap(dev, pci_resource_start(pdev, PCI_BAR_2), + pci_resource_len(pdev, PCI_BAR_2)); + if (!vfqm->io_base) { + ret = -EIO; + goto err_ioremap; + } + + val = readl(vfqm->io_base + QM_QUE_ISO_CFG_V); + val = val & BIT(0); + if (val) { + vfqm->db_phys_base = pci_resource_start(pdev, PCI_BAR_4); + vfqm->db_io_base = devm_ioremap(dev, pci_resource_start(pdev, + PCI_BAR_4), pci_resource_len(pdev, PCI_BAR_4)); + if (!vfqm->db_io_base) { + ret = -EIO; + goto err_db_ioremap; + } + } else { + vfqm->db_phys_base = vfqm->phys_base; + vfqm->db_io_base = vfqm->io_base; + } + + vfqm->pdev = pdev; + mutex_init(&vfqm->mailbox_lock); + + /* + * Allow VF devices to be loaded in VM when + * it loaded in migration driver + */ + pci_release_mem_regions(pdev); + + return 0; + +err_db_ioremap: + devm_iounmap(dev, vfqm->io_base); +err_ioremap: + pci_release_mem_regions(pdev); + return ret; +} + +static int acc_vf_dev_init(struct pci_dev *pdev, struct hisi_qm *pf_qm, + struct acc_vf_migration *acc_vf_dev) +{ + struct hisi_qm *vf_qm; + int ret; + + vf_qm = kzalloc(sizeof(struct hisi_qm), GFP_KERNEL); + if (!vf_qm) + return -ENOMEM; + + /* get vf qm dev name from pf */ + vf_qm->dev_name = pf_qm->dev_name; + vf_qm->fun_type = QM_HW_VF; + acc_vf_dev->vf_qm = vf_qm; + acc_vf_dev->pf_qm = pf_qm; + + ret = vf_qm_pci_init(pdev, vf_qm); + if (ret) + goto init_qm_error; + + ret = qm_acc_type_init(acc_vf_dev); + if (ret) + goto init_qm_error; + + return 0; + +init_qm_error: + kfree(vf_qm); + return -ENOMEM; +} + +static void *acc_vf_probe(struct pci_dev *pdev) +{ + struct acc_vf_migration *acc_vf_dev; + struct pci_dev *pf_dev, *vf_dev; + struct hisi_qm *pf_qm; + int vf_id, ret; + + pf_dev = pdev->physfn; + vf_dev = pdev; + /* + * the VF driver have been remove after unbind + * the PF driver have probe + */ + pf_qm = pci_get_drvdata(pf_dev); + if (!pf_qm) { + dev_err(&pdev->dev, "host qm driver not insmod!\n"); + return ERR_PTR(-EINVAL); + } + if (pf_qm->ver < QM_HW_V3) { + dev_err(&pdev->dev, + "device can't support migration! version: 0x%x\n", + pf_qm->ver); + return ERR_PTR(-EINVAL); + } + + vf_id = PCI_FUNC(vf_dev->devfn); + if (vf_id < 0) { + dev_info(&pdev->dev, "vf device: %s, vf id: %d\n", + pf_qm->dev_name, vf_id); + return ERR_PTR(-EINVAL); + } + + acc_vf_dev = kzalloc(sizeof(*acc_vf_dev), GFP_KERNEL); + if (!acc_vf_dev) + return ERR_PTR(-ENOMEM); + + ret = acc_vf_dev_init(pdev, pf_qm, acc_vf_dev); + if (ret) { + kfree(acc_vf_dev); + return ERR_PTR(-ENOMEM); + } + + acc_vf_dev->vf_id = vf_id; + acc_vf_dev->vf_vendor = pdev->vendor; + acc_vf_dev->vf_device = pdev->device; + acc_vf_dev->pf_dev = pf_dev; + acc_vf_dev->vf_dev = vf_dev; + acc_vf_dev->mig_ignore = false; + mutex_init(&acc_vf_dev->reflock); + + vf_debugfs_init(acc_vf_dev); + + return acc_vf_dev; +} + +static void acc_vf_remove(void *vendor_data) +{ + struct acc_vf_migration *acc_vf_dev = vendor_data; + struct device *dev = &acc_vf_dev->vf_dev->dev; + struct hisi_qm *qm = acc_vf_dev->vf_qm; + + vf_debugfs_exit(acc_vf_dev); + + devm_iounmap(dev, qm->io_base); + + kfree(qm); + kfree(acc_vf_dev); +} + +static struct vfio_pci_vendor_driver_ops sec_vf_mig_ops = { + .owner = THIS_MODULE, + .name = "hisi_sec2", + .probe = acc_vf_probe, + .remove = acc_vf_remove, + .device_ops = &acc_vf_device_ops_node, +}; + +static struct vfio_pci_vendor_driver_ops hpre_vf_mig_ops = { + .owner = THIS_MODULE, + .name = "hisi_hpre", + .probe = acc_vf_probe, + .remove = acc_vf_remove, + .device_ops = &acc_vf_device_ops_node, +}; + +static struct vfio_pci_vendor_driver_ops zip_vf_mig_ops = { + .owner = THIS_MODULE, + .name = "hisi_zip", + .probe = acc_vf_probe, + .remove = acc_vf_remove, + .device_ops = &acc_vf_device_ops_node, +}; + +static int __init acc_vf_module_init(void) +{ + __vfio_pci_register_vendor_driver(&sec_vf_mig_ops); + + __vfio_pci_register_vendor_driver(&hpre_vf_mig_ops); + + __vfio_pci_register_vendor_driver(&zip_vf_mig_ops); + + return 0; +}; + +static void __exit acc_vf_module_exit(void) +{ + vfio_pci_unregister_vendor_driver(&acc_vf_device_ops_node); +}; +module_init(acc_vf_module_init); +module_exit(acc_vf_module_exit); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Longfang Liu "); +MODULE_DESCRIPTION("HiSilicon Accelerator VF live migration driver"); diff --git a/kae_driver/hisilicon/migration/acc_vf_migration.h b/kae_driver/hisilicon/migration/acc_vf_migration.h new file mode 100755 index 0000000..981a273 --- /dev/null +++ b/kae_driver/hisilicon/migration/acc_vf_migration.h @@ -0,0 +1,242 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2021 HiSilicon Limited. */ + +#ifndef ACC_MIG_H +#define ACC_MIG_H + +#include +#include +#include + +#include "../hisi_acc_qm.h" + +#define VFIO_PCI_OFFSET_SHIFT 40 +#define VFIO_PCI_OFFSET_TO_INDEX(off) ((off) >> VFIO_PCI_OFFSET_SHIFT) +#define VFIO_PCI_INDEX_TO_OFFSET(index) ((u64)(index) << VFIO_PCI_OFFSET_SHIFT) +#define VFIO_PCI_OFFSET_MASK (((u64)(1) << VFIO_PCI_OFFSET_SHIFT) - 1) + +#define MIGRATION_REGION_SZ (sizeof(struct acc_vf_data) + \ + sizeof(struct vfio_device_migration_info)) +#define VFIO_DEV_DBG_LEN 256 +#define VFIO_DBG_LOG_LEN 16 +#define VFIO_DEVFN_MASK 0xFF + +#define PCI_BAR_2 2 +#define PCI_BAR_4 4 +#define POLL_PERIOD 10 +#define POLL_TIMEOUT 1000 +#define QM_CACHE_WB_START 0x204 +#define QM_CACHE_WB_DONE 0x208 +#define QM_MB_CMD_PAUSE_QM 0xe +#define QM_ABNORMAL_INT_STATUS 0x100008 +#define QM_IFC_INT_STATUS 0x0028 +#define SEC_CORE_INT_STATUS 0x301008 +#define HPRE_HAC_INT_STATUS 0x301800 +#define HZIP_CORE_INT_STATUS 0x3010AC + +#define QM_VFT_CFG_RDY 0x10006c +#define QM_VFT_CFG_OP_WR 0x100058 +#define QM_VFT_CFG_TYPE 0x10005c +#define QM_VFT_CFG 0x100060 +#define QM_VFT_CFG_OP_ENABLE 0x100054 +#define QM_VFT_CFG_DATA_L 0x100064 +#define QM_VFT_CFG_DATA_H 0x100068 + +#define ERROR_CHECK_TIMEOUT 100 +#define CHECK_DELAY_TIME 100 + +#define QM_SQC_VFT_BASE_SHIFT_V2 28 +#define QM_SQC_VFT_BASE_MASK_V2 GENMASK(15, 0) +#define QM_SQC_VFT_NUM_SHIFT_V2 45 +#define QM_SQC_VFT_NUM_MASK_V2 GENMASK(9, 0) + +/* mailbox */ +#define QM_MB_CMD_SQC_BT 0x4 +#define QM_MB_CMD_CQC_BT 0x5 +#define QM_MB_CMD_SQC_VFT_V2 0x6 + +#define QM_MB_CMD_SEND_BASE 0x300 +#define QM_MB_BUSY_SHIFT 13 +#define QM_MB_OP_SHIFT 14 +#define QM_MB_CMD_DATA_ADDR_L 0x304 +#define QM_MB_CMD_DATA_ADDR_H 0x308 +#define QM_MB_MAX_WAIT_CNT 6000 + +/* doorbell */ +#define QM_DOORBELL_CMD_SQ 0 +#define QM_DOORBELL_CMD_CQ 1 +#define QM_DOORBELL_SQ_CQ_BASE_V2 0x1000 +#define QM_DOORBELL_EQ_AEQ_BASE_V2 0x2000 +#define QM_DB_CMD_SHIFT_V2 12 +#define QM_DB_RAND_SHIFT_V2 16 +#define QM_DB_INDEX_SHIFT_V2 32 +#define QM_DB_PRIORITY_SHIFT_V2 48 + +/* RW regs */ +#define QM_REGS_MAX_LEN 7 +#define QM_REG_ADDR_OFFSET 0x0004 + +#define QM_XQC_ADDR_OFFSET 32U +#define QM_VF_AEQ_INT_MASK 0x0004 +#define QM_VF_EQ_INT_MASK 0x000c +#define QM_IFC_INT_SOURCE_V 0x0020 +#define QM_IFC_INT_MASK 0x0024 +#define QM_IFC_INT_SET_V 0x002c +#define QM_QUE_ISO_CFG_V 0x0030 +#define QM_PAGE_SIZE 0x0034 + +#define QM_EQC_DW0 0X8000 +#define QM_AEQC_DW0 0X8020 + +struct qm_mailbox { + __le16 w0; + __le16 queue_num; + __le32 base_l; + __le32 base_h; + __le32 rsvd; +}; + +enum acc_type { + HISI_SEC = 0x1, + HISI_HPRE = 0x2, + HISI_ZIP = 0x3, +}; + +struct vf_acc_type { + const char *name; + u32 type; +}; + +static struct vf_acc_type vf_acc_types[] = { + {"hisi_sec2", HISI_SEC}, + {"hisi_hpre", HISI_HPRE}, + {"hisi_zip", HISI_ZIP}, +}; + +enum mig_debug_cmd { + STATE_SAVE, + STATE_RESUME, + MB_TEST, + MIG_DATA_DUMP, + MIG_DEV_SHOW, +}; + +static const char * const vf_dev_state[] = { + "Stop", + "Running", + "Saving", + "Running & Saving", + "Resuming", +}; + +#define QM_MATCH_SIZE 32L +struct acc_vf_data { + /* QM match information */ + u32 qp_num; + u32 acc_type; + u32 que_iso_cfg; + u32 qp_base; + /* QM reserved 4 match information */ + u32 qm_rsv_state[4]; + + /* QM RW regs */ + u32 aeq_int_mask; + u32 eq_int_mask; + u32 ifc_int_source; + u32 ifc_int_mask; + u32 ifc_int_set; + u32 page_size; + u32 vf_state; + + /* + * QM_VF_MB has 4 regs don't need to migration + * mailbox regs writeback value will cause + * hardware to perform command operations + */ + + /* QM_EQC_DW has 7 regs */ + u32 qm_eqc_dw[7]; + + /* QM_AEQC_DW has 7 regs */ + u32 qm_aeqc_dw[7]; + + /* QM reserved 5 regs */ + u32 qm_rsv_regs[5]; + + /* qm memory init information */ + dma_addr_t eqe_dma; + dma_addr_t aeqe_dma; + dma_addr_t sqc_dma; + dma_addr_t cqc_dma; +}; + +struct acc_vf_remap_irq_ctx { + struct eventfd_ctx *trigger; + struct virqfd *sync; + atomic_t cnt; + wait_queue_head_t waitq; + bool init; +}; + +struct acc_vf_migration { + __u32 vf_vendor; + __u32 vf_device; + __u32 handle; + struct pci_dev *pf_dev; + struct pci_dev *vf_dev; + struct hisi_qm *pf_qm; + struct hisi_qm *vf_qm; + int vf_id; + int refcnt; + u8 acc_type; + bool mig_ignore; + struct mutex reflock; + + struct vfio_device_migration_info *mig_ctl; + struct acc_vf_data *vf_data; + bool in_dirty_track; + struct acc_vf_remap_irq_ctx remap_irq_ctx; + struct acc_vf_region *regions; + int num_regions; + struct dentry *debug_root; +}; + +struct acc_vf_region_ops { + int (*rw)(struct acc_vf_migration *acc_vf_dev, + char __user *buf, size_t count, + loff_t *ppos, bool iswrite); + void (*release)(struct acc_vf_migration *acc_vf_dev, + struct acc_vf_region *region); + int (*mmap)(struct acc_vf_migration *acc_vf_dev, + struct acc_vf_region *region, + struct vm_area_struct *vma); + int (*add_cap)(struct acc_vf_migration *acc_vf_dev, + struct acc_vf_region *region, + struct vfio_info_cap *caps); +}; + +struct acc_vf_region { + u32 type; + u32 subtype; + size_t size; + u32 flags; + const struct acc_vf_region_ops *ops; + void *data; +}; + +struct acc_vf_irqops { + int (*set_irqs)(struct acc_vf_migration *acc_vf_dev, + u32 flags, unsigned int index, + unsigned int start, unsigned int count, + void *data); +}; + +struct acc_vf_irq { + u32 type; + u32 subtype; + u32 flags; + u32 count; + const struct acc_vf_irqops *ops; +}; + +#endif /* ACC_MIG_H */ diff --git a/kae_driver/hisilicon/qm.c b/kae_driver/hisilicon/qm.c new file mode 100755 index 0000000..bac7bc9 --- /dev/null +++ b/kae_driver/hisilicon/qm.c @@ -0,0 +1,6630 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "../include_linux/uacce.h" +#include +#include "hisi_qm.h" +#include "hisi_acc_qm.h" + +/* eq/aeq irq enable */ +#define QM_VF_AEQ_INT_SOURCE 0x0 +#define QM_VF_AEQ_INT_MASK 0x4 +#define QM_VF_EQ_INT_SOURCE 0x8 +#define QM_VF_EQ_INT_MASK 0xc + +#define QM_IRQ_VECTOR_MASK GENMASK(15, 0) +#define QM_IRQ_TYPE_MASK GENMASK(15, 0) +#define QM_IRQ_TYPE_SHIFT 16 +#define QM_ABN_IRQ_TYPE_MASK GENMASK(7, 0) + +/* mailbox */ +#define QM_MB_PING_ALL_VFS 0xffff +#define QM_MB_CMD_DATA_SHIFT 32 +#define QM_MB_CMD_DATA_MASK GENMASK(31, 0) +#define QM_MB_STATUS_MASK GENMASK(12, 9) + +/* sqc shift */ +#define QM_SQ_HOP_NUM_SHIFT 0 +#define QM_SQ_PAGE_SIZE_SHIFT 4 +#define QM_SQ_BUF_SIZE_SHIFT 8 +#define QM_SQ_SQE_SIZE_SHIFT 12 +#define QM_SQ_PRIORITY_SHIFT 0 +#define QM_SQ_ORDERS_SHIFT 4 +#define QM_SQ_TYPE_SHIFT 8 +#define QM_QC_PASID_ENABLE 0x1 +#define QM_QC_PASID_ENABLE_SHIFT 7 + +#define QM_SQ_TYPE_MASK GENMASK(3, 0) +#define QM_SQ_TAIL_IDX(sqc) ((le16_to_cpu((sqc)->w11) >> 6) & 0x1) + +/* cqc shift */ +#define QM_CQ_HOP_NUM_SHIFT 0 +#define QM_CQ_PAGE_SIZE_SHIFT 4 +#define QM_CQ_BUF_SIZE_SHIFT 8 +#define QM_CQ_CQE_SIZE_SHIFT 12 +#define QM_CQ_PHASE_SHIFT 0 +#define QM_CQ_FLAG_SHIFT 1 + +#define QM_CQE_PHASE(cqe) (le16_to_cpu((cqe)->w7) & 0x1) +#define QM_QC_CQE_SIZE 4 +#define QM_CQ_TAIL_IDX(cqc) ((le16_to_cpu((cqc)->w11) >> 6) & 0x1) + +/* eqc shift */ +#define QM_EQE_AEQE_SIZE (2UL << 12) +#define QM_EQC_PHASE_SHIFT 16 + +#define QM_EQE_PHASE(eqe) ((le32_to_cpu((eqe)->dw0) >> 16) & 0x1) +#define QM_EQE_CQN_MASK GENMASK(15, 0) + +#define QM_AEQE_PHASE(aeqe) ((le32_to_cpu((aeqe)->dw0) >> 16) & 0x1) +#define QM_AEQE_TYPE_SHIFT 17 +#define QM_AEQE_CQN_MASK GENMASK(15, 0) +#define QM_CQ_OVERFLOW 0 +#define QM_EQ_OVERFLOW 1 +#define QM_CQE_ERROR 2 + +#define QM_XQ_DEPTH_SHIFT 16 +#define QM_XQ_DEPTH_MASK GENMASK(15, 0) + +#define QM_DOORBELL_CMD_SQ 0 +#define QM_DOORBELL_CMD_CQ 1 +#define QM_DOORBELL_CMD_EQ 2 +#define QM_DOORBELL_CMD_AEQ 3 + +#define QM_DOORBELL_BASE_V1 0x340 +#define QM_DB_CMD_SHIFT_V1 16 +#define QM_DB_INDEX_SHIFT_V1 32 +#define QM_DB_PRIORITY_SHIFT_V1 48 +#define QM_PAGE_SIZE 0x0034 +#define QM_QP_DB_INTERVAL 0x10000 + +#define QM_MEM_START_INIT 0x100040 +#define QM_MEM_INIT_DONE 0x100044 +#define QM_VFT_CFG_RDY 0x10006c +#define QM_VFT_CFG_OP_WR 0x100058 +#define QM_VFT_CFG_TYPE 0x10005c +#define QM_SQC_VFT 0x0 +#define QM_CQC_VFT 0x1 +#define QM_VFT_CFG 0x100060 +#define QM_VFT_CFG_OP_ENABLE 0x100054 +#define QM_PM_CTRL 0x100148 +#define QM_IDLE_DISABLE BIT(9) + +#define QM_VFT_CFG_DATA_L 0x100064 +#define QM_VFT_CFG_DATA_H 0x100068 +#define QM_SQC_VFT_BUF_SIZE (7ULL << 8) +#define QM_SQC_VFT_SQC_SIZE (5ULL << 12) +#define QM_SQC_VFT_INDEX_NUMBER (1ULL << 16) +#define QM_SQC_VFT_START_SQN_SHIFT 28 +#define QM_SQC_VFT_VALID (1ULL << 44) +#define QM_SQC_VFT_SQN_SHIFT 45 +#define QM_CQC_VFT_BUF_SIZE (7ULL << 8) +#define QM_CQC_VFT_SQC_SIZE (5ULL << 12) +#define QM_CQC_VFT_INDEX_NUMBER (1ULL << 16) +#define QM_CQC_VFT_VALID (1ULL << 28) + +#define QM_SQC_VFT_BASE_SHIFT_V2 28 +#define QM_SQC_VFT_BASE_MASK_V2 GENMASK(15, 0) +#define QM_SQC_VFT_NUM_SHIFT_V2 45 +#define QM_SQC_VFT_NUM_MASK_v2 GENMASK(9, 0) + +#define QM_DFX_CNT_CLR_CE 0x100118 + +#define QM_ABNORMAL_INT_SOURCE 0x100000 +#define QM_ABNORMAL_INT_MASK 0x100004 +#define QM_ABNORMAL_INT_MASK_VALUE 0x7fff +#define QM_ABNORMAL_INT_STATUS 0x100008 +#define QM_ABNORMAL_INT_SET 0x10000c +#define QM_ABNORMAL_INF00 0x100010 +#define QM_FIFO_OVERFLOW_TYPE 0xc0 +#define QM_FIFO_OVERFLOW_TYPE_SHIFT 6 +#define QM_FIFO_OVERFLOW_VF 0x3f +#define QM_ABNORMAL_INF01 0x100014 +#define QM_DB_TIMEOUT_TYPE 0xc0 +#define QM_DB_TIMEOUT_TYPE_SHIFT 6 +#define QM_DB_TIMEOUT_VF 0x3f +#define QM_RAS_CE_ENABLE 0x1000ec +#define QM_RAS_FE_ENABLE 0x1000f0 +#define QM_RAS_NFE_ENABLE 0x1000f4 +#define QM_RAS_CE_THRESHOLD 0x1000f8 +#define QM_RAS_CE_TIMES_PER_IRQ 1 +#define QM_OOO_SHUTDOWN_SEL 0x1040f8 +#define QM_ECC_MBIT BIT(2) +#define QM_DB_TIMEOUT BIT(10) +#define QM_OF_FIFO_OF BIT(11) + +#define QM_RESET_WAIT_TIMEOUT 400 +#define QM_PEH_VENDOR_ID 0x1000d8 +#define ACC_VENDOR_ID_VALUE 0x5a5a +#define QM_PEH_DFX_INFO0 0x1000fc +#define QM_PEH_DFX_INFO1 0x100100 +#define QM_PEH_DFX_MASK (BIT(0) | BIT(2)) +#define QM_PEH_MSI_FINISH_MASK GENMASK(19, 16) +#define ACC_PEH_SRIOV_CTRL_VF_MSE_SHIFT 3 +#define ACC_PEH_MSI_DISABLE GENMASK(31, 0) +#define ACC_MASTER_GLOBAL_CTRL_SHUTDOWN 0x1 +#define ACC_MASTER_TRANS_RETURN_RW 3 +#define ACC_MASTER_TRANS_RETURN 0x300150 +#define ACC_MASTER_GLOBAL_CTRL 0x300000 +#define ACC_AM_CFG_PORT_WR_EN 0x30001c +#define QM_RAS_NFE_MBIT_DISABLE ~QM_ECC_MBIT +#define ACC_AM_ROB_ECC_INT_STS 0x300104 +#define ACC_ROB_ECC_ERR_MULTPL BIT(1) +#define QM_MSI_CAP_ENABLE BIT(16) + +/* interfunction communication */ +#define QM_IFC_READY_STATUS 0x100128 +#define QM_IFC_C_STS_M 0x10012C +#define QM_IFC_INT_SET_P 0x100130 +#define QM_IFC_INT_CFG 0x100134 +#define QM_IFC_INT_SOURCE_P 0x100138 +#define QM_IFC_INT_SOURCE_V 0x0020 +#define QM_IFC_INT_MASK 0x0024 +#define QM_IFC_INT_STATUS 0x0028 +#define QM_IFC_INT_SET_V 0x002C +#define QM_IFC_SEND_ALL_VFS GENMASK(6, 0) +#define QM_IFC_INT_SOURCE_CLR GENMASK(63, 0) +#define QM_IFC_INT_SOURCE_MASK BIT(0) +#define QM_IFC_INT_DISABLE BIT(0) +#define QM_IFC_INT_STATUS_MASK BIT(0) +#define QM_IFC_INT_SET_MASK BIT(0) +#define QM_WAIT_DST_ACK 10 +#define QM_MAX_PF_WAIT_COUNT 10 +#define QM_MAX_VF_WAIT_COUNT 40 +#define QM_VF_RESET_WAIT_US 20000 +#define QM_VF_RESET_WAIT_CNT 3000 +#define QM_VF_RESET_WAIT_TIMEOUT_US \ + (QM_VF_RESET_WAIT_US * QM_VF_RESET_WAIT_CNT) + +#define QM_DFX_MB_CNT_VF 0x104010 +#define QM_DFX_DB_CNT_VF 0x104020 +#define QM_DFX_SQE_CNT_VF_SQN 0x104030 +#define QM_DFX_CQE_CNT_VF_CQN 0x104040 +#define QM_DFX_QN_SHIFT 16 +#define CURRENT_FUN_MASK GENMASK(5, 0) +#define CURRENT_Q_MASK GENMASK(31, 16) + +#define POLL_PERIOD 10 +#define POLL_TIMEOUT 1000 +#define WAIT_PERIOD_US_MAX 200 +#define WAIT_PERIOD_US_MIN 100 +#define MAX_WAIT_COUNTS 1000 +#define QM_CACHE_WB_START 0x204 +#define QM_CACHE_WB_DONE 0x208 +#define QM_FUNC_CAPS_REG 0x3100 +#define QM_CAPBILITY_VERSION GENMASK(7, 0) + +#define PCI_BAR_2 2 +#define PCI_BAR_4 4 +#define QM_SQE_DATA_ALIGN_MASK GENMASK(6, 0) +#define QMC_ALIGN(sz) ALIGN(sz, 32) + +#define QM_DBG_READ_LEN 256 +#define QM_DBG_WRITE_LEN 1024 +#define QM_DBG_TMP_BUF_LEN 22 +#define QM_PCI_COMMAND_INVALID ~0 +#define QM_RESET_STOP_TX_OFFSET 1 +#define QM_RESET_STOP_RX_OFFSET 2 + +#define WAIT_PERIOD 20 +#define REMOVE_WAIT_DELAY 10 +#define QM_SQE_ADDR_MASK GENMASK(7, 0) + +#define QM_DRIVER_REMOVING 0 +#define QM_RST_SCHED 1 +#define QM_RESETTING 2 +#define QM_QOS_PARAM_NUM 2 +#define QM_QOS_VAL_NUM 1 +#define QM_QOS_BDF_PARAM_NUM 4 +#define QM_QOS_MAX_VAL 1000 +#define QM_QOS_RATE 100 +#define QM_QOS_EXPAND_RATE 1000 +#define QM_SHAPER_CIR_B_MASK GENMASK(7, 0) +#define QM_SHAPER_CIR_U_MASK GENMASK(10, 8) +#define QM_SHAPER_CIR_S_MASK GENMASK(14, 11) +#define QM_SHAPER_FACTOR_CIR_U_SHIFT 8 +#define QM_SHAPER_FACTOR_CIR_S_SHIFT 11 +#define QM_SHAPER_FACTOR_CBS_B_SHIFT 15 +#define QM_SHAPER_FACTOR_CBS_S_SHIFT 19 +#define QM_SHAPER_CBS_B 1 +#define QM_SHAPER_CBS_S 16 +#define QM_SHAPER_VFT_OFFSET 6 +#define WAIT_FOR_QOS_VF 100 +#define QM_QOS_MIN_ERROR_RATE 5 +#define QM_QOS_TYPICAL_NUM 8 +#define QM_SHAPER_MIN_CBS_S 8 +#define QM_QOS_TICK 0x300U +#define QM_QOS_DIVISOR_CLK 0x1f40U +#define QM_QOS_MAX_CIR_B 200 +#define QM_QOS_MIN_CIR_B 100 +#define QM_QOS_MAX_CIR_U 6 +#define QM_QOS_MAX_CIR_S 11 +#define QM_DFX_BASE 0x0100000 +#define QM_DFX_STATE1 0x0104000 +#define QM_DFX_STATE2 0x01040C8 +#define QM_DFX_COMMON 0x0000 +#define QM_DFX_BASE_LEN 0x5A +#define QM_DFX_STATE1_LEN 0x2E +#define QM_DFX_STATE2_LEN 0x11 +#define QM_DFX_COMMON_LEN 0xC3 +#define QM_DFX_REGS_LEN 4UL +#define QM_AUTOSUSPEND_DELAY 3000 + +#define QM_MK_CQC_DW3_V1(hop_num, pg_sz, buf_sz, cqe_sz) \ + (((hop_num) << QM_CQ_HOP_NUM_SHIFT) | \ + ((pg_sz) << QM_CQ_PAGE_SIZE_SHIFT) | \ + ((buf_sz) << QM_CQ_BUF_SIZE_SHIFT) | \ + ((cqe_sz) << QM_CQ_CQE_SIZE_SHIFT)) + +#define QM_MK_CQC_DW3_V2(cqe_sz, cq_depth) \ + ((((u32)cq_depth) - 1) | ((cqe_sz) << QM_CQ_CQE_SIZE_SHIFT)) + +#define QM_MK_SQC_W13(priority, orders, alg_type) \ + (((priority) << QM_SQ_PRIORITY_SHIFT) | \ + ((orders) << QM_SQ_ORDERS_SHIFT) | \ + (((alg_type) & QM_SQ_TYPE_MASK) << QM_SQ_TYPE_SHIFT)) + +#define QM_MK_SQC_DW3_V1(hop_num, pg_sz, buf_sz, sqe_sz) \ + (((hop_num) << QM_SQ_HOP_NUM_SHIFT) | \ + ((pg_sz) << QM_SQ_PAGE_SIZE_SHIFT) | \ + ((buf_sz) << QM_SQ_BUF_SIZE_SHIFT) | \ + ((u32)ilog2(sqe_sz) << QM_SQ_SQE_SIZE_SHIFT)) + +#define QM_MK_SQC_DW3_V2(sqe_sz, sq_depth) \ + ((((u32)sq_depth) - 1) | ((u32)ilog2(sqe_sz) << QM_SQ_SQE_SIZE_SHIFT)) + +#define INIT_QC_COMMON(qc, base, pasid) do { \ + (qc)->head = 0; \ + (qc)->tail = 0; \ + (qc)->base_l = cpu_to_le32(lower_32_bits(base)); \ + (qc)->base_h = cpu_to_le32(upper_32_bits(base)); \ + (qc)->dw3 = 0; \ + (qc)->w8 = 0; \ + (qc)->rsvd0 = 0; \ + (qc)->pasid = cpu_to_le16(pasid); \ + (qc)->w11 = 0; \ + (qc)->rsvd1 = 0; \ +} while (0) + +enum vft_type { + SQC_VFT = 0, + CQC_VFT, + SHAPER_VFT, +}; + +enum acc_err_result { + ACC_ERR_NONE, + ACC_ERR_NEED_RESET, + ACC_ERR_RECOVERED, +}; + +enum qm_alg_type { + ALG_TYPE_0, + ALG_TYPE_1, +}; + +enum qm_mb_cmd { + QM_PF_FLR_PREPARE = 0x01, + QM_PF_SRST_PREPARE, + QM_PF_RESET_DONE, + QM_VF_PREPARE_DONE, + QM_VF_PREPARE_FAIL, + QM_VF_START_DONE, + QM_VF_START_FAIL, + QM_PF_SET_QOS, + QM_VF_GET_QOS, +}; + +enum qm_basic_type { + QM_TOTAL_QP_NUM_CAP = 0x0, + QM_FUNC_MAX_QP_CAP, + QM_XEQ_DEPTH_CAP, + QM_QP_DEPTH_CAP, + QM_EQ_IRQ_TYPE_CAP, + QM_AEQ_IRQ_TYPE_CAP, + QM_ABN_IRQ_TYPE_CAP, + QM_PF2VF_IRQ_TYPE_CAP, + QM_PF_IRQ_NUM_CAP, + QM_VF_IRQ_NUM_CAP, +}; + +static const struct hisi_qm_cap_info qm_cap_info_comm[] = { + {QM_SUPPORT_DB_ISOLATION, 0x30, 0, BIT(0), 0x0, 0x0, 0x0}, + {QM_SUPPORT_FUNC_QOS, 0x3100, 0, BIT(8), 0x0, 0x0, 0x1}, + {QM_SUPPORT_STOP_QP, 0x3100, 0, BIT(9), 0x0, 0x0, 0x1}, + {QM_SUPPORT_MB_COMMAND, 0x3100, 0, BIT(11), 0x0, 0x0, 0x1}, + {QM_SUPPORT_SVA_PREFETCH, 0x3100, 0, BIT(14), 0x0, 0x0, 0x1}, +}; + +static const struct hisi_qm_cap_info qm_cap_info_pf[] = { + {QM_SUPPORT_RPM, 0x3100, 0, BIT(13), 0x0, 0x0, 0x1}, +}; + +static const struct hisi_qm_cap_info qm_cap_info_vf[] = { + {QM_SUPPORT_RPM, 0x3100, 0, BIT(12), 0x0, 0x0, 0x0}, +}; + +static const struct hisi_qm_cap_info qm_basic_info[] = { + {QM_TOTAL_QP_NUM_CAP, 0x100158, 0, GENMASK(10, 0), 0x1000, 0x400, 0x400}, + {QM_FUNC_MAX_QP_CAP, 0x100158, 11, GENMASK(10, 0), 0x1000, 0x400, 0x400}, + {QM_XEQ_DEPTH_CAP, 0x3104, 0, GENMASK(15, 0), 0x800, 0x4000800, 0x4000800}, + {QM_QP_DEPTH_CAP, 0x3108, 0, GENMASK(31, 0), 0x4000400, 0x4000400, 0x4000400}, + {QM_EQ_IRQ_TYPE_CAP, 0x310c, 0, GENMASK(31, 0), 0x10000, 0x10000, 0x10000}, + {QM_AEQ_IRQ_TYPE_CAP, 0x3110, 0, GENMASK(31, 0), 0x0, 0x10001, 0x10001}, + {QM_ABN_IRQ_TYPE_CAP, 0x3114, 0, GENMASK(31, 0), 0x0, 0x10003, 0x10003}, + {QM_PF2VF_IRQ_TYPE_CAP, 0x3118, 0, GENMASK(31, 0), 0x0, 0x0, 0x10002}, + {QM_PF_IRQ_NUM_CAP, 0x311c, 16, GENMASK(15, 0), 0x1, 0x4, 0x4}, + {QM_VF_IRQ_NUM_CAP, 0x311c, 0, GENMASK(15, 0), 0x1, 0x2, 0x3}, +}; + +struct qm_cqe { + __le32 rsvd0; + __le16 cmd_id; + __le16 rsvd1; + __le16 sq_head; + __le16 sq_num; + __le16 rsvd2; + __le16 w7; +}; + +struct qm_eqe { + __le32 dw0; +}; + +struct qm_aeqe { + __le32 dw0; +}; + +struct qm_sqc { + __le16 head; + __le16 tail; + __le32 base_l; + __le32 base_h; + __le32 dw3; + __le16 w8; + __le16 rsvd0; + __le16 pasid; + __le16 w11; + __le16 cq_num; + __le16 w13; + __le32 rsvd1; +}; + +struct qm_cqc { + __le16 head; + __le16 tail; + __le32 base_l; + __le32 base_h; + __le32 dw3; + __le16 w8; + __le16 rsvd0; + __le16 pasid; + __le16 w11; + __le32 dw6; + __le32 rsvd1; +}; + +struct qm_eqc { + __le16 head; + __le16 tail; + __le32 base_l; + __le32 base_h; + __le32 dw3; + __le32 rsvd[2]; + __le32 dw6; +}; + +struct qm_aeqc { + __le16 head; + __le16 tail; + __le32 base_l; + __le32 base_h; + __le32 dw3; + __le32 rsvd[2]; + __le32 dw6; +}; + +struct qm_mailbox { + __le16 w0; + __le16 queue_num; + __le32 base_l; + __le32 base_h; + __le32 rsvd; +}; + +struct qm_doorbell { + __le16 queue_num; + __le16 cmd; + __le16 index; + __le16 priority; +}; + +struct hisi_qm_resource { + struct hisi_qm *qm; + int distance; + struct list_head list; +}; + +struct hisi_qm_hw_ops { + int (*get_vft)(struct hisi_qm *qm, u32 *base, u32 *number); + void (*qm_db)(struct hisi_qm *qm, u16 qn, + u8 cmd, u16 index, u8 priority); + int (*debug_init)(struct hisi_qm *qm); + void (*hw_error_init)(struct hisi_qm *qm); + void (*hw_error_uninit)(struct hisi_qm *qm); + enum acc_err_result (*hw_error_handle)(struct hisi_qm *qm); + int (*set_msi)(struct hisi_qm *qm, bool set); +}; + +struct qm_dfx_item { + const char *name; + u32 offset; +}; + +static struct qm_dfx_item qm_dfx_files[] = { + {"err_irq", offsetof(struct qm_dfx, err_irq_cnt)}, + {"aeq_irq", offsetof(struct qm_dfx, aeq_irq_cnt)}, + {"abnormal_irq", offsetof(struct qm_dfx, abnormal_irq_cnt)}, + {"create_qp_err", offsetof(struct qm_dfx, create_qp_err_cnt)}, + {"mb_err", offsetof(struct qm_dfx, mb_err_cnt)}, +}; + +static const char * const qm_debug_file_name[] = { + [CURRENT_QM] = "current_qm", + [CURRENT_Q] = "current_q", + [CLEAR_ENABLE] = "clear_enable", +}; + +struct hisi_qm_hw_error { + u32 int_msk; + const char *msg; +}; + +static const struct hisi_qm_hw_error qm_hw_error[] = { + { .int_msk = BIT(0), .msg = "qm_axi_rresp" }, + { .int_msk = BIT(1), .msg = "qm_axi_bresp" }, + { .int_msk = BIT(2), .msg = "qm_ecc_mbit" }, + { .int_msk = BIT(3), .msg = "qm_ecc_1bit" }, + { .int_msk = BIT(4), .msg = "qm_acc_get_task_timeout" }, + { .int_msk = BIT(5), .msg = "qm_acc_do_task_timeout" }, + { .int_msk = BIT(6), .msg = "qm_acc_wb_not_ready_timeout" }, + { .int_msk = BIT(7), .msg = "qm_sq_cq_vf_invalid" }, + { .int_msk = BIT(8), .msg = "qm_cq_vf_invalid" }, + { .int_msk = BIT(9), .msg = "qm_sq_vf_invalid" }, + { .int_msk = BIT(10), .msg = "qm_db_timeout" }, + { .int_msk = BIT(11), .msg = "qm_of_fifo_of" }, + { .int_msk = BIT(12), .msg = "qm_db_random_invalid" }, + { .int_msk = BIT(13), .msg = "qm_mailbox_timeout" }, + { .int_msk = BIT(14), .msg = "qm_flr_timeout" }, + { /* sentinel */ } +}; + +/* define the QM's dfx regs region and region length */ +static struct dfx_diff_registers qm_diff_regs[] = { + { + .reg_offset = QM_DFX_BASE, + .reg_len = QM_DFX_BASE_LEN, + }, { + .reg_offset = QM_DFX_STATE1, + .reg_len = QM_DFX_STATE1_LEN, + }, { + .reg_offset = QM_DFX_STATE2, + .reg_len = QM_DFX_STATE2_LEN, + }, { + .reg_offset = QM_DFX_COMMON, + .reg_len = QM_DFX_COMMON_LEN, + }, +}; + +static const char * const qm_db_timeout[] = { + "sq", "cq", "eq", "aeq", +}; + +static const char * const qm_fifo_overflow[] = { + "cq", "eq", "aeq", +}; + +static const char * const qm_s[] = { + "init", "start", "close", "stop", +}; + +static const char * const qp_s[] = { + "none", "init", "start", "stop", "close", +}; + +struct qm_typical_qos_table { + u32 start; + u32 end; + u32 val; +}; + +/* the qos step is 100 */ +static struct qm_typical_qos_table shaper_cir_s[] = { + {100, 100, 4}, + {200, 200, 3}, + {300, 500, 2}, + {600, 1000, 1}, + {1100, 100000, 0}, +}; + +static struct qm_typical_qos_table shaper_cbs_s[] = { + {100, 200, 9}, + {300, 500, 11}, + {600, 1000, 12}, + {1100, 10000, 16}, + {10100, 25000, 17}, + {25100, 50000, 18}, + {50100, 100000, 19} +}; + +static void qm_irqs_unregister(struct hisi_qm *qm); + +static bool qm_avail_state(struct hisi_qm *qm, enum qm_state new) +{ + enum qm_state curr = atomic_read(&qm->status.flags); + bool avail = false; + + switch (curr) { + case QM_INIT: + if (new == QM_START || new == QM_CLOSE) + avail = true; + break; + case QM_START: + if (new == QM_STOP) + avail = true; + break; + case QM_STOP: + if (new == QM_CLOSE || new == QM_START) + avail = true; + break; + default: + break; + } + + dev_dbg(&qm->pdev->dev, "change qm state from %s to %s\n", + qm_s[curr], qm_s[new]); + + if (!avail) + dev_warn(&qm->pdev->dev, "Can not change qm state from %s to %s\n", + qm_s[curr], qm_s[new]); + + return avail; +} + +static bool qm_qp_avail_state(struct hisi_qm *qm, struct hisi_qp *qp, + enum qp_state new) +{ + enum qm_state qm_curr = atomic_read(&qm->status.flags); + enum qp_state qp_curr = 0; + bool avail = false; + + if (qp) + qp_curr = atomic_read(&qp->qp_status.flags); + + switch (new) { + case QP_INIT: + if (qm_curr == QM_START || qm_curr == QM_INIT) + avail = true; + break; + case QP_START: + if ((qm_curr == QM_START && qp_curr == QP_INIT) || + (qm_curr == QM_START && qp_curr == QP_STOP)) + avail = true; + break; + case QP_STOP: + if ((qm_curr == QM_START && qp_curr == QP_START) || + (qp_curr == QP_INIT)) + avail = true; + break; + case QP_CLOSE: + if ((qm_curr == QM_START && qp_curr == QP_INIT) || + (qm_curr == QM_START && qp_curr == QP_STOP) || + (qm_curr == QM_STOP && qp_curr == QP_STOP) || + (qm_curr == QM_STOP && qp_curr == QP_INIT)) + avail = true; + break; + default: + break; + } + + dev_dbg(&qm->pdev->dev, "change qp state from %s to %s in QM %s\n", + qp_s[qp_curr], qp_s[new], qm_s[qm_curr]); + + if (!avail) + dev_warn(&qm->pdev->dev, + "Can not change qp state from %s to %s in QM %s\n", + qp_s[qp_curr], qp_s[new], qm_s[qm_curr]); + + return avail; +} + +static u32 qm_get_hw_error_status(struct hisi_qm *qm) +{ + return readl(qm->io_base + QM_ABNORMAL_INT_STATUS); +} + +static u32 qm_get_dev_err_status(struct hisi_qm *qm) +{ + return qm->err_ini->get_dev_hw_err_status(qm); +} + +/* Check if the error causes the master ooo block */ +static bool qm_check_dev_error(struct hisi_qm *qm) +{ + u32 val, dev_val; + + if (qm->fun_type == QM_HW_VF) + return false; + + val = qm_get_hw_error_status(qm) & qm->err_info.qm_shutdown_mask; + dev_val = qm_get_dev_err_status(qm) & qm->err_info.dev_shutdown_mask; + + return val || dev_val; +} + +static int qm_wait_reset_finish(struct hisi_qm *qm) +{ + int delay = 0; + + /* All reset requests need to be queued for processing */ + while (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) { + msleep(++delay); + if (delay > QM_RESET_WAIT_TIMEOUT) + return -EBUSY; + } + + return 0; +} + +static int qm_reset_prepare_ready(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(pdev)); + + /* + * PF and VF on host doesnot support resetting at the + * same time on Kunpeng920. + */ + if (qm->ver < QM_HW_V3) + return qm_wait_reset_finish(pf_qm); + + return qm_wait_reset_finish(qm); +} + +static void qm_reset_bit_clear(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(pdev)); + + if (qm->ver < QM_HW_V3) + clear_bit(QM_RESETTING, &pf_qm->misc_ctl); + + clear_bit(QM_RESETTING, &qm->misc_ctl); +} + +static void qm_mb_pre_init(struct qm_mailbox *mailbox, u8 cmd, + u64 base, u16 queue, bool op) +{ + mailbox->w0 = cpu_to_le16((cmd) | + ((op) ? 0x1 << QM_MB_OP_SHIFT : 0) | + (0x1 << QM_MB_BUSY_SHIFT)); + mailbox->queue_num = cpu_to_le16(queue); + mailbox->base_l = cpu_to_le32(lower_32_bits(base)); + mailbox->base_h = cpu_to_le32(upper_32_bits(base)); + mailbox->rsvd = 0; +} + +/* return 0 mailbox ready, -ETIMEDOUT hardware timeout */ +int hisi_qm_wait_mb_ready(struct hisi_qm *qm) +{ + u32 val; + + return readl_relaxed_poll_timeout(qm->io_base + QM_MB_CMD_SEND_BASE, + val, !((val >> QM_MB_BUSY_SHIFT) & + 0x1), POLL_PERIOD, POLL_TIMEOUT); +} +EXPORT_SYMBOL_GPL(hisi_qm_wait_mb_ready); + +/* 128 bit should be written to hardware at one time to trigger a mailbox */ +static void qm_mb_write(struct hisi_qm *qm, const void *src) +{ + void __iomem *fun_base = qm->io_base + QM_MB_CMD_SEND_BASE; + unsigned long tmp0 = 0, tmp1 = 0; + + if (!IS_ENABLED(CONFIG_ARM64)) { + memcpy_toio(fun_base, src, 16); + dma_wmb(); + return; + } + + asm volatile("ldp %0, %1, %3\n" + "stp %0, %1, %2\n" + "dmb oshst\n" + : "=&r" (tmp0), + "=&r" (tmp1), + "+Q" (*((char __iomem *)fun_base)) + : "Q" (*((char *)src)) + : "memory"); +} + +static int qm_mb_nolock(struct hisi_qm *qm, struct qm_mailbox *mailbox) +{ + int ret; + u32 val; + + if (unlikely(hisi_qm_wait_mb_ready(qm))) { + dev_err(&qm->pdev->dev, "QM mailbox is busy to start!\n"); + ret = -EBUSY; + goto mb_busy; + } + + qm_mb_write(qm, mailbox); + + if (unlikely(hisi_qm_wait_mb_ready(qm))) { + dev_err(&qm->pdev->dev, "QM mailbox operation timeout!\n"); + ret = -ETIMEDOUT; + goto mb_busy; + } + + val = readl(qm->io_base + QM_MB_CMD_SEND_BASE); + if (val & QM_MB_STATUS_MASK) { + dev_err(&qm->pdev->dev, "QM mailbox operation failed!\n"); + ret = -EIO; + goto mb_busy; + } + + return 0; + +mb_busy: + atomic64_inc(&qm->debug.dfx.mb_err_cnt); + return ret; +} + +int hisi_qm_mb(struct hisi_qm *qm, u8 cmd, dma_addr_t dma_addr, u16 queue, + bool op) +{ + struct qm_mailbox mailbox; + int ret; + + dev_dbg(&qm->pdev->dev, "QM mailbox request to q%u: %u-%llx\n", + queue, cmd, (unsigned long long)dma_addr); + + qm_mb_pre_init(&mailbox, cmd, dma_addr, queue, op); + + mutex_lock(&qm->mailbox_lock); + ret = qm_mb_nolock(qm, &mailbox); + mutex_unlock(&qm->mailbox_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_mb); + +static void qm_db_v1(struct hisi_qm *qm, u16 qn, u8 cmd, u16 index, u8 priority) +{ + u64 doorbell; + + doorbell = qn | ((u64)cmd << QM_DB_CMD_SHIFT_V1) | + ((u64)index << QM_DB_INDEX_SHIFT_V1) | + ((u64)priority << QM_DB_PRIORITY_SHIFT_V1); + + writeq(doorbell, qm->io_base + QM_DOORBELL_BASE_V1); +} + +static void qm_db_v2(struct hisi_qm *qm, u16 qn, u8 cmd, u16 index, u8 priority) +{ + void __iomem *io_base = qm->io_base; + u16 randata = 0; + u64 doorbell; + + if (cmd == QM_DOORBELL_CMD_SQ || cmd == QM_DOORBELL_CMD_CQ) + io_base = qm->db_io_base + (u64)qn * qm->db_interval + + QM_DOORBELL_SQ_CQ_BASE_V2; + else + io_base += QM_DOORBELL_EQ_AEQ_BASE_V2; + + doorbell = qn | ((u64)cmd << QM_DB_CMD_SHIFT_V2) | + ((u64)randata << QM_DB_RAND_SHIFT_V2) | + ((u64)index << QM_DB_INDEX_SHIFT_V2) | + ((u64)priority << QM_DB_PRIORITY_SHIFT_V2); + + writeq(doorbell, io_base); +} + +static void qm_db(struct hisi_qm *qm, u16 qn, u8 cmd, u16 index, u8 priority) +{ + dev_dbg(&qm->pdev->dev, "QM doorbell request: qn=%u, cmd=%u, index=%u\n", + qn, cmd, index); + + qm->ops->qm_db(qm, qn, cmd, index, priority); +} + +static void qm_disable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + /* if qm enables clock gating in Kunpeng930, qos will be inaccurate. */ + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + QM_PM_CTRL); + val |= QM_IDLE_DISABLE; + writel(val, qm->io_base + QM_PM_CTRL); +} + +static int qm_dev_mem_reset(struct hisi_qm *qm) +{ + u32 val; + + writel(0x1, qm->io_base + QM_MEM_START_INIT); + return readl_relaxed_poll_timeout(qm->io_base + QM_MEM_INIT_DONE, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); +} + +/** + * hisi_qm_get_hw_info() - Get device information. + * @qm: The qm which want to get information. + * @info_table: Array for storing device information. + * @index: Index in info_table. + * @is_read: Whether read from reg, 0: not support read from reg. + * + * This function returns device information the caller needs. + */ +u32 hisi_qm_get_hw_info(struct hisi_qm *qm, + const struct hisi_qm_cap_info *info_table, + u32 index, bool is_read) +{ + u32 val; + + switch (qm->ver) { + case QM_HW_V1: + return info_table[index].v1_val; + case QM_HW_V2: + return info_table[index].v2_val; + default: + if (!is_read) + return info_table[index].v3_val; + + val = readl(qm->io_base + info_table[index].offset); + return (val >> info_table[index].shift) & info_table[index].mask; + } +} +EXPORT_SYMBOL_GPL(hisi_qm_get_hw_info); + +static void qm_get_xqc_depth(struct hisi_qm *qm, u16 *low_bits, + u16 *high_bits, enum qm_basic_type type) +{ + u32 depth; + + depth = hisi_qm_get_hw_info(qm, qm_basic_info, type, qm->cap_ver); + *low_bits = depth & QM_XQ_DEPTH_MASK; + *high_bits = (depth >> QM_XQ_DEPTH_SHIFT) & QM_XQ_DEPTH_MASK; +} + +static u32 qm_get_irq_num(struct hisi_qm *qm) +{ + if (qm->fun_type == QM_HW_PF) + return hisi_qm_get_hw_info(qm, qm_basic_info, QM_PF_IRQ_NUM_CAP, qm->cap_ver); + + return hisi_qm_get_hw_info(qm, qm_basic_info, QM_VF_IRQ_NUM_CAP, qm->cap_ver); +} + +static int qm_pm_get_sync(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + if (!test_bit(QM_SUPPORT_RPM, &qm->caps)) + return 0; + + ret = pm_runtime_resume_and_get(dev); + if (ret < 0) { + dev_err(dev, "failed to get_sync(%d).\n", ret); + return ret; + } + + return 0; +} + +static void qm_pm_put_sync(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + + if (!test_bit(QM_SUPPORT_RPM, &qm->caps)) + return; + + pm_runtime_mark_last_busy(dev); + pm_runtime_put_autosuspend(dev); +} + +static void qm_cq_head_update(struct hisi_qp *qp) +{ + if (qp->qp_status.cq_head == qp->cq_depth - 1) { + qp->qp_status.cqc_phase = !qp->qp_status.cqc_phase; + qp->qp_status.cq_head = 0; + } else { + qp->qp_status.cq_head++; + } +} + +static void qm_poll_req_cb(struct hisi_qp *qp) +{ + struct qm_cqe *cqe = qp->cqe + qp->qp_status.cq_head; + struct hisi_qm *qm = qp->qm; + + while (QM_CQE_PHASE(cqe) == qp->qp_status.cqc_phase) { + dma_rmb(); + qp->req_cb(qp, qp->sqe + qm->sqe_size * + le16_to_cpu(cqe->sq_head)); + qm_cq_head_update(qp); + cqe = qp->cqe + qp->qp_status.cq_head; + qm_db(qm, qp->qp_id, QM_DOORBELL_CMD_CQ, + qp->qp_status.cq_head, 0); + atomic_dec(&qp->qp_status.used); + } + + /* set c_flag */ + qm_db(qm, qp->qp_id, QM_DOORBELL_CMD_CQ, qp->qp_status.cq_head, 1); +} + +static int qm_get_complete_eqe_num(struct hisi_qm_poll_data *poll_data) +{ + struct hisi_qm *qm = poll_data->qm; + struct qm_eqe *eqe = qm->eqe + qm->status.eq_head; + u16 eq_depth = qm->eq_depth; + int eqe_num = 0; + u16 cqn; + + while (QM_EQE_PHASE(eqe) == qm->status.eqc_phase) { + cqn = le32_to_cpu(eqe->dw0) & QM_EQE_CQN_MASK; + poll_data->qp_finish_id[eqe_num] = cqn; + eqe_num++; + + if (qm->status.eq_head == eq_depth - 1) { + qm->status.eqc_phase = !qm->status.eqc_phase; + eqe = qm->eqe; + qm->status.eq_head = 0; + } else { + eqe++; + qm->status.eq_head++; + } + + if (eqe_num == (eq_depth >> 1) - 1) + break; + } + + qm_db(qm, 0, QM_DOORBELL_CMD_EQ, qm->status.eq_head, 0); + + return eqe_num; +} + +static void qm_work_process(struct work_struct *work) +{ + struct hisi_qm_poll_data *poll_data = + container_of(work, struct hisi_qm_poll_data, work); + struct hisi_qm *qm = poll_data->qm; + struct hisi_qp *qp; + int eqe_num, i; + + /* Get qp id of completed tasks and re-enable the interrupt. */ + eqe_num = qm_get_complete_eqe_num(poll_data); + for (i = eqe_num - 1; i >= 0; i--) { + qp = &qm->qp_array[poll_data->qp_finish_id[i]]; + if (unlikely(atomic_read(&qp->qp_status.flags) == QP_STOP)) + continue; + + if (qp->event_cb) { + qp->event_cb(qp); + continue; + } + + if (likely(qp->req_cb)) + qm_poll_req_cb(qp); + } +} + +static bool do_qm_irq(struct hisi_qm *qm) +{ + struct qm_eqe *eqe = qm->eqe + qm->status.eq_head; + struct hisi_qm_poll_data *poll_data; + u16 cqn; + + if (!readl(qm->io_base + QM_VF_EQ_INT_SOURCE)) + return false; + + if (QM_EQE_PHASE(eqe) == qm->status.eqc_phase) { + cqn = le32_to_cpu(eqe->dw0) & QM_EQE_CQN_MASK; + poll_data = &qm->poll_data[cqn]; + queue_work(qm->wq, &poll_data->work); + + return true; + } + + return false; +} + +static irqreturn_t qm_irq(int irq, void *data) +{ + struct hisi_qm *qm = data; + bool ret; + + ret = do_qm_irq(qm); + if (ret) + return IRQ_HANDLED; + + atomic64_inc(&qm->debug.dfx.err_irq_cnt); + qm_db(qm, 0, QM_DOORBELL_CMD_EQ, qm->status.eq_head, 0); + + return IRQ_NONE; +} + +static irqreturn_t qm_mb_cmd_irq(int irq, void *data) +{ + struct hisi_qm *qm = data; + u32 val; + + val = readl(qm->io_base + QM_IFC_INT_STATUS); + val &= QM_IFC_INT_STATUS_MASK; + if (!val) + return IRQ_NONE; + + schedule_work(&qm->cmd_process); + + return IRQ_HANDLED; +} + +static void qm_set_qp_disable(struct hisi_qp *qp, int offset) +{ + u32 *addr; + + if (qp->is_in_kernel) + return; + + addr = (u32 *)(qp->qdma.va + qp->qdma.size) - offset; + *addr = 1; + + /* make sure setup is completed */ + smp_wmb(); +} + +static void qm_disable_qp(struct hisi_qm *qm, u32 qp_id) +{ + struct hisi_qp *qp = &qm->qp_array[qp_id]; + + qm_set_qp_disable(qp, QM_RESET_STOP_TX_OFFSET); + hisi_qm_stop_qp(qp); + qm_set_qp_disable(qp, QM_RESET_STOP_RX_OFFSET); +} + +static void qm_reset_function(struct hisi_qm *qm) +{ + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(qm->pdev)); + struct device *dev = &qm->pdev->dev; + int ret; + + if (qm_check_dev_error(pf_qm)) + return; + + ret = qm_reset_prepare_ready(qm); + if (ret) { + dev_err(dev, "reset function not ready\n"); + return; + } + + ret = hisi_qm_stop(qm, QM_FLR); + if (ret) { + dev_err(dev, "failed to stop qm when reset function\n"); + goto clear_bit; + } + + ret = hisi_qm_start(qm); + if (ret) + dev_err(dev, "failed to start qm when reset function\n"); + +clear_bit: + qm_reset_bit_clear(qm); +} + +static irqreturn_t qm_aeq_thread(int irq, void *data) +{ + struct hisi_qm *qm = data; + struct qm_aeqe *aeqe = qm->aeqe + qm->status.aeq_head; + u16 aeq_depth = qm->aeq_depth; + u32 type, qp_id; + + while (QM_AEQE_PHASE(aeqe) == qm->status.aeqc_phase) { + type = le32_to_cpu(aeqe->dw0) >> QM_AEQE_TYPE_SHIFT; + qp_id = le32_to_cpu(aeqe->dw0) & QM_AEQE_CQN_MASK; + + switch (type) { + case QM_EQ_OVERFLOW: + dev_err(&qm->pdev->dev, "eq overflow, reset function\n"); + qm_reset_function(qm); + return IRQ_HANDLED; + case QM_CQ_OVERFLOW: + dev_err(&qm->pdev->dev, "cq overflow, stop qp(%u)\n", + qp_id); + fallthrough; + case QM_CQE_ERROR: + qm_disable_qp(qm, qp_id); + break; + default: + dev_err(&qm->pdev->dev, "unknown error type %u\n", + type); + break; + } + + if (qm->status.aeq_head == aeq_depth - 1) { + qm->status.aeqc_phase = !qm->status.aeqc_phase; + aeqe = qm->aeqe; + qm->status.aeq_head = 0; + } else { + aeqe++; + qm->status.aeq_head++; + } + } + + qm_db(qm, 0, QM_DOORBELL_CMD_AEQ, qm->status.aeq_head, 0); + + return IRQ_HANDLED; +} + +static irqreturn_t qm_aeq_irq(int irq, void *data) +{ + struct hisi_qm *qm = data; + + atomic64_inc(&qm->debug.dfx.aeq_irq_cnt); + if (!readl(qm->io_base + QM_VF_AEQ_INT_SOURCE)) + return IRQ_NONE; + + return IRQ_WAKE_THREAD; +} + +static void qm_init_qp_status(struct hisi_qp *qp) +{ + struct hisi_qp_status *qp_status = &qp->qp_status; + + qp_status->sq_tail = 0; + qp_status->cq_head = 0; + qp_status->cqc_phase = true; + atomic_set(&qp_status->used, 0); +} + +static void qm_init_prefetch(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u32 page_type = 0x0; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + switch (PAGE_SIZE) { + case SZ_4K: + page_type = 0x0; + break; + case SZ_16K: + page_type = 0x1; + break; + case SZ_64K: + page_type = 0x2; + break; + default: + dev_err(dev, "system page size is not support: %lu, default set to 4KB", + PAGE_SIZE); + } + + writel(page_type, qm->io_base + QM_PAGE_SIZE); +} + +/* + * acc_shaper_para_calc() Get the IR value by the qos formula, the return value + * is the expected qos calculated. + * the formula: + * IR = X Mbps if ir = 1 means IR = 100 Mbps, if ir = 10000 means = 10Gbps + * + * IR_b * (2 ^ IR_u) * 8000 + * IR(Mbps) = ------------------------- + * Tick * (2 ^ IR_s) + */ +static u32 acc_shaper_para_calc(u64 cir_b, u64 cir_u, u64 cir_s) +{ + return ((cir_b * QM_QOS_DIVISOR_CLK) * (1 << cir_u)) / + (QM_QOS_TICK * (1 << cir_s)); +} + +static u32 acc_shaper_calc_cbs_s(u32 ir) +{ + int table_size = ARRAY_SIZE(shaper_cbs_s); + int i; + + for (i = 0; i < table_size; i++) { + if (ir >= shaper_cbs_s[i].start && ir <= shaper_cbs_s[i].end) + return shaper_cbs_s[i].val; + } + + return QM_SHAPER_MIN_CBS_S; +} + +static u32 acc_shaper_calc_cir_s(u32 ir) +{ + int table_size = ARRAY_SIZE(shaper_cir_s); + int i; + + for (i = 0; i < table_size; i++) { + if (ir >= shaper_cir_s[i].start && ir <= shaper_cir_s[i].end) + return shaper_cir_s[i].val; + } + + return 0; +} + +static int qm_get_shaper_para(u32 ir, struct qm_shaper_factor *factor) +{ + u32 cir_b, cir_u, cir_s, ir_calc; + u32 error_rate; + + factor->cbs_s = acc_shaper_calc_cbs_s(ir); + cir_s = acc_shaper_calc_cir_s(ir); + + for (cir_b = QM_QOS_MIN_CIR_B; cir_b <= QM_QOS_MAX_CIR_B; cir_b++) { + for (cir_u = 0; cir_u <= QM_QOS_MAX_CIR_U; cir_u++) { + ir_calc = acc_shaper_para_calc(cir_b, cir_u, cir_s); + + error_rate = QM_QOS_EXPAND_RATE * (u32)abs(ir_calc - ir) / ir; + if (error_rate <= QM_QOS_MIN_ERROR_RATE) { + factor->cir_b = cir_b; + factor->cir_u = cir_u; + factor->cir_s = cir_s; + return 0; + } + } + } + + return -EINVAL; +} + +static void qm_vft_data_cfg(struct hisi_qm *qm, enum vft_type type, u32 base, + u32 number, struct qm_shaper_factor *factor) +{ + u64 tmp = 0; + + if (number > 0) { + switch (type) { + case SQC_VFT: + if (qm->ver == QM_HW_V1) { + tmp = QM_SQC_VFT_BUF_SIZE | + QM_SQC_VFT_SQC_SIZE | + QM_SQC_VFT_INDEX_NUMBER | + QM_SQC_VFT_VALID | + (u64)base << QM_SQC_VFT_START_SQN_SHIFT; + } else { + tmp = (u64)base << QM_SQC_VFT_START_SQN_SHIFT | + QM_SQC_VFT_VALID | + (u64)(number - 1) << QM_SQC_VFT_SQN_SHIFT; + } + break; + case CQC_VFT: + if (qm->ver == QM_HW_V1) { + tmp = QM_CQC_VFT_BUF_SIZE | + QM_CQC_VFT_SQC_SIZE | + QM_CQC_VFT_INDEX_NUMBER | + QM_CQC_VFT_VALID; + } else { + tmp = QM_CQC_VFT_VALID; + } + break; + case SHAPER_VFT: + if (factor) { + tmp = factor->cir_b | + (factor->cir_u << QM_SHAPER_FACTOR_CIR_U_SHIFT) | + (factor->cir_s << QM_SHAPER_FACTOR_CIR_S_SHIFT) | + (QM_SHAPER_CBS_B << QM_SHAPER_FACTOR_CBS_B_SHIFT) | + (factor->cbs_s << QM_SHAPER_FACTOR_CBS_S_SHIFT); + } + break; + } + } + + writel(lower_32_bits(tmp), qm->io_base + QM_VFT_CFG_DATA_L); + writel(upper_32_bits(tmp), qm->io_base + QM_VFT_CFG_DATA_H); +} + +static int qm_set_vft_common(struct hisi_qm *qm, enum vft_type type, + u32 fun_num, u32 base, u32 number) +{ + struct qm_shaper_factor *factor = NULL; + unsigned int val; + int ret; + + if (type == SHAPER_VFT && test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) + factor = &qm->factor[fun_num]; + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); + if (ret) + return ret; + + writel(0x0, qm->io_base + QM_VFT_CFG_OP_WR); + writel(type, qm->io_base + QM_VFT_CFG_TYPE); + if (type == SHAPER_VFT) + fun_num |= base << QM_SHAPER_VFT_OFFSET; + + writel(fun_num, qm->io_base + QM_VFT_CFG); + + qm_vft_data_cfg(qm, type, base, number, factor); + + writel(0x0, qm->io_base + QM_VFT_CFG_RDY); + writel(0x1, qm->io_base + QM_VFT_CFG_OP_ENABLE); + + return readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); +} + +static int qm_shaper_init_vft(struct hisi_qm *qm, u32 fun_num) +{ + u32 qos = qm->factor[fun_num].func_qos; + int ret, i; + + ret = qm_get_shaper_para(qos * QM_QOS_RATE, &qm->factor[fun_num]); + if (ret) { + dev_err(&qm->pdev->dev, "failed to calculate shaper parameter!\n"); + return ret; + } + writel(qm->type_rate, qm->io_base + QM_SHAPER_CFG); + for (i = ALG_TYPE_0; i <= ALG_TYPE_1; i++) { + /* The base number of queue reuse for different alg type */ + ret = qm_set_vft_common(qm, SHAPER_VFT, fun_num, i, 1); + if (ret) + return ret; + } + + return 0; +} + +/* The config should be conducted after qm_dev_mem_reset() */ +static int qm_set_sqc_cqc_vft(struct hisi_qm *qm, u32 fun_num, u32 base, + u32 number) +{ + int ret, i; + + for (i = SQC_VFT; i <= CQC_VFT; i++) { + ret = qm_set_vft_common(qm, i, fun_num, base, number); + if (ret) + return ret; + } + + /* init default shaper qos val */ + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) { + ret = qm_shaper_init_vft(qm, fun_num); + if (ret) + goto back_sqc_cqc; + } + + return 0; +back_sqc_cqc: + for (i = SQC_VFT; i <= CQC_VFT; i++) + qm_set_vft_common(qm, i, fun_num, 0, 0); + + return ret; +} + +static int qm_get_vft_v2(struct hisi_qm *qm, u32 *base, u32 *number) +{ + u64 sqc_vft; + int ret; + + ret = hisi_qm_mb(qm, QM_MB_CMD_SQC_VFT_V2, 0, 0, 1); + if (ret) + return ret; + + sqc_vft = readl(qm->io_base + QM_MB_CMD_DATA_ADDR_L) | + ((u64)readl(qm->io_base + QM_MB_CMD_DATA_ADDR_H) << 32); + *base = QM_SQC_VFT_BASE_MASK_V2 & (sqc_vft >> QM_SQC_VFT_BASE_SHIFT_V2); + *number = (QM_SQC_VFT_NUM_MASK_v2 & + (sqc_vft >> QM_SQC_VFT_NUM_SHIFT_V2)) + 1; + + return 0; +} + +static int qm_get_vf_qp_num(struct hisi_qm *qm, u32 fun_num) +{ + u32 remain_q_num, vfq_num; + u32 num_vfs = qm->vfs_num; + + vfq_num = (qm->ctrl_qp_num - qm->qp_num) / num_vfs; + if (vfq_num >= qm->max_qp_num) + return qm->max_qp_num; + + remain_q_num = (qm->ctrl_qp_num - qm->qp_num) % num_vfs; + if (vfq_num + remain_q_num <= qm->max_qp_num) + return fun_num == num_vfs ? vfq_num + remain_q_num : vfq_num; + + /* + * if vfq_num + remain_q_num > max_qp_num, the last VFs, + * each with one more queue. + */ + return fun_num + remain_q_num > num_vfs ? vfq_num + 1 : vfq_num; +} + +static struct hisi_qm *file_to_qm(struct debugfs_file *file) +{ + struct qm_debug *debug = file->debug; + + return container_of(debug, struct hisi_qm, debug); +} + +static u32 current_q_read(struct hisi_qm *qm) +{ + return readl(qm->io_base + QM_DFX_SQE_CNT_VF_SQN) >> QM_DFX_QN_SHIFT; +} + +static int current_q_write(struct hisi_qm *qm, u32 val) +{ + u32 tmp; + + if (val >= qm->debug.curr_qm_qp_num) + return -EINVAL; + + tmp = val << QM_DFX_QN_SHIFT | + (readl(qm->io_base + QM_DFX_SQE_CNT_VF_SQN) & CURRENT_FUN_MASK); + writel(tmp, qm->io_base + QM_DFX_SQE_CNT_VF_SQN); + + tmp = val << QM_DFX_QN_SHIFT | + (readl(qm->io_base + QM_DFX_CQE_CNT_VF_CQN) & CURRENT_FUN_MASK); + writel(tmp, qm->io_base + QM_DFX_CQE_CNT_VF_CQN); + + return 0; +} + +static u32 clear_enable_read(struct hisi_qm *qm) +{ + return readl(qm->io_base + QM_DFX_CNT_CLR_CE); +} + +/* rd_clr_ctrl 1 enable read clear, otherwise 0 disable it */ +static int clear_enable_write(struct hisi_qm *qm, u32 rd_clr_ctrl) +{ + if (rd_clr_ctrl > 1) + return -EINVAL; + + writel(rd_clr_ctrl, qm->io_base + QM_DFX_CNT_CLR_CE); + + return 0; +} + +static u32 current_qm_read(struct hisi_qm *qm) +{ + return readl(qm->io_base + QM_DFX_MB_CNT_VF); +} + +static int current_qm_write(struct hisi_qm *qm, u32 val) +{ + u32 tmp; + + if (val > qm->vfs_num) + return -EINVAL; + + /* According PF or VF Dev ID to calculation curr_qm_qp_num and store */ + if (!val) + qm->debug.curr_qm_qp_num = qm->qp_num; + else + qm->debug.curr_qm_qp_num = qm_get_vf_qp_num(qm, val); + + writel(val, qm->io_base + QM_DFX_MB_CNT_VF); + writel(val, qm->io_base + QM_DFX_DB_CNT_VF); + + tmp = val | + (readl(qm->io_base + QM_DFX_SQE_CNT_VF_SQN) & CURRENT_Q_MASK); + writel(tmp, qm->io_base + QM_DFX_SQE_CNT_VF_SQN); + + tmp = val | + (readl(qm->io_base + QM_DFX_CQE_CNT_VF_CQN) & CURRENT_Q_MASK); + writel(tmp, qm->io_base + QM_DFX_CQE_CNT_VF_CQN); + + return 0; +} + +static ssize_t qm_debug_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct debugfs_file *file = filp->private_data; + enum qm_debug_file index = file->index; + struct hisi_qm *qm = file_to_qm(file); + char tbuf[QM_DBG_TMP_BUF_LEN]; + u32 val; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + mutex_lock(&file->lock); + switch (index) { + case CURRENT_QM: + val = current_qm_read(qm); + break; + case CURRENT_Q: + val = current_q_read(qm); + break; + case CLEAR_ENABLE: + val = clear_enable_read(qm); + break; + default: + goto err_input; + } + mutex_unlock(&file->lock); + + hisi_qm_put_dfx_access(qm); + ret = scnprintf(tbuf, QM_DBG_TMP_BUF_LEN, "%u\n", val); + return simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_input: + mutex_unlock(&file->lock); + hisi_qm_put_dfx_access(qm); + return -EINVAL; +} + +static ssize_t qm_debug_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct debugfs_file *file = filp->private_data; + enum qm_debug_file index = file->index; + struct hisi_qm *qm = file_to_qm(file); + unsigned long val; + char tbuf[QM_DBG_TMP_BUF_LEN]; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= QM_DBG_TMP_BUF_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, QM_DBG_TMP_BUF_LEN - 1, pos, buf, + count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + if (kstrtoul(tbuf, 0, &val)) + return -EFAULT; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + mutex_lock(&file->lock); + switch (index) { + case CURRENT_QM: + ret = current_qm_write(qm, val); + break; + case CURRENT_Q: + ret = current_q_write(qm, val); + break; + case CLEAR_ENABLE: + ret = clear_enable_write(qm, val); + break; + default: + ret = -EINVAL; + } + mutex_unlock(&file->lock); + + hisi_qm_put_dfx_access(qm); + + if (ret) + return ret; + + return count; +} + +static const struct file_operations qm_debug_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = qm_debug_read, + .write = qm_debug_write, +}; + +#define CNT_CYC_REGS_NUM 10 +static const struct debugfs_reg32 qm_dfx_regs[] = { + /* XXX_CNT are reading clear register */ + {"QM_ECC_1BIT_CNT ", 0x104000ull}, + {"QM_ECC_MBIT_CNT ", 0x104008ull}, + {"QM_DFX_MB_CNT ", 0x104018ull}, + {"QM_DFX_DB_CNT ", 0x104028ull}, + {"QM_DFX_SQE_CNT ", 0x104038ull}, + {"QM_DFX_CQE_CNT ", 0x104048ull}, + {"QM_DFX_SEND_SQE_TO_ACC_CNT ", 0x104050ull}, + {"QM_DFX_WB_SQE_FROM_ACC_CNT ", 0x104058ull}, + {"QM_DFX_ACC_FINISH_CNT ", 0x104060ull}, + {"QM_DFX_CQE_ERR_CNT ", 0x1040b4ull}, + {"QM_DFX_FUNS_ACTIVE_ST ", 0x200ull}, + {"QM_ECC_1BIT_INF ", 0x104004ull}, + {"QM_ECC_MBIT_INF ", 0x10400cull}, + {"QM_DFX_ACC_RDY_VLD0 ", 0x1040a0ull}, + {"QM_DFX_ACC_RDY_VLD1 ", 0x1040a4ull}, + {"QM_DFX_AXI_RDY_VLD ", 0x1040a8ull}, + {"QM_DFX_FF_ST0 ", 0x1040c8ull}, + {"QM_DFX_FF_ST1 ", 0x1040ccull}, + {"QM_DFX_FF_ST2 ", 0x1040d0ull}, + {"QM_DFX_FF_ST3 ", 0x1040d4ull}, + {"QM_DFX_FF_ST4 ", 0x1040d8ull}, + {"QM_DFX_FF_ST5 ", 0x1040dcull}, + {"QM_DFX_FF_ST6 ", 0x1040e0ull}, + {"QM_IN_IDLE_ST ", 0x1040e4ull}, +}; + +static const struct debugfs_reg32 qm_vf_dfx_regs[] = { + {"QM_DFX_FUNS_ACTIVE_ST ", 0x200ull}, +}; + +/** + * hisi_qm_regs_dump() - Dump registers's value. + * @s: debugfs file handle. + * @regset: accelerator registers information. + * + * Dump accelerator registers. + */ +void hisi_qm_regs_dump(struct seq_file *s, struct debugfs_regset32 *regset) +{ + struct pci_dev *pdev = to_pci_dev(regset->dev); + struct hisi_qm *qm = pci_get_drvdata(pdev); + const struct debugfs_reg32 *regs = regset->regs; + int regs_len = regset->nregs; + int i, ret; + u32 val; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return; + + for (i = 0; i < regs_len; i++) { + val = readl(regset->base + regs[i].offset); + seq_printf(s, "%s= 0x%08x\n", regs[i].name, val); + } + + hisi_qm_put_dfx_access(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_regs_dump); + +static int qm_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + struct debugfs_regset32 regset; + + if (qm->fun_type == QM_HW_PF) { + regset.regs = qm_dfx_regs; + regset.nregs = ARRAY_SIZE(qm_dfx_regs); + } else { + regset.regs = qm_vf_dfx_regs; + regset.nregs = ARRAY_SIZE(qm_vf_dfx_regs); + } + + regset.base = qm->io_base; + regset.dev = &qm->pdev->dev; + + hisi_qm_regs_dump(s, ®set); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(qm_regs); + +static struct dfx_diff_registers *dfx_regs_init(struct hisi_qm *qm, + const struct dfx_diff_registers *cregs, int reg_len) +{ + struct dfx_diff_registers *diff_regs; + u32 j, base_offset; + int i; + + diff_regs = kcalloc(reg_len, sizeof(*diff_regs), GFP_KERNEL); + if (!diff_regs) + return ERR_PTR(-ENOMEM); + + for (i = 0; i < reg_len; i++) { + if (!cregs[i].reg_len) + continue; + + diff_regs[i].reg_offset = cregs[i].reg_offset; + diff_regs[i].reg_len = cregs[i].reg_len; + diff_regs[i].regs = kcalloc(QM_DFX_REGS_LEN, cregs[i].reg_len, + GFP_KERNEL); + if (!diff_regs[i].regs) + goto alloc_error; + + for (j = 0; j < diff_regs[i].reg_len; j++) { + base_offset = diff_regs[i].reg_offset + + j * QM_DFX_REGS_LEN; + diff_regs[i].regs[j] = readl(qm->io_base + base_offset); + } + } + + return diff_regs; + +alloc_error: + while (i > 0) { + i--; + kfree(diff_regs[i].regs); + } + kfree(diff_regs); + return ERR_PTR(-ENOMEM); +} + +static void dfx_regs_uninit(struct hisi_qm *qm, + struct dfx_diff_registers *dregs, int reg_len) +{ + int i; + + /* Setting the pointer is NULL to prevent double free */ + for (i = 0; i < reg_len; i++) { + kfree(dregs[i].regs); + dregs[i].regs = NULL; + } + kfree(dregs); + dregs = NULL; +} + +/** + * hisi_qm_diff_regs_init() - Allocate memory for registers. + * @qm: device qm handle. + * @dregs: diff registers handle. + * @reg_len: diff registers region length. + */ +int hisi_qm_diff_regs_init(struct hisi_qm *qm, + struct dfx_diff_registers *dregs, int reg_len) +{ + if (!qm || !dregs || reg_len <= 0) + return -EINVAL; + + if (qm->fun_type != QM_HW_PF) + return 0; + + qm->debug.qm_diff_regs = dfx_regs_init(qm, qm_diff_regs, + ARRAY_SIZE(qm_diff_regs)); + if (IS_ERR(qm->debug.qm_diff_regs)) + return PTR_ERR(qm->debug.qm_diff_regs); + + qm->debug.acc_diff_regs = dfx_regs_init(qm, dregs, reg_len); + if (IS_ERR(qm->debug.acc_diff_regs)) { + dfx_regs_uninit(qm, qm->debug.qm_diff_regs, + ARRAY_SIZE(qm_diff_regs)); + return PTR_ERR(qm->debug.acc_diff_regs); + } + + return 0; +} +EXPORT_SYMBOL_GPL(hisi_qm_diff_regs_init); + +/** + * hisi_qm_diff_regs_uninit() - Free memory for registers. + * @qm: device qm handle. + * @reg_len: diff registers region length. + */ +void hisi_qm_diff_regs_uninit(struct hisi_qm *qm, int reg_len) +{ + if (!qm || reg_len <= 0 || qm->fun_type != QM_HW_PF) + return; + + dfx_regs_uninit(qm, qm->debug.acc_diff_regs, reg_len); + dfx_regs_uninit(qm, qm->debug.qm_diff_regs, ARRAY_SIZE(qm_diff_regs)); +} +EXPORT_SYMBOL_GPL(hisi_qm_diff_regs_uninit); + +/** + * hisi_qm_acc_diff_regs_dump() - Dump registers's value. + * @qm: device qm handle. + * @s: Debugfs file handle. + * @dregs: diff registers handle. + * @regs_len: diff registers region length. + */ +void hisi_qm_acc_diff_regs_dump(struct hisi_qm *qm, struct seq_file *s, + struct dfx_diff_registers *dregs, int regs_len) +{ + u32 j, val, base_offset; + int i, ret; + + if (!qm || !s || !dregs || regs_len <= 0) + return; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return; + + down_read(&qm->qps_lock); + for (i = 0; i < regs_len; i++) { + if (!dregs[i].reg_len) + continue; + + for (j = 0; j < dregs[i].reg_len; j++) { + base_offset = dregs[i].reg_offset + j * QM_DFX_REGS_LEN; + val = readl(qm->io_base + base_offset); + if (val != dregs[i].regs[j]) + seq_printf(s, "0x%08x = 0x%08x ---> 0x%08x\n", + base_offset, dregs[i].regs[j], val); + } + } + up_read(&qm->qps_lock); + + hisi_qm_put_dfx_access(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_acc_diff_regs_dump); + +static int qm_diff_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + + hisi_qm_acc_diff_regs_dump(qm, s, qm->debug.qm_diff_regs, + ARRAY_SIZE(qm_diff_regs)); + + return 0; +} +DEFINE_SHOW_ATTRIBUTE(qm_diff_regs); + +static ssize_t qm_cmd_read(struct file *filp, char __user *buffer, + size_t count, loff_t *pos) +{ + char buf[QM_DBG_READ_LEN]; + int len; + + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", + "Please echo help to cmd to get help information"); + + return simple_read_from_buffer(buffer, count, pos, buf, len); +} + +static void *qm_ctx_alloc(struct hisi_qm *qm, size_t ctx_size, + dma_addr_t *dma_addr) +{ + struct device *dev = &qm->pdev->dev; + void *ctx_addr; + + ctx_addr = kzalloc(ctx_size, GFP_KERNEL); + if (!ctx_addr) + return ERR_PTR(-ENOMEM); + + *dma_addr = dma_map_single(dev, ctx_addr, ctx_size, DMA_FROM_DEVICE); + if (dma_mapping_error(dev, *dma_addr)) { + dev_err(dev, "DMA mapping error!\n"); + kfree(ctx_addr); + return ERR_PTR(-ENOMEM); + } + + return ctx_addr; +} + +static void qm_ctx_free(struct hisi_qm *qm, size_t ctx_size, + const void *ctx_addr, dma_addr_t *dma_addr) +{ + struct device *dev = &qm->pdev->dev; + + dma_unmap_single(dev, *dma_addr, ctx_size, DMA_FROM_DEVICE); + kfree(ctx_addr); +} + +static void dump_show(struct hisi_qm *qm, void *info, + unsigned int info_size, char *info_name) +{ + struct device *dev = &qm->pdev->dev; + u8 *info_curr = info; + u32 i; +#define BYTE_PER_DW 4 + + dev_info(dev, "%s DUMP\n", info_name); + for (i = 0; i < info_size; i += BYTE_PER_DW, info_curr += BYTE_PER_DW) { + pr_info("DW%u: %02X%02X %02X%02X\n", i / BYTE_PER_DW, + *(info_curr + 3), *(info_curr + 2), *(info_curr + 1), *(info_curr)); + } +} + +static int qm_dump_sqc_raw(struct hisi_qm *qm, dma_addr_t dma_addr, u16 qp_id) +{ + return hisi_qm_mb(qm, QM_MB_CMD_SQC, dma_addr, qp_id, 1); +} + +static int qm_dump_cqc_raw(struct hisi_qm *qm, dma_addr_t dma_addr, u16 qp_id) +{ + return hisi_qm_mb(qm, QM_MB_CMD_CQC, dma_addr, qp_id, 1); +} + +static int qm_sqc_dump(struct hisi_qm *qm, const char *s) +{ + struct device *dev = &qm->pdev->dev; + struct qm_sqc *sqc, *sqc_curr; + dma_addr_t sqc_dma; + u32 qp_id; + int ret; + + if (!s) + return -EINVAL; + + ret = kstrtou32(s, 0, &qp_id); + if (ret || qp_id >= qm->qp_num) { + dev_err(dev, "Please input qp num (0-%u)", qm->qp_num - 1); + return -EINVAL; + } + + sqc = qm_ctx_alloc(qm, sizeof(*sqc), &sqc_dma); + if (IS_ERR(sqc)) + return PTR_ERR(sqc); + + ret = qm_dump_sqc_raw(qm, sqc_dma, qp_id); + if (ret) { + down_read(&qm->qps_lock); + if (qm->sqc) { + sqc_curr = qm->sqc + qp_id; + + dump_show(qm, sqc_curr, sizeof(*sqc), "SOFT SQC"); + } + up_read(&qm->qps_lock); + + goto free_ctx; + } + + dump_show(qm, sqc, sizeof(*sqc), "SQC"); + +free_ctx: + qm_ctx_free(qm, sizeof(*sqc), sqc, &sqc_dma); + return 0; +} + +static int qm_cqc_dump(struct hisi_qm *qm, const char *s) +{ + struct device *dev = &qm->pdev->dev; + struct qm_cqc *cqc, *cqc_curr; + dma_addr_t cqc_dma; + u32 qp_id; + int ret; + + if (!s) + return -EINVAL; + + ret = kstrtou32(s, 0, &qp_id); + if (ret || qp_id >= qm->qp_num) { + dev_err(dev, "Please input qp num (0-%u)", qm->qp_num - 1); + return -EINVAL; + } + + cqc = qm_ctx_alloc(qm, sizeof(*cqc), &cqc_dma); + if (IS_ERR(cqc)) + return PTR_ERR(cqc); + + ret = qm_dump_cqc_raw(qm, cqc_dma, qp_id); + if (ret) { + down_read(&qm->qps_lock); + if (qm->cqc) { + cqc_curr = qm->cqc + qp_id; + + dump_show(qm, cqc_curr, sizeof(*cqc), "SOFT CQC"); + } + up_read(&qm->qps_lock); + + goto free_ctx; + } + + dump_show(qm, cqc, sizeof(*cqc), "CQC"); + +free_ctx: + qm_ctx_free(qm, sizeof(*cqc), cqc, &cqc_dma); + return 0; +} + +static int qm_eqc_aeqc_dump(struct hisi_qm *qm, char *s, size_t size, + int cmd, char *name) +{ + struct device *dev = &qm->pdev->dev; + dma_addr_t xeqc_dma; + void *xeqc; + int ret; + + if (strsep(&s, " ")) { + dev_err(dev, "Please do not input extra characters!\n"); + return -EINVAL; + } + + xeqc = qm_ctx_alloc(qm, size, &xeqc_dma); + if (IS_ERR(xeqc)) + return PTR_ERR(xeqc); + + ret = hisi_qm_mb(qm, cmd, xeqc_dma, 0, 1); + if (ret) + goto err_free_ctx; + + dump_show(qm, xeqc, size, name); + +err_free_ctx: + qm_ctx_free(qm, size, xeqc, &xeqc_dma); + return ret; +} + +static int q_dump_param_parse(struct hisi_qm *qm, char *s, + u32 *e_id, u32 *q_id, u16 q_depth) +{ + struct device *dev = &qm->pdev->dev; + unsigned int qp_num = qm->qp_num; + char *presult; + int ret; + + presult = strsep(&s, " "); + if (!presult) { + dev_err(dev, "Please input qp number!\n"); + return -EINVAL; + } + + ret = kstrtou32(presult, 0, q_id); + if (ret || *q_id >= qp_num) { + dev_err(dev, "Please input qp num (0-%u)", qp_num - 1); + return -EINVAL; + } + + presult = strsep(&s, " "); + if (!presult) { + dev_err(dev, "Please input sqe number!\n"); + return -EINVAL; + } + + ret = kstrtou32(presult, 0, e_id); + if (ret || *e_id >= q_depth) { + dev_err(dev, "Please input sqe num (0-%u)", q_depth - 1); + return -EINVAL; + } + + if (strsep(&s, " ")) { + dev_err(dev, "Please do not input extra characters!\n"); + return -EINVAL; + } + + return 0; +} + +static int qm_sq_dump(struct hisi_qm *qm, char *s) +{ + u16 sq_depth = qm->qp_array->cq_depth; + void *sqe, *sqe_curr; + struct hisi_qp *qp; + u32 qp_id, sqe_id; + int ret; + + ret = q_dump_param_parse(qm, s, &sqe_id, &qp_id, sq_depth); + if (ret) + return ret; + + sqe = kzalloc(qm->sqe_size * sq_depth, GFP_KERNEL); + if (!sqe) + return -ENOMEM; + + qp = &qm->qp_array[qp_id]; + memcpy(sqe, qp->sqe, qm->sqe_size * sq_depth); + sqe_curr = sqe + (u32)(sqe_id * qm->sqe_size); + memset(sqe_curr + qm->debug.sqe_mask_offset, QM_SQE_ADDR_MASK, + qm->debug.sqe_mask_len); + + dump_show(qm, sqe_curr, qm->sqe_size, "SQE"); + + kfree(sqe); + + return 0; +} + +static int qm_cq_dump(struct hisi_qm *qm, char *s) +{ + struct qm_cqe *cqe_curr; + struct hisi_qp *qp; + u32 qp_id, cqe_id; + int ret; + + ret = q_dump_param_parse(qm, s, &cqe_id, &qp_id, qm->qp_array->cq_depth); + if (ret) + return ret; + + qp = &qm->qp_array[qp_id]; + cqe_curr = qp->cqe + cqe_id; + dump_show(qm, cqe_curr, sizeof(struct qm_cqe), "CQE"); + + return 0; +} + +static int qm_eq_aeq_dump(struct hisi_qm *qm, const char *s, + size_t size, char *name) +{ + struct device *dev = &qm->pdev->dev; + void *xeqe; + u32 xeqe_id; + int ret; + + if (!s) + return -EINVAL; + + ret = kstrtou32(s, 0, &xeqe_id); + if (ret) + return -EINVAL; + + if (!strcmp(name, "EQE") && xeqe_id >= qm->eq_depth) { + dev_err(dev, "Please input eqe num (0-%u)", qm->eq_depth - 1); + return -EINVAL; + } else if (!strcmp(name, "AEQE") && xeqe_id >= qm->aeq_depth) { + dev_err(dev, "Please input aeqe num (0-%u)", qm->eq_depth - 1); + return -EINVAL; + } + + down_read(&qm->qps_lock); + + if (qm->eqe && !strcmp(name, "EQE")) { + xeqe = qm->eqe + xeqe_id; + } else if (qm->aeqe && !strcmp(name, "AEQE")) { + xeqe = qm->aeqe + xeqe_id; + } else { + ret = -EINVAL; + goto err_unlock; + } + + dump_show(qm, xeqe, size, name); + +err_unlock: + up_read(&qm->qps_lock); + return ret; +} + +static int qm_dbg_help(struct hisi_qm *qm, char *s) +{ + struct device *dev = &qm->pdev->dev; + + if (strsep(&s, " ")) { + dev_err(dev, "Please do not input extra characters!\n"); + return -EINVAL; + } + + dev_info(dev, "available commands:\n"); + dev_info(dev, "sqc \n"); + dev_info(dev, "cqc \n"); + dev_info(dev, "eqc\n"); + dev_info(dev, "aeqc\n"); + dev_info(dev, "sq \n"); + dev_info(dev, "cq \n"); + dev_info(dev, "eq \n"); + dev_info(dev, "aeq \n"); + + return 0; +} + +static int qm_cmd_write_dump(struct hisi_qm *qm, const char *cmd_buf) +{ + struct device *dev = &qm->pdev->dev; + char *presult, *s, *s_tmp; + int ret; + + s = kstrdup(cmd_buf, GFP_KERNEL); + if (!s) + return -ENOMEM; + + s_tmp = s; + presult = strsep(&s, " "); + if (!presult) { + ret = -EINVAL; + goto err_buffer_free; + } + + if (!strcmp(presult, "sqc")) + ret = qm_sqc_dump(qm, s); + else if (!strcmp(presult, "cqc")) + ret = qm_cqc_dump(qm, s); + else if (!strcmp(presult, "eqc")) + ret = qm_eqc_aeqc_dump(qm, s, sizeof(struct qm_eqc), + QM_MB_CMD_EQC, "EQC"); + else if (!strcmp(presult, "aeqc")) + ret = qm_eqc_aeqc_dump(qm, s, sizeof(struct qm_aeqc), + QM_MB_CMD_AEQC, "AEQC"); + else if (!strcmp(presult, "sq")) + ret = qm_sq_dump(qm, s); + else if (!strcmp(presult, "cq")) + ret = qm_cq_dump(qm, s); + else if (!strcmp(presult, "eq")) + ret = qm_eq_aeq_dump(qm, s, sizeof(struct qm_eqe), "EQE"); + else if (!strcmp(presult, "aeq")) + ret = qm_eq_aeq_dump(qm, s, sizeof(struct qm_aeqe), "AEQE"); + else if (!strcmp(presult, "help")) + ret = qm_dbg_help(qm, s); + else + ret = -EINVAL; + + if (ret) + dev_info(dev, "Please echo help\n"); + +err_buffer_free: + kfree(s_tmp); + + return ret; +} + +static ssize_t qm_cmd_write(struct file *filp, const char __user *buffer, + size_t count, loff_t *pos) +{ + struct hisi_qm *qm = filp->private_data; + char *cmd_buf, *cmd_buf_tmp; + int ret; + + if (*pos) + return 0; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + /* Judge if the instance is being reset. */ + if (unlikely(atomic_read(&qm->status.flags) == QM_STOP)) { + ret = 0; + goto put_dfx_access; + } + + if (count > QM_DBG_WRITE_LEN) { + ret = -ENOSPC; + goto put_dfx_access; + } + + cmd_buf = memdup_user_nul(buffer, count); + if (IS_ERR(cmd_buf)) { + ret = PTR_ERR(cmd_buf); + goto put_dfx_access; + } + + cmd_buf_tmp = strchr(cmd_buf, '\n'); + if (cmd_buf_tmp) { + *cmd_buf_tmp = '\0'; + count = cmd_buf_tmp - cmd_buf + 1; + } + + ret = qm_cmd_write_dump(qm, cmd_buf); + if (ret) { + kfree(cmd_buf); + goto put_dfx_access; + } + + kfree(cmd_buf); + + ret = count; + +put_dfx_access: + hisi_qm_put_dfx_access(qm); + return ret; +} + +static const struct file_operations qm_cmd_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = qm_cmd_read, + .write = qm_cmd_write, +}; + +static void qm_create_debugfs_file(struct hisi_qm *qm, struct dentry *dir, + enum qm_debug_file index) +{ + struct debugfs_file *file = qm->debug.files + index; + + debugfs_create_file(qm_debug_file_name[index], 0600, dir, file, + &qm_debug_fops); + + file->index = index; + mutex_init(&file->lock); + file->debug = &qm->debug; +} + +static void qm_hw_error_init_v1(struct hisi_qm *qm) +{ + writel(QM_ABNORMAL_INT_MASK_VALUE, qm->io_base + QM_ABNORMAL_INT_MASK); +} + +static void qm_hw_error_cfg(struct hisi_qm *qm) +{ + struct hisi_qm_err_info *err_info = &qm->err_info; + + qm->error_mask = err_info->nfe | err_info->ce | err_info->fe; + /* clear QM hw residual error source */ + writel(qm->error_mask, qm->io_base + QM_ABNORMAL_INT_SOURCE); + + /* configure error type */ + writel(err_info->ce, qm->io_base + QM_RAS_CE_ENABLE); + writel(QM_RAS_CE_TIMES_PER_IRQ, qm->io_base + QM_RAS_CE_THRESHOLD); + writel(err_info->nfe, qm->io_base + QM_RAS_NFE_ENABLE); + writel(err_info->fe, qm->io_base + QM_RAS_FE_ENABLE); +} + +static void qm_hw_error_init_v2(struct hisi_qm *qm) +{ + u32 irq_unmask; + + qm_hw_error_cfg(qm); + + irq_unmask = ~qm->error_mask; + irq_unmask &= readl(qm->io_base + QM_ABNORMAL_INT_MASK); + writel(irq_unmask, qm->io_base + QM_ABNORMAL_INT_MASK); +} + +static void qm_hw_error_uninit_v2(struct hisi_qm *qm) +{ + u32 irq_mask = qm->error_mask; + + irq_mask |= readl(qm->io_base + QM_ABNORMAL_INT_MASK); + writel(irq_mask, qm->io_base + QM_ABNORMAL_INT_MASK); +} + +static void qm_hw_error_init_v3(struct hisi_qm *qm) +{ + u32 irq_unmask; + + qm_hw_error_cfg(qm); + + /* enable close master ooo when hardware error happened */ + writel(qm->err_info.qm_shutdown_mask, qm->io_base + QM_OOO_SHUTDOWN_SEL); + + irq_unmask = ~qm->error_mask; + irq_unmask &= readl(qm->io_base + QM_ABNORMAL_INT_MASK); + writel(irq_unmask, qm->io_base + QM_ABNORMAL_INT_MASK); +} + +static void qm_hw_error_uninit_v3(struct hisi_qm *qm) +{ + u32 irq_mask = qm->error_mask; + + irq_mask |= readl(qm->io_base + QM_ABNORMAL_INT_MASK); + writel(irq_mask, qm->io_base + QM_ABNORMAL_INT_MASK); + + /* disable close master ooo when hardware error happened */ + writel(0x0, qm->io_base + QM_OOO_SHUTDOWN_SEL); +} + +static void qm_log_hw_error(struct hisi_qm *qm, u32 error_status) +{ + const struct hisi_qm_hw_error *err; + struct device *dev = &qm->pdev->dev; + u32 reg_val, type, vf_num; + int i; + + for (i = 0; i < ARRAY_SIZE(qm_hw_error); i++) { + err = &qm_hw_error[i]; + if (!(err->int_msk & error_status)) + continue; + + dev_err(dev, "%s [error status=0x%x] found\n", + err->msg, err->int_msk); + + if (err->int_msk & QM_DB_TIMEOUT) { + reg_val = readl(qm->io_base + QM_ABNORMAL_INF01); + type = (reg_val & QM_DB_TIMEOUT_TYPE) >> + QM_DB_TIMEOUT_TYPE_SHIFT; + vf_num = reg_val & QM_DB_TIMEOUT_VF; + dev_err(dev, "qm %s doorbell timeout in function %u\n", + qm_db_timeout[type], vf_num); + } else if (err->int_msk & QM_OF_FIFO_OF) { + reg_val = readl(qm->io_base + QM_ABNORMAL_INF00); + type = (reg_val & QM_FIFO_OVERFLOW_TYPE) >> + QM_FIFO_OVERFLOW_TYPE_SHIFT; + vf_num = reg_val & QM_FIFO_OVERFLOW_VF; + + if (type < ARRAY_SIZE(qm_fifo_overflow)) + dev_err(dev, "qm %s fifo overflow in function %u\n", + qm_fifo_overflow[type], vf_num); + else + dev_err(dev, "unknown error type\n"); + } + } +} + +static enum acc_err_result qm_hw_error_handle_v2(struct hisi_qm *qm) +{ + u32 error_status, tmp; + + /* read err sts */ + tmp = readl(qm->io_base + QM_ABNORMAL_INT_STATUS); + error_status = qm->error_mask & tmp; + + if (error_status) { + if (error_status & QM_ECC_MBIT) + qm->err_status.is_qm_ecc_mbit = true; + + qm_log_hw_error(qm, error_status); + if (error_status & qm->err_info.qm_reset_mask) + return ACC_ERR_NEED_RESET; + + writel(error_status, qm->io_base + QM_ABNORMAL_INT_SOURCE); + writel(qm->err_info.nfe, qm->io_base + QM_RAS_NFE_ENABLE); + } + + return ACC_ERR_RECOVERED; +} + +static int qm_get_mb_cmd(struct hisi_qm *qm, u64 *msg, u16 fun_num) +{ + struct qm_mailbox mailbox; + int ret; + + qm_mb_pre_init(&mailbox, QM_MB_CMD_DST, 0, fun_num, 0); + mutex_lock(&qm->mailbox_lock); + ret = qm_mb_nolock(qm, &mailbox); + if (ret) + goto err_unlock; + + *msg = readl(qm->io_base + QM_MB_CMD_DATA_ADDR_L) | + ((u64)readl(qm->io_base + QM_MB_CMD_DATA_ADDR_H) << 32); + +err_unlock: + mutex_unlock(&qm->mailbox_lock); + return ret; +} + +static void qm_clear_cmd_interrupt(struct hisi_qm *qm, u64 vf_mask) +{ + u32 val; + + if (qm->fun_type == QM_HW_PF) + writeq(vf_mask, qm->io_base + QM_IFC_INT_SOURCE_P); + + val = readl(qm->io_base + QM_IFC_INT_SOURCE_V); + val |= QM_IFC_INT_SOURCE_MASK; + writel(val, qm->io_base + QM_IFC_INT_SOURCE_V); +} + +static void qm_handle_vf_msg(struct hisi_qm *qm, u32 vf_id) +{ + struct device *dev = &qm->pdev->dev; + u32 cmd; + u64 msg; + int ret; + + ret = qm_get_mb_cmd(qm, &msg, vf_id); + if (ret) { + dev_err(dev, "failed to get msg from VF(%u)!\n", vf_id); + return; + } + + cmd = msg & QM_MB_CMD_DATA_MASK; + switch (cmd) { + case QM_VF_PREPARE_FAIL: + dev_err(dev, "failed to stop VF(%u)!\n", vf_id); + break; + case QM_VF_START_FAIL: + dev_err(dev, "failed to start VF(%u)!\n", vf_id); + break; + case QM_VF_PREPARE_DONE: + case QM_VF_START_DONE: + break; + default: + dev_err(dev, "unsupported cmd %u sent by VF(%u)!\n", cmd, vf_id); + break; + } +} + +static int qm_wait_vf_prepare_finish(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u32 vfs_num = qm->vfs_num; + int cnt = 0; + int ret = 0; + u64 val; + u32 i; + + if (!qm->vfs_num || !test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) + return 0; + + while (true) { + val = readq(qm->io_base + QM_IFC_INT_SOURCE_P); + /* All VFs send command to PF, break */ + if ((val & GENMASK(vfs_num, 1)) == GENMASK(vfs_num, 1)) + break; + + if (++cnt > QM_MAX_PF_WAIT_COUNT) { + ret = -EBUSY; + break; + } + + msleep(QM_WAIT_DST_ACK); + } + + /* PF check VFs msg */ + for (i = 1; i <= vfs_num; i++) { + if (val & BIT(i)) + qm_handle_vf_msg(qm, i); + else + dev_err(dev, "VF(%u) not ping PF!\n", i); + } + + /* PF clear interrupt to ack VFs */ + qm_clear_cmd_interrupt(qm, val); + + return ret; +} + +static void qm_trigger_vf_interrupt(struct hisi_qm *qm, u32 fun_num) +{ + u32 val; + + val = readl(qm->io_base + QM_IFC_INT_CFG); + val &= ~QM_IFC_SEND_ALL_VFS; + val |= fun_num; + writel(val, qm->io_base + QM_IFC_INT_CFG); + + val = readl(qm->io_base + QM_IFC_INT_SET_P); + val |= QM_IFC_INT_SET_MASK; + writel(val, qm->io_base + QM_IFC_INT_SET_P); +} + +static void qm_trigger_pf_interrupt(struct hisi_qm *qm) +{ + u32 val; + + val = readl(qm->io_base + QM_IFC_INT_SET_V); + val |= QM_IFC_INT_SET_MASK; + writel(val, qm->io_base + QM_IFC_INT_SET_V); +} + +static int qm_ping_single_vf(struct hisi_qm *qm, u64 cmd, u32 fun_num) +{ + struct device *dev = &qm->pdev->dev; + struct qm_mailbox mailbox; + int cnt = 0; + u64 val; + int ret; + + qm_mb_pre_init(&mailbox, QM_MB_CMD_SRC, cmd, fun_num, 0); + mutex_lock(&qm->mailbox_lock); + ret = qm_mb_nolock(qm, &mailbox); + if (ret) { + dev_err(dev, "failed to send command to vf(%u)!\n", fun_num); + goto err_unlock; + } + + qm_trigger_vf_interrupt(qm, fun_num); + while (true) { + msleep(QM_WAIT_DST_ACK); + val = readq(qm->io_base + QM_IFC_READY_STATUS); + /* if VF respond, PF notifies VF successfully. */ + if (!(val & BIT(fun_num))) + goto err_unlock; + + if (++cnt > QM_MAX_PF_WAIT_COUNT) { + dev_err(dev, "failed to get response from VF(%u)!\n", fun_num); + ret = -ETIMEDOUT; + break; + } + } + +err_unlock: + mutex_unlock(&qm->mailbox_lock); + return ret; +} + +static int qm_ping_all_vfs(struct hisi_qm *qm, u64 cmd) +{ + struct device *dev = &qm->pdev->dev; + u32 vfs_num = qm->vfs_num; + struct qm_mailbox mailbox; + u64 val = 0; + int cnt = 0; + int ret; + u32 i; + + qm_mb_pre_init(&mailbox, QM_MB_CMD_SRC, cmd, QM_MB_PING_ALL_VFS, 0); + mutex_lock(&qm->mailbox_lock); + /* PF sends command to all VFs by mailbox */ + ret = qm_mb_nolock(qm, &mailbox); + if (ret) { + dev_err(dev, "failed to send command to VFs!\n"); + mutex_unlock(&qm->mailbox_lock); + return ret; + } + + qm_trigger_vf_interrupt(qm, QM_IFC_SEND_ALL_VFS); + while (true) { + msleep(QM_WAIT_DST_ACK); + val = readq(qm->io_base + QM_IFC_READY_STATUS); + /* If all VFs acked, PF notifies VFs successfully. */ + if (!(val & GENMASK(vfs_num, 1))) { + mutex_unlock(&qm->mailbox_lock); + return 0; + } + + if (++cnt > QM_MAX_PF_WAIT_COUNT) + break; + } + + mutex_unlock(&qm->mailbox_lock); + + /* Check which vf respond timeout. */ + for (i = 1; i <= vfs_num; i++) { + if (val & BIT(i)) + dev_err(dev, "failed to get response from VF(%u)!\n", i); + } + + return -ETIMEDOUT; +} + +static int qm_ping_pf(struct hisi_qm *qm, u64 cmd) +{ + struct qm_mailbox mailbox; + int cnt = 0; + u32 val; + int ret; + + qm_mb_pre_init(&mailbox, QM_MB_CMD_SRC, cmd, 0, 0); + mutex_lock(&qm->mailbox_lock); + ret = qm_mb_nolock(qm, &mailbox); + if (ret) { + dev_err(&qm->pdev->dev, "failed to send command to PF!\n"); + goto unlock; + } + + qm_trigger_pf_interrupt(qm); + /* Waiting for PF response */ + while (true) { + msleep(QM_WAIT_DST_ACK); + val = readl(qm->io_base + QM_IFC_INT_SET_V); + if (!(val & QM_IFC_INT_STATUS_MASK)) + break; + + if (++cnt > QM_MAX_VF_WAIT_COUNT) { + ret = -ETIMEDOUT; + break; + } + } + +unlock: + mutex_unlock(&qm->mailbox_lock); + return ret; +} + +static int qm_stop_qp(struct hisi_qp *qp) +{ + return hisi_qm_mb(qp->qm, QM_MB_CMD_STOP_QP, 0, qp->qp_id, 0); +} + +static int qm_set_msi(struct hisi_qm *qm, bool set) +{ + struct pci_dev *pdev = qm->pdev; + + if (set) { + pci_write_config_dword(pdev, pdev->msi_cap + PCI_MSI_MASK_64, + 0); + } else { + pci_write_config_dword(pdev, pdev->msi_cap + PCI_MSI_MASK_64, + ACC_PEH_MSI_DISABLE); + if (qm->err_status.is_qm_ecc_mbit || + qm->err_status.is_dev_ecc_mbit) + return 0; + + mdelay(1); + if (readl(qm->io_base + QM_PEH_DFX_INFO0)) + return -EFAULT; + } + + return 0; +} + +static void qm_wait_msi_finish(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 cmd = ~0; + int cnt = 0; + u32 val; + int ret; + + while (true) { + pci_read_config_dword(pdev, pdev->msi_cap + + PCI_MSI_PENDING_64, &cmd); + if (!cmd) + break; + + if (++cnt > MAX_WAIT_COUNTS) { + pci_warn(pdev, "failed to empty MSI PENDING!\n"); + break; + } + + udelay(1); + } + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_PEH_DFX_INFO0, + val, !(val & QM_PEH_DFX_MASK), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) + pci_warn(pdev, "failed to empty PEH MSI!\n"); + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_PEH_DFX_INFO1, + val, !(val & QM_PEH_MSI_FINISH_MASK), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) + pci_warn(pdev, "failed to finish MSI operation!\n"); +} + +static int qm_set_msi_v3(struct hisi_qm *qm, bool set) +{ + struct pci_dev *pdev = qm->pdev; + int ret = -ETIMEDOUT; + u32 cmd, i; + + pci_read_config_dword(pdev, pdev->msi_cap, &cmd); + if (set) + cmd |= QM_MSI_CAP_ENABLE; + else + cmd &= ~QM_MSI_CAP_ENABLE; + + pci_write_config_dword(pdev, pdev->msi_cap, cmd); + if (set) { + for (i = 0; i < MAX_WAIT_COUNTS; i++) { + pci_read_config_dword(pdev, pdev->msi_cap, &cmd); + if (cmd & QM_MSI_CAP_ENABLE) + return 0; + + udelay(1); + } + } else { + udelay(WAIT_PERIOD_US_MIN); + qm_wait_msi_finish(qm); + ret = 0; + } + + return ret; +} + +static const struct hisi_qm_hw_ops qm_hw_ops_v1 = { + .qm_db = qm_db_v1, + .hw_error_init = qm_hw_error_init_v1, + .set_msi = qm_set_msi, +}; + +static const struct hisi_qm_hw_ops qm_hw_ops_v2 = { + .get_vft = qm_get_vft_v2, + .qm_db = qm_db_v2, + .hw_error_init = qm_hw_error_init_v2, + .hw_error_uninit = qm_hw_error_uninit_v2, + .hw_error_handle = qm_hw_error_handle_v2, + .set_msi = qm_set_msi, +}; + +static const struct hisi_qm_hw_ops qm_hw_ops_v3 = { + .get_vft = qm_get_vft_v2, + .qm_db = qm_db_v2, + .hw_error_init = qm_hw_error_init_v3, + .hw_error_uninit = qm_hw_error_uninit_v3, + .hw_error_handle = qm_hw_error_handle_v2, + .set_msi = qm_set_msi_v3, +}; + +static void *qm_get_avail_sqe(struct hisi_qp *qp) +{ + struct hisi_qp_status *qp_status = &qp->qp_status; + u16 sq_tail = qp_status->sq_tail; + + if (unlikely(atomic_read(&qp->qp_status.used) == qp->sq_depth - 1)) + return NULL; + + return qp->sqe + sq_tail * qp->qm->sqe_size; +} + +static void hisi_qm_unset_hw_reset(struct hisi_qp *qp) +{ + u64 *addr; + + /* Use last 64 bits of DUS to reset status. */ + addr = (u64 *)(qp->qdma.va + qp->qdma.size) - QM_RESET_STOP_TX_OFFSET; + *addr = 0; +} + +static struct hisi_qp *qm_create_qp_nolock(struct hisi_qm *qm, u8 alg_type) +{ + struct device *dev = &qm->pdev->dev; + struct hisi_qp *qp; + int qp_id; + + if (!qm_qp_avail_state(qm, NULL, QP_INIT)) + return ERR_PTR(-EPERM); + + if (qm->qp_in_used == qm->qp_num) { + dev_info_ratelimited(dev, "All %u queues of QM are busy!\n", + qm->qp_num); + atomic64_inc(&qm->debug.dfx.create_qp_err_cnt); + return ERR_PTR(-EBUSY); + } + + qp_id = idr_alloc_cyclic(&qm->qp_idr, NULL, 0, qm->qp_num, GFP_ATOMIC); + if (qp_id < 0) { + dev_info_ratelimited(dev, "All %u queues of QM are busy!\n", + qm->qp_num); + atomic64_inc(&qm->debug.dfx.create_qp_err_cnt); + return ERR_PTR(-EBUSY); + } + + qp = &qm->qp_array[qp_id]; + hisi_qm_unset_hw_reset(qp); + memset(qp->cqe, 0, sizeof(struct qm_cqe) * qp->cq_depth); + + qp->event_cb = NULL; + qp->req_cb = NULL; + qp->qp_id = qp_id; + qp->alg_type = alg_type; + qp->is_in_kernel = true; + qm->qp_in_used++; + atomic_set(&qp->qp_status.flags, QP_INIT); + + return qp; +} + +/** + * hisi_qm_create_qp() - Create a queue pair from qm. + * @qm: The qm we create a qp from. + * @alg_type: Accelerator specific algorithm type in sqc. + * + * return created qp, -EBUSY if all qps in qm allocated, -ENOMEM if allocating + * qp memory fails. + */ +static struct hisi_qp *hisi_qm_create_qp(struct hisi_qm *qm, u8 alg_type) +{ + struct hisi_qp *qp; + int ret; + + ret = qm_pm_get_sync(qm); + if (ret) + return ERR_PTR(ret); + + down_write(&qm->qps_lock); + qp = qm_create_qp_nolock(qm, alg_type); + up_write(&qm->qps_lock); + + if (IS_ERR(qp)) + qm_pm_put_sync(qm); + + return qp; +} + +/** + * hisi_qm_release_qp() - Release a qp back to its qm. + * @qp: The qp we want to release. + * + * This function releases the resource of a qp. + */ +static void hisi_qm_release_qp(struct hisi_qp *qp) +{ + struct hisi_qm *qm = qp->qm; + + down_write(&qm->qps_lock); + + if (!qm_qp_avail_state(qm, qp, QP_CLOSE)) { + up_write(&qm->qps_lock); + return; + } + + qm->qp_in_used--; + idr_remove(&qm->qp_idr, qp->qp_id); + + up_write(&qm->qps_lock); + + qm_pm_put_sync(qm); +} + +static int qm_sq_ctx_cfg(struct hisi_qp *qp, int qp_id, u32 pasid) +{ + struct hisi_qm *qm = qp->qm; + struct device *dev = &qm->pdev->dev; + enum qm_hw_ver ver = qm->ver; + struct qm_sqc *sqc; + dma_addr_t sqc_dma; + int ret; + + sqc = kzalloc(sizeof(struct qm_sqc), GFP_KERNEL); + if (!sqc) + return -ENOMEM; + + INIT_QC_COMMON(sqc, qp->sqe_dma, pasid); + if (ver == QM_HW_V1) { + sqc->dw3 = cpu_to_le32(QM_MK_SQC_DW3_V1(0, 0, 0, qm->sqe_size)); + sqc->w8 = cpu_to_le16(qp->sq_depth - 1); + } else { + sqc->dw3 = cpu_to_le32(QM_MK_SQC_DW3_V2(qm->sqe_size, qp->sq_depth)); + sqc->w8 = 0; /* rand_qc */ + } + sqc->cq_num = cpu_to_le16(qp_id); + sqc->w13 = cpu_to_le16(QM_MK_SQC_W13(0, 1, qp->alg_type)); + + if (ver >= QM_HW_V3 && qm->use_sva && !qp->is_in_kernel) + sqc->w11 = cpu_to_le16(QM_QC_PASID_ENABLE << + QM_QC_PASID_ENABLE_SHIFT); + + sqc_dma = dma_map_single(dev, sqc, sizeof(struct qm_sqc), + DMA_TO_DEVICE); + if (dma_mapping_error(dev, sqc_dma)) { + kfree(sqc); + return -ENOMEM; + } + + ret = hisi_qm_mb(qm, QM_MB_CMD_SQC, sqc_dma, qp_id, 0); + dma_unmap_single(dev, sqc_dma, sizeof(struct qm_sqc), DMA_TO_DEVICE); + kfree(sqc); + + return ret; +} + +static int qm_cq_ctx_cfg(struct hisi_qp *qp, int qp_id, u32 pasid) +{ + struct hisi_qm *qm = qp->qm; + struct device *dev = &qm->pdev->dev; + enum qm_hw_ver ver = qm->ver; + struct qm_cqc *cqc; + dma_addr_t cqc_dma; + int ret; + + cqc = kzalloc(sizeof(struct qm_cqc), GFP_KERNEL); + if (!cqc) + return -ENOMEM; + + INIT_QC_COMMON(cqc, qp->cqe_dma, pasid); + if (ver == QM_HW_V1) { + cqc->dw3 = cpu_to_le32(QM_MK_CQC_DW3_V1(0, 0, 0, + QM_QC_CQE_SIZE)); + cqc->w8 = cpu_to_le16(qp->cq_depth - 1); + } else { + cqc->dw3 = cpu_to_le32(QM_MK_CQC_DW3_V2(QM_QC_CQE_SIZE, qp->cq_depth)); + cqc->w8 = 0; /* rand_qc */ + } + cqc->dw6 = cpu_to_le32(1 << QM_CQ_PHASE_SHIFT | 1 << QM_CQ_FLAG_SHIFT); + + if (ver >= QM_HW_V3 && qm->use_sva && !qp->is_in_kernel) + cqc->w11 = cpu_to_le16(QM_QC_PASID_ENABLE); + + cqc_dma = dma_map_single(dev, cqc, sizeof(struct qm_cqc), + DMA_TO_DEVICE); + if (dma_mapping_error(dev, cqc_dma)) { + kfree(cqc); + return -ENOMEM; + } + + ret = hisi_qm_mb(qm, QM_MB_CMD_CQC, cqc_dma, qp_id, 0); + dma_unmap_single(dev, cqc_dma, sizeof(struct qm_cqc), DMA_TO_DEVICE); + kfree(cqc); + + return ret; +} + +static int qm_qp_ctx_cfg(struct hisi_qp *qp, int qp_id, u32 pasid) +{ + int ret; + + qm_init_qp_status(qp); + + ret = qm_sq_ctx_cfg(qp, qp_id, pasid); + if (ret) + return ret; + + return qm_cq_ctx_cfg(qp, qp_id, pasid); +} + +static int qm_start_qp_nolock(struct hisi_qp *qp, unsigned long arg) +{ + struct hisi_qm *qm = qp->qm; + struct device *dev = &qm->pdev->dev; + int qp_id = qp->qp_id; + u32 pasid = arg; + int ret; + + if (!qm_qp_avail_state(qm, qp, QP_START)) + return -EPERM; + + ret = qm_qp_ctx_cfg(qp, qp_id, pasid); + if (ret) + return ret; + + atomic_set(&qp->qp_status.flags, QP_START); + dev_dbg(dev, "queue %d started\n", qp_id); + + return 0; +} + +/** + * hisi_qm_start_qp() - Start a qp into running. + * @qp: The qp we want to start to run. + * @arg: Accelerator specific argument. + * + * After this function, qp can receive request from user. Return 0 if + * successful, Return -EBUSY if failed. + */ +int hisi_qm_start_qp(struct hisi_qp *qp, unsigned long arg) +{ + struct hisi_qm *qm = qp->qm; + int ret; + + down_write(&qm->qps_lock); + ret = qm_start_qp_nolock(qp, arg); + up_write(&qm->qps_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_start_qp); + +/** + * qp_stop_fail_cb() - call request cb. + * @qp: stopped failed qp. + * + * Callback function should be called whether task completed or not. + */ +static void qp_stop_fail_cb(struct hisi_qp *qp) +{ + int qp_used = atomic_read(&qp->qp_status.used); + u16 cur_tail = qp->qp_status.sq_tail; + u16 sq_depth = qp->sq_depth; + u16 cur_head = (cur_tail + sq_depth - qp_used) % sq_depth; + struct hisi_qm *qm = qp->qm; + u16 pos; + int i; + + for (i = 0; i < qp_used; i++) { + pos = (i + cur_head) % sq_depth; + qp->req_cb(qp, qp->sqe + (u32)(qm->sqe_size * pos)); + atomic_dec(&qp->qp_status.used); + } +} + +/** + * qm_drain_qp() - Drain a qp. + * @qp: The qp we want to drain. + * + * Determine whether the queue is cleared by judging the tail pointers of + * sq and cq. + */ +static int qm_drain_qp(struct hisi_qp *qp) +{ + size_t size = sizeof(struct qm_sqc) + sizeof(struct qm_cqc); + struct hisi_qm *qm = qp->qm; + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(qm->pdev)); + struct device *dev = &qm->pdev->dev; + struct qm_sqc *sqc; + struct qm_cqc *cqc; + dma_addr_t dma_addr; + int ret = 0, i = 0; + void *addr; + + /* No need to judge if master OOO is blocked. */ + if (qm_check_dev_error(pf_qm)) + return 0; + + /* Kunpeng930 supports drain qp by device */ + if (test_bit(QM_SUPPORT_STOP_QP, &qm->caps)) { + ret = qm_stop_qp(qp); + if (ret) + dev_err(dev, "Failed to stop qp(%u)!\n", qp->qp_id); + return ret; + } + + addr = qm_ctx_alloc(qm, size, &dma_addr); + if (IS_ERR(addr)) { + dev_err(dev, "Failed to alloc ctx for sqc and cqc!\n"); + return -ENOMEM; + } + + while (++i) { + ret = qm_dump_sqc_raw(qm, dma_addr, qp->qp_id); + if (ret) { + dev_err_ratelimited(dev, "Failed to dump sqc!\n"); + break; + } + sqc = addr; + + ret = qm_dump_cqc_raw(qm, (dma_addr + sizeof(struct qm_sqc)), + qp->qp_id); + if (ret) { + dev_err_ratelimited(dev, "Failed to dump cqc!\n"); + break; + } + cqc = addr + sizeof(struct qm_sqc); + + if ((sqc->tail == cqc->tail) && + (QM_SQ_TAIL_IDX(sqc) == QM_CQ_TAIL_IDX(cqc))) + break; + + if (i == MAX_WAIT_COUNTS) { + dev_err(dev, "Fail to empty queue %u!\n", qp->qp_id); + ret = -EBUSY; + break; + } + + usleep_range(WAIT_PERIOD_US_MIN, WAIT_PERIOD_US_MAX); + } + + qm_ctx_free(qm, size, addr, &dma_addr); + + return ret; +} + +static int qm_stop_qp_nolock(struct hisi_qp *qp) +{ + struct device *dev = &qp->qm->pdev->dev; + int ret; + + /* + * It is allowed to stop and release qp when reset, If the qp is + * stopped when reset but still want to be released then, the + * is_resetting flag should be set negative so that this qp will not + * be restarted after reset. + */ + if (atomic_read(&qp->qp_status.flags) == QP_STOP) { + qp->is_resetting = false; + return 0; + } + + if (!qm_qp_avail_state(qp->qm, qp, QP_STOP)) + return -EPERM; + + atomic_set(&qp->qp_status.flags, QP_STOP); + + ret = qm_drain_qp(qp); + if (ret) + dev_err(dev, "Failed to drain out data for stopping!\n"); + + + flush_workqueue(qp->qm->wq); + if (unlikely(qp->is_resetting && atomic_read(&qp->qp_status.used))) + qp_stop_fail_cb(qp); + + dev_dbg(dev, "stop queue %u!", qp->qp_id); + + return 0; +} + +/** + * hisi_qm_stop_qp() - Stop a qp in qm. + * @qp: The qp we want to stop. + * + * This function is reverse of hisi_qm_start_qp. Return 0 if successful. + */ +int hisi_qm_stop_qp(struct hisi_qp *qp) +{ + int ret; + + down_write(&qp->qm->qps_lock); + ret = qm_stop_qp_nolock(qp); + up_write(&qp->qm->qps_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_stop_qp); + +/** + * hisi_qp_send() - Queue up a task in the hardware queue. + * @qp: The qp in which to put the message. + * @msg: The message. + * + * This function will return -EBUSY if qp is currently full, and -EAGAIN + * if qp related qm is resetting. + * + * Note: This function may run with qm_irq_thread and ACC reset at same time. + * It has no race with qm_irq_thread. However, during hisi_qp_send, ACC + * reset may happen, we have no lock here considering performance. This + * causes current qm_db sending fail or can not receive sended sqe. QM + * sync/async receive function should handle the error sqe. ACC reset + * done function should clear used sqe to 0. + */ +int hisi_qp_send(struct hisi_qp *qp, const void *msg) +{ + struct hisi_qp_status *qp_status = &qp->qp_status; + u16 sq_tail = qp_status->sq_tail; + u16 sq_tail_next = (sq_tail + 1) % qp->sq_depth; + void *sqe = qm_get_avail_sqe(qp); + + if (unlikely(atomic_read(&qp->qp_status.flags) == QP_STOP || + atomic_read(&qp->qm->status.flags) == QM_STOP || + qp->is_resetting)) { + dev_info_ratelimited(&qp->qm->pdev->dev, "QP is stopped or resetting\n"); + return -EAGAIN; + } + + if (!sqe) + return -EBUSY; + + memcpy(sqe, msg, qp->qm->sqe_size); + + qm_db(qp->qm, qp->qp_id, QM_DOORBELL_CMD_SQ, sq_tail_next, 0); + atomic_inc(&qp->qp_status.used); + qp_status->sq_tail = sq_tail_next; + + return 0; +} +EXPORT_SYMBOL_GPL(hisi_qp_send); + +static void hisi_qm_cache_wb(struct hisi_qm *qm) +{ + unsigned int val; + + if (qm->ver == QM_HW_V1) + return; + + writel(0x1, qm->io_base + QM_CACHE_WB_START); + if (readl_relaxed_poll_timeout(qm->io_base + QM_CACHE_WB_DONE, + val, val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT)) + dev_err(&qm->pdev->dev, "QM writeback sqc cache fail!\n"); +} + +static void qm_qp_event_notifier(struct hisi_qp *qp) +{ + wake_up_interruptible(&qp->uacce_q->wait); +} + + /* This function returns free number of qp in qm. */ +static int hisi_qm_get_available_instances(struct uacce_device *uacce) +{ + struct hisi_qm *qm = uacce->priv; + int ret; + + down_read(&qm->qps_lock); + ret = qm->qp_num - qm->qp_in_used; + up_read(&qm->qps_lock); + + return ret; +} + +static void hisi_qm_set_hw_reset(struct hisi_qm *qm, int offset) +{ + int i; + + for (i = 0; i < qm->qp_num; i++) + qm_set_qp_disable(&qm->qp_array[i], offset); +} + +static int hisi_qm_uacce_get_queue(struct uacce_device *uacce, + unsigned long arg, + struct uacce_queue *q) +{ + struct hisi_qm *qm = uacce->priv; + struct hisi_qp *qp; + u8 alg_type = 0; + + qp = hisi_qm_create_qp(qm, alg_type); + if (IS_ERR(qp)) + return PTR_ERR(qp); + + q->priv = qp; + q->uacce = uacce; + qp->uacce_q = q; + qp->event_cb = qm_qp_event_notifier; + qp->pasid = arg; + qp->is_in_kernel = false; + + return 0; +} + +static void hisi_qm_uacce_put_queue(struct uacce_queue *q) +{ + struct hisi_qp *qp = q->priv; + + hisi_qm_release_qp(qp); +} + +/* map sq/cq/doorbell to user space */ +static int hisi_qm_uacce_mmap(struct uacce_queue *q, + struct vm_area_struct *vma, + struct uacce_qfile_region *qfr) +{ + struct hisi_qp *qp = q->priv; + struct hisi_qm *qm = qp->qm; + resource_size_t phys_base = qm->db_phys_base + + qp->qp_id * qm->db_interval; + size_t sz = vma->vm_end - vma->vm_start; + struct pci_dev *pdev = qm->pdev; + struct device *dev = &pdev->dev; + unsigned long vm_pgoff; + int ret; + + switch (qfr->type) { + case UACCE_QFRT_MMIO: + if (qm->ver == QM_HW_V1) { + if (sz > PAGE_SIZE * QM_DOORBELL_PAGE_NR) + return -EINVAL; + } else if (!test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps)) { + if (sz > PAGE_SIZE * (QM_DOORBELL_PAGE_NR + + QM_DOORBELL_SQ_CQ_BASE_V2 / PAGE_SIZE)) + return -EINVAL; + } else { + if (sz > qm->db_interval) + return -EINVAL; + } + + vma->vm_flags |= VM_IO; + + return remap_pfn_range(vma, vma->vm_start, + phys_base >> PAGE_SHIFT, + sz, pgprot_noncached(vma->vm_page_prot)); + case UACCE_QFRT_DUS: + if (sz != qp->qdma.size) + return -EINVAL; + + /* + * dma_mmap_coherent() requires vm_pgoff as 0 + * restore vm_pfoff to initial value for mmap() + */ + vm_pgoff = vma->vm_pgoff; + vma->vm_pgoff = 0; + ret = dma_mmap_coherent(dev, vma, qp->qdma.va, + qp->qdma.dma, sz); + vma->vm_pgoff = vm_pgoff; + return ret; + + default: + return -EINVAL; + } +} + +static int hisi_qm_uacce_start_queue(struct uacce_queue *q) +{ + struct hisi_qp *qp = q->priv; + + return hisi_qm_start_qp(qp, qp->pasid); +} + +static void hisi_qm_uacce_stop_queue(struct uacce_queue *q) +{ + hisi_qm_stop_qp(q->priv); +} + +static int hisi_qm_is_q_updated(struct uacce_queue *q) +{ + struct hisi_qp *qp = q->priv; + struct qm_cqe *cqe = qp->cqe + qp->qp_status.cq_head; + int updated = 0; + + while (QM_CQE_PHASE(cqe) == qp->qp_status.cqc_phase) { + /* make sure to read data from memory */ + dma_rmb(); + qm_cq_head_update(qp); + cqe = qp->cqe + qp->qp_status.cq_head; + updated = 1; + } + + return updated; +} + +static void qm_set_sqctype(struct uacce_queue *q, u16 type) +{ + struct hisi_qm *qm = q->uacce->priv; + struct hisi_qp *qp = q->priv; + + down_write(&qm->qps_lock); + qp->alg_type = type; + up_write(&qm->qps_lock); +} + +static long hisi_qm_uacce_ioctl(struct uacce_queue *q, unsigned int cmd, + unsigned long arg) +{ + struct hisi_qp *qp = q->priv; + struct hisi_qp_info qp_info; + struct hisi_qp_ctx qp_ctx; + + if (cmd == UACCE_CMD_QM_SET_QP_CTX) { + if (copy_from_user(&qp_ctx, (void __user *)arg, + sizeof(struct hisi_qp_ctx))) + return -EFAULT; + + if (qp_ctx.qc_type != 0 && qp_ctx.qc_type != 1) + return -EINVAL; + + qm_set_sqctype(q, qp_ctx.qc_type); + qp_ctx.id = qp->qp_id; + + if (copy_to_user((void __user *)arg, &qp_ctx, + sizeof(struct hisi_qp_ctx))) + return -EFAULT; + + return 0; + } else if (cmd == UACCE_CMD_QM_SET_QP_INFO) { + if (copy_from_user(&qp_info, (void __user *)arg, + sizeof(struct hisi_qp_info))) + return -EFAULT; + + qp_info.sqe_size = qp->qm->sqe_size; + qp_info.sq_depth = qp->sq_depth; + qp_info.cq_depth = qp->cq_depth; + + if (copy_to_user((void __user *)arg, &qp_info, + sizeof(struct hisi_qp_info))) + return -EFAULT; + + return 0; + } + + return -EINVAL; +} + +static const struct uacce_ops uacce_qm_ops = { + .get_available_instances = hisi_qm_get_available_instances, + .get_queue = hisi_qm_uacce_get_queue, + .put_queue = hisi_qm_uacce_put_queue, + .start_queue = hisi_qm_uacce_start_queue, + .stop_queue = hisi_qm_uacce_stop_queue, + .mmap = hisi_qm_uacce_mmap, + .ioctl = hisi_qm_uacce_ioctl, + .is_q_updated = hisi_qm_is_q_updated, +}; + +static int qm_alloc_uacce(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct uacce_device *uacce; + unsigned long mmio_page_nr; + unsigned long dus_page_nr; + u16 sq_depth, cq_depth; + struct uacce_interface interface = { + .flags = UACCE_DEV_SVA, + .ops = &uacce_qm_ops, + }; + int ret; + + ret = strscpy(interface.name, dev_driver_string(&pdev->dev), + sizeof(interface.name)); + if (ret < 0) + return -ENAMETOOLONG; + + uacce = uacce_alloc(&pdev->dev, &interface); + if (IS_ERR(uacce)) + return PTR_ERR(uacce); + + if (uacce->flags & UACCE_DEV_SVA) { + qm->use_sva = true; + } else { + /* only consider sva case */ + uacce_remove(uacce); + qm->uacce = NULL; + return -EINVAL; + } + + uacce->is_vf = pdev->is_virtfn; + uacce->priv = qm; + + if (qm->ver == QM_HW_V1) + uacce->api_ver = HISI_QM_API_VER_BASE; + else if (qm->ver == QM_HW_V2) + uacce->api_ver = HISI_QM_API_VER2_BASE; + else + uacce->api_ver = HISI_QM_API_VER3_BASE; + + if (qm->ver == QM_HW_V1) + mmio_page_nr = QM_DOORBELL_PAGE_NR; + else if (!test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps)) + mmio_page_nr = QM_DOORBELL_PAGE_NR + + QM_DOORBELL_SQ_CQ_BASE_V2 / PAGE_SIZE; + else + mmio_page_nr = qm->db_interval / PAGE_SIZE; + + qm_get_xqc_depth(qm, &sq_depth, &cq_depth, QM_QP_DEPTH_CAP); + + /* Add one more page for device or qp status */ + dus_page_nr = (PAGE_SIZE - 1 + qm->sqe_size * sq_depth + + sizeof(struct qm_cqe) * cq_depth + PAGE_SIZE) >> + PAGE_SHIFT; + + uacce->qf_pg_num[UACCE_QFRT_MMIO] = mmio_page_nr; + uacce->qf_pg_num[UACCE_QFRT_DUS] = dus_page_nr; + + qm->uacce = uacce; + + return 0; +} + +/** + * qm_frozen() - Try to froze QM to cut continuous queue request. If + * there is user on the QM, return failure without doing anything. + * @qm: The qm needed to be fronzen. + * + * This function frozes QM, then we can do SRIOV disabling. + */ +static int qm_frozen(struct hisi_qm *qm) +{ + if (test_bit(QM_DRIVER_REMOVING, &qm->misc_ctl)) + return 0; + + down_write(&qm->qps_lock); + + if (!qm->qp_in_used) { + qm->qp_in_used = qm->qp_num; + up_write(&qm->qps_lock); + set_bit(QM_DRIVER_REMOVING, &qm->misc_ctl); + return 0; + } + + up_write(&qm->qps_lock); + + return -EBUSY; +} + +static int qm_try_frozen_vfs(struct pci_dev *pdev, + struct hisi_qm_list *qm_list) +{ + struct hisi_qm *qm, *vf_qm; + struct pci_dev *dev; + int ret = 0; + + if (!qm_list || !pdev) + return -EINVAL; + + /* Try to frozen all the VFs as disable SRIOV */ + mutex_lock(&qm_list->lock); + list_for_each_entry(qm, &qm_list->list, list) { + dev = qm->pdev; + if (dev == pdev) + continue; + if (pci_physfn(dev) == pdev) { + vf_qm = pci_get_drvdata(dev); + ret = qm_frozen(vf_qm); + if (ret) + goto frozen_fail; + } + } + +frozen_fail: + mutex_unlock(&qm_list->lock); + + return ret; +} + +/** + * hisi_qm_wait_task_finish() - Wait until the task is finished + * when removing the driver. + * @qm: The qm needed to wait for the task to finish. + * @qm_list: The list of all available devices. + */ +void hisi_qm_wait_task_finish(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +{ + while (qm_frozen(qm) || + ((qm->fun_type == QM_HW_PF) && + qm_try_frozen_vfs(qm->pdev, qm_list))) { + msleep(WAIT_PERIOD); + } + + while (test_bit(QM_RST_SCHED, &qm->misc_ctl) || + test_bit(QM_RESETTING, &qm->misc_ctl)) + msleep(WAIT_PERIOD); + + udelay(REMOVE_WAIT_DELAY); +} +EXPORT_SYMBOL_GPL(hisi_qm_wait_task_finish); + +static void hisi_qp_memory_uninit(struct hisi_qm *qm, int num) +{ + struct device *dev = &qm->pdev->dev; + struct qm_dma *qdma; + int i; + + for (i = num - 1; i >= 0; i--) { + qdma = &qm->qp_array[i].qdma; + dma_free_coherent(dev, qdma->size, qdma->va, qdma->dma); + kfree(qm->poll_data[i].qp_finish_id); + } + + kfree(qm->poll_data); + kfree(qm->qp_array); +} + +static int hisi_qp_memory_init(struct hisi_qm *qm, size_t dma_size, int id, + u16 sq_depth, u16 cq_depth) +{ + struct device *dev = &qm->pdev->dev; + size_t off = qm->sqe_size * sq_depth; + struct hisi_qp *qp; + int ret = -ENOMEM; + + qm->poll_data[id].qp_finish_id = kcalloc(qm->qp_num, sizeof(u16), + GFP_KERNEL); + if (!qm->poll_data[id].qp_finish_id) + return -ENOMEM; + + qp = &qm->qp_array[id]; + qp->qdma.va = dma_alloc_coherent(dev, dma_size, &qp->qdma.dma, + GFP_KERNEL); + if (!qp->qdma.va) + goto err_free_qp_finish_id; + + qp->sqe = qp->qdma.va; + qp->sqe_dma = qp->qdma.dma; + qp->cqe = qp->qdma.va + off; + qp->cqe_dma = qp->qdma.dma + off; + qp->qdma.size = dma_size; + qp->sq_depth = sq_depth; + qp->cq_depth = cq_depth; + qp->qm = qm; + qp->qp_id = id; + + return 0; + +err_free_qp_finish_id: + kfree(qm->poll_data[id].qp_finish_id); + return ret; +} + +static void hisi_qm_pre_init(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + + if (qm->ver == QM_HW_V1) + qm->ops = &qm_hw_ops_v1; + else if (qm->ver == QM_HW_V2) + qm->ops = &qm_hw_ops_v2; + else + qm->ops = &qm_hw_ops_v3; + + pci_set_drvdata(pdev, qm); + mutex_init(&qm->mailbox_lock); + init_rwsem(&qm->qps_lock); + qm->qp_in_used = 0; + qm->misc_ctl = false; + if (test_bit(QM_SUPPORT_RPM, &qm->caps)) { + if (!acpi_device_power_manageable(ACPI_COMPANION(&pdev->dev))) + dev_info(&pdev->dev, "_PS0 and _PR0 are not defined"); + } +} + +static void qm_cmd_uninit(struct hisi_qm *qm) +{ + u32 val; + + if (!test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) + return; + + val = readl(qm->io_base + QM_IFC_INT_MASK); + val |= QM_IFC_INT_DISABLE; + writel(val, qm->io_base + QM_IFC_INT_MASK); +} + +static void qm_cmd_init(struct hisi_qm *qm) +{ + u32 val; + + if (!test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) + return; + + /* Clear communication interrupt source */ + qm_clear_cmd_interrupt(qm, QM_IFC_INT_SOURCE_CLR); + + /* Enable pf to vf communication reg. */ + val = readl(qm->io_base + QM_IFC_INT_MASK); + val &= ~QM_IFC_INT_DISABLE; + writel(val, qm->io_base + QM_IFC_INT_MASK); +} + +static void qm_put_pci_res(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + + if (test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps)) + iounmap(qm->db_io_base); + + iounmap(qm->io_base); + pci_release_mem_regions(pdev); +} + +static void hisi_qm_pci_uninit(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + + pci_free_irq_vectors(pdev); + qm_put_pci_res(qm); + pci_disable_device(pdev); +} + +static void hisi_qm_set_state(struct hisi_qm *qm, enum vf_state state) +{ + /* set vf driver state */ + if (qm->ver > QM_HW_V2) + writel(state, qm->io_base + QM_VF_STATE); +} + +static void qm_last_regs_uninit(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + + if (qm->fun_type == QM_HW_VF || !debug->qm_last_words) + return; + + kfree(debug->qm_last_words); + debug->qm_last_words = NULL; +} + +static void hisi_qm_unint_work(struct hisi_qm *qm) +{ + destroy_workqueue(qm->wq); +} + +static void hisi_qm_memory_uninit(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + + hisi_qp_memory_uninit(qm, qm->qp_num); + if (qm->qdma.va) { + hisi_qm_cache_wb(qm); + dma_free_coherent(dev, qm->qdma.size, + qm->qdma.va, qm->qdma.dma); + } + + idr_destroy(&qm->qp_idr); + + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) + kfree(qm->factor); +} + +/** + * hisi_qm_uninit() - Uninitialize qm. + * @qm: The qm needed uninit. + * + * This function uninits qm related device resources. + */ +void hisi_qm_uninit(struct hisi_qm *qm) +{ + qm_last_regs_uninit(qm); + + qm_cmd_uninit(qm); + hisi_qm_unint_work(qm); + down_write(&qm->qps_lock); + + if (!qm_avail_state(qm, QM_CLOSE)) { + up_write(&qm->qps_lock); + return; + } + + hisi_qm_memory_uninit(qm); + hisi_qm_set_state(qm, VF_NOT_READY); + up_write(&qm->qps_lock); + + qm_irqs_unregister(qm); + hisi_qm_pci_uninit(qm); + if (qm->use_sva) { + uacce_remove(qm->uacce); + qm->uacce = NULL; + } +} +EXPORT_SYMBOL_GPL(hisi_qm_uninit); + +/** + * hisi_qm_get_vft() - Get vft from a qm. + * @qm: The qm we want to get its vft. + * @base: The base number of queue in vft. + * @number: The number of queues in vft. + * + * We can allocate multiple queues to a qm by configuring virtual function + * function in VF driver to get the queue information. + * + * qm hw v1 does not support this interface. + */ +static int hisi_qm_get_vft(struct hisi_qm *qm, u32 *base, u32 *number) +{ + if (!base || !number) + return -EINVAL; + + if (!qm->ops->get_vft) { + dev_err(&qm->pdev->dev, "Don't support vft read!\n"); + return -EINVAL; + } + + return qm->ops->get_vft(qm, base, number); +} + +/** + * hisi_qm_set_vft() - Set vft to a qm. + * @qm: The qm we want to set its vft. + * @fun_num: The function number. + * @base: The base number of queue in vft. + * @number: The number of queues in vft. + * + * This function is alway called in PF driver, it is used to assign queues + * among PF and VFs. + * + * Assign queues A~B to PF: hisi_qm_set_vft(qm, 0, A, B - A + 1) + * Assign queues A~B to VF: hisi_qm_set_vft(qm, 2, A, B - A + 1) + * (VF function number 0x2) + */ +static int hisi_qm_set_vft(struct hisi_qm *qm, u32 fun_num, u32 base, + u32 number) +{ + u32 max_q_num = qm->ctrl_qp_num; + + if (base >= max_q_num || number > max_q_num || + (base + number) > max_q_num) + return -EINVAL; + + return qm_set_sqc_cqc_vft(qm, fun_num, base, number); +} + +static void qm_init_eq_aeq_status(struct hisi_qm *qm) +{ + struct hisi_qm_status *status = &qm->status; + + status->eq_head = 0; + status->aeq_head = 0; + status->eqc_phase = true; + status->aeqc_phase = true; +} + +static void qm_enable_eq_aeq_interrupts(struct hisi_qm *qm) +{ + /* Clear eq/aeq interrupt source */ + qm_db(qm, 0, QM_DOORBELL_CMD_AEQ, qm->status.aeq_head, 0); + qm_db(qm, 0, QM_DOORBELL_CMD_EQ, qm->status.eq_head, 0); + + writel(0x0, qm->io_base + QM_VF_EQ_INT_MASK); + writel(0x0, qm->io_base + QM_VF_AEQ_INT_MASK); +} + +static void qm_disable_eq_aeq_interrupts(struct hisi_qm *qm) +{ + writel(0x1, qm->io_base + QM_VF_EQ_INT_MASK); + writel(0x1, qm->io_base + QM_VF_AEQ_INT_MASK); +} + +static int qm_eq_ctx_cfg(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct qm_eqc *eqc; + dma_addr_t eqc_dma; + int ret; + + eqc = kzalloc(sizeof(struct qm_eqc), GFP_KERNEL); + if (!eqc) + return -ENOMEM; + + eqc->base_l = cpu_to_le32(lower_32_bits(qm->eqe_dma)); + eqc->base_h = cpu_to_le32(upper_32_bits(qm->eqe_dma)); + if (qm->ver == QM_HW_V1) + eqc->dw3 = cpu_to_le32(QM_EQE_AEQE_SIZE); + eqc->dw6 = cpu_to_le32(((u32)qm->eq_depth - 1) | (1 << QM_EQC_PHASE_SHIFT)); + + eqc_dma = dma_map_single(dev, eqc, sizeof(struct qm_eqc), + DMA_TO_DEVICE); + if (dma_mapping_error(dev, eqc_dma)) { + kfree(eqc); + return -ENOMEM; + } + + ret = hisi_qm_mb(qm, QM_MB_CMD_EQC, eqc_dma, 0, 0); + dma_unmap_single(dev, eqc_dma, sizeof(struct qm_eqc), DMA_TO_DEVICE); + kfree(eqc); + + return ret; +} + +static int qm_aeq_ctx_cfg(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct qm_aeqc *aeqc; + dma_addr_t aeqc_dma; + int ret; + + aeqc = kzalloc(sizeof(struct qm_aeqc), GFP_KERNEL); + if (!aeqc) + return -ENOMEM; + + aeqc->base_l = cpu_to_le32(lower_32_bits(qm->aeqe_dma)); + aeqc->base_h = cpu_to_le32(upper_32_bits(qm->aeqe_dma)); + aeqc->dw6 = cpu_to_le32(((u32)qm->aeq_depth - 1) | (1 << QM_EQC_PHASE_SHIFT)); + + aeqc_dma = dma_map_single(dev, aeqc, sizeof(struct qm_aeqc), + DMA_TO_DEVICE); + if (dma_mapping_error(dev, aeqc_dma)) { + kfree(aeqc); + return -ENOMEM; + } + + ret = hisi_qm_mb(qm, QM_MB_CMD_AEQC, aeqc_dma, 0, 0); + dma_unmap_single(dev, aeqc_dma, sizeof(struct qm_aeqc), DMA_TO_DEVICE); + kfree(aeqc); + + return ret; +} + +static int qm_eq_aeq_ctx_cfg(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + qm_init_eq_aeq_status(qm); + + ret = qm_eq_ctx_cfg(qm); + if (ret) { + dev_err(dev, "Set eqc failed!\n"); + return ret; + } + + return qm_aeq_ctx_cfg(qm); +} + +static int __hisi_qm_start(struct hisi_qm *qm) +{ + int ret; + + WARN_ON(!qm->qdma.va); + + if (qm->fun_type == QM_HW_PF) { + ret = hisi_qm_set_vft(qm, 0, qm->qp_base, qm->qp_num); + if (ret) + return ret; + } + + ret = qm_eq_aeq_ctx_cfg(qm); + if (ret) + return ret; + + ret = hisi_qm_mb(qm, QM_MB_CMD_SQC_BT, qm->sqc_dma, 0, 0); + if (ret) + return ret; + + ret = hisi_qm_mb(qm, QM_MB_CMD_CQC_BT, qm->cqc_dma, 0, 0); + if (ret) + return ret; + + qm_init_prefetch(qm); + qm_enable_eq_aeq_interrupts(qm); + + return 0; +} + +/** + * hisi_qm_start() - start qm + * @qm: The qm to be started. + * + * This function starts a qm, then we can allocate qp from this qm. + */ +int hisi_qm_start(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret = 0; + + down_write(&qm->qps_lock); + + if (!qm_avail_state(qm, QM_START)) { + up_write(&qm->qps_lock); + return -EPERM; + } + + dev_dbg(dev, "qm start with %u queue pairs\n", qm->qp_num); + + if (!qm->qp_num) { + dev_err(dev, "qp_num should not be 0\n"); + ret = -EINVAL; + goto err_unlock; + } + + ret = __hisi_qm_start(qm); + if (!ret) + atomic_set(&qm->status.flags, QM_START); + + hisi_qm_set_state(qm, VF_READY); + +err_unlock: + up_write(&qm->qps_lock); + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_start); + +static int qm_restart(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct hisi_qp *qp; + int ret, i; + + ret = hisi_qm_start(qm); + if (ret < 0) + return ret; + + down_write(&qm->qps_lock); + for (i = 0; i < qm->qp_num; i++) { + qp = &qm->qp_array[i]; + if (atomic_read(&qp->qp_status.flags) == QP_STOP && + qp->is_resetting == true) { + ret = qm_start_qp_nolock(qp, 0); + if (ret < 0) { + dev_err(dev, "Failed to start qp%d!\n", i); + + up_write(&qm->qps_lock); + return ret; + } + qp->is_resetting = false; + } + } + up_write(&qm->qps_lock); + + return 0; +} + +/* Stop started qps in reset flow */ +static int qm_stop_started_qp(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct hisi_qp *qp; + int i, ret; + + for (i = 0; i < qm->qp_num; i++) { + qp = &qm->qp_array[i]; + if (qp && atomic_read(&qp->qp_status.flags) == QP_START) { + qp->is_resetting = true; + ret = qm_stop_qp_nolock(qp); + if (ret < 0) { + dev_err(dev, "Failed to stop qp%d!\n", i); + return ret; + } + } + } + + return 0; +} + + +/** + * qm_clear_queues() - Clear all queues memory in a qm. + * @qm: The qm in which the queues will be cleared. + * + * This function clears all queues memory in a qm. Reset of accelerator can + * use this to clear queues. + */ +static void qm_clear_queues(struct hisi_qm *qm) +{ + struct hisi_qp *qp; + int i; + + for (i = 0; i < qm->qp_num; i++) { + qp = &qm->qp_array[i]; + if (qp->is_in_kernel && qp->is_resetting) + memset(qp->qdma.va, 0, qp->qdma.size); + } + + memset(qm->qdma.va, 0, qm->qdma.size); +} + +/** + * hisi_qm_stop() - Stop a qm. + * @qm: The qm which will be stopped. + * @r: The reason to stop qm. + * + * This function stops qm and its qps, then qm can not accept request. + * Related resources are not released at this state, we can use hisi_qm_start + * to let qm start again. + */ +int hisi_qm_stop(struct hisi_qm *qm, enum qm_stop_reason r) +{ + struct device *dev = &qm->pdev->dev; + int ret = 0; + + hisi_qm_set_state(qm, VF_PREPARE); + + down_write(&qm->qps_lock); + + qm->status.stop_reason = r; + if (!qm_avail_state(qm, QM_STOP)) { + ret = -EPERM; + goto err_unlock; + } + + if (qm->status.stop_reason == QM_SOFT_RESET || + qm->status.stop_reason == QM_FLR) { + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_TX_OFFSET); + ret = qm_stop_started_qp(qm); + if (ret < 0) { + dev_err(dev, "Failed to stop started qp!\n"); + goto err_unlock; + } + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_RX_OFFSET); + } + + qm_disable_eq_aeq_interrupts(qm); + if (qm->fun_type == QM_HW_PF) { + ret = hisi_qm_set_vft(qm, 0, 0, 0); + if (ret < 0) { + dev_err(dev, "Failed to set vft!\n"); + ret = -EBUSY; + goto err_unlock; + } + } + + qm_clear_queues(qm); + atomic_set(&qm->status.flags, QM_STOP); + +err_unlock: + up_write(&qm->qps_lock); + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_stop); + +static ssize_t qm_status_read(struct file *filp, char __user *buffer, + size_t count, loff_t *pos) +{ + struct hisi_qm *qm = filp->private_data; + char buf[QM_DBG_READ_LEN]; + int val, len; + + val = atomic_read(&qm->status.flags); + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", qm_s[val]); + + return simple_read_from_buffer(buffer, count, pos, buf, len); +} + +static const struct file_operations qm_status_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = qm_status_read, +}; + +static int qm_debugfs_atomic64_set(void *data, u64 val) +{ + if (val) + return -EINVAL; + + atomic64_set((atomic64_t *)data, 0); + + return 0; +} + +static int qm_debugfs_atomic64_get(void *data, u64 *val) +{ + *val = atomic64_read((atomic64_t *)data); + + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(qm_atomic64_ops, qm_debugfs_atomic64_get, + qm_debugfs_atomic64_set, "%llu\n"); + +static void qm_hw_error_init(struct hisi_qm *qm) +{ + if (!qm->ops->hw_error_init) { + dev_err(&qm->pdev->dev, "QM doesn't support hw error handling!\n"); + return; + } + + qm->ops->hw_error_init(qm); +} + +static void qm_hw_error_uninit(struct hisi_qm *qm) +{ + if (!qm->ops->hw_error_uninit) { + dev_err(&qm->pdev->dev, "Unexpected QM hw error uninit!\n"); + return; + } + + qm->ops->hw_error_uninit(qm); +} + +static enum acc_err_result qm_hw_error_handle(struct hisi_qm *qm) +{ + if (!qm->ops->hw_error_handle) { + dev_err(&qm->pdev->dev, "QM doesn't support hw error report!\n"); + return ACC_ERR_NONE; + } + + return qm->ops->hw_error_handle(qm); +} + +/** + * hisi_qm_dev_err_init() - Initialize device error configuration. + * @qm: The qm for which we want to do error initialization. + * + * Initialize QM and device error related configuration. + */ +void hisi_qm_dev_err_init(struct hisi_qm *qm) +{ + if (qm->fun_type == QM_HW_VF) + return; + + qm_hw_error_init(qm); + + if (!qm->err_ini->hw_err_enable) { + dev_err(&qm->pdev->dev, "Device doesn't support hw error init!\n"); + return; + } + qm->err_ini->hw_err_enable(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_dev_err_init); + +/** + * hisi_qm_dev_err_uninit() - Uninitialize device error configuration. + * @qm: The qm for which we want to do error uninitialization. + * + * Uninitialize QM and device error related configuration. + */ +void hisi_qm_dev_err_uninit(struct hisi_qm *qm) +{ + if (qm->fun_type == QM_HW_VF) + return; + + qm_hw_error_uninit(qm); + + if (!qm->err_ini->hw_err_disable) { + dev_err(&qm->pdev->dev, "Unexpected device hw error uninit!\n"); + return; + } + qm->err_ini->hw_err_disable(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_dev_err_uninit); + +/** + * hisi_qm_free_qps() - free multiple queue pairs. + * @qps: The queue pairs need to be freed. + * @qp_num: The num of queue pairs. + */ +void hisi_qm_free_qps(struct hisi_qp **qps, int qp_num) +{ + int i; + + if (!qps || qp_num <= 0) + return; + + for (i = qp_num - 1; i >= 0; i--) + hisi_qm_release_qp(qps[i]); +} +EXPORT_SYMBOL_GPL(hisi_qm_free_qps); + +static void free_list(struct list_head *head) +{ + struct hisi_qm_resource *res, *tmp; + + list_for_each_entry_safe(res, tmp, head, list) { + list_del(&res->list); + kfree(res); + } +} + +static int hisi_qm_sort_devices(int node, struct list_head *head, + struct hisi_qm_list *qm_list) +{ + struct hisi_qm_resource *res, *tmp; + struct hisi_qm *qm; + struct list_head *n; + struct device *dev; + int dev_node; + + list_for_each_entry(qm, &qm_list->list, list) { + dev = &qm->pdev->dev; + + dev_node = dev_to_node(dev); + if (dev_node < 0) + dev_node = 0; + + res = kzalloc(sizeof(*res), GFP_KERNEL); + if (!res) + return -ENOMEM; + + res->qm = qm; + res->distance = node_distance(dev_node, node); + n = head; + list_for_each_entry(tmp, head, list) { + if (res->distance < tmp->distance) { + n = &tmp->list; + break; + } + } + list_add_tail(&res->list, n); + } + + return 0; +} + +/** + * hisi_qm_alloc_qps_node() - Create multiple queue pairs. + * @qm_list: The list of all available devices. + * @qp_num: The number of queue pairs need created. + * @alg_type: The algorithm type. + * @node: The numa node. + * @qps: The queue pairs need created. + * + * This function will sort all available device according to numa distance. + * Then try to create all queue pairs from one device, if all devices do + * not meet the requirements will return error. + */ +int hisi_qm_alloc_qps_node(struct hisi_qm_list *qm_list, int qp_num, + u8 alg_type, int node, struct hisi_qp **qps) +{ + struct hisi_qm_resource *tmp; + int ret = -ENODEV; + LIST_HEAD(head); + int i; + + if (!qps || !qm_list || qp_num <= 0) + return -EINVAL; + + mutex_lock(&qm_list->lock); + if (hisi_qm_sort_devices(node, &head, qm_list)) { + mutex_unlock(&qm_list->lock); + goto err; + } + + list_for_each_entry(tmp, &head, list) { + for (i = 0; i < qp_num; i++) { + qps[i] = hisi_qm_create_qp(tmp->qm, alg_type); + if (IS_ERR(qps[i])) { + hisi_qm_free_qps(qps, i); + break; + } + } + + if (i == qp_num) { + ret = 0; + break; + } + } + + mutex_unlock(&qm_list->lock); + if (ret) + pr_info("Failed to create qps, node[%d], alg[%u], qp[%d]!\n", + node, alg_type, qp_num); + +err: + free_list(&head); + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_alloc_qps_node); + +static int qm_vf_q_assign(struct hisi_qm *qm, u32 num_vfs) +{ + u32 remain_q_num, vfs_q_num, act_q_num, q_num, i, j; + u32 max_qp_num = qm->max_qp_num; + u32 q_base = qm->qp_num; + int ret; + + if (!num_vfs) + return -EINVAL; + + vfs_q_num = qm->ctrl_qp_num - qm->qp_num; + + /* If vfs_q_num is less than num_vfs, return error. */ + if (vfs_q_num < num_vfs) + return -EINVAL; + + q_num = vfs_q_num / num_vfs; + remain_q_num = vfs_q_num % num_vfs; + + for (i = num_vfs; i > 0; i--) { + /* + * if q_num + remain_q_num > max_qp_num in last vf, divide the + * remaining queues equally. + */ + if (i == num_vfs && q_num + remain_q_num <= max_qp_num) { + act_q_num = q_num + remain_q_num; + remain_q_num = 0; + } else if (remain_q_num > 0) { + act_q_num = q_num + 1; + remain_q_num--; + } else { + act_q_num = q_num; + } + + act_q_num = min_t(int, act_q_num, max_qp_num); + ret = hisi_qm_set_vft(qm, i, q_base, act_q_num); + if (ret) { + for (j = num_vfs; j > i; j--) + hisi_qm_set_vft(qm, j, 0, 0); + return ret; + } + q_base += act_q_num; + } + + return 0; +} + +static int qm_clear_vft_config(struct hisi_qm *qm) +{ + int ret; + u32 i; + + for (i = 1; i <= qm->vfs_num; i++) { + ret = hisi_qm_set_vft(qm, i, 0, 0); + if (ret) + return ret; + } + qm->vfs_num = 0; + + return 0; +} + +static int qm_func_shaper_enable(struct hisi_qm *qm, u32 fun_index, u32 qos) +{ + struct device *dev = &qm->pdev->dev; + u32 ir = qos * QM_QOS_RATE; + int ret, total_vfs, i; + + total_vfs = pci_sriov_get_totalvfs(qm->pdev); + if (fun_index > total_vfs) + return -EINVAL; + + qm->factor[fun_index].func_qos = qos; + + ret = qm_get_shaper_para(ir, &qm->factor[fun_index]); + if (ret) { + dev_err(dev, "failed to calculate shaper parameter!\n"); + return -EINVAL; + } + + for (i = ALG_TYPE_0; i <= ALG_TYPE_1; i++) { + /* The base number of queue reuse for different alg type */ + ret = qm_set_vft_common(qm, SHAPER_VFT, fun_index, i, 1); + if (ret) { + dev_err(dev, "type: %d, failed to set shaper vft!\n", i); + return -EINVAL; + } + } + + return 0; +} + +static u32 qm_get_shaper_vft_qos(struct hisi_qm *qm, u32 fun_index) +{ + u64 cir_u = 0, cir_b = 0, cir_s = 0; + u64 shaper_vft, ir_calc, ir; + unsigned int val; + u32 error_rate; + int ret; + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); + if (ret) + return 0; + + writel(0x1, qm->io_base + QM_VFT_CFG_OP_WR); + writel(SHAPER_VFT, qm->io_base + QM_VFT_CFG_TYPE); + writel(fun_index, qm->io_base + QM_VFT_CFG); + + writel(0x0, qm->io_base + QM_VFT_CFG_RDY); + writel(0x1, qm->io_base + QM_VFT_CFG_OP_ENABLE); + + ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val, + val & BIT(0), POLL_PERIOD, + POLL_TIMEOUT); + if (ret) + return 0; + + shaper_vft = readl(qm->io_base + QM_VFT_CFG_DATA_L) | + ((u64)readl(qm->io_base + QM_VFT_CFG_DATA_H) << 32); + + cir_b = shaper_vft & QM_SHAPER_CIR_B_MASK; + cir_u = shaper_vft & QM_SHAPER_CIR_U_MASK; + cir_u = cir_u >> QM_SHAPER_FACTOR_CIR_U_SHIFT; + + cir_s = shaper_vft & QM_SHAPER_CIR_S_MASK; + cir_s = cir_s >> QM_SHAPER_FACTOR_CIR_S_SHIFT; + + ir_calc = acc_shaper_para_calc(cir_b, cir_u, cir_s); + + ir = qm->factor[fun_index].func_qos * QM_QOS_RATE; + + error_rate = QM_QOS_EXPAND_RATE * (u32)abs(ir_calc - ir) / ir; + if (error_rate > QM_QOS_MIN_ERROR_RATE) { + pci_err(qm->pdev, "error_rate: %u, get function qos is error!\n", error_rate); + return 0; + } + + return ir; +} + +static void qm_vf_get_qos(struct hisi_qm *qm, u32 fun_num) +{ + struct device *dev = &qm->pdev->dev; + u64 mb_cmd; + u32 qos; + int ret; + + qos = qm_get_shaper_vft_qos(qm, fun_num); + if (!qos) { + dev_err(dev, "function(%u) failed to get qos by PF!\n", fun_num); + return; + } + + mb_cmd = QM_PF_SET_QOS | (u64)qos << QM_MB_CMD_DATA_SHIFT; + ret = qm_ping_single_vf(qm, mb_cmd, fun_num); + if (ret) + dev_err(dev, "failed to send cmd to VF(%u)!\n", fun_num); +} + +static int qm_vf_read_qos(struct hisi_qm *qm) +{ + int cnt = 0; + int ret = -EINVAL; + + /* reset mailbox qos val */ + qm->mb_qos = 0; + + /* vf ping pf to get function qos */ + ret = qm_ping_pf(qm, QM_VF_GET_QOS); + if (ret) { + pci_err(qm->pdev, "failed to send cmd to PF to get qos!\n"); + return ret; + } + + while (true) { + msleep(QM_WAIT_DST_ACK); + if (qm->mb_qos) + break; + + if (++cnt > QM_MAX_VF_WAIT_COUNT) { + pci_err(qm->pdev, "PF ping VF timeout!\n"); + return -ETIMEDOUT; + } + } + + return ret; +} + +static ssize_t qm_algqos_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct hisi_qm *qm = filp->private_data; + char tbuf[QM_DBG_READ_LEN]; + u32 qos_val, ir; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + /* Mailbox and reset cannot be operated at the same time */ + if (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) { + pci_err(qm->pdev, "dev resetting, read alg qos failed!\n"); + ret = -EAGAIN; + goto err_put_dfx_access; + } + + if (qm->fun_type == QM_HW_PF) { + ir = qm_get_shaper_vft_qos(qm, 0); + } else { + ret = qm_vf_read_qos(qm); + if (ret) + goto err_get_status; + ir = qm->mb_qos; + } + + qos_val = ir / QM_QOS_RATE; + ret = scnprintf(tbuf, QM_DBG_READ_LEN, "%u\n", qos_val); + + ret = simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_get_status: + clear_bit(QM_RESETTING, &qm->misc_ctl); +err_put_dfx_access: + hisi_qm_put_dfx_access(qm); + return ret; +} + +static ssize_t qm_get_qos_value(struct hisi_qm *qm, const char *buf, + unsigned long *val, + unsigned int *fun_index) +{ + struct bus_type *bus_type = qm->pdev->dev.bus; + char tbuf_bdf[QM_DBG_READ_LEN] = {0}; + char val_buf[QM_DBG_READ_LEN] = {0}; + struct pci_dev *pdev; + struct device *dev; + int ret; + + ret = sscanf(buf, "%s %s", tbuf_bdf, val_buf); + if (ret != QM_QOS_PARAM_NUM) + return -EINVAL; + + ret = kstrtoul(val_buf, 10, val); + if (ret || *val == 0 || *val > QM_QOS_MAX_VAL) { + pci_err(qm->pdev, "input qos value is error, please set 1~1000!\n"); + return -EINVAL; + } + + dev = bus_find_device_by_name(bus_type, NULL, tbuf_bdf); + if (!dev) { + pci_err(qm->pdev, "input pci bdf number is error!\n"); + return -ENODEV; + } + + pdev = container_of(dev, struct pci_dev, dev); + + *fun_index = pdev->devfn; + + return 0; +} + +static ssize_t qm_algqos_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct hisi_qm *qm = filp->private_data; + char tbuf[QM_DBG_READ_LEN]; + unsigned int fun_index; + unsigned long val; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= QM_DBG_READ_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, QM_DBG_READ_LEN - 1, pos, buf, count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + ret = qm_get_qos_value(qm, tbuf, &val, &fun_index); + if (ret) + return ret; + + /* Mailbox and reset cannot be operated at the same time */ + if (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) { + pci_err(qm->pdev, "dev resetting, write alg qos failed!\n"); + return -EAGAIN; + } + + ret = qm_pm_get_sync(qm); + if (ret) { + ret = -EINVAL; + goto err_get_status; + } + + ret = qm_func_shaper_enable(qm, fun_index, val); + if (ret) { + pci_err(qm->pdev, "failed to enable function shaper!\n"); + ret = -EINVAL; + goto err_put_sync; + } + + pci_info(qm->pdev, "the qos value of function%u is set to %lu.\n", + fun_index, val); + ret = count; + +err_put_sync: + qm_pm_put_sync(qm); +err_get_status: + clear_bit(QM_RESETTING, &qm->misc_ctl); + return ret; +} + +static const struct file_operations qm_algqos_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = qm_algqos_read, + .write = qm_algqos_write, +}; + +/** + * hisi_qm_set_algqos_init() - Initialize function qos debugfs files. + * @qm: The qm for which we want to add debugfs files. + * + * Create function qos debugfs files, VF ping PF to get function qos. + */ +static void hisi_qm_set_algqos_init(struct hisi_qm *qm) +{ + if (qm->fun_type == QM_HW_PF) + debugfs_create_file("alg_qos", 0644, qm->debug.debug_root, + qm, &qm_algqos_fops); + else if (test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) + debugfs_create_file("alg_qos", 0444, qm->debug.debug_root, + qm, &qm_algqos_fops); +} + +/** + * hisi_qm_debug_init() - Initialize qm related debugfs files. + * @qm: The qm for which we want to add debugfs files. + * + * Create qm related debugfs files. + */ +void hisi_qm_debug_init(struct hisi_qm *qm) +{ + struct dfx_diff_registers *qm_regs = qm->debug.qm_diff_regs; + struct qm_dfx *dfx = &qm->debug.dfx; + struct dentry *qm_d; + void *data; + int i; + + qm_d = debugfs_create_dir("qm", qm->debug.debug_root); + qm->debug.qm_d = qm_d; + + /* only show this in PF */ + if (qm->fun_type == QM_HW_PF) { + qm_create_debugfs_file(qm, qm->debug.debug_root, CURRENT_QM); + for (i = CURRENT_Q; i < DEBUG_FILE_NUM; i++) + qm_create_debugfs_file(qm, qm->debug.qm_d, i); + } + + if (qm_regs) + debugfs_create_file("diff_regs", 0444, qm->debug.qm_d, + qm, &qm_diff_regs_fops); + + debugfs_create_file("regs", 0444, qm->debug.qm_d, qm, &qm_regs_fops); + + debugfs_create_file("cmd", 0600, qm->debug.qm_d, qm, &qm_cmd_fops); + + debugfs_create_file("status", 0444, qm->debug.qm_d, qm, + &qm_status_fops); + for (i = 0; i < ARRAY_SIZE(qm_dfx_files); i++) { + data = (atomic64_t *)((uintptr_t)dfx + qm_dfx_files[i].offset); + debugfs_create_file(qm_dfx_files[i].name, + 0644, + qm_d, + data, + &qm_atomic64_ops); + } + + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) + hisi_qm_set_algqos_init(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_debug_init); + +/** + * hisi_qm_debug_regs_clear() - clear qm debug related registers. + * @qm: The qm for which we want to clear its debug registers. + */ +void hisi_qm_debug_regs_clear(struct hisi_qm *qm) +{ + const struct debugfs_reg32 *regs; + int i; + + /* clear current_qm */ + writel(0x0, qm->io_base + QM_DFX_MB_CNT_VF); + writel(0x0, qm->io_base + QM_DFX_DB_CNT_VF); + + /* clear current_q */ + writel(0x0, qm->io_base + QM_DFX_SQE_CNT_VF_SQN); + writel(0x0, qm->io_base + QM_DFX_CQE_CNT_VF_CQN); + + /* + * these registers are reading and clearing, so clear them after + * reading them. + */ + writel(0x1, qm->io_base + QM_DFX_CNT_CLR_CE); + + regs = qm_dfx_regs; + for (i = 0; i < CNT_CYC_REGS_NUM; i++) { + readl(qm->io_base + regs->offset); + regs++; + } + + /* clear clear_enable */ + writel(0x0, qm->io_base + QM_DFX_CNT_CLR_CE); +} +EXPORT_SYMBOL_GPL(hisi_qm_debug_regs_clear); + +static void hisi_qm_init_vf_qos(struct hisi_qm *qm, int total_func) +{ + int i; + + for (i = 1; i <= total_func; i++) + qm->factor[i].func_qos = QM_QOS_MAX_VAL; +} + +/** + * hisi_qm_sriov_enable() - enable virtual functions + * @pdev: the PCIe device + * @max_vfs: the number of virtual functions to enable + * + * Returns the number of enabled VFs. If there are VFs enabled already or + * max_vfs is more than the total number of device can be enabled, returns + * failure. + */ +int hisi_qm_sriov_enable(struct pci_dev *pdev, int max_vfs) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + int pre_existing_vfs, num_vfs, total_vfs, ret; + + ret = qm_pm_get_sync(qm); + if (ret) + return ret; + + total_vfs = pci_sriov_get_totalvfs(pdev); + pre_existing_vfs = pci_num_vf(pdev); + if (pre_existing_vfs) { + pci_err(pdev, "%d VFs already enabled. Please disable pre-enabled VFs!\n", + pre_existing_vfs); + goto err_put_sync; + } + + if (max_vfs > total_vfs) { + pci_err(pdev, "%d VFs is more than total VFs %d!\n", max_vfs, total_vfs); + ret = -ERANGE; + goto err_put_sync; + } + + num_vfs = max_vfs; + + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) + hisi_qm_init_vf_qos(qm, num_vfs); + + ret = qm_vf_q_assign(qm, num_vfs); + if (ret) { + pci_err(pdev, "Can't assign queues for VF!\n"); + goto err_put_sync; + } + + qm->vfs_num = num_vfs; + + ret = pci_enable_sriov(pdev, num_vfs); + if (ret) { + pci_err(pdev, "Can't enable VF!\n"); + qm_clear_vft_config(qm); + goto err_put_sync; + } + + pci_info(pdev, "VF enabled, vfs_num(=%d)!\n", num_vfs); + + return num_vfs; + +err_put_sync: + qm_pm_put_sync(qm); + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_sriov_enable); + +/** + * hisi_qm_sriov_disable - disable virtual functions + * @pdev: the PCI device. + * @is_frozen: true when all the VFs are frozen. + * + * Return failure if there are VFs assigned already or VF is in used. + */ +int hisi_qm_sriov_disable(struct pci_dev *pdev, bool is_frozen) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + if (pci_vfs_assigned(pdev)) { + pci_err(pdev, "Failed to disable VFs as VFs are assigned!\n"); + return -EPERM; + } + + /* While VF is in used, SRIOV cannot be disabled. */ + if (!is_frozen && qm_try_frozen_vfs(pdev, qm->qm_list)) { + pci_err(pdev, "Task is using its VF!\n"); + return -EBUSY; + } + + pci_disable_sriov(pdev); + + ret = qm_clear_vft_config(qm); + if (ret) + return ret; + + qm_pm_put_sync(qm); + + return 0; +} +EXPORT_SYMBOL_GPL(hisi_qm_sriov_disable); + +/** + * hisi_qm_sriov_configure - configure the number of VFs + * @pdev: The PCI device + * @num_vfs: The number of VFs need enabled + * + * Enable SR-IOV according to num_vfs, 0 means disable. + */ +int hisi_qm_sriov_configure(struct pci_dev *pdev, int num_vfs) +{ + if (num_vfs == 0) + return hisi_qm_sriov_disable(pdev, false); + else + return hisi_qm_sriov_enable(pdev, num_vfs); +} +EXPORT_SYMBOL_GPL(hisi_qm_sriov_configure); + +static enum acc_err_result qm_dev_err_handle(struct hisi_qm *qm) +{ + u32 err_sts; + + if (!qm->err_ini->get_dev_hw_err_status) { + dev_err(&qm->pdev->dev, "Device doesn't support get hw error status!\n"); + return ACC_ERR_NONE; + } + + /* get device hardware error status */ + err_sts = qm->err_ini->get_dev_hw_err_status(qm); + if (err_sts) { + if (err_sts & qm->err_info.ecc_2bits_mask) + qm->err_status.is_dev_ecc_mbit = true; + + if (qm->err_ini->log_dev_hw_err) + qm->err_ini->log_dev_hw_err(qm, err_sts); + + if (err_sts & qm->err_info.dev_reset_mask) + return ACC_ERR_NEED_RESET; + + if (qm->err_ini->clear_dev_hw_err_status) + qm->err_ini->clear_dev_hw_err_status(qm, err_sts); + } + + return ACC_ERR_RECOVERED; +} + +static enum acc_err_result qm_process_dev_error(struct hisi_qm *qm) +{ + enum acc_err_result qm_ret, dev_ret; + + /* log qm error */ + qm_ret = qm_hw_error_handle(qm); + + /* log device error */ + dev_ret = qm_dev_err_handle(qm); + + return (qm_ret == ACC_ERR_NEED_RESET || + dev_ret == ACC_ERR_NEED_RESET) ? + ACC_ERR_NEED_RESET : ACC_ERR_RECOVERED; +} + +/** + * hisi_qm_dev_err_detected() - Get device and qm error status then log it. + * @pdev: The PCI device which need report error. + * @state: The connectivity between CPU and device. + * + * We register this function into PCIe AER handlers, It will report device or + * qm hardware error status when error occur. + */ +pci_ers_result_t hisi_qm_dev_err_detected(struct pci_dev *pdev, + pci_channel_state_t state) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + enum acc_err_result ret; + + if (pdev->is_virtfn) + return PCI_ERS_RESULT_NONE; + + pci_info(pdev, "PCI error detected, state(=%u)!!\n", state); + if (state == pci_channel_io_perm_failure) + return PCI_ERS_RESULT_DISCONNECT; + + ret = qm_process_dev_error(qm); + if (ret == ACC_ERR_NEED_RESET) + return PCI_ERS_RESULT_NEED_RESET; + + return PCI_ERS_RESULT_RECOVERED; +} +EXPORT_SYMBOL_GPL(hisi_qm_dev_err_detected); + +static int qm_check_req_recv(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + u32 val; + + if (qm->ver >= QM_HW_V3) + return 0; + + writel(ACC_VENDOR_ID_VALUE, qm->io_base + QM_PEH_VENDOR_ID); + ret = readl_relaxed_poll_timeout(qm->io_base + QM_PEH_VENDOR_ID, val, + (val == ACC_VENDOR_ID_VALUE), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) { + dev_err(&pdev->dev, "Fails to read QM reg!\n"); + return ret; + } + + writel(PCI_VENDOR_ID_HUAWEI, qm->io_base + QM_PEH_VENDOR_ID); + ret = readl_relaxed_poll_timeout(qm->io_base + QM_PEH_VENDOR_ID, val, + (val == PCI_VENDOR_ID_HUAWEI), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) + dev_err(&pdev->dev, "Fails to read QM reg in the second time!\n"); + + return ret; +} + +static int qm_set_pf_mse(struct hisi_qm *qm, bool set) +{ + struct pci_dev *pdev = qm->pdev; + u16 cmd; + int i; + + pci_read_config_word(pdev, PCI_COMMAND, &cmd); + if (set) + cmd |= PCI_COMMAND_MEMORY; + else + cmd &= ~PCI_COMMAND_MEMORY; + + pci_write_config_word(pdev, PCI_COMMAND, cmd); + for (i = 0; i < MAX_WAIT_COUNTS; i++) { + pci_read_config_word(pdev, PCI_COMMAND, &cmd); + if (set == ((cmd & PCI_COMMAND_MEMORY) >> 1)) + return 0; + + udelay(1); + } + + return -ETIMEDOUT; +} + +static int qm_set_vf_mse(struct hisi_qm *qm, bool set) +{ + struct pci_dev *pdev = qm->pdev; + u16 sriov_ctrl; + int pos; + int i; + + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV); + pci_read_config_word(pdev, pos + PCI_SRIOV_CTRL, &sriov_ctrl); + if (set) + sriov_ctrl |= PCI_SRIOV_CTRL_MSE; + else + sriov_ctrl &= ~PCI_SRIOV_CTRL_MSE; + pci_write_config_word(pdev, pos + PCI_SRIOV_CTRL, sriov_ctrl); + + for (i = 0; i < MAX_WAIT_COUNTS; i++) { + pci_read_config_word(pdev, pos + PCI_SRIOV_CTRL, &sriov_ctrl); + if (set == (sriov_ctrl & PCI_SRIOV_CTRL_MSE) >> + ACC_PEH_SRIOV_CTRL_VF_MSE_SHIFT) + return 0; + + udelay(1); + } + + return -ETIMEDOUT; +} + +static int qm_vf_reset_prepare(struct hisi_qm *qm, + enum qm_stop_reason stop_reason) +{ + struct hisi_qm_list *qm_list = qm->qm_list; + struct pci_dev *pdev = qm->pdev; + struct pci_dev *virtfn; + struct hisi_qm *vf_qm; + int ret = 0; + + mutex_lock(&qm_list->lock); + list_for_each_entry(vf_qm, &qm_list->list, list) { + virtfn = vf_qm->pdev; + if (virtfn == pdev) + continue; + + if (pci_physfn(virtfn) == pdev) { + /* save VFs PCIE BAR configuration */ + pci_save_state(virtfn); + + ret = hisi_qm_stop(vf_qm, stop_reason); + if (ret) + goto stop_fail; + } + } + +stop_fail: + mutex_unlock(&qm_list->lock); + return ret; +} + +static int qm_try_stop_vfs(struct hisi_qm *qm, u64 cmd, + enum qm_stop_reason stop_reason) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + if (!qm->vfs_num) + return 0; + + /* Kunpeng930 supports to notify VFs to stop before PF reset */ + if (test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) { + ret = qm_ping_all_vfs(qm, cmd); + if (ret) + pci_err(pdev, "failed to send cmd to all VFs before PF reset!\n"); + } else { + ret = qm_vf_reset_prepare(qm, stop_reason); + if (ret) + pci_err(pdev, "failed to prepare reset, ret = %d.\n", ret); + } + + return ret; +} + +static int qm_controller_reset_prepare(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + ret = qm_reset_prepare_ready(qm); + if (ret) { + pci_err(pdev, "Controller reset not ready!\n"); + return ret; + } + + /* PF obtains the information of VF by querying the register. */ + qm_cmd_uninit(qm); + + /* Whether VFs stop successfully, soft reset will continue. */ + ret = qm_try_stop_vfs(qm, QM_PF_SRST_PREPARE, QM_SOFT_RESET); + if (ret) + pci_err(pdev, "failed to stop vfs by pf in soft reset.\n"); + + ret = hisi_qm_stop(qm, QM_SOFT_RESET); + if (ret) { + pci_err(pdev, "Fails to stop QM!\n"); + qm_reset_bit_clear(qm); + return ret; + } + + ret = qm_wait_vf_prepare_finish(qm); + if (ret) + pci_err(pdev, "failed to stop by vfs in soft reset!\n"); + + clear_bit(QM_RST_SCHED, &qm->misc_ctl); + + return 0; +} + +static void qm_dev_ecc_mbit_handle(struct hisi_qm *qm) +{ + u32 nfe_enb = 0; + + /* Kunpeng930 hardware automatically close master ooo when NFE occurs */ + if (qm->ver >= QM_HW_V3) + return; + + if (!qm->err_status.is_dev_ecc_mbit && + qm->err_status.is_qm_ecc_mbit && + qm->err_ini->close_axi_master_ooo) { + + qm->err_ini->close_axi_master_ooo(qm); + + } else if (qm->err_status.is_dev_ecc_mbit && + !qm->err_status.is_qm_ecc_mbit && + !qm->err_ini->close_axi_master_ooo) { + + nfe_enb = readl(qm->io_base + QM_RAS_NFE_ENABLE); + writel(nfe_enb & QM_RAS_NFE_MBIT_DISABLE, + qm->io_base + QM_RAS_NFE_ENABLE); + writel(QM_ECC_MBIT, qm->io_base + QM_ABNORMAL_INT_SET); + } +} + +static int qm_soft_reset(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + u32 val; + + /* Ensure all doorbells and mailboxes received by QM */ + ret = qm_check_req_recv(qm); + if (ret) + return ret; + + if (qm->vfs_num) { + ret = qm_set_vf_mse(qm, false); + if (ret) { + pci_err(pdev, "Fails to disable vf MSE bit.\n"); + return ret; + } + } + + ret = qm->ops->set_msi(qm, false); + if (ret) { + pci_err(pdev, "Fails to disable PEH MSI bit.\n"); + return ret; + } + + qm_dev_ecc_mbit_handle(qm); + + /* OOO register set and check */ + writel(ACC_MASTER_GLOBAL_CTRL_SHUTDOWN, + qm->io_base + ACC_MASTER_GLOBAL_CTRL); + + /* If bus lock, reset chip */ + ret = readl_relaxed_poll_timeout(qm->io_base + ACC_MASTER_TRANS_RETURN, + val, + (val == ACC_MASTER_TRANS_RETURN_RW), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) { + pci_emerg(pdev, "Bus lock! Please reset system.\n"); + return ret; + } + + if (qm->err_ini->close_sva_prefetch) + qm->err_ini->close_sva_prefetch(qm); + + ret = qm_set_pf_mse(qm, false); + if (ret) { + pci_err(pdev, "Fails to disable pf MSE bit.\n"); + return ret; + } + + /* The reset related sub-control registers are not in PCI BAR */ + if (ACPI_HANDLE(&pdev->dev)) { + unsigned long long value = 0; + acpi_status s; + + s = acpi_evaluate_integer(ACPI_HANDLE(&pdev->dev), + qm->err_info.acpi_rst, + NULL, &value); + if (ACPI_FAILURE(s)) { + pci_err(pdev, "NO controller reset method!\n"); + return -EIO; + } + + if (value) { + pci_err(pdev, "Reset step %llu failed!\n", value); + return -EIO; + } + } else { + pci_err(pdev, "No reset method!\n"); + return -EINVAL; + } + + return 0; +} + +static int qm_vf_reset_done(struct hisi_qm *qm) +{ + struct hisi_qm_list *qm_list = qm->qm_list; + struct pci_dev *pdev = qm->pdev; + struct pci_dev *virtfn; + struct hisi_qm *vf_qm; + int ret = 0; + + mutex_lock(&qm_list->lock); + list_for_each_entry(vf_qm, &qm_list->list, list) { + virtfn = vf_qm->pdev; + if (virtfn == pdev) + continue; + + if (pci_physfn(virtfn) == pdev) { + /* enable VFs PCIE BAR configuration */ + pci_restore_state(virtfn); + + ret = qm_restart(vf_qm); + if (ret) + goto restart_fail; + } + } + +restart_fail: + mutex_unlock(&qm_list->lock); + return ret; +} + +static int qm_try_start_vfs(struct hisi_qm *qm, enum qm_mb_cmd cmd) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + if (!qm->vfs_num) + return 0; + + ret = qm_vf_q_assign(qm, qm->vfs_num); + if (ret) { + pci_err(pdev, "failed to assign VFs, ret = %d.\n", ret); + return ret; + } + + /* Kunpeng930 supports to notify VFs to start after PF reset. */ + if (test_bit(QM_SUPPORT_MB_COMMAND, &qm->caps)) { + ret = qm_ping_all_vfs(qm, cmd); + if (ret) + pci_warn(pdev, "failed to send cmd to all VFs after PF reset!\n"); + } else { + ret = qm_vf_reset_done(qm); + if (ret) + pci_warn(pdev, "failed to start vfs, ret = %d.\n", ret); + } + + return ret; +} + +static int qm_dev_hw_init(struct hisi_qm *qm) +{ + return qm->err_ini->hw_init(qm); +} + +static void qm_restart_prepare(struct hisi_qm *qm) +{ + u32 value; + + if (qm->err_ini->open_sva_prefetch) + qm->err_ini->open_sva_prefetch(qm); + + if (qm->ver >= QM_HW_V3) + return; + + if (!qm->err_status.is_qm_ecc_mbit && + !qm->err_status.is_dev_ecc_mbit) + return; + + /* temporarily close the OOO port used for PEH to write out MSI */ + value = readl(qm->io_base + ACC_AM_CFG_PORT_WR_EN); + writel(value & ~qm->err_info.msi_wr_port, + qm->io_base + ACC_AM_CFG_PORT_WR_EN); + + /* clear dev ecc 2bit error source if having */ + value = qm_get_dev_err_status(qm) & qm->err_info.ecc_2bits_mask; + if (value && qm->err_ini->clear_dev_hw_err_status) + qm->err_ini->clear_dev_hw_err_status(qm, value); + + /* clear QM ecc mbit error source */ + writel(QM_ECC_MBIT, qm->io_base + QM_ABNORMAL_INT_SOURCE); + + /* clear AM Reorder Buffer ecc mbit source */ + writel(ACC_ROB_ECC_ERR_MULTPL, qm->io_base + ACC_AM_ROB_ECC_INT_STS); +} + +static void qm_restart_done(struct hisi_qm *qm) +{ + u32 value; + + if (qm->ver >= QM_HW_V3) + goto clear_flags; + + if (!qm->err_status.is_qm_ecc_mbit && + !qm->err_status.is_dev_ecc_mbit) + return; + + /* open the OOO port for PEH to write out MSI */ + value = readl(qm->io_base + ACC_AM_CFG_PORT_WR_EN); + value |= qm->err_info.msi_wr_port; + writel(value, qm->io_base + ACC_AM_CFG_PORT_WR_EN); + +clear_flags: + qm->err_status.is_qm_ecc_mbit = false; + qm->err_status.is_dev_ecc_mbit = false; +} + +static int qm_controller_reset_done(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + ret = qm->ops->set_msi(qm, true); + if (ret) { + pci_err(pdev, "Fails to enable PEH MSI bit!\n"); + return ret; + } + + ret = qm_set_pf_mse(qm, true); + if (ret) { + pci_err(pdev, "Fails to enable pf MSE bit!\n"); + return ret; + } + + if (qm->vfs_num) { + ret = qm_set_vf_mse(qm, true); + if (ret) { + pci_err(pdev, "Fails to enable vf MSE bit!\n"); + return ret; + } + } + + ret = qm_dev_hw_init(qm); + if (ret) { + pci_err(pdev, "Failed to init device\n"); + return ret; + } + + qm_restart_prepare(qm); + hisi_qm_dev_err_init(qm); + if (qm->err_ini->open_axi_master_ooo) + qm->err_ini->open_axi_master_ooo(qm); + + ret = qm_dev_mem_reset(qm); + if (ret) { + pci_err(pdev, "failed to reset device memory\n"); + return ret; + } + + ret = qm_restart(qm); + if (ret) { + pci_err(pdev, "Failed to start QM!\n"); + return ret; + } + + ret = qm_try_start_vfs(qm, QM_PF_RESET_DONE); + if (ret) + pci_err(pdev, "failed to start vfs by pf in soft reset.\n"); + + ret = qm_wait_vf_prepare_finish(qm); + if (ret) + pci_err(pdev, "failed to start by vfs in soft reset!\n"); + + qm_cmd_init(qm); + qm_restart_done(qm); + + qm_reset_bit_clear(qm); + + return 0; +} + +static void qm_show_last_dfx_regs(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + struct pci_dev *pdev = qm->pdev; + u32 val; + int i; + + if (qm->fun_type == QM_HW_VF || !debug->qm_last_words) + return; + + for (i = 0; i < ARRAY_SIZE(qm_dfx_regs); i++) { + val = readl_relaxed(qm->io_base + qm_dfx_regs[i].offset); + if (debug->qm_last_words[i] != val) + pci_info(pdev, "%s \t= 0x%08x => 0x%08x\n", + qm_dfx_regs[i].name, debug->qm_last_words[i], val); + } +} + +static int qm_controller_reset(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + pci_info(pdev, "Controller resetting...\n"); + + ret = qm_controller_reset_prepare(qm); + if (ret) { + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_TX_OFFSET); + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_RX_OFFSET); + clear_bit(QM_RST_SCHED, &qm->misc_ctl); + return ret; + } + + qm_show_last_dfx_regs(qm); + if (qm->err_ini->show_last_dfx_regs) + qm->err_ini->show_last_dfx_regs(qm); + + ret = qm_soft_reset(qm); + if (ret) { + pci_err(pdev, "Controller reset failed (%d)\n", ret); + qm_reset_bit_clear(qm); + return ret; + } + + ret = qm_controller_reset_done(qm); + if (ret) { + qm_reset_bit_clear(qm); + return ret; + } + + pci_info(pdev, "Controller reset complete\n"); + + return 0; +} + +/** + * hisi_qm_dev_slot_reset() - slot reset + * @pdev: the PCIe device + * + * This function offers QM relate PCIe device reset interface. Drivers which + * use QM can use this function as slot_reset in its struct pci_error_handlers. + */ +pci_ers_result_t hisi_qm_dev_slot_reset(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + if (pdev->is_virtfn) + return PCI_ERS_RESULT_RECOVERED; + + /* reset pcie device controller */ + ret = qm_controller_reset(qm); + if (ret) { + pci_err(pdev, "Controller reset failed (%d)\n", ret); + return PCI_ERS_RESULT_DISCONNECT; + } + + return PCI_ERS_RESULT_RECOVERED; +} +EXPORT_SYMBOL_GPL(hisi_qm_dev_slot_reset); + +void hisi_qm_reset_prepare(struct pci_dev *pdev) +{ + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(pdev)); + struct hisi_qm *qm = pci_get_drvdata(pdev); + u32 delay = 0; + int ret; + + hisi_qm_dev_err_uninit(pf_qm); + + /* + * Check whether there is an ECC mbit error, If it occurs, need to + * wait for soft reset to fix it. + */ + while (qm_check_dev_error(pf_qm)) { + msleep(++delay); + if (delay > QM_RESET_WAIT_TIMEOUT) + return; + } + + ret = qm_reset_prepare_ready(qm); + if (ret) { + pci_err(pdev, "FLR not ready!\n"); + return; + } + + /* PF obtains the information of VF by querying the register. */ + if (qm->fun_type == QM_HW_PF) + qm_cmd_uninit(qm); + + ret = qm_try_stop_vfs(qm, QM_PF_FLR_PREPARE, QM_FLR); + if (ret) + pci_err(pdev, "failed to stop vfs by pf in FLR.\n"); + + ret = hisi_qm_stop(qm, QM_FLR); + if (ret) { + pci_err(pdev, "Failed to stop QM, ret = %d.\n", ret); + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_TX_OFFSET); + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_RX_OFFSET); + return; + } + + ret = qm_wait_vf_prepare_finish(qm); + if (ret) + pci_err(pdev, "failed to stop by vfs in FLR!\n"); + + pci_info(pdev, "FLR resetting...\n"); +} +EXPORT_SYMBOL_GPL(hisi_qm_reset_prepare); + +static bool qm_flr_reset_complete(struct pci_dev *pdev) +{ + struct pci_dev *pf_pdev = pci_physfn(pdev); + struct hisi_qm *qm = pci_get_drvdata(pf_pdev); + u32 id; + + pci_read_config_dword(qm->pdev, PCI_COMMAND, &id); + if (id == QM_PCI_COMMAND_INVALID) { + pci_err(pdev, "Device can not be used!\n"); + return false; + } + + return true; +} + +void hisi_qm_reset_done(struct pci_dev *pdev) +{ + struct hisi_qm *pf_qm = pci_get_drvdata(pci_physfn(pdev)); + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + if (qm->fun_type == QM_HW_PF) { + ret = qm_dev_hw_init(qm); + if (ret) { + pci_err(pdev, "Failed to init PF, ret = %d.\n", ret); + goto flr_done; + } + } + + hisi_qm_dev_err_init(pf_qm); + + ret = qm_restart(qm); + if (ret) { + pci_err(pdev, "Failed to start QM, ret = %d.\n", ret); + goto flr_done; + } + + ret = qm_try_start_vfs(qm, QM_PF_RESET_DONE); + if (ret) + pci_err(pdev, "failed to start vfs by pf in FLR.\n"); + + ret = qm_wait_vf_prepare_finish(qm); + if (ret) + pci_err(pdev, "failed to start by vfs in FLR!\n"); + +flr_done: + if (qm->fun_type == QM_HW_PF) + qm_cmd_init(qm); + + if (qm_flr_reset_complete(pdev)) + pci_info(pdev, "FLR reset complete\n"); + + qm_reset_bit_clear(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_reset_done); + +static irqreturn_t qm_abnormal_irq(int irq, void *data) +{ + struct hisi_qm *qm = data; + enum acc_err_result ret; + + atomic64_inc(&qm->debug.dfx.abnormal_irq_cnt); + ret = qm_process_dev_error(qm); + if (ret == ACC_ERR_NEED_RESET && + !test_bit(QM_DRIVER_REMOVING, &qm->misc_ctl) && + !test_and_set_bit(QM_RST_SCHED, &qm->misc_ctl)) + schedule_work(&qm->rst_work); + + return IRQ_HANDLED; +} + + +/** + * hisi_qm_dev_shutdown() - Shutdown device. + * @pdev: The device will be shutdown. + * + * This function will stop qm when OS shutdown or rebooting. + */ +void hisi_qm_dev_shutdown(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + ret = hisi_qm_stop(qm, QM_NORMAL); + if (ret) + dev_err(&pdev->dev, "Fail to stop qm in shutdown!\n"); +} +EXPORT_SYMBOL_GPL(hisi_qm_dev_shutdown); + +static void hisi_qm_controller_reset(struct work_struct *rst_work) +{ + struct hisi_qm *qm = container_of(rst_work, struct hisi_qm, rst_work); + int ret; + + ret = qm_pm_get_sync(qm); + if (ret) { + clear_bit(QM_RST_SCHED, &qm->misc_ctl); + return; + } + + /* reset pcie device controller */ + ret = qm_controller_reset(qm); + if (ret) + dev_err(&qm->pdev->dev, "controller reset failed (%d)\n", ret); + + qm_pm_put_sync(qm); +} + +static void qm_pf_reset_vf_prepare(struct hisi_qm *qm, + enum qm_stop_reason stop_reason) +{ + enum qm_mb_cmd cmd = QM_VF_PREPARE_DONE; + struct pci_dev *pdev = qm->pdev; + int ret; + + ret = qm_reset_prepare_ready(qm); + if (ret) { + dev_err(&pdev->dev, "reset prepare not ready!\n"); + atomic_set(&qm->status.flags, QM_STOP); + cmd = QM_VF_PREPARE_FAIL; + goto err_prepare; + } + + ret = hisi_qm_stop(qm, stop_reason); + if (ret) { + dev_err(&pdev->dev, "failed to stop QM, ret = %d.\n", ret); + atomic_set(&qm->status.flags, QM_STOP); + cmd = QM_VF_PREPARE_FAIL; + goto err_prepare; + } else { + goto out; + } + +err_prepare: + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_TX_OFFSET); + hisi_qm_set_hw_reset(qm, QM_RESET_STOP_RX_OFFSET); +out: + pci_save_state(pdev); + ret = qm_ping_pf(qm, cmd); + if (ret) + dev_warn(&pdev->dev, "PF responds timeout in reset prepare!\n"); +} + +static void qm_pf_reset_vf_done(struct hisi_qm *qm) +{ + enum qm_mb_cmd cmd = QM_VF_START_DONE; + struct pci_dev *pdev = qm->pdev; + int ret; + + pci_restore_state(pdev); + ret = hisi_qm_start(qm); + if (ret) { + dev_err(&pdev->dev, "failed to start QM, ret = %d.\n", ret); + cmd = QM_VF_START_FAIL; + } + + qm_cmd_init(qm); + ret = qm_ping_pf(qm, cmd); + if (ret) + dev_warn(&pdev->dev, "PF responds timeout in reset done!\n"); + + qm_reset_bit_clear(qm); +} + +static int qm_wait_pf_reset_finish(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u32 val, cmd; + u64 msg; + int ret; + + /* Wait for reset to finish */ + ret = readl_relaxed_poll_timeout(qm->io_base + QM_IFC_INT_SOURCE_V, val, + val == BIT(0), QM_VF_RESET_WAIT_US, + QM_VF_RESET_WAIT_TIMEOUT_US); + /* hardware completion status should be available by this time */ + if (ret) { + dev_err(dev, "couldn't get reset done status from PF, timeout!\n"); + return -ETIMEDOUT; + } + + /* + * Whether message is got successfully, + * VF needs to ack PF by clearing the interrupt. + */ + ret = qm_get_mb_cmd(qm, &msg, 0); + qm_clear_cmd_interrupt(qm, 0); + if (ret) { + dev_err(dev, "failed to get msg from PF in reset done!\n"); + return ret; + } + + cmd = msg & QM_MB_CMD_DATA_MASK; + if (cmd != QM_PF_RESET_DONE) { + dev_err(dev, "the cmd(%u) is not reset done!\n", cmd); + ret = -EINVAL; + } + + return ret; +} + +static void qm_pf_reset_vf_process(struct hisi_qm *qm, + enum qm_stop_reason stop_reason) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + dev_info(dev, "device reset start...\n"); + + /* The message is obtained by querying the register during resetting */ + qm_cmd_uninit(qm); + qm_pf_reset_vf_prepare(qm, stop_reason); + + ret = qm_wait_pf_reset_finish(qm); + if (ret) + goto err_get_status; + + qm_pf_reset_vf_done(qm); + + dev_info(dev, "device reset done.\n"); + + return; + +err_get_status: + qm_cmd_init(qm); + qm_reset_bit_clear(qm); +} + +static void qm_handle_cmd_msg(struct hisi_qm *qm, u32 fun_num) +{ + struct device *dev = &qm->pdev->dev; + u64 msg; + u32 cmd; + int ret; + + /* + * Get the msg from source by sending mailbox. Whether message is got + * successfully, destination needs to ack source by clearing the interrupt. + */ + ret = qm_get_mb_cmd(qm, &msg, fun_num); + qm_clear_cmd_interrupt(qm, BIT(fun_num)); + if (ret) { + dev_err(dev, "failed to get msg from source!\n"); + return; + } + + cmd = msg & QM_MB_CMD_DATA_MASK; + switch (cmd) { + case QM_PF_FLR_PREPARE: + qm_pf_reset_vf_process(qm, QM_FLR); + break; + case QM_PF_SRST_PREPARE: + qm_pf_reset_vf_process(qm, QM_SOFT_RESET); + break; + case QM_VF_GET_QOS: + qm_vf_get_qos(qm, fun_num); + break; + case QM_PF_SET_QOS: + qm->mb_qos = msg >> QM_MB_CMD_DATA_SHIFT; + break; + default: + dev_err(dev, "unsupported cmd %u sent by function(%u)!\n", cmd, fun_num); + break; + } +} + +static void qm_cmd_process(struct work_struct *cmd_process) +{ + struct hisi_qm *qm = container_of(cmd_process, + struct hisi_qm, cmd_process); + u32 vfs_num = qm->vfs_num; + u64 val; + u32 i; + + if (qm->fun_type == QM_HW_PF) { + val = readq(qm->io_base + QM_IFC_INT_SOURCE_P); + if (!val) + return; + + for (i = 1; i <= vfs_num; i++) { + if (val & BIT(i)) + qm_handle_cmd_msg(qm, i); + } + + return; + } + + qm_handle_cmd_msg(qm, 0); +} + +/** + * hisi_qm_alg_register() - Register alg to crypto and add qm to qm_list. + * @qm: The qm needs add. + * @qm_list: The qm list. + * + * This function adds qm to qm list, and will register algorithm to + * crypto when the qm list is empty. + */ +int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +{ + struct device *dev = &qm->pdev->dev; + int flag = 0; + int ret = 0; + + mutex_lock(&qm_list->lock); + if (list_empty(&qm_list->list)) + flag = 1; + list_add_tail(&qm->list, &qm_list->list); + mutex_unlock(&qm_list->lock); + + if (qm->ver <= QM_HW_V2 && qm->use_sva) { + dev_info(dev, "HW V2 not both use uacce sva mode and hardware crypto algs.\n"); + return 0; + } + + if (flag) { + ret = qm_list->register_to_crypto(qm); + if (ret) { + mutex_lock(&qm_list->lock); + list_del(&qm->list); + mutex_unlock(&qm_list->lock); + } + } + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_alg_register); + +/** + * hisi_qm_alg_unregister() - Unregister alg from crypto and delete qm from + * qm list. + * @qm: The qm needs delete. + * @qm_list: The qm list. + * + * This function deletes qm from qm list, and will unregister algorithm + * from crypto when the qm list is empty. + */ +void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +{ + mutex_lock(&qm_list->lock); + list_del(&qm->list); + mutex_unlock(&qm_list->lock); + + if (qm->ver <= QM_HW_V2 && qm->use_sva) + return; + + if (list_empty(&qm_list->list)) + qm_list->unregister_from_crypto(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_alg_unregister); + +static void qm_unregister_abnormal_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + + if (qm->fun_type == QM_HW_VF) + return; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_ABN_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_ABN_IRQ_TYPE_MASK)) + return; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + free_irq(pci_irq_vector(pdev, irq_vector), qm); +} + +static int qm_register_abnormal_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + int ret; + + if (qm->fun_type == QM_HW_VF) + return 0; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_ABN_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_ABN_IRQ_TYPE_MASK)) + return 0; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + ret = request_irq(pci_irq_vector(pdev, irq_vector), qm_abnormal_irq, 0, qm->dev_name, qm); + if (ret) + dev_err(&qm->pdev->dev, "failed to request abnormal irq, ret = %d", ret); + + return ret; +} + +static void qm_unregister_mb_cmd_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_PF2VF_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + free_irq(pci_irq_vector(pdev, irq_vector), qm); +} + +static int qm_register_mb_cmd_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + int ret; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_PF2VF_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return 0; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + ret = request_irq(pci_irq_vector(pdev, irq_vector), qm_mb_cmd_irq, 0, qm->dev_name, qm); + if (ret) + dev_err(&pdev->dev, "failed to request function communication irq, ret = %d", ret); + + return ret; +} + +static void qm_unregister_aeq_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_AEQ_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + free_irq(pci_irq_vector(pdev, irq_vector), qm); +} + +static int qm_register_aeq_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + int ret; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_AEQ_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return 0; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + ret = request_threaded_irq(pci_irq_vector(pdev, irq_vector), qm_aeq_irq, + qm_aeq_thread, 0, qm->dev_name, qm); + if (ret) + dev_err(&pdev->dev, "failed to request eq irq, ret = %d", ret); + + return ret; +} + +static void qm_unregister_eq_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_EQ_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + free_irq(pci_irq_vector(pdev, irq_vector), qm); +} + +static int qm_register_eq_irq(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + u32 irq_vector, val; + int ret; + + val = hisi_qm_get_hw_info(qm, qm_basic_info, QM_EQ_IRQ_TYPE_CAP, qm->cap_ver); + if (!((val >> QM_IRQ_TYPE_SHIFT) & QM_IRQ_TYPE_MASK)) + return 0; + + irq_vector = val & QM_IRQ_VECTOR_MASK; + ret = request_irq(pci_irq_vector(pdev, irq_vector), qm_irq, 0, qm->dev_name, qm); + if (ret) + dev_err(&pdev->dev, "failed to request eq irq, ret = %d", ret); + + return ret; +} + +static void qm_irqs_unregister(struct hisi_qm *qm) +{ + qm_unregister_mb_cmd_irq(qm); + qm_unregister_abnormal_irq(qm); + qm_unregister_aeq_irq(qm); + qm_unregister_eq_irq(qm); +} + +static int qm_irqs_register(struct hisi_qm *qm) +{ + int ret; + + ret = qm_register_eq_irq(qm); + if (ret) + return ret; + + ret = qm_register_aeq_irq(qm); + if (ret) + goto free_eq_irq; + + ret = qm_register_abnormal_irq(qm); + if (ret) + goto free_aeq_irq; + + ret = qm_register_mb_cmd_irq(qm); + if (ret) + goto free_abnormal_irq; + + return 0; + +free_abnormal_irq: + qm_unregister_abnormal_irq(qm); +free_aeq_irq: + qm_unregister_aeq_irq(qm); +free_eq_irq: + qm_unregister_eq_irq(qm); + return ret; +} + +static int qm_get_qp_num(struct hisi_qm *qm) +{ + bool is_db_isolation; + + /* VF's qp_num assigned by PF in v2, and VF can get qp_num by vft. */ + if (qm->fun_type == QM_HW_VF) { + if (qm->ver != QM_HW_V1) + /* v2 starts to support get vft by mailbox */ + return hisi_qm_get_vft(qm, &qm->qp_base, &qm->qp_num); + + return 0; + } + + is_db_isolation = test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps); + qm->ctrl_qp_num = hisi_qm_get_hw_info(qm, qm_basic_info, QM_TOTAL_QP_NUM_CAP, true); + qm->max_qp_num = hisi_qm_get_hw_info(qm, qm_basic_info, + QM_FUNC_MAX_QP_CAP, is_db_isolation); + + /* check if qp number is valid */ + if (qm->qp_num > qm->max_qp_num) { + dev_err(&qm->pdev->dev, "qp num(%u) is more than max qp num(%u)!\n", + qm->qp_num, qm->max_qp_num); + return -EINVAL; + } + + return 0; +} + +static void qm_get_hw_caps(struct hisi_qm *qm) +{ + const struct hisi_qm_cap_info *cap_info = qm->fun_type == QM_HW_PF ? + qm_cap_info_pf : qm_cap_info_vf; + u32 size = qm->fun_type == QM_HW_PF ? ARRAY_SIZE(qm_cap_info_pf) : + ARRAY_SIZE(qm_cap_info_vf); + u32 val, i; + + /* Doorbell isolate register is a independent register. */ + val = hisi_qm_get_hw_info(qm, qm_cap_info_comm, QM_SUPPORT_DB_ISOLATION, true); + if (val) + set_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps); + + if (qm->ver >= QM_HW_V3) { + val = readl(qm->io_base + QM_FUNC_CAPS_REG); + qm->cap_ver = val & QM_CAPBILITY_VERSION; + } + + /* Get PF/VF common capbility */ + for (i = 1; i < ARRAY_SIZE(qm_cap_info_comm); i++) { + val = hisi_qm_get_hw_info(qm, qm_cap_info_comm, i, qm->cap_ver); + if (val) + set_bit(qm_cap_info_comm[i].type, &qm->caps); + } + + /* Get PF/VF different capbility */ + for (i = 0; i < size; i++) { + val = hisi_qm_get_hw_info(qm, cap_info, i, qm->cap_ver); + if (val) + set_bit(cap_info[i].type, &qm->caps); + } +} + +static int qm_get_pci_res(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct device *dev = &pdev->dev; + int ret; + + ret = pci_request_mem_regions(pdev, qm->dev_name); + if (ret < 0) { + dev_err(dev, "Failed to request mem regions!\n"); + return ret; + } + + qm->phys_base = pci_resource_start(pdev, PCI_BAR_2); + qm->io_base = ioremap(qm->phys_base, pci_resource_len(pdev, PCI_BAR_2)); + if (!qm->io_base) { + ret = -EIO; + goto err_request_mem_regions; + } + + qm_get_hw_caps(qm); + if (test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps)) { + qm->db_interval = QM_QP_DB_INTERVAL; + qm->db_phys_base = pci_resource_start(pdev, PCI_BAR_4); + qm->db_io_base = ioremap(qm->db_phys_base, + pci_resource_len(pdev, PCI_BAR_4)); + if (!qm->db_io_base) { + ret = -EIO; + goto err_ioremap; + } + } else { + qm->db_phys_base = qm->phys_base; + qm->db_io_base = qm->io_base; + qm->db_interval = 0; + } + + ret = qm_get_qp_num(qm); + if (ret) + goto err_db_ioremap; + + return 0; + +err_db_ioremap: + if (test_bit(QM_SUPPORT_DB_ISOLATION, &qm->caps)) + iounmap(qm->db_io_base); +err_ioremap: + iounmap(qm->io_base); +err_request_mem_regions: + pci_release_mem_regions(pdev); + return ret; +} + +static int hisi_qm_pci_init(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct device *dev = &pdev->dev; + unsigned int num_vec; + int ret; + + ret = pci_enable_device_mem(pdev); + if (ret < 0) { + dev_err(dev, "Failed to enable device mem!\n"); + return ret; + } + + ret = qm_get_pci_res(qm); + if (ret) + goto err_disable_pcidev; + + ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); + if (ret < 0) + goto err_get_pci_res; + pci_set_master(pdev); + + hisi_qm_set_state(qm, VF_PREPARE); + + num_vec = qm_get_irq_num(qm); + ret = pci_alloc_irq_vectors(pdev, num_vec, num_vec, PCI_IRQ_MSI); + if (ret < 0) { + dev_err(dev, "Failed to enable MSI vectors!\n"); + goto err_get_pci_res; + } + + return 0; + +err_get_pci_res: + qm_put_pci_res(qm); +err_disable_pcidev: + pci_disable_device(pdev); + return ret; +} + +static int hisi_qm_init_work(struct hisi_qm *qm) +{ + int i; + + for (i = 0; i < qm->qp_num; i++) + INIT_WORK(&qm->poll_data[i].work, qm_work_process); + + if (qm->fun_type == QM_HW_PF) + INIT_WORK(&qm->rst_work, hisi_qm_controller_reset); + + if (qm->ver > QM_HW_V2) + INIT_WORK(&qm->cmd_process, qm_cmd_process); + + qm->wq = alloc_workqueue("%s", WQ_HIGHPRI | WQ_MEM_RECLAIM | + WQ_UNBOUND, num_online_cpus(), + pci_name(qm->pdev)); + if (!qm->wq) { + pci_err(qm->pdev, "failed to alloc workqueue!\n"); + return -ENOMEM; + } + + return 0; +} + +static int hisi_qp_alloc_memory(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u16 sq_depth, cq_depth; + size_t qp_dma_size; + int i, ret; + + qm->qp_array = kcalloc(qm->qp_num, sizeof(struct hisi_qp), GFP_KERNEL); + if (!qm->qp_array) + return -ENOMEM; + + qm->poll_data = kcalloc(qm->qp_num, sizeof(struct hisi_qm_poll_data), GFP_KERNEL); + if (!qm->poll_data) { + kfree(qm->qp_array); + return -ENOMEM; + } + + qm_get_xqc_depth(qm, &sq_depth, &cq_depth, QM_QP_DEPTH_CAP); + + /* one more page for device or qp statuses */ + qp_dma_size = qm->sqe_size * sq_depth + sizeof(struct qm_cqe) * cq_depth; + qp_dma_size = PAGE_ALIGN(qp_dma_size) + PAGE_SIZE; + for (i = 0; i < qm->qp_num; i++) { + qm->poll_data[i].qm = qm; + ret = hisi_qp_memory_init(qm, qp_dma_size, i, sq_depth, cq_depth); + if (ret) + goto err_init_qp_mem; + + dev_dbg(dev, "allocate qp dma buf size=%zx)\n", qp_dma_size); + } + + return 0; +err_init_qp_mem: + hisi_qp_memory_uninit(qm, i); + + return ret; +} + +static int hisi_qm_memory_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret, total_func; + size_t off = 0; + + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) { + total_func = pci_sriov_get_totalvfs(qm->pdev) + 1; + qm->factor = kcalloc(total_func, sizeof(struct qm_shaper_factor), GFP_KERNEL); + if (!qm->factor) + return -ENOMEM; + + /* Only the PF value needs to be initialized */ + qm->factor[0].func_qos = QM_QOS_MAX_VAL; + } + +#define QM_INIT_BUF(qm, type, num) do { \ + (qm)->type = ((qm)->qdma.va + (off)); \ + (qm)->type##_dma = (qm)->qdma.dma + (off); \ + off += QMC_ALIGN(sizeof(struct qm_##type) * (num)); \ +} while (0) + + idr_init(&qm->qp_idr); + qm_get_xqc_depth(qm, &qm->eq_depth, &qm->aeq_depth, QM_XEQ_DEPTH_CAP); + qm->qdma.size = QMC_ALIGN(sizeof(struct qm_eqe) * qm->eq_depth) + + QMC_ALIGN(sizeof(struct qm_aeqe) * qm->aeq_depth) + + QMC_ALIGN(sizeof(struct qm_sqc) * qm->qp_num) + + QMC_ALIGN(sizeof(struct qm_cqc) * qm->qp_num); + qm->qdma.va = dma_alloc_coherent(dev, qm->qdma.size, &qm->qdma.dma, + GFP_ATOMIC); + dev_dbg(dev, "allocate qm dma buf size=%zx)\n", qm->qdma.size); + if (!qm->qdma.va) { + ret = -ENOMEM; + goto err_destroy_idr; + } + + QM_INIT_BUF(qm, eqe, qm->eq_depth); + QM_INIT_BUF(qm, aeqe, qm->aeq_depth); + QM_INIT_BUF(qm, sqc, qm->qp_num); + QM_INIT_BUF(qm, cqc, qm->qp_num); + + ret = hisi_qp_alloc_memory(qm); + if (ret) + goto err_alloc_qp_array; + + return 0; + +err_alloc_qp_array: + dma_free_coherent(dev, qm->qdma.size, qm->qdma.va, qm->qdma.dma); +err_destroy_idr: + idr_destroy(&qm->qp_idr); + if (test_bit(QM_SUPPORT_FUNC_QOS, &qm->caps)) + kfree(qm->factor); + + return ret; +} + +static void qm_last_regs_init(struct hisi_qm *qm) +{ + int dfx_regs_num = ARRAY_SIZE(qm_dfx_regs); + struct qm_debug *debug = &qm->debug; + int i; + + if (qm->fun_type == QM_HW_VF) + return; + + debug->qm_last_words = kcalloc(dfx_regs_num, sizeof(unsigned int), + GFP_KERNEL); + if (!debug->qm_last_words) + return; + + for (i = 0; i < dfx_regs_num; i++) { + debug->qm_last_words[i] = readl_relaxed(qm->io_base + + qm_dfx_regs[i].offset); + } +} + +/** + * hisi_qm_init() - Initialize configures about qm. + * @qm: The qm needing init. + * + * This function init qm, then we can call hisi_qm_start to put qm into work. + */ +int hisi_qm_init(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + struct device *dev = &pdev->dev; + int ret; + + hisi_qm_pre_init(qm); + + ret = hisi_qm_pci_init(qm); + if (ret) + return ret; + + ret = qm_irqs_register(qm); + if (ret) + goto err_pci_init; + + if (qm->fun_type == QM_HW_PF) { + qm_disable_clock_gate(qm); + ret = qm_dev_mem_reset(qm); + if (ret) { + dev_err(dev, "failed to reset device memory\n"); + goto err_irq_register; + } + } + + if (qm->mode == UACCE_MODE_SVA) { + ret = qm_alloc_uacce(qm); + if (ret < 0) + dev_warn(dev, "fail to alloc uacce (%d)\n", ret); + } + + ret = hisi_qm_memory_init(qm); + if (ret) + goto err_alloc_uacce; + + ret = hisi_qm_init_work(qm); + if (ret) + goto err_free_qm_memory; + + qm_cmd_init(qm); + atomic_set(&qm->status.flags, QM_INIT); + + qm_last_regs_init(qm); + + return 0; + +err_free_qm_memory: + hisi_qm_memory_uninit(qm); +err_alloc_uacce: + if (qm->use_sva) { + uacce_remove(qm->uacce); + qm->uacce = NULL; + } +err_irq_register: + qm_irqs_unregister(qm); +err_pci_init: + hisi_qm_pci_uninit(qm); + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_init); + +/** + * hisi_qm_get_dfx_access() - Try to get dfx access. + * @qm: pointer to accelerator device. + * + * Try to get dfx access, then user can get message. + * + * If device is in suspended, return failure, otherwise + * bump up the runtime PM usage counter. + */ +int hisi_qm_get_dfx_access(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + + if (pm_runtime_suspended(dev)) { + dev_info(dev, "can not read/write - device in suspended.\n"); + return -EAGAIN; + } + + return qm_pm_get_sync(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_get_dfx_access); + +/** + * hisi_qm_put_dfx_access() - Put dfx access. + * @qm: pointer to accelerator device. + * + * Put dfx access, drop runtime PM usage counter. + */ +void hisi_qm_put_dfx_access(struct hisi_qm *qm) +{ + qm_pm_put_sync(qm); +} +EXPORT_SYMBOL_GPL(hisi_qm_put_dfx_access); + +/** + * hisi_qm_pm_init() - Initialize qm runtime PM. + * @qm: pointer to accelerator device. + * + * Function that initialize qm runtime PM. + */ +void hisi_qm_pm_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + + if (!test_bit(QM_SUPPORT_RPM, &qm->caps)) + return; + + pm_runtime_set_autosuspend_delay(dev, QM_AUTOSUSPEND_DELAY); + pm_runtime_use_autosuspend(dev); + pm_runtime_put_noidle(dev); +} +EXPORT_SYMBOL_GPL(hisi_qm_pm_init); + +/** + * hisi_qm_pm_uninit() - Uninitialize qm runtime PM. + * @qm: pointer to accelerator device. + * + * Function that uninitialize qm runtime PM. + */ +void hisi_qm_pm_uninit(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + + if (!test_bit(QM_SUPPORT_RPM, &qm->caps)) + return; + + pm_runtime_get_noresume(dev); + pm_runtime_dont_use_autosuspend(dev); +} +EXPORT_SYMBOL_GPL(hisi_qm_pm_uninit); + +static int qm_prepare_for_suspend(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + u32 val; + + ret = qm->ops->set_msi(qm, false); + if (ret) { + pci_err(pdev, "failed to disable MSI before suspending!\n"); + return ret; + } + + /* shutdown OOO register */ + writel(ACC_MASTER_GLOBAL_CTRL_SHUTDOWN, + qm->io_base + ACC_MASTER_GLOBAL_CTRL); + + ret = readl_relaxed_poll_timeout(qm->io_base + ACC_MASTER_TRANS_RETURN, + val, + (val == ACC_MASTER_TRANS_RETURN_RW), + POLL_PERIOD, POLL_TIMEOUT); + if (ret) { + pci_emerg(pdev, "Bus lock! Please reset system.\n"); + return ret; + } + + ret = qm_set_pf_mse(qm, false); + if (ret) + pci_err(pdev, "failed to disable MSE before suspending!\n"); + + return ret; +} + +static int qm_rebuild_for_resume(struct hisi_qm *qm) +{ + struct pci_dev *pdev = qm->pdev; + int ret; + + ret = qm_set_pf_mse(qm, true); + if (ret) { + pci_err(pdev, "failed to enable MSE after resuming!\n"); + return ret; + } + + ret = qm->ops->set_msi(qm, true); + if (ret) { + pci_err(pdev, "failed to enable MSI after resuming!\n"); + return ret; + } + + ret = qm_dev_hw_init(qm); + if (ret) { + pci_err(pdev, "failed to init device after resuming\n"); + return ret; + } + + qm_cmd_init(qm); + hisi_qm_dev_err_init(qm); + qm_disable_clock_gate(qm); + ret = qm_dev_mem_reset(qm); + if (ret) + pci_err(pdev, "failed to reset device memory\n"); + + return ret; +} + +/** + * hisi_qm_suspend() - Runtime suspend of given device. + * @dev: device to suspend. + * + * Function that suspend the device. + */ +int hisi_qm_suspend(struct device *dev) +{ + struct pci_dev *pdev = to_pci_dev(dev); + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + pci_info(pdev, "entering suspended state\n"); + + ret = hisi_qm_stop(qm, QM_NORMAL); + if (ret) { + pci_err(pdev, "failed to stop qm(%d)\n", ret); + return ret; + } + + ret = qm_prepare_for_suspend(qm); + if (ret) + pci_err(pdev, "failed to prepare suspended(%d)\n", ret); + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_suspend); + +/** + * hisi_qm_resume() - Runtime resume of given device. + * @dev: device to resume. + * + * Function that resume the device. + */ +int hisi_qm_resume(struct device *dev) +{ + struct pci_dev *pdev = to_pci_dev(dev); + struct hisi_qm *qm = pci_get_drvdata(pdev); + int ret; + + pci_info(pdev, "resuming from suspend state\n"); + + ret = qm_rebuild_for_resume(qm); + if (ret) { + pci_err(pdev, "failed to rebuild resume(%d)\n", ret); + return ret; + } + + ret = hisi_qm_start(qm); + if (ret) + pci_err(pdev, "failed to start qm(%d)\n", ret); + + return ret; +} +EXPORT_SYMBOL_GPL(hisi_qm_resume); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Zhou Wang "); +MODULE_DESCRIPTION("HiSilicon Accelerator queue manager driver"); diff --git a/kae_driver/hisilicon/sec/Makefile b/kae_driver/hisilicon/sec/Makefile new file mode 100755 index 0000000..a55b698 --- /dev/null +++ b/kae_driver/hisilicon/sec/Makefile @@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0 +obj-$(CONFIG_CRYPTO_DEV_HISI_SEC) += hisi_sec.o +hisi_sec-y = sec_algs.o sec_drv.o diff --git a/kae_driver/hisilicon/sec/sec_algs.c b/kae_driver/hisilicon/sec/sec_algs.c new file mode 100755 index 0000000..490e154 --- /dev/null +++ b/kae_driver/hisilicon/sec/sec_algs.c @@ -0,0 +1,1122 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2016-2017 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +#include "sec_drv.h" + +#define SEC_MAX_CIPHER_KEY 64 +#define SEC_REQ_LIMIT SZ_32M + +struct sec_c_alg_cfg { + unsigned c_alg : 3; + unsigned c_mode : 3; + unsigned key_len : 2; + unsigned c_width : 2; +}; + +static const struct sec_c_alg_cfg sec_c_alg_cfgs[] = { + [SEC_C_DES_ECB_64] = { + .c_alg = SEC_C_ALG_DES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_DES, + }, + [SEC_C_DES_CBC_64] = { + .c_alg = SEC_C_ALG_DES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_DES, + }, + [SEC_C_3DES_ECB_192_3KEY] = { + .c_alg = SEC_C_ALG_3DES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_3DES_3_KEY, + }, + [SEC_C_3DES_ECB_192_2KEY] = { + .c_alg = SEC_C_ALG_3DES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_3DES_2_KEY, + }, + [SEC_C_3DES_CBC_192_3KEY] = { + .c_alg = SEC_C_ALG_3DES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_3DES_3_KEY, + }, + [SEC_C_3DES_CBC_192_2KEY] = { + .c_alg = SEC_C_ALG_3DES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_3DES_2_KEY, + }, + [SEC_C_AES_ECB_128] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_AES_128, + }, + [SEC_C_AES_ECB_192] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_AES_192, + }, + [SEC_C_AES_ECB_256] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_ECB, + .key_len = SEC_KEY_LEN_AES_256, + }, + [SEC_C_AES_CBC_128] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_AES_128, + }, + [SEC_C_AES_CBC_192] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_AES_192, + }, + [SEC_C_AES_CBC_256] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CBC, + .key_len = SEC_KEY_LEN_AES_256, + }, + [SEC_C_AES_CTR_128] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CTR, + .key_len = SEC_KEY_LEN_AES_128, + }, + [SEC_C_AES_CTR_192] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CTR, + .key_len = SEC_KEY_LEN_AES_192, + }, + [SEC_C_AES_CTR_256] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_CTR, + .key_len = SEC_KEY_LEN_AES_256, + }, + [SEC_C_AES_XTS_128] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_XTS, + .key_len = SEC_KEY_LEN_AES_128, + }, + [SEC_C_AES_XTS_256] = { + .c_alg = SEC_C_ALG_AES, + .c_mode = SEC_C_MODE_XTS, + .key_len = SEC_KEY_LEN_AES_256, + }, + [SEC_C_NULL] = { + }, +}; + +/* + * Mutex used to ensure safe operation of reference count of + * alg providers + */ +static DEFINE_MUTEX(algs_lock); +static unsigned int active_devs; + +static void sec_alg_skcipher_init_template(struct sec_alg_tfm_ctx *ctx, + struct sec_bd_info *req, + enum sec_cipher_alg alg) +{ + const struct sec_c_alg_cfg *cfg = &sec_c_alg_cfgs[alg]; + + memset(req, 0, sizeof(*req)); + req->w0 |= cfg->c_mode << SEC_BD_W0_C_MODE_S; + req->w1 |= cfg->c_alg << SEC_BD_W1_C_ALG_S; + req->w3 |= cfg->key_len << SEC_BD_W3_C_KEY_LEN_S; + req->w0 |= cfg->c_width << SEC_BD_W0_C_WIDTH_S; + + req->cipher_key_addr_lo = lower_32_bits(ctx->pkey); + req->cipher_key_addr_hi = upper_32_bits(ctx->pkey); +} + +static void sec_alg_skcipher_init_context(struct crypto_skcipher *atfm, + const u8 *key, + unsigned int keylen, + enum sec_cipher_alg alg) +{ + struct crypto_tfm *tfm = crypto_skcipher_tfm(atfm); + struct sec_alg_tfm_ctx *ctx = crypto_tfm_ctx(tfm); + + ctx->cipher_alg = alg; + memcpy(ctx->key, key, keylen); + sec_alg_skcipher_init_template(ctx, &ctx->req_template, + ctx->cipher_alg); +} + +static void sec_free_hw_sgl(struct sec_hw_sgl *hw_sgl, + dma_addr_t psec_sgl, struct sec_dev_info *info) +{ + struct sec_hw_sgl *sgl_current, *sgl_next; + dma_addr_t sgl_next_dma; + + sgl_current = hw_sgl; + while (sgl_current) { + sgl_next = sgl_current->next; + sgl_next_dma = sgl_current->next_sgl; + + dma_pool_free(info->hw_sgl_pool, sgl_current, psec_sgl); + + sgl_current = sgl_next; + psec_sgl = sgl_next_dma; + } +} + +static int sec_alloc_and_fill_hw_sgl(struct sec_hw_sgl **sec_sgl, + dma_addr_t *psec_sgl, + struct scatterlist *sgl, + int count, + struct sec_dev_info *info, + gfp_t gfp) +{ + struct sec_hw_sgl *sgl_current = NULL; + struct sec_hw_sgl *sgl_next; + dma_addr_t sgl_next_dma; + struct scatterlist *sg; + int ret, sge_index, i; + + if (!count) + return -EINVAL; + + for_each_sg(sgl, sg, count, i) { + sge_index = i % SEC_MAX_SGE_NUM; + if (sge_index == 0) { + sgl_next = dma_pool_zalloc(info->hw_sgl_pool, + gfp, &sgl_next_dma); + if (!sgl_next) { + ret = -ENOMEM; + goto err_free_hw_sgls; + } + + if (!sgl_current) { /* First one */ + *psec_sgl = sgl_next_dma; + *sec_sgl = sgl_next; + } else { /* Chained */ + sgl_current->entry_sum_in_sgl = SEC_MAX_SGE_NUM; + sgl_current->next_sgl = sgl_next_dma; + sgl_current->next = sgl_next; + } + sgl_current = sgl_next; + } + sgl_current->sge_entries[sge_index].buf = sg_dma_address(sg); + sgl_current->sge_entries[sge_index].len = sg_dma_len(sg); + sgl_current->data_bytes_in_sgl += sg_dma_len(sg); + } + sgl_current->entry_sum_in_sgl = count % SEC_MAX_SGE_NUM; + sgl_current->next_sgl = 0; + (*sec_sgl)->entry_sum_in_chain = count; + + return 0; + +err_free_hw_sgls: + sec_free_hw_sgl(*sec_sgl, *psec_sgl, info); + *psec_sgl = 0; + + return ret; +} + +static int sec_alg_skcipher_setkey(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen, + enum sec_cipher_alg alg) +{ + struct sec_alg_tfm_ctx *ctx = crypto_skcipher_ctx(tfm); + struct device *dev = ctx->queue->dev_info->dev; + + mutex_lock(&ctx->lock); + if (ctx->key) { + /* rekeying */ + memset(ctx->key, 0, SEC_MAX_CIPHER_KEY); + } else { + /* new key */ + ctx->key = dma_alloc_coherent(dev, SEC_MAX_CIPHER_KEY, + &ctx->pkey, GFP_KERNEL); + if (!ctx->key) { + mutex_unlock(&ctx->lock); + return -ENOMEM; + } + } + mutex_unlock(&ctx->lock); + sec_alg_skcipher_init_context(tfm, key, keylen, alg); + + return 0; +} + +static int sec_alg_skcipher_setkey_aes_ecb(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + enum sec_cipher_alg alg; + + switch (keylen) { + case AES_KEYSIZE_128: + alg = SEC_C_AES_ECB_128; + break; + case AES_KEYSIZE_192: + alg = SEC_C_AES_ECB_192; + break; + case AES_KEYSIZE_256: + alg = SEC_C_AES_ECB_256; + break; + default: + return -EINVAL; + } + + return sec_alg_skcipher_setkey(tfm, key, keylen, alg); +} + +static int sec_alg_skcipher_setkey_aes_cbc(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + enum sec_cipher_alg alg; + + switch (keylen) { + case AES_KEYSIZE_128: + alg = SEC_C_AES_CBC_128; + break; + case AES_KEYSIZE_192: + alg = SEC_C_AES_CBC_192; + break; + case AES_KEYSIZE_256: + alg = SEC_C_AES_CBC_256; + break; + default: + return -EINVAL; + } + + return sec_alg_skcipher_setkey(tfm, key, keylen, alg); +} + +static int sec_alg_skcipher_setkey_aes_ctr(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + enum sec_cipher_alg alg; + + switch (keylen) { + case AES_KEYSIZE_128: + alg = SEC_C_AES_CTR_128; + break; + case AES_KEYSIZE_192: + alg = SEC_C_AES_CTR_192; + break; + case AES_KEYSIZE_256: + alg = SEC_C_AES_CTR_256; + break; + default: + return -EINVAL; + } + + return sec_alg_skcipher_setkey(tfm, key, keylen, alg); +} + +static int sec_alg_skcipher_setkey_aes_xts(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + enum sec_cipher_alg alg; + int ret; + + ret = xts_verify_key(tfm, key, keylen); + if (ret) + return ret; + + switch (keylen) { + case AES_KEYSIZE_128 * 2: + alg = SEC_C_AES_XTS_128; + break; + case AES_KEYSIZE_256 * 2: + alg = SEC_C_AES_XTS_256; + break; + default: + return -EINVAL; + } + + return sec_alg_skcipher_setkey(tfm, key, keylen, alg); +} + +static int sec_alg_skcipher_setkey_des_ecb(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + return verify_skcipher_des_key(tfm, key) ?: + sec_alg_skcipher_setkey(tfm, key, keylen, SEC_C_DES_ECB_64); +} + +static int sec_alg_skcipher_setkey_des_cbc(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + return verify_skcipher_des_key(tfm, key) ?: + sec_alg_skcipher_setkey(tfm, key, keylen, SEC_C_DES_CBC_64); +} + +static int sec_alg_skcipher_setkey_3des_ecb(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + return verify_skcipher_des3_key(tfm, key) ?: + sec_alg_skcipher_setkey(tfm, key, keylen, + SEC_C_3DES_ECB_192_3KEY); +} + +static int sec_alg_skcipher_setkey_3des_cbc(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + return verify_skcipher_des3_key(tfm, key) ?: + sec_alg_skcipher_setkey(tfm, key, keylen, + SEC_C_3DES_CBC_192_3KEY); +} + +static void sec_alg_free_el(struct sec_request_el *el, + struct sec_dev_info *info) +{ + sec_free_hw_sgl(el->out, el->dma_out, info); + sec_free_hw_sgl(el->in, el->dma_in, info); + kfree(el->sgl_in); + kfree(el->sgl_out); + kfree(el); +} + +/* queuelock must be held */ +static int sec_send_request(struct sec_request *sec_req, struct sec_queue *queue) +{ + struct sec_request_el *el, *temp; + int ret = 0; + + mutex_lock(&sec_req->lock); + list_for_each_entry_safe(el, temp, &sec_req->elements, head) { + /* + * Add to hardware queue only under following circumstances + * 1) Software and hardware queue empty so no chain dependencies + * 2) No dependencies as new IV - (check software queue empty + * to maintain order) + * 3) No dependencies because the mode does no chaining. + * + * In other cases first insert onto the software queue which + * is then emptied as requests complete + */ + if (!queue->havesoftqueue || + (kfifo_is_empty(&queue->softqueue) && + sec_queue_empty(queue))) { + ret = sec_queue_send(queue, &el->req, sec_req); + if (ret == -EAGAIN) { + /* Wait unti we can send then try again */ + /* DEAD if here - should not happen */ + ret = -EBUSY; + goto err_unlock; + } + } else { + kfifo_put(&queue->softqueue, el); + } + } +err_unlock: + mutex_unlock(&sec_req->lock); + + return ret; +} + +static void sec_skcipher_alg_callback(struct sec_bd_info *sec_resp, + struct crypto_async_request *req_base) +{ + struct skcipher_request *skreq = container_of(req_base, + struct skcipher_request, + base); + struct sec_request *sec_req = skcipher_request_ctx(skreq); + struct sec_request *backlog_req; + struct sec_request_el *sec_req_el, *nextrequest; + struct sec_alg_tfm_ctx *ctx = sec_req->tfm_ctx; + struct crypto_skcipher *atfm = crypto_skcipher_reqtfm(skreq); + struct device *dev = ctx->queue->dev_info->dev; + int icv_or_skey_en, ret; + bool done; + + sec_req_el = list_first_entry(&sec_req->elements, struct sec_request_el, + head); + icv_or_skey_en = (sec_resp->w0 & SEC_BD_W0_ICV_OR_SKEY_EN_M) >> + SEC_BD_W0_ICV_OR_SKEY_EN_S; + if (sec_resp->w1 & SEC_BD_W1_BD_INVALID || icv_or_skey_en == 3) { + dev_err(dev, "Got an invalid answer %lu %d\n", + sec_resp->w1 & SEC_BD_W1_BD_INVALID, + icv_or_skey_en); + sec_req->err = -EINVAL; + /* + * We need to muddle on to avoid getting stuck with elements + * on the queue. Error will be reported so requester so + * it should be able to handle appropriately. + */ + } + + spin_lock_bh(&ctx->queue->queuelock); + /* Put the IV in place for chained cases */ + switch (ctx->cipher_alg) { + case SEC_C_AES_CBC_128: + case SEC_C_AES_CBC_192: + case SEC_C_AES_CBC_256: + if (sec_req_el->req.w0 & SEC_BD_W0_DE) + sg_pcopy_to_buffer(sec_req_el->sgl_out, + sg_nents(sec_req_el->sgl_out), + skreq->iv, + crypto_skcipher_ivsize(atfm), + sec_req_el->el_length - + crypto_skcipher_ivsize(atfm)); + else + sg_pcopy_to_buffer(sec_req_el->sgl_in, + sg_nents(sec_req_el->sgl_in), + skreq->iv, + crypto_skcipher_ivsize(atfm), + sec_req_el->el_length - + crypto_skcipher_ivsize(atfm)); + /* No need to sync to the device as coherent DMA */ + break; + case SEC_C_AES_CTR_128: + case SEC_C_AES_CTR_192: + case SEC_C_AES_CTR_256: + crypto_inc(skreq->iv, 16); + break; + default: + /* Do not update */ + break; + } + + if (ctx->queue->havesoftqueue && + !kfifo_is_empty(&ctx->queue->softqueue) && + sec_queue_empty(ctx->queue)) { + ret = kfifo_get(&ctx->queue->softqueue, &nextrequest); + if (ret <= 0) + dev_err(dev, + "Error getting next element from kfifo %d\n", + ret); + else + /* We know there is space so this cannot fail */ + sec_queue_send(ctx->queue, &nextrequest->req, + nextrequest->sec_req); + } else if (!list_empty(&ctx->backlog)) { + /* Need to verify there is room first */ + backlog_req = list_first_entry(&ctx->backlog, + typeof(*backlog_req), + backlog_head); + if (sec_queue_can_enqueue(ctx->queue, + backlog_req->num_elements) || + (ctx->queue->havesoftqueue && + kfifo_avail(&ctx->queue->softqueue) > + backlog_req->num_elements)) { + sec_send_request(backlog_req, ctx->queue); + backlog_req->req_base->complete(backlog_req->req_base, + -EINPROGRESS); + list_del(&backlog_req->backlog_head); + } + } + spin_unlock_bh(&ctx->queue->queuelock); + + mutex_lock(&sec_req->lock); + list_del(&sec_req_el->head); + mutex_unlock(&sec_req->lock); + sec_alg_free_el(sec_req_el, ctx->queue->dev_info); + + /* + * Request is done. + * The dance is needed as the lock is freed in the completion + */ + mutex_lock(&sec_req->lock); + done = list_empty(&sec_req->elements); + mutex_unlock(&sec_req->lock); + if (done) { + if (crypto_skcipher_ivsize(atfm)) { + dma_unmap_single(dev, sec_req->dma_iv, + crypto_skcipher_ivsize(atfm), + DMA_TO_DEVICE); + } + dma_unmap_sg(dev, skreq->src, sec_req->len_in, + DMA_BIDIRECTIONAL); + if (skreq->src != skreq->dst) + dma_unmap_sg(dev, skreq->dst, sec_req->len_out, + DMA_BIDIRECTIONAL); + skreq->base.complete(&skreq->base, sec_req->err); + } +} + +void sec_alg_callback(struct sec_bd_info *resp, void *shadow) +{ + struct sec_request *sec_req = shadow; + + sec_req->cb(resp, sec_req->req_base); +} + +static int sec_alg_alloc_and_calc_split_sizes(int length, size_t **split_sizes, + int *steps, gfp_t gfp) +{ + size_t *sizes; + int i; + + /* Split into suitable sized blocks */ + *steps = roundup(length, SEC_REQ_LIMIT) / SEC_REQ_LIMIT; + sizes = kcalloc(*steps, sizeof(*sizes), gfp); + if (!sizes) + return -ENOMEM; + + for (i = 0; i < *steps - 1; i++) + sizes[i] = SEC_REQ_LIMIT; + sizes[*steps - 1] = length - SEC_REQ_LIMIT * (*steps - 1); + *split_sizes = sizes; + + return 0; +} + +static int sec_map_and_split_sg(struct scatterlist *sgl, size_t *split_sizes, + int steps, struct scatterlist ***splits, + int **splits_nents, + int sgl_len_in, + struct device *dev, gfp_t gfp) +{ + int ret, count; + + count = dma_map_sg(dev, sgl, sgl_len_in, DMA_BIDIRECTIONAL); + if (!count) + return -EINVAL; + + *splits = kcalloc(steps, sizeof(struct scatterlist *), gfp); + if (!*splits) { + ret = -ENOMEM; + goto err_unmap_sg; + } + *splits_nents = kcalloc(steps, sizeof(int), gfp); + if (!*splits_nents) { + ret = -ENOMEM; + goto err_free_splits; + } + + /* output the scatter list before and after this */ + ret = sg_split(sgl, count, 0, steps, split_sizes, + *splits, *splits_nents, gfp); + if (ret) { + ret = -ENOMEM; + goto err_free_splits_nents; + } + + return 0; + +err_free_splits_nents: + kfree(*splits_nents); +err_free_splits: + kfree(*splits); +err_unmap_sg: + dma_unmap_sg(dev, sgl, sgl_len_in, DMA_BIDIRECTIONAL); + + return ret; +} + +/* + * Reverses the sec_map_and_split_sg call for messages not yet added to + * the queues. + */ +static void sec_unmap_sg_on_err(struct scatterlist *sgl, int steps, + struct scatterlist **splits, int *splits_nents, + int sgl_len_in, struct device *dev) +{ + int i; + + for (i = 0; i < steps; i++) + kfree(splits[i]); + kfree(splits_nents); + kfree(splits); + + dma_unmap_sg(dev, sgl, sgl_len_in, DMA_BIDIRECTIONAL); +} + +static struct sec_request_el +*sec_alg_alloc_and_fill_el(struct sec_bd_info *template, int encrypt, + int el_size, bool different_dest, + struct scatterlist *sgl_in, int n_ents_in, + struct scatterlist *sgl_out, int n_ents_out, + struct sec_dev_info *info, gfp_t gfp) +{ + struct sec_request_el *el; + struct sec_bd_info *req; + int ret; + + el = kzalloc(sizeof(*el), gfp); + if (!el) + return ERR_PTR(-ENOMEM); + el->el_length = el_size; + req = &el->req; + memcpy(req, template, sizeof(*req)); + + req->w0 &= ~SEC_BD_W0_CIPHER_M; + if (encrypt) + req->w0 |= SEC_CIPHER_ENCRYPT << SEC_BD_W0_CIPHER_S; + else + req->w0 |= SEC_CIPHER_DECRYPT << SEC_BD_W0_CIPHER_S; + + req->w0 &= ~SEC_BD_W0_C_GRAN_SIZE_19_16_M; + req->w0 |= ((el_size >> 16) << SEC_BD_W0_C_GRAN_SIZE_19_16_S) & + SEC_BD_W0_C_GRAN_SIZE_19_16_M; + + req->w0 &= ~SEC_BD_W0_C_GRAN_SIZE_21_20_M; + req->w0 |= ((el_size >> 20) << SEC_BD_W0_C_GRAN_SIZE_21_20_S) & + SEC_BD_W0_C_GRAN_SIZE_21_20_M; + + /* Writing whole u32 so no need to take care of masking */ + req->w2 = ((1 << SEC_BD_W2_GRAN_NUM_S) & SEC_BD_W2_GRAN_NUM_M) | + ((el_size << SEC_BD_W2_C_GRAN_SIZE_15_0_S) & + SEC_BD_W2_C_GRAN_SIZE_15_0_M); + + req->w3 &= ~SEC_BD_W3_CIPHER_LEN_OFFSET_M; + req->w1 |= SEC_BD_W1_ADDR_TYPE; + + el->sgl_in = sgl_in; + + ret = sec_alloc_and_fill_hw_sgl(&el->in, &el->dma_in, el->sgl_in, + n_ents_in, info, gfp); + if (ret) + goto err_free_el; + + req->data_addr_lo = lower_32_bits(el->dma_in); + req->data_addr_hi = upper_32_bits(el->dma_in); + + if (different_dest) { + el->sgl_out = sgl_out; + ret = sec_alloc_and_fill_hw_sgl(&el->out, &el->dma_out, + el->sgl_out, + n_ents_out, info, gfp); + if (ret) + goto err_free_hw_sgl_in; + + req->w0 |= SEC_BD_W0_DE; + req->cipher_destin_addr_lo = lower_32_bits(el->dma_out); + req->cipher_destin_addr_hi = upper_32_bits(el->dma_out); + + } else { + req->w0 &= ~SEC_BD_W0_DE; + req->cipher_destin_addr_lo = lower_32_bits(el->dma_in); + req->cipher_destin_addr_hi = upper_32_bits(el->dma_in); + } + + return el; + +err_free_hw_sgl_in: + sec_free_hw_sgl(el->in, el->dma_in, info); +err_free_el: + kfree(el); + + return ERR_PTR(ret); +} + +static int sec_alg_skcipher_crypto(struct skcipher_request *skreq, + bool encrypt) +{ + struct crypto_skcipher *atfm = crypto_skcipher_reqtfm(skreq); + struct crypto_tfm *tfm = crypto_skcipher_tfm(atfm); + struct sec_alg_tfm_ctx *ctx = crypto_tfm_ctx(tfm); + struct sec_queue *queue = ctx->queue; + struct sec_request *sec_req = skcipher_request_ctx(skreq); + struct sec_dev_info *info = queue->dev_info; + int i, ret, steps; + size_t *split_sizes; + struct scatterlist **splits_in; + struct scatterlist **splits_out = NULL; + int *splits_in_nents; + int *splits_out_nents = NULL; + struct sec_request_el *el, *temp; + bool split = skreq->src != skreq->dst; + gfp_t gfp = skreq->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP ? GFP_KERNEL : GFP_ATOMIC; + + mutex_init(&sec_req->lock); + sec_req->req_base = &skreq->base; + sec_req->err = 0; + /* SGL mapping out here to allow us to break it up as necessary */ + sec_req->len_in = sg_nents(skreq->src); + + ret = sec_alg_alloc_and_calc_split_sizes(skreq->cryptlen, &split_sizes, + &steps, gfp); + if (ret) + return ret; + sec_req->num_elements = steps; + ret = sec_map_and_split_sg(skreq->src, split_sizes, steps, &splits_in, + &splits_in_nents, sec_req->len_in, + info->dev, gfp); + if (ret) + goto err_free_split_sizes; + + if (split) { + sec_req->len_out = sg_nents(skreq->dst); + ret = sec_map_and_split_sg(skreq->dst, split_sizes, steps, + &splits_out, &splits_out_nents, + sec_req->len_out, info->dev, gfp); + if (ret) + goto err_unmap_in_sg; + } + /* Shared info stored in seq_req - applies to all BDs */ + sec_req->tfm_ctx = ctx; + sec_req->cb = sec_skcipher_alg_callback; + INIT_LIST_HEAD(&sec_req->elements); + + /* + * Future optimization. + * In the chaining case we can't use a dma pool bounce buffer + * but in the case where we know there is no chaining we can + */ + if (crypto_skcipher_ivsize(atfm)) { + sec_req->dma_iv = dma_map_single(info->dev, skreq->iv, + crypto_skcipher_ivsize(atfm), + DMA_TO_DEVICE); + if (dma_mapping_error(info->dev, sec_req->dma_iv)) { + ret = -ENOMEM; + goto err_unmap_out_sg; + } + } + + /* Set them all up then queue - cleaner error handling. */ + for (i = 0; i < steps; i++) { + el = sec_alg_alloc_and_fill_el(&ctx->req_template, + encrypt ? 1 : 0, + split_sizes[i], + skreq->src != skreq->dst, + splits_in[i], splits_in_nents[i], + split ? splits_out[i] : NULL, + split ? splits_out_nents[i] : 0, + info, gfp); + if (IS_ERR(el)) { + ret = PTR_ERR(el); + goto err_free_elements; + } + el->req.cipher_iv_addr_lo = lower_32_bits(sec_req->dma_iv); + el->req.cipher_iv_addr_hi = upper_32_bits(sec_req->dma_iv); + el->sec_req = sec_req; + list_add_tail(&el->head, &sec_req->elements); + } + + /* + * Only attempt to queue if the whole lot can fit in the queue - + * we can't successfully cleanup after a partial queing so this + * must succeed or fail atomically. + * + * Big hammer test of both software and hardware queues - could be + * more refined but this is unlikely to happen so no need. + */ + + /* Grab a big lock for a long time to avoid concurrency issues */ + spin_lock_bh(&queue->queuelock); + + /* + * Can go on to queue if we have space in either: + * 1) The hardware queue and no software queue + * 2) The software queue + * AND there is nothing in the backlog. If there is backlog we + * have to only queue to the backlog queue and return busy. + */ + if ((!sec_queue_can_enqueue(queue, steps) && + (!queue->havesoftqueue || + kfifo_avail(&queue->softqueue) > steps)) || + !list_empty(&ctx->backlog)) { + ret = -EBUSY; + if ((skreq->base.flags & CRYPTO_TFM_REQ_MAY_BACKLOG)) { + list_add_tail(&sec_req->backlog_head, &ctx->backlog); + spin_unlock_bh(&queue->queuelock); + goto out; + } + + spin_unlock_bh(&queue->queuelock); + goto err_free_elements; + } + ret = sec_send_request(sec_req, queue); + spin_unlock_bh(&queue->queuelock); + if (ret) + goto err_free_elements; + + ret = -EINPROGRESS; +out: + /* Cleanup - all elements in pointer arrays have been copied */ + kfree(splits_in_nents); + kfree(splits_in); + kfree(splits_out_nents); + kfree(splits_out); + kfree(split_sizes); + return ret; + +err_free_elements: + list_for_each_entry_safe(el, temp, &sec_req->elements, head) { + list_del(&el->head); + sec_alg_free_el(el, info); + } + if (crypto_skcipher_ivsize(atfm)) + dma_unmap_single(info->dev, sec_req->dma_iv, + crypto_skcipher_ivsize(atfm), + DMA_BIDIRECTIONAL); +err_unmap_out_sg: + if (split) + sec_unmap_sg_on_err(skreq->dst, steps, splits_out, + splits_out_nents, sec_req->len_out, + info->dev); +err_unmap_in_sg: + sec_unmap_sg_on_err(skreq->src, steps, splits_in, splits_in_nents, + sec_req->len_in, info->dev); +err_free_split_sizes: + kfree(split_sizes); + + return ret; +} + +static int sec_alg_skcipher_encrypt(struct skcipher_request *req) +{ + return sec_alg_skcipher_crypto(req, true); +} + +static int sec_alg_skcipher_decrypt(struct skcipher_request *req) +{ + return sec_alg_skcipher_crypto(req, false); +} + +static int sec_alg_skcipher_init(struct crypto_skcipher *tfm) +{ + struct sec_alg_tfm_ctx *ctx = crypto_skcipher_ctx(tfm); + + mutex_init(&ctx->lock); + INIT_LIST_HEAD(&ctx->backlog); + crypto_skcipher_set_reqsize(tfm, sizeof(struct sec_request)); + + ctx->queue = sec_queue_alloc_start_safe(); + if (IS_ERR(ctx->queue)) + return PTR_ERR(ctx->queue); + + spin_lock_init(&ctx->queue->queuelock); + ctx->queue->havesoftqueue = false; + + return 0; +} + +static void sec_alg_skcipher_exit(struct crypto_skcipher *tfm) +{ + struct sec_alg_tfm_ctx *ctx = crypto_skcipher_ctx(tfm); + struct device *dev = ctx->queue->dev_info->dev; + + if (ctx->key) { + memzero_explicit(ctx->key, SEC_MAX_CIPHER_KEY); + dma_free_coherent(dev, SEC_MAX_CIPHER_KEY, ctx->key, + ctx->pkey); + } + sec_queue_stop_release(ctx->queue); +} + +static int sec_alg_skcipher_init_with_queue(struct crypto_skcipher *tfm) +{ + struct sec_alg_tfm_ctx *ctx = crypto_skcipher_ctx(tfm); + int ret; + + ret = sec_alg_skcipher_init(tfm); + if (ret) + return ret; + + INIT_KFIFO(ctx->queue->softqueue); + ret = kfifo_alloc(&ctx->queue->softqueue, 512, GFP_KERNEL); + if (ret) { + sec_alg_skcipher_exit(tfm); + return ret; + } + ctx->queue->havesoftqueue = true; + + return 0; +} + +static void sec_alg_skcipher_exit_with_queue(struct crypto_skcipher *tfm) +{ + struct sec_alg_tfm_ctx *ctx = crypto_skcipher_ctx(tfm); + + kfifo_free(&ctx->queue->softqueue); + sec_alg_skcipher_exit(tfm); +} + +static struct skcipher_alg sec_algs[] = { + { + .base = { + .cra_name = "ecb(aes)", + .cra_driver_name = "hisi_sec_aes_ecb", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init, + .exit = sec_alg_skcipher_exit, + .setkey = sec_alg_skcipher_setkey_aes_ecb, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = 0, + }, { + .base = { + .cra_name = "cbc(aes)", + .cra_driver_name = "hisi_sec_aes_cbc", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init_with_queue, + .exit = sec_alg_skcipher_exit_with_queue, + .setkey = sec_alg_skcipher_setkey_aes_cbc, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + }, { + .base = { + .cra_name = "ctr(aes)", + .cra_driver_name = "hisi_sec_aes_ctr", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init_with_queue, + .exit = sec_alg_skcipher_exit_with_queue, + .setkey = sec_alg_skcipher_setkey_aes_ctr, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + }, { + .base = { + .cra_name = "xts(aes)", + .cra_driver_name = "hisi_sec_aes_xts", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init, + .exit = sec_alg_skcipher_exit, + .setkey = sec_alg_skcipher_setkey_aes_xts, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = 2 * AES_MIN_KEY_SIZE, + .max_keysize = 2 * AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + }, { + /* Unable to find any test vectors so untested */ + .base = { + .cra_name = "ecb(des)", + .cra_driver_name = "hisi_sec_des_ecb", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = DES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init, + .exit = sec_alg_skcipher_exit, + .setkey = sec_alg_skcipher_setkey_des_ecb, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = DES_KEY_SIZE, + .max_keysize = DES_KEY_SIZE, + .ivsize = 0, + }, { + .base = { + .cra_name = "cbc(des)", + .cra_driver_name = "hisi_sec_des_cbc", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = DES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init_with_queue, + .exit = sec_alg_skcipher_exit_with_queue, + .setkey = sec_alg_skcipher_setkey_des_cbc, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = DES_KEY_SIZE, + .max_keysize = DES_KEY_SIZE, + .ivsize = DES_BLOCK_SIZE, + }, { + .base = { + .cra_name = "cbc(des3_ede)", + .cra_driver_name = "hisi_sec_3des_cbc", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = DES3_EDE_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init_with_queue, + .exit = sec_alg_skcipher_exit_with_queue, + .setkey = sec_alg_skcipher_setkey_3des_cbc, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = DES3_EDE_KEY_SIZE, + .max_keysize = DES3_EDE_KEY_SIZE, + .ivsize = DES3_EDE_BLOCK_SIZE, + }, { + .base = { + .cra_name = "ecb(des3_ede)", + .cra_driver_name = "hisi_sec_3des_ecb", + .cra_priority = 4001, + .cra_flags = CRYPTO_ALG_ASYNC | + CRYPTO_ALG_ALLOCATES_MEMORY, + .cra_blocksize = DES3_EDE_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sec_alg_tfm_ctx), + .cra_alignmask = 0, + .cra_module = THIS_MODULE, + }, + .init = sec_alg_skcipher_init, + .exit = sec_alg_skcipher_exit, + .setkey = sec_alg_skcipher_setkey_3des_ecb, + .decrypt = sec_alg_skcipher_decrypt, + .encrypt = sec_alg_skcipher_encrypt, + .min_keysize = DES3_EDE_KEY_SIZE, + .max_keysize = DES3_EDE_KEY_SIZE, + .ivsize = 0, + } +}; + +int sec_algs_register(void) +{ + int ret = 0; + + mutex_lock(&algs_lock); + if (++active_devs != 1) + goto unlock; + + ret = crypto_register_skciphers(sec_algs, ARRAY_SIZE(sec_algs)); + if (ret) + --active_devs; +unlock: + mutex_unlock(&algs_lock); + + return ret; +} + +void sec_algs_unregister(void) +{ + mutex_lock(&algs_lock); + if (--active_devs != 0) + goto unlock; + crypto_unregister_skciphers(sec_algs, ARRAY_SIZE(sec_algs)); + +unlock: + mutex_unlock(&algs_lock); +} diff --git a/kae_driver/hisilicon/sec/sec_drv.c b/kae_driver/hisilicon/sec/sec_drv.c new file mode 100755 index 0000000..e758513 --- /dev/null +++ b/kae_driver/hisilicon/sec/sec_drv.c @@ -0,0 +1,1321 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Driver for the HiSilicon SEC units found on Hip06 Hip07 + * + * Copyright (c) 2016-2017 HiSilicon Limited. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "sec_drv.h" + +#define SEC_QUEUE_AR_FROCE_ALLOC 0 +#define SEC_QUEUE_AR_FROCE_NOALLOC 1 +#define SEC_QUEUE_AR_FROCE_DIS 2 + +#define SEC_QUEUE_AW_FROCE_ALLOC 0 +#define SEC_QUEUE_AW_FROCE_NOALLOC 1 +#define SEC_QUEUE_AW_FROCE_DIS 2 + +/* SEC_ALGSUB registers */ +#define SEC_ALGSUB_CLK_EN_REG 0x03b8 +#define SEC_ALGSUB_CLK_DIS_REG 0x03bc +#define SEC_ALGSUB_CLK_ST_REG 0x535c +#define SEC_ALGSUB_RST_REQ_REG 0x0aa8 +#define SEC_ALGSUB_RST_DREQ_REG 0x0aac +#define SEC_ALGSUB_RST_ST_REG 0x5a54 +#define SEC_ALGSUB_RST_ST_IS_RST BIT(0) + +#define SEC_ALGSUB_BUILD_RST_REQ_REG 0x0ab8 +#define SEC_ALGSUB_BUILD_RST_DREQ_REG 0x0abc +#define SEC_ALGSUB_BUILD_RST_ST_REG 0x5a5c +#define SEC_ALGSUB_BUILD_RST_ST_IS_RST BIT(0) + +#define SEC_SAA_BASE 0x00001000UL + +/* SEC_SAA registers */ +#define SEC_SAA_CTRL_REG(x) ((x) * SEC_SAA_ADDR_SIZE) +#define SEC_SAA_CTRL_GET_QM_EN BIT(0) + +#define SEC_ST_INTMSK1_REG 0x0200 +#define SEC_ST_RINT1_REG 0x0400 +#define SEC_ST_INTSTS1_REG 0x0600 +#define SEC_BD_MNG_STAT_REG 0x0800 +#define SEC_PARSING_STAT_REG 0x0804 +#define SEC_LOAD_TIME_OUT_CNT_REG 0x0808 +#define SEC_CORE_WORK_TIME_OUT_CNT_REG 0x080c +#define SEC_BACK_TIME_OUT_CNT_REG 0x0810 +#define SEC_BD1_PARSING_RD_TIME_OUT_CNT_REG 0x0814 +#define SEC_BD1_PARSING_WR_TIME_OUT_CNT_REG 0x0818 +#define SEC_BD2_PARSING_RD_TIME_OUT_CNT_REG 0x081c +#define SEC_BD2_PARSING_WR_TIME_OUT_CNT_REG 0x0820 +#define SEC_SAA_ACC_REG 0x083c +#define SEC_BD_NUM_CNT_IN_SEC_REG 0x0858 +#define SEC_LOAD_WORK_TIME_CNT_REG 0x0860 +#define SEC_CORE_WORK_WORK_TIME_CNT_REG 0x0864 +#define SEC_BACK_WORK_TIME_CNT_REG 0x0868 +#define SEC_SAA_IDLE_TIME_CNT_REG 0x086c +#define SEC_SAA_CLK_CNT_REG 0x0870 + +/* SEC_COMMON registers */ +#define SEC_CLK_EN_REG 0x0000 +#define SEC_CTRL_REG 0x0004 + +#define SEC_COMMON_CNT_CLR_CE_REG 0x0008 +#define SEC_COMMON_CNT_CLR_CE_CLEAR BIT(0) +#define SEC_COMMON_CNT_CLR_CE_SNAP_EN BIT(1) + +#define SEC_SECURE_CTRL_REG 0x000c +#define SEC_AXI_CACHE_CFG_REG 0x0010 +#define SEC_AXI_QOS_CFG_REG 0x0014 +#define SEC_IPV4_MASK_TABLE_REG 0x0020 +#define SEC_IPV6_MASK_TABLE_X_REG(x) (0x0024 + (x) * 4) +#define SEC_FSM_MAX_CNT_REG 0x0064 + +#define SEC_CTRL2_REG 0x0068 +#define SEC_CTRL2_DATA_AXI_RD_OTSD_CFG_M GENMASK(3, 0) +#define SEC_CTRL2_DATA_AXI_RD_OTSD_CFG_S 0 +#define SEC_CTRL2_DATA_AXI_WR_OTSD_CFG_M GENMASK(6, 4) +#define SEC_CTRL2_DATA_AXI_WR_OTSD_CFG_S 4 +#define SEC_CTRL2_CLK_GATE_EN BIT(7) +#define SEC_CTRL2_ENDIAN_BD BIT(8) +#define SEC_CTRL2_ENDIAN_BD_TYPE BIT(9) + +#define SEC_CNT_PRECISION_CFG_REG 0x006c +#define SEC_DEBUG_BD_CFG_REG 0x0070 +#define SEC_DEBUG_BD_CFG_WB_NORMAL BIT(0) +#define SEC_DEBUG_BD_CFG_WB_EN BIT(1) + +#define SEC_Q_SIGHT_SEL 0x0074 +#define SEC_Q_SIGHT_HIS_CLR 0x0078 +#define SEC_Q_VMID_CFG_REG(q) (0x0100 + (q) * 4) +#define SEC_Q_WEIGHT_CFG_REG(q) (0x200 + (q) * 4) +#define SEC_STAT_CLR_REG 0x0a00 +#define SEC_SAA_IDLE_CNT_CLR_REG 0x0a04 +#define SEC_QM_CPL_Q_IDBUF_DFX_CFG_REG 0x0b00 +#define SEC_QM_CPL_Q_IDBUF_DFX_RESULT_REG 0x0b04 +#define SEC_QM_BD_DFX_CFG_REG 0x0b08 +#define SEC_QM_BD_DFX_RESULT_REG 0x0b0c +#define SEC_QM_BDID_DFX_RESULT_REG 0x0b10 +#define SEC_QM_BD_DFIFO_STATUS_REG 0x0b14 +#define SEC_QM_BD_DFX_CFG2_REG 0x0b1c +#define SEC_QM_BD_DFX_RESULT2_REG 0x0b20 +#define SEC_QM_BD_IDFIFO_STATUS_REG 0x0b18 +#define SEC_QM_BD_DFIFO_STATUS2_REG 0x0b28 +#define SEC_QM_BD_IDFIFO_STATUS2_REG 0x0b2c + +#define SEC_HASH_IPV4_MASK 0xfff00000 +#define SEC_MAX_SAA_NUM 0xa +#define SEC_SAA_ADDR_SIZE 0x1000 + +#define SEC_Q_INIT_REG 0x0 +#define SEC_Q_INIT_WO_STAT_CLEAR 0x2 +#define SEC_Q_INIT_AND_STAT_CLEAR 0x3 + +#define SEC_Q_CFG_REG 0x8 +#define SEC_Q_CFG_REORDER BIT(0) + +#define SEC_Q_PROC_NUM_CFG_REG 0x10 +#define SEC_QUEUE_ENB_REG 0x18 + +#define SEC_Q_DEPTH_CFG_REG 0x50 +#define SEC_Q_DEPTH_CFG_DEPTH_M GENMASK(11, 0) +#define SEC_Q_DEPTH_CFG_DEPTH_S 0 + +#define SEC_Q_BASE_HADDR_REG 0x54 +#define SEC_Q_BASE_LADDR_REG 0x58 +#define SEC_Q_WR_PTR_REG 0x5c +#define SEC_Q_OUTORDER_BASE_HADDR_REG 0x60 +#define SEC_Q_OUTORDER_BASE_LADDR_REG 0x64 +#define SEC_Q_OUTORDER_RD_PTR_REG 0x68 +#define SEC_Q_OT_TH_REG 0x6c + +#define SEC_Q_ARUSER_CFG_REG 0x70 +#define SEC_Q_ARUSER_CFG_FA BIT(0) +#define SEC_Q_ARUSER_CFG_FNA BIT(1) +#define SEC_Q_ARUSER_CFG_RINVLD BIT(2) +#define SEC_Q_ARUSER_CFG_PKG BIT(3) + +#define SEC_Q_AWUSER_CFG_REG 0x74 +#define SEC_Q_AWUSER_CFG_FA BIT(0) +#define SEC_Q_AWUSER_CFG_FNA BIT(1) +#define SEC_Q_AWUSER_CFG_PKG BIT(2) + +#define SEC_Q_ERR_BASE_HADDR_REG 0x7c +#define SEC_Q_ERR_BASE_LADDR_REG 0x80 +#define SEC_Q_CFG_VF_NUM_REG 0x84 +#define SEC_Q_SOFT_PROC_PTR_REG 0x88 +#define SEC_Q_FAIL_INT_MSK_REG 0x300 +#define SEC_Q_FLOW_INT_MKS_REG 0x304 +#define SEC_Q_FAIL_RINT_REG 0x400 +#define SEC_Q_FLOW_RINT_REG 0x404 +#define SEC_Q_FAIL_INT_STATUS_REG 0x500 +#define SEC_Q_FLOW_INT_STATUS_REG 0x504 +#define SEC_Q_STATUS_REG 0x600 +#define SEC_Q_RD_PTR_REG 0x604 +#define SEC_Q_PRO_PTR_REG 0x608 +#define SEC_Q_OUTORDER_WR_PTR_REG 0x60c +#define SEC_Q_OT_CNT_STATUS_REG 0x610 +#define SEC_Q_INORDER_BD_NUM_ST_REG 0x650 +#define SEC_Q_INORDER_GET_FLAG_ST_REG 0x654 +#define SEC_Q_INORDER_ADD_FLAG_ST_REG 0x658 +#define SEC_Q_INORDER_TASK_INT_NUM_LEFT_ST_REG 0x65c +#define SEC_Q_RD_DONE_PTR_REG 0x660 +#define SEC_Q_CPL_Q_BD_NUM_ST_REG 0x700 +#define SEC_Q_CPL_Q_PTR_ST_REG 0x704 +#define SEC_Q_CPL_Q_H_ADDR_ST_REG 0x708 +#define SEC_Q_CPL_Q_L_ADDR_ST_REG 0x70c +#define SEC_Q_CPL_TASK_INT_NUM_LEFT_ST_REG 0x710 +#define SEC_Q_WRR_ID_CHECK_REG 0x714 +#define SEC_Q_CPLQ_FULL_CHECK_REG 0x718 +#define SEC_Q_SUCCESS_BD_CNT_REG 0x800 +#define SEC_Q_FAIL_BD_CNT_REG 0x804 +#define SEC_Q_GET_BD_CNT_REG 0x808 +#define SEC_Q_IVLD_CNT_REG 0x80c +#define SEC_Q_BD_PROC_GET_CNT_REG 0x810 +#define SEC_Q_BD_PROC_DONE_CNT_REG 0x814 +#define SEC_Q_LAT_CLR_REG 0x850 +#define SEC_Q_PKT_LAT_MAX_REG 0x854 +#define SEC_Q_PKT_LAT_AVG_REG 0x858 +#define SEC_Q_PKT_LAT_MIN_REG 0x85c +#define SEC_Q_ID_CLR_CFG_REG 0x900 +#define SEC_Q_1ST_BD_ERR_ID_REG 0x904 +#define SEC_Q_1ST_AUTH_FAIL_ID_REG 0x908 +#define SEC_Q_1ST_RD_ERR_ID_REG 0x90c +#define SEC_Q_1ST_ECC2_ERR_ID_REG 0x910 +#define SEC_Q_1ST_IVLD_ID_REG 0x914 +#define SEC_Q_1ST_BD_WR_ERR_ID_REG 0x918 +#define SEC_Q_1ST_ERR_BD_WR_ERR_ID_REG 0x91c +#define SEC_Q_1ST_BD_MAC_WR_ERR_ID_REG 0x920 + +struct sec_debug_bd_info { +#define SEC_DEBUG_BD_INFO_SOFT_ERR_CHECK_M GENMASK(22, 0) + u32 soft_err_check; +#define SEC_DEBUG_BD_INFO_HARD_ERR_CHECK_M GENMASK(9, 0) + u32 hard_err_check; + u32 icv_mac1st_word; +#define SEC_DEBUG_BD_INFO_GET_ID_M GENMASK(19, 0) + u32 sec_get_id; + /* W4---W15 */ + u32 reserv_left[12]; +}; + +struct sec_out_bd_info { +#define SEC_OUT_BD_INFO_Q_ID_M GENMASK(11, 0) +#define SEC_OUT_BD_INFO_ECC_2BIT_ERR BIT(14) + u16 data; +}; + +#define SEC_MAX_DEVICES 8 +static struct sec_dev_info *sec_devices[SEC_MAX_DEVICES]; +static DEFINE_MUTEX(sec_id_lock); + +static int sec_queue_map_io(struct sec_queue *queue) +{ + struct device *dev = queue->dev_info->dev; + struct resource *res; + + res = platform_get_resource(to_platform_device(dev), + IORESOURCE_MEM, + 2 + queue->queue_id); + if (!res) { + dev_err(dev, "Failed to get queue %u memory resource\n", + queue->queue_id); + return -ENOMEM; + } + queue->regs = ioremap(res->start, resource_size(res)); + if (!queue->regs) + return -ENOMEM; + + return 0; +} + +static void sec_queue_unmap_io(struct sec_queue *queue) +{ + iounmap(queue->regs); +} + +static int sec_queue_ar_pkgattr(struct sec_queue *queue, u32 ar_pkg) +{ + void __iomem *addr = queue->regs + SEC_Q_ARUSER_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (ar_pkg) + regval |= SEC_Q_ARUSER_CFG_PKG; + else + regval &= ~SEC_Q_ARUSER_CFG_PKG; + writel_relaxed(regval, addr); + + return 0; +} + +static int sec_queue_aw_pkgattr(struct sec_queue *queue, u32 aw_pkg) +{ + void __iomem *addr = queue->regs + SEC_Q_AWUSER_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + regval |= SEC_Q_AWUSER_CFG_PKG; + writel_relaxed(regval, addr); + + return 0; +} + +static int sec_clk_en(struct sec_dev_info *info) +{ + void __iomem *base = info->regs[SEC_COMMON]; + u32 i = 0; + + writel_relaxed(0x7, base + SEC_ALGSUB_CLK_EN_REG); + do { + usleep_range(1000, 10000); + if ((readl_relaxed(base + SEC_ALGSUB_CLK_ST_REG) & 0x7) == 0x7) + return 0; + i++; + } while (i < 10); + dev_err(info->dev, "sec clock enable fail!\n"); + + return -EIO; +} + +static int sec_clk_dis(struct sec_dev_info *info) +{ + void __iomem *base = info->regs[SEC_COMMON]; + u32 i = 0; + + writel_relaxed(0x7, base + SEC_ALGSUB_CLK_DIS_REG); + do { + usleep_range(1000, 10000); + if ((readl_relaxed(base + SEC_ALGSUB_CLK_ST_REG) & 0x7) == 0) + return 0; + i++; + } while (i < 10); + dev_err(info->dev, "sec clock disable fail!\n"); + + return -EIO; +} + +static int sec_reset_whole_module(struct sec_dev_info *info) +{ + void __iomem *base = info->regs[SEC_COMMON]; + bool is_reset, b_is_reset; + u32 i = 0; + + writel_relaxed(1, base + SEC_ALGSUB_RST_REQ_REG); + writel_relaxed(1, base + SEC_ALGSUB_BUILD_RST_REQ_REG); + while (1) { + usleep_range(1000, 10000); + is_reset = readl_relaxed(base + SEC_ALGSUB_RST_ST_REG) & + SEC_ALGSUB_RST_ST_IS_RST; + b_is_reset = readl_relaxed(base + SEC_ALGSUB_BUILD_RST_ST_REG) & + SEC_ALGSUB_BUILD_RST_ST_IS_RST; + if (is_reset && b_is_reset) + break; + i++; + if (i > 10) { + dev_err(info->dev, "Reset req failed\n"); + return -EIO; + } + } + + i = 0; + writel_relaxed(1, base + SEC_ALGSUB_RST_DREQ_REG); + writel_relaxed(1, base + SEC_ALGSUB_BUILD_RST_DREQ_REG); + while (1) { + usleep_range(1000, 10000); + is_reset = readl_relaxed(base + SEC_ALGSUB_RST_ST_REG) & + SEC_ALGSUB_RST_ST_IS_RST; + b_is_reset = readl_relaxed(base + SEC_ALGSUB_BUILD_RST_ST_REG) & + SEC_ALGSUB_BUILD_RST_ST_IS_RST; + if (!is_reset && !b_is_reset) + break; + + i++; + if (i > 10) { + dev_err(info->dev, "Reset dreq failed\n"); + return -EIO; + } + } + + return 0; +} + +static void sec_bd_endian_little(struct sec_dev_info *info) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_CTRL2_REG; + u32 regval; + + regval = readl_relaxed(addr); + regval &= ~(SEC_CTRL2_ENDIAN_BD | SEC_CTRL2_ENDIAN_BD_TYPE); + writel_relaxed(regval, addr); +} + +/* + * sec_cache_config - configure optimum cache placement + */ +static void sec_cache_config(struct sec_dev_info *info) +{ + struct iommu_domain *domain; + void __iomem *addr = info->regs[SEC_SAA] + SEC_CTRL_REG; + + domain = iommu_get_domain_for_dev(info->dev); + + /* Check that translation is occurring */ + if (domain && (domain->type & __IOMMU_DOMAIN_PAGING)) + writel_relaxed(0x44cf9e, addr); + else + writel_relaxed(0x4cfd9, addr); +} + +static void sec_data_axiwr_otsd_cfg(struct sec_dev_info *info, u32 cfg) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_CTRL2_REG; + u32 regval; + + regval = readl_relaxed(addr); + regval &= ~SEC_CTRL2_DATA_AXI_WR_OTSD_CFG_M; + regval |= (cfg << SEC_CTRL2_DATA_AXI_WR_OTSD_CFG_S) & + SEC_CTRL2_DATA_AXI_WR_OTSD_CFG_M; + writel_relaxed(regval, addr); +} + +static void sec_data_axird_otsd_cfg(struct sec_dev_info *info, u32 cfg) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_CTRL2_REG; + u32 regval; + + regval = readl_relaxed(addr); + regval &= ~SEC_CTRL2_DATA_AXI_RD_OTSD_CFG_M; + regval |= (cfg << SEC_CTRL2_DATA_AXI_RD_OTSD_CFG_S) & + SEC_CTRL2_DATA_AXI_RD_OTSD_CFG_M; + writel_relaxed(regval, addr); +} + +static void sec_clk_gate_en(struct sec_dev_info *info, bool clkgate) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_CTRL2_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (clkgate) + regval |= SEC_CTRL2_CLK_GATE_EN; + else + regval &= ~SEC_CTRL2_CLK_GATE_EN; + writel_relaxed(regval, addr); +} + +static void sec_comm_cnt_cfg(struct sec_dev_info *info, bool clr_ce) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_COMMON_CNT_CLR_CE_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (clr_ce) + regval |= SEC_COMMON_CNT_CLR_CE_CLEAR; + else + regval &= ~SEC_COMMON_CNT_CLR_CE_CLEAR; + writel_relaxed(regval, addr); +} + +static void sec_commsnap_en(struct sec_dev_info *info, bool snap_en) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_COMMON_CNT_CLR_CE_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (snap_en) + regval |= SEC_COMMON_CNT_CLR_CE_SNAP_EN; + else + regval &= ~SEC_COMMON_CNT_CLR_CE_SNAP_EN; + writel_relaxed(regval, addr); +} + +static void sec_ipv6_hashmask(struct sec_dev_info *info, u32 hash_mask[]) +{ + void __iomem *base = info->regs[SEC_SAA]; + int i; + + for (i = 0; i < 10; i++) + writel_relaxed(hash_mask[0], + base + SEC_IPV6_MASK_TABLE_X_REG(i)); +} + +static int sec_ipv4_hashmask(struct sec_dev_info *info, u32 hash_mask) +{ + if (hash_mask & SEC_HASH_IPV4_MASK) { + dev_err(info->dev, "Sec Ipv4 Hash Mask Input Error!\n "); + return -EINVAL; + } + + writel_relaxed(hash_mask, + info->regs[SEC_SAA] + SEC_IPV4_MASK_TABLE_REG); + + return 0; +} + +static void sec_set_dbg_bd_cfg(struct sec_dev_info *info, u32 cfg) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_DEBUG_BD_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + /* Always disable write back of normal bd */ + regval &= ~SEC_DEBUG_BD_CFG_WB_NORMAL; + + if (cfg) + regval &= ~SEC_DEBUG_BD_CFG_WB_EN; + else + regval |= SEC_DEBUG_BD_CFG_WB_EN; + + writel_relaxed(regval, addr); +} + +static void sec_saa_getqm_en(struct sec_dev_info *info, u32 saa_indx, u32 en) +{ + void __iomem *addr = info->regs[SEC_SAA] + SEC_SAA_BASE + + SEC_SAA_CTRL_REG(saa_indx); + u32 regval; + + regval = readl_relaxed(addr); + if (en) + regval |= SEC_SAA_CTRL_GET_QM_EN; + else + regval &= ~SEC_SAA_CTRL_GET_QM_EN; + writel_relaxed(regval, addr); +} + +static void sec_saa_int_mask(struct sec_dev_info *info, u32 saa_indx, + u32 saa_int_mask) +{ + writel_relaxed(saa_int_mask, + info->regs[SEC_SAA] + SEC_SAA_BASE + SEC_ST_INTMSK1_REG + + saa_indx * SEC_SAA_ADDR_SIZE); +} + +static void sec_streamid(struct sec_dev_info *info, int i) +{ + #define SEC_SID 0x600 + #define SEC_VMID 0 + + writel_relaxed((SEC_VMID | ((SEC_SID & 0xffff) << 8)), + info->regs[SEC_SAA] + SEC_Q_VMID_CFG_REG(i)); +} + +static void sec_queue_ar_alloc(struct sec_queue *queue, u32 alloc) +{ + void __iomem *addr = queue->regs + SEC_Q_ARUSER_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (alloc == SEC_QUEUE_AR_FROCE_ALLOC) { + regval |= SEC_Q_ARUSER_CFG_FA; + regval &= ~SEC_Q_ARUSER_CFG_FNA; + } else { + regval &= ~SEC_Q_ARUSER_CFG_FA; + regval |= SEC_Q_ARUSER_CFG_FNA; + } + + writel_relaxed(regval, addr); +} + +static void sec_queue_aw_alloc(struct sec_queue *queue, u32 alloc) +{ + void __iomem *addr = queue->regs + SEC_Q_AWUSER_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + if (alloc == SEC_QUEUE_AW_FROCE_ALLOC) { + regval |= SEC_Q_AWUSER_CFG_FA; + regval &= ~SEC_Q_AWUSER_CFG_FNA; + } else { + regval &= ~SEC_Q_AWUSER_CFG_FA; + regval |= SEC_Q_AWUSER_CFG_FNA; + } + + writel_relaxed(regval, addr); +} + +static void sec_queue_reorder(struct sec_queue *queue, bool reorder) +{ + void __iomem *base = queue->regs; + u32 regval; + + regval = readl_relaxed(base + SEC_Q_CFG_REG); + if (reorder) + regval |= SEC_Q_CFG_REORDER; + else + regval &= ~SEC_Q_CFG_REORDER; + writel_relaxed(regval, base + SEC_Q_CFG_REG); +} + +static void sec_queue_depth(struct sec_queue *queue, u32 depth) +{ + void __iomem *addr = queue->regs + SEC_Q_DEPTH_CFG_REG; + u32 regval; + + regval = readl_relaxed(addr); + regval &= ~SEC_Q_DEPTH_CFG_DEPTH_M; + regval |= (depth << SEC_Q_DEPTH_CFG_DEPTH_S) & SEC_Q_DEPTH_CFG_DEPTH_M; + + writel_relaxed(regval, addr); +} + +static void sec_queue_cmdbase_addr(struct sec_queue *queue, u64 addr) +{ + writel_relaxed(upper_32_bits(addr), queue->regs + SEC_Q_BASE_HADDR_REG); + writel_relaxed(lower_32_bits(addr), queue->regs + SEC_Q_BASE_LADDR_REG); +} + +static void sec_queue_outorder_addr(struct sec_queue *queue, u64 addr) +{ + writel_relaxed(upper_32_bits(addr), + queue->regs + SEC_Q_OUTORDER_BASE_HADDR_REG); + writel_relaxed(lower_32_bits(addr), + queue->regs + SEC_Q_OUTORDER_BASE_LADDR_REG); +} + +static void sec_queue_errbase_addr(struct sec_queue *queue, u64 addr) +{ + writel_relaxed(upper_32_bits(addr), + queue->regs + SEC_Q_ERR_BASE_HADDR_REG); + writel_relaxed(lower_32_bits(addr), + queue->regs + SEC_Q_ERR_BASE_LADDR_REG); +} + +static void sec_queue_irq_disable(struct sec_queue *queue) +{ + writel_relaxed((u32)~0, queue->regs + SEC_Q_FLOW_INT_MKS_REG); +} + +static void sec_queue_irq_enable(struct sec_queue *queue) +{ + writel_relaxed(0, queue->regs + SEC_Q_FLOW_INT_MKS_REG); +} + +static void sec_queue_abn_irq_disable(struct sec_queue *queue) +{ + writel_relaxed((u32)~0, queue->regs + SEC_Q_FAIL_INT_MSK_REG); +} + +static void sec_queue_stop(struct sec_queue *queue) +{ + disable_irq(queue->task_irq); + sec_queue_irq_disable(queue); + writel_relaxed(0x0, queue->regs + SEC_QUEUE_ENB_REG); +} + +static void sec_queue_start(struct sec_queue *queue) +{ + sec_queue_irq_enable(queue); + enable_irq(queue->task_irq); + queue->expected = 0; + writel_relaxed(SEC_Q_INIT_AND_STAT_CLEAR, queue->regs + SEC_Q_INIT_REG); + writel_relaxed(0x1, queue->regs + SEC_QUEUE_ENB_REG); +} + +static struct sec_queue *sec_alloc_queue(struct sec_dev_info *info) +{ + int i; + + mutex_lock(&info->dev_lock); + + /* Get the first idle queue in SEC device */ + for (i = 0; i < SEC_Q_NUM; i++) + if (!info->queues[i].in_use) { + info->queues[i].in_use = true; + info->queues_in_use++; + mutex_unlock(&info->dev_lock); + + return &info->queues[i]; + } + mutex_unlock(&info->dev_lock); + + return ERR_PTR(-ENODEV); +} + +static int sec_queue_free(struct sec_queue *queue) +{ + struct sec_dev_info *info = queue->dev_info; + + if (queue->queue_id >= SEC_Q_NUM) { + dev_err(info->dev, "No queue %u\n", queue->queue_id); + return -ENODEV; + } + + if (!queue->in_use) { + dev_err(info->dev, "Queue %u is idle\n", queue->queue_id); + return -ENODEV; + } + + mutex_lock(&info->dev_lock); + queue->in_use = false; + info->queues_in_use--; + mutex_unlock(&info->dev_lock); + + return 0; +} + +static irqreturn_t sec_isr_handle_th(int irq, void *q) +{ + sec_queue_irq_disable(q); + return IRQ_WAKE_THREAD; +} + +static irqreturn_t sec_isr_handle(int irq, void *q) +{ + struct sec_queue *queue = q; + struct sec_queue_ring_cmd *msg_ring = &queue->ring_cmd; + struct sec_queue_ring_cq *cq_ring = &queue->ring_cq; + struct sec_out_bd_info *outorder_msg; + struct sec_bd_info *msg; + u32 ooo_read, ooo_write; + void __iomem *base = queue->regs; + int q_id; + + ooo_read = readl(base + SEC_Q_OUTORDER_RD_PTR_REG); + ooo_write = readl(base + SEC_Q_OUTORDER_WR_PTR_REG); + outorder_msg = cq_ring->vaddr + ooo_read; + q_id = outorder_msg->data & SEC_OUT_BD_INFO_Q_ID_M; + msg = msg_ring->vaddr + q_id; + + while ((ooo_write != ooo_read) && msg->w0 & SEC_BD_W0_DONE) { + /* + * Must be before callback otherwise blocks adding other chained + * elements + */ + set_bit(q_id, queue->unprocessed); + if (q_id == queue->expected) + while (test_bit(queue->expected, queue->unprocessed)) { + clear_bit(queue->expected, queue->unprocessed); + msg = msg_ring->vaddr + queue->expected; + msg->w0 &= ~SEC_BD_W0_DONE; + msg_ring->callback(msg, + queue->shadow[queue->expected]); + queue->shadow[queue->expected] = NULL; + queue->expected = (queue->expected + 1) % + SEC_QUEUE_LEN; + atomic_dec(&msg_ring->used); + } + + ooo_read = (ooo_read + 1) % SEC_QUEUE_LEN; + writel(ooo_read, base + SEC_Q_OUTORDER_RD_PTR_REG); + ooo_write = readl(base + SEC_Q_OUTORDER_WR_PTR_REG); + outorder_msg = cq_ring->vaddr + ooo_read; + q_id = outorder_msg->data & SEC_OUT_BD_INFO_Q_ID_M; + msg = msg_ring->vaddr + q_id; + } + + sec_queue_irq_enable(queue); + + return IRQ_HANDLED; +} + +static int sec_queue_irq_init(struct sec_queue *queue) +{ + struct sec_dev_info *info = queue->dev_info; + int irq = queue->task_irq; + int ret; + + ret = request_threaded_irq(irq, sec_isr_handle_th, sec_isr_handle, + IRQF_TRIGGER_RISING, queue->name, queue); + if (ret) { + dev_err(info->dev, "request irq(%d) failed %d\n", irq, ret); + return ret; + } + disable_irq(irq); + + return 0; +} + +static int sec_queue_irq_uninit(struct sec_queue *queue) +{ + free_irq(queue->task_irq, queue); + + return 0; +} + +static struct sec_dev_info *sec_device_get(void) +{ + struct sec_dev_info *sec_dev = NULL; + struct sec_dev_info *this_sec_dev; + int least_busy_n = SEC_Q_NUM + 1; + int i; + + /* Find which one is least busy and use that first */ + for (i = 0; i < SEC_MAX_DEVICES; i++) { + this_sec_dev = sec_devices[i]; + if (this_sec_dev && + this_sec_dev->queues_in_use < least_busy_n) { + least_busy_n = this_sec_dev->queues_in_use; + sec_dev = this_sec_dev; + } + } + + return sec_dev; +} + +static struct sec_queue *sec_queue_alloc_start(struct sec_dev_info *info) +{ + struct sec_queue *queue; + + queue = sec_alloc_queue(info); + if (IS_ERR(queue)) { + dev_err(info->dev, "alloc sec queue failed! %ld\n", + PTR_ERR(queue)); + return queue; + } + + sec_queue_start(queue); + + return queue; +} + +/** + * sec_queue_alloc_start_safe - get a hw queue from appropriate instance + * + * This function does extremely simplistic load balancing. It does not take into + * account NUMA locality of the accelerator, or which cpu has requested the + * queue. Future work may focus on optimizing this in order to improve full + * machine throughput. + */ +struct sec_queue *sec_queue_alloc_start_safe(void) +{ + struct sec_dev_info *info; + struct sec_queue *queue = ERR_PTR(-ENODEV); + + mutex_lock(&sec_id_lock); + info = sec_device_get(); + if (!info) + goto unlock; + + queue = sec_queue_alloc_start(info); + +unlock: + mutex_unlock(&sec_id_lock); + + return queue; +} + +/** + * sec_queue_stop_release() - free up a hw queue for reuse + * @queue: The queue we are done with. + * + * This will stop the current queue, terminanting any transactions + * that are inflight an return it to the pool of available hw queuess + */ +int sec_queue_stop_release(struct sec_queue *queue) +{ + struct device *dev = queue->dev_info->dev; + int ret; + + sec_queue_stop(queue); + + ret = sec_queue_free(queue); + if (ret) + dev_err(dev, "Releasing queue failed %d\n", ret); + + return ret; +} + +/** + * sec_queue_empty() - Is this hardware queue currently empty. + * @queue: The queue to test + * + * We need to know if we have an empty queue for some of the chaining modes + * as if it is not empty we may need to hold the message in a software queue + * until the hw queue is drained. + */ +bool sec_queue_empty(struct sec_queue *queue) +{ + struct sec_queue_ring_cmd *msg_ring = &queue->ring_cmd; + + return !atomic_read(&msg_ring->used); +} + +/** + * sec_queue_send() - queue up a single operation in the hw queue + * @queue: The queue in which to put the message + * @msg: The message + * @ctx: Context to be put in the shadow array and passed back to cb on result. + * + * This function will return -EAGAIN if the queue is currently full. + */ +int sec_queue_send(struct sec_queue *queue, struct sec_bd_info *msg, void *ctx) +{ + struct sec_queue_ring_cmd *msg_ring = &queue->ring_cmd; + void __iomem *base = queue->regs; + u32 write, read; + + mutex_lock(&msg_ring->lock); + read = readl(base + SEC_Q_RD_PTR_REG); + write = readl(base + SEC_Q_WR_PTR_REG); + if (write == read && atomic_read(&msg_ring->used) == SEC_QUEUE_LEN) { + mutex_unlock(&msg_ring->lock); + return -EAGAIN; + } + memcpy(msg_ring->vaddr + write, msg, sizeof(*msg)); + queue->shadow[write] = ctx; + write = (write + 1) % SEC_QUEUE_LEN; + + /* Ensure content updated before queue advance */ + wmb(); + writel(write, base + SEC_Q_WR_PTR_REG); + + atomic_inc(&msg_ring->used); + mutex_unlock(&msg_ring->lock); + + return 0; +} + +bool sec_queue_can_enqueue(struct sec_queue *queue, int num) +{ + struct sec_queue_ring_cmd *msg_ring = &queue->ring_cmd; + + return SEC_QUEUE_LEN - atomic_read(&msg_ring->used) >= num; +} + +static void sec_queue_hw_init(struct sec_queue *queue) +{ + sec_queue_ar_alloc(queue, SEC_QUEUE_AR_FROCE_NOALLOC); + sec_queue_aw_alloc(queue, SEC_QUEUE_AW_FROCE_NOALLOC); + sec_queue_ar_pkgattr(queue, 1); + sec_queue_aw_pkgattr(queue, 1); + + /* Enable out of order queue */ + sec_queue_reorder(queue, true); + + /* Interrupt after a single complete element */ + writel_relaxed(1, queue->regs + SEC_Q_PROC_NUM_CFG_REG); + + sec_queue_depth(queue, SEC_QUEUE_LEN - 1); + + sec_queue_cmdbase_addr(queue, queue->ring_cmd.paddr); + + sec_queue_outorder_addr(queue, queue->ring_cq.paddr); + + sec_queue_errbase_addr(queue, queue->ring_db.paddr); + + writel_relaxed(0x100, queue->regs + SEC_Q_OT_TH_REG); + + sec_queue_abn_irq_disable(queue); + sec_queue_irq_disable(queue); + writel_relaxed(SEC_Q_INIT_AND_STAT_CLEAR, queue->regs + SEC_Q_INIT_REG); +} + +static int sec_hw_init(struct sec_dev_info *info) +{ + struct iommu_domain *domain; + u32 sec_ipv4_mask = 0; + u32 sec_ipv6_mask[10] = {}; + u32 i, ret; + + domain = iommu_get_domain_for_dev(info->dev); + + /* + * Enable all available processing unit clocks. + * Only the first cluster is usable with translations. + */ + if (domain && (domain->type & __IOMMU_DOMAIN_PAGING)) + info->num_saas = 5; + + else + info->num_saas = 10; + + writel_relaxed(GENMASK(info->num_saas - 1, 0), + info->regs[SEC_SAA] + SEC_CLK_EN_REG); + + /* 32 bit little endian */ + sec_bd_endian_little(info); + + sec_cache_config(info); + + /* Data axi port write and read outstanding config as per datasheet */ + sec_data_axiwr_otsd_cfg(info, 0x7); + sec_data_axird_otsd_cfg(info, 0x7); + + /* Enable clock gating */ + sec_clk_gate_en(info, true); + + /* Set CNT_CYC register not read clear */ + sec_comm_cnt_cfg(info, false); + + /* Enable CNT_CYC */ + sec_commsnap_en(info, false); + + writel_relaxed((u32)~0, info->regs[SEC_SAA] + SEC_FSM_MAX_CNT_REG); + + ret = sec_ipv4_hashmask(info, sec_ipv4_mask); + if (ret) { + dev_err(info->dev, "Failed to set ipv4 hashmask %d\n", ret); + return -EIO; + } + + sec_ipv6_hashmask(info, sec_ipv6_mask); + + /* do not use debug bd */ + sec_set_dbg_bd_cfg(info, 0); + + if (domain && (domain->type & __IOMMU_DOMAIN_PAGING)) { + for (i = 0; i < SEC_Q_NUM; i++) { + sec_streamid(info, i); + /* Same QoS for all queues */ + writel_relaxed(0x3f, + info->regs[SEC_SAA] + + SEC_Q_WEIGHT_CFG_REG(i)); + } + } + + for (i = 0; i < info->num_saas; i++) { + sec_saa_getqm_en(info, i, 1); + sec_saa_int_mask(info, i, 0); + } + + return 0; +} + +static void sec_hw_exit(struct sec_dev_info *info) +{ + int i; + + for (i = 0; i < SEC_MAX_SAA_NUM; i++) { + sec_saa_int_mask(info, i, (u32)~0); + sec_saa_getqm_en(info, i, 0); + } +} + +static void sec_queue_base_init(struct sec_dev_info *info, + struct sec_queue *queue, int queue_id) +{ + queue->dev_info = info; + queue->queue_id = queue_id; + snprintf(queue->name, sizeof(queue->name), + "%s_%d", dev_name(info->dev), queue->queue_id); +} + +static int sec_map_io(struct sec_dev_info *info, struct platform_device *pdev) +{ + struct resource *res; + int i; + + for (i = 0; i < SEC_NUM_ADDR_REGIONS; i++) { + res = platform_get_resource(pdev, IORESOURCE_MEM, i); + + if (!res) { + dev_err(info->dev, "Memory resource %d not found\n", i); + return -EINVAL; + } + + info->regs[i] = devm_ioremap(info->dev, res->start, + resource_size(res)); + if (!info->regs[i]) { + dev_err(info->dev, + "Memory resource %d could not be remapped\n", + i); + return -EINVAL; + } + } + + return 0; +} + +static int sec_base_init(struct sec_dev_info *info, + struct platform_device *pdev) +{ + int ret; + + ret = sec_map_io(info, pdev); + if (ret) + return ret; + + ret = sec_clk_en(info); + if (ret) + return ret; + + ret = sec_reset_whole_module(info); + if (ret) + goto sec_clk_disable; + + ret = sec_hw_init(info); + if (ret) + goto sec_clk_disable; + + return 0; + +sec_clk_disable: + sec_clk_dis(info); + + return ret; +} + +static void sec_base_exit(struct sec_dev_info *info) +{ + sec_hw_exit(info); + sec_clk_dis(info); +} + +#define SEC_Q_CMD_SIZE \ + round_up(SEC_QUEUE_LEN * sizeof(struct sec_bd_info), PAGE_SIZE) +#define SEC_Q_CQ_SIZE \ + round_up(SEC_QUEUE_LEN * sizeof(struct sec_out_bd_info), PAGE_SIZE) +#define SEC_Q_DB_SIZE \ + round_up(SEC_QUEUE_LEN * sizeof(struct sec_debug_bd_info), PAGE_SIZE) + +static int sec_queue_res_cfg(struct sec_queue *queue) +{ + struct device *dev = queue->dev_info->dev; + struct sec_queue_ring_cmd *ring_cmd = &queue->ring_cmd; + struct sec_queue_ring_cq *ring_cq = &queue->ring_cq; + struct sec_queue_ring_db *ring_db = &queue->ring_db; + int ret; + + ring_cmd->vaddr = dma_alloc_coherent(dev, SEC_Q_CMD_SIZE, + &ring_cmd->paddr, GFP_KERNEL); + if (!ring_cmd->vaddr) + return -ENOMEM; + + atomic_set(&ring_cmd->used, 0); + mutex_init(&ring_cmd->lock); + ring_cmd->callback = sec_alg_callback; + + ring_cq->vaddr = dma_alloc_coherent(dev, SEC_Q_CQ_SIZE, + &ring_cq->paddr, GFP_KERNEL); + if (!ring_cq->vaddr) { + ret = -ENOMEM; + goto err_free_ring_cmd; + } + + ring_db->vaddr = dma_alloc_coherent(dev, SEC_Q_DB_SIZE, + &ring_db->paddr, GFP_KERNEL); + if (!ring_db->vaddr) { + ret = -ENOMEM; + goto err_free_ring_cq; + } + queue->task_irq = platform_get_irq(to_platform_device(dev), + queue->queue_id * 2 + 1); + if (queue->task_irq <= 0) { + ret = -EINVAL; + goto err_free_ring_db; + } + + return 0; + +err_free_ring_db: + dma_free_coherent(dev, SEC_Q_DB_SIZE, queue->ring_db.vaddr, + queue->ring_db.paddr); +err_free_ring_cq: + dma_free_coherent(dev, SEC_Q_CQ_SIZE, queue->ring_cq.vaddr, + queue->ring_cq.paddr); +err_free_ring_cmd: + dma_free_coherent(dev, SEC_Q_CMD_SIZE, queue->ring_cmd.vaddr, + queue->ring_cmd.paddr); + + return ret; +} + +static void sec_queue_free_ring_pages(struct sec_queue *queue) +{ + struct device *dev = queue->dev_info->dev; + + dma_free_coherent(dev, SEC_Q_DB_SIZE, queue->ring_db.vaddr, + queue->ring_db.paddr); + dma_free_coherent(dev, SEC_Q_CQ_SIZE, queue->ring_cq.vaddr, + queue->ring_cq.paddr); + dma_free_coherent(dev, SEC_Q_CMD_SIZE, queue->ring_cmd.vaddr, + queue->ring_cmd.paddr); +} + +static int sec_queue_config(struct sec_dev_info *info, struct sec_queue *queue, + int queue_id) +{ + int ret; + + sec_queue_base_init(info, queue, queue_id); + + ret = sec_queue_res_cfg(queue); + if (ret) + return ret; + + ret = sec_queue_map_io(queue); + if (ret) { + dev_err(info->dev, "Queue map failed %d\n", ret); + sec_queue_free_ring_pages(queue); + return ret; + } + + sec_queue_hw_init(queue); + + return 0; +} + +static void sec_queue_unconfig(struct sec_dev_info *info, + struct sec_queue *queue) +{ + sec_queue_unmap_io(queue); + sec_queue_free_ring_pages(queue); +} + +static int sec_id_alloc(struct sec_dev_info *info) +{ + int ret = 0; + int i; + + mutex_lock(&sec_id_lock); + + for (i = 0; i < SEC_MAX_DEVICES; i++) + if (!sec_devices[i]) + break; + if (i == SEC_MAX_DEVICES) { + ret = -ENOMEM; + goto unlock; + } + info->sec_id = i; + sec_devices[info->sec_id] = info; + +unlock: + mutex_unlock(&sec_id_lock); + + return ret; +} + +static void sec_id_free(struct sec_dev_info *info) +{ + mutex_lock(&sec_id_lock); + sec_devices[info->sec_id] = NULL; + mutex_unlock(&sec_id_lock); +} + +static int sec_probe(struct platform_device *pdev) +{ + struct sec_dev_info *info; + struct device *dev = &pdev->dev; + int i, j; + int ret; + + ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); + if (ret) { + dev_err(dev, "Failed to set 64 bit dma mask %d", ret); + return -ENODEV; + } + + info = devm_kzalloc(dev, (sizeof(*info)), GFP_KERNEL); + if (!info) + return -ENOMEM; + + info->dev = dev; + mutex_init(&info->dev_lock); + + info->hw_sgl_pool = dmam_pool_create("sgl", dev, + sizeof(struct sec_hw_sgl), 64, 0); + if (!info->hw_sgl_pool) { + dev_err(dev, "Failed to create sec sgl dma pool\n"); + return -ENOMEM; + } + + ret = sec_base_init(info, pdev); + if (ret) { + dev_err(dev, "Base initialization fail! %d\n", ret); + return ret; + } + + for (i = 0; i < SEC_Q_NUM; i++) { + ret = sec_queue_config(info, &info->queues[i], i); + if (ret) + goto queues_unconfig; + + ret = sec_queue_irq_init(&info->queues[i]); + if (ret) { + sec_queue_unconfig(info, &info->queues[i]); + goto queues_unconfig; + } + } + + ret = sec_algs_register(); + if (ret) { + dev_err(dev, "Failed to register algorithms with crypto %d\n", + ret); + goto queues_unconfig; + } + + platform_set_drvdata(pdev, info); + + ret = sec_id_alloc(info); + if (ret) + goto algs_unregister; + + return 0; + +algs_unregister: + sec_algs_unregister(); +queues_unconfig: + for (j = i - 1; j >= 0; j--) { + sec_queue_irq_uninit(&info->queues[j]); + sec_queue_unconfig(info, &info->queues[j]); + } + sec_base_exit(info); + + return ret; +} + +static int sec_remove(struct platform_device *pdev) +{ + struct sec_dev_info *info = platform_get_drvdata(pdev); + int i; + + /* Unexpose as soon as possible, reuse during remove is fine */ + sec_id_free(info); + + sec_algs_unregister(); + + for (i = 0; i < SEC_Q_NUM; i++) { + sec_queue_irq_uninit(&info->queues[i]); + sec_queue_unconfig(info, &info->queues[i]); + } + + sec_base_exit(info); + + return 0; +} + +static const __maybe_unused struct of_device_id sec_match[] = { + { .compatible = "hisilicon,hip06-sec" }, + { .compatible = "hisilicon,hip07-sec" }, + {} +}; +MODULE_DEVICE_TABLE(of, sec_match); + +static const __maybe_unused struct acpi_device_id sec_acpi_match[] = { + { "HISI02C1", 0 }, + { } +}; +MODULE_DEVICE_TABLE(acpi, sec_acpi_match); + +static struct platform_driver sec_driver = { + .probe = sec_probe, + .remove = sec_remove, + .driver = { + .name = "hisi_sec_platform_driver", + .of_match_table = sec_match, + .acpi_match_table = ACPI_PTR(sec_acpi_match), + }, +}; +module_platform_driver(sec_driver); + +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("HiSilicon Security Accelerators"); +MODULE_AUTHOR("Zaibo Xu "); diff --git a/kae_driver/hisilicon/sec/sec_drv.h b/kae_driver/hisilicon/sec/sec_drv.h new file mode 100755 index 0000000..e2a50bf --- /dev/null +++ b/kae_driver/hisilicon/sec/sec_drv.h @@ -0,0 +1,428 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2016-2017 HiSilicon Limited. */ + +#ifndef _SEC_DRV_H_ +#define _SEC_DRV_H_ + +#include +#include + +#define SEC_MAX_SGE_NUM 64 +#define SEC_HW_RING_NUM 3 + +#define SEC_CMD_RING 0 +#define SEC_OUTORDER_RING 1 +#define SEC_DBG_RING 2 + +/* A reasonable length to balance memory use against flexibility */ +#define SEC_QUEUE_LEN 512 + +#define SEC_MAX_SGE_NUM 64 + +struct sec_bd_info { +#define SEC_BD_W0_T_LEN_M GENMASK(4, 0) +#define SEC_BD_W0_T_LEN_S 0 + +#define SEC_BD_W0_C_WIDTH_M GENMASK(6, 5) +#define SEC_BD_W0_C_WIDTH_S 5 +#define SEC_C_WIDTH_AES_128BIT 0 +#define SEC_C_WIDTH_AES_8BIT 1 +#define SEC_C_WIDTH_AES_1BIT 2 +#define SEC_C_WIDTH_DES_64BIT 0 +#define SEC_C_WIDTH_DES_8BIT 1 +#define SEC_C_WIDTH_DES_1BIT 2 + +#define SEC_BD_W0_C_MODE_M GENMASK(9, 7) +#define SEC_BD_W0_C_MODE_S 7 +#define SEC_C_MODE_ECB 0 +#define SEC_C_MODE_CBC 1 +#define SEC_C_MODE_CTR 4 +#define SEC_C_MODE_CCM 5 +#define SEC_C_MODE_GCM 6 +#define SEC_C_MODE_XTS 7 + +#define SEC_BD_W0_SEQ BIT(10) +#define SEC_BD_W0_DE BIT(11) +#define SEC_BD_W0_DAT_SKIP_M GENMASK(13, 12) +#define SEC_BD_W0_DAT_SKIP_S 12 +#define SEC_BD_W0_C_GRAN_SIZE_19_16_M GENMASK(17, 14) +#define SEC_BD_W0_C_GRAN_SIZE_19_16_S 14 + +#define SEC_BD_W0_CIPHER_M GENMASK(19, 18) +#define SEC_BD_W0_CIPHER_S 18 +#define SEC_CIPHER_NULL 0 +#define SEC_CIPHER_ENCRYPT 1 +#define SEC_CIPHER_DECRYPT 2 + +#define SEC_BD_W0_AUTH_M GENMASK(21, 20) +#define SEC_BD_W0_AUTH_S 20 +#define SEC_AUTH_NULL 0 +#define SEC_AUTH_MAC 1 +#define SEC_AUTH_VERIF 2 + +#define SEC_BD_W0_AI_GEN BIT(22) +#define SEC_BD_W0_CI_GEN BIT(23) +#define SEC_BD_W0_NO_HPAD BIT(24) +#define SEC_BD_W0_HM_M GENMASK(26, 25) +#define SEC_BD_W0_HM_S 25 +#define SEC_BD_W0_ICV_OR_SKEY_EN_M GENMASK(28, 27) +#define SEC_BD_W0_ICV_OR_SKEY_EN_S 27 + +/* Multi purpose field - gran size bits for send, flag for recv */ +#define SEC_BD_W0_FLAG_M GENMASK(30, 29) +#define SEC_BD_W0_C_GRAN_SIZE_21_20_M GENMASK(30, 29) +#define SEC_BD_W0_FLAG_S 29 +#define SEC_BD_W0_C_GRAN_SIZE_21_20_S 29 + +#define SEC_BD_W0_DONE BIT(31) + u32 w0; + +#define SEC_BD_W1_AUTH_GRAN_SIZE_M GENMASK(21, 0) +#define SEC_BD_W1_AUTH_GRAN_SIZE_S 0 +#define SEC_BD_W1_M_KEY_EN BIT(22) +#define SEC_BD_W1_BD_INVALID BIT(23) +#define SEC_BD_W1_ADDR_TYPE BIT(24) + +#define SEC_BD_W1_A_ALG_M GENMASK(28, 25) +#define SEC_BD_W1_A_ALG_S 25 +#define SEC_A_ALG_SHA1 0 +#define SEC_A_ALG_SHA256 1 +#define SEC_A_ALG_MD5 2 +#define SEC_A_ALG_SHA224 3 +#define SEC_A_ALG_HMAC_SHA1 8 +#define SEC_A_ALG_HMAC_SHA224 10 +#define SEC_A_ALG_HMAC_SHA256 11 +#define SEC_A_ALG_HMAC_MD5 12 +#define SEC_A_ALG_AES_XCBC 13 +#define SEC_A_ALG_AES_CMAC 14 + +#define SEC_BD_W1_C_ALG_M GENMASK(31, 29) +#define SEC_BD_W1_C_ALG_S 29 +#define SEC_C_ALG_DES 0 +#define SEC_C_ALG_3DES 1 +#define SEC_C_ALG_AES 2 + + u32 w1; + +#define SEC_BD_W2_C_GRAN_SIZE_15_0_M GENMASK(15, 0) +#define SEC_BD_W2_C_GRAN_SIZE_15_0_S 0 +#define SEC_BD_W2_GRAN_NUM_M GENMASK(31, 16) +#define SEC_BD_W2_GRAN_NUM_S 16 + u32 w2; + +#define SEC_BD_W3_AUTH_LEN_OFFSET_M GENMASK(9, 0) +#define SEC_BD_W3_AUTH_LEN_OFFSET_S 0 +#define SEC_BD_W3_CIPHER_LEN_OFFSET_M GENMASK(19, 10) +#define SEC_BD_W3_CIPHER_LEN_OFFSET_S 10 +#define SEC_BD_W3_MAC_LEN_M GENMASK(24, 20) +#define SEC_BD_W3_MAC_LEN_S 20 +#define SEC_BD_W3_A_KEY_LEN_M GENMASK(29, 25) +#define SEC_BD_W3_A_KEY_LEN_S 25 +#define SEC_BD_W3_C_KEY_LEN_M GENMASK(31, 30) +#define SEC_BD_W3_C_KEY_LEN_S 30 +#define SEC_KEY_LEN_AES_128 0 +#define SEC_KEY_LEN_AES_192 1 +#define SEC_KEY_LEN_AES_256 2 +#define SEC_KEY_LEN_DES 1 +#define SEC_KEY_LEN_3DES_3_KEY 1 +#define SEC_KEY_LEN_3DES_2_KEY 3 + u32 w3; + + /* W4,5 */ + union { + u32 authkey_addr_lo; + u32 authiv_addr_lo; + }; + union { + u32 authkey_addr_hi; + u32 authiv_addr_hi; + }; + + /* W6,7 */ + u32 cipher_key_addr_lo; + u32 cipher_key_addr_hi; + + /* W8,9 */ + u32 cipher_iv_addr_lo; + u32 cipher_iv_addr_hi; + + /* W10,11 */ + u32 data_addr_lo; + u32 data_addr_hi; + + /* W12,13 */ + u32 mac_addr_lo; + u32 mac_addr_hi; + + /* W14,15 */ + u32 cipher_destin_addr_lo; + u32 cipher_destin_addr_hi; +}; + +enum sec_mem_region { + SEC_COMMON = 0, + SEC_SAA, + SEC_NUM_ADDR_REGIONS +}; + +#define SEC_NAME_SIZE 64 +#define SEC_Q_NUM 16 + + +/** + * struct sec_queue_ring_cmd - store information about a SEC HW cmd ring + * @used: Local counter used to cheaply establish if the ring is empty. + * @lock: Protect against simultaneous adjusting of the read and write pointers. + * @vaddr: Virtual address for the ram pages used for the ring. + * @paddr: Physical address of the dma mapped region of ram used for the ring. + * @callback: Callback function called on a ring element completing. + */ +struct sec_queue_ring_cmd { + atomic_t used; + struct mutex lock; + struct sec_bd_info *vaddr; + dma_addr_t paddr; + void (*callback)(struct sec_bd_info *resp, void *ctx); +}; + +struct sec_debug_bd_info; +struct sec_queue_ring_db { + struct sec_debug_bd_info *vaddr; + dma_addr_t paddr; +}; + +struct sec_out_bd_info; +struct sec_queue_ring_cq { + struct sec_out_bd_info *vaddr; + dma_addr_t paddr; +}; + +struct sec_dev_info; + +enum sec_cipher_alg { + SEC_C_DES_ECB_64, + SEC_C_DES_CBC_64, + + SEC_C_3DES_ECB_192_3KEY, + SEC_C_3DES_ECB_192_2KEY, + + SEC_C_3DES_CBC_192_3KEY, + SEC_C_3DES_CBC_192_2KEY, + + SEC_C_AES_ECB_128, + SEC_C_AES_ECB_192, + SEC_C_AES_ECB_256, + + SEC_C_AES_CBC_128, + SEC_C_AES_CBC_192, + SEC_C_AES_CBC_256, + + SEC_C_AES_CTR_128, + SEC_C_AES_CTR_192, + SEC_C_AES_CTR_256, + + SEC_C_AES_XTS_128, + SEC_C_AES_XTS_256, + + SEC_C_NULL, +}; + +/** + * struct sec_alg_tfm_ctx - hardware specific tranformation context + * @cipher_alg: Cipher algorithm enabled include encryption mode. + * @key: Key storage if required. + * @pkey: DMA address for the key storage. + * @req_template: Request template to save time on setup. + * @queue: The hardware queue associated with this tfm context. + * @lock: Protect key and pkey to ensure they are consistent + * @auth_buf: Current context buffer for auth operations. + * @backlog: The backlog queue used for cases where our buffers aren't + * large enough. + */ +struct sec_alg_tfm_ctx { + enum sec_cipher_alg cipher_alg; + u8 *key; + dma_addr_t pkey; + struct sec_bd_info req_template; + struct sec_queue *queue; + struct mutex lock; + u8 *auth_buf; + struct list_head backlog; +}; + +/** + * struct sec_request - data associate with a single crypto request + * @elements: List of subparts of this request (hardware size restriction) + * @num_elements: The number of subparts (used as an optimization) + * @lock: Protect elements of this structure against concurrent change. + * @tfm_ctx: hardware specific context. + * @len_in: length of in sgl from upper layers + * @len_out: length of out sgl from upper layers + * @dma_iv: initialization vector - phsyical address + * @err: store used to track errors across subelements of this request. + * @req_base: pointer to base element of associate crypto context. + * This is needed to allow shared handling skcipher, ahash etc. + * @cb: completion callback. + * @backlog_head: list head to allow backlog maintenance. + * + * The hardware is limited in the maximum size of data that it can + * process from a single BD. Typically this is fairly large (32MB) + * but still requires the complexity of splitting the incoming + * skreq up into a number of elements complete with appropriate + * iv chaining. + */ +struct sec_request { + struct list_head elements; + int num_elements; + struct mutex lock; + struct sec_alg_tfm_ctx *tfm_ctx; + int len_in; + int len_out; + dma_addr_t dma_iv; + int err; + struct crypto_async_request *req_base; + void (*cb)(struct sec_bd_info *resp, struct crypto_async_request *req); + struct list_head backlog_head; +}; + +/** + * struct sec_request_el - A subpart of a request. + * @head: allow us to attach this to the list in the sec_request + * @req: hardware block descriptor corresponding to this request subpart + * @in: hardware sgl for input - virtual address + * @dma_in: hardware sgl for input - physical address + * @sgl_in: scatterlist for this request subpart + * @out: hardware sgl for output - virtual address + * @dma_out: hardware sgl for output - physical address + * @sgl_out: scatterlist for this request subpart + * @sec_req: The request which this subpart forms a part of + * @el_length: Number of bytes in this subpart. Needed to locate + * last ivsize chunk for iv chaining. + */ +struct sec_request_el { + struct list_head head; + struct sec_bd_info req; + struct sec_hw_sgl *in; + dma_addr_t dma_in; + struct scatterlist *sgl_in; + struct sec_hw_sgl *out; + dma_addr_t dma_out; + struct scatterlist *sgl_out; + struct sec_request *sec_req; + size_t el_length; +}; + +/** + * struct sec_queue - All the information about a HW queue + * @dev_info: The parent SEC device to which this queue belongs. + * @task_irq: Completion interrupt for the queue. + * @name: Human readable queue description also used as irq name. + * @ring: The several HW rings associated with one queue. + * @regs: The iomapped device registers + * @queue_id: Index of the queue used for naming and resource selection. + * @in_use: Flag to say if the queue is in use. + * @expected: The next expected element to finish assuming we were in order. + * @uprocessed: A bitmap to track which OoO elements are done but not handled. + * @softqueue: A software queue used when chaining requirements prevent direct + * use of the hardware queues. + * @havesoftqueue: A flag to say we have a queues - as we may need one for the + * current mode. + * @queuelock: Protect the soft queue from concurrent changes to avoid some + * potential loss of data races. + * @shadow: Pointers back to the shadow copy of the hardware ring element + * need because we can't store any context reference in the bd element. + */ +struct sec_queue { + struct sec_dev_info *dev_info; + int task_irq; + char name[SEC_NAME_SIZE]; + struct sec_queue_ring_cmd ring_cmd; + struct sec_queue_ring_cq ring_cq; + struct sec_queue_ring_db ring_db; + void __iomem *regs; + u32 queue_id; + bool in_use; + int expected; + + DECLARE_BITMAP(unprocessed, SEC_QUEUE_LEN); + DECLARE_KFIFO_PTR(softqueue, typeof(struct sec_request_el *)); + bool havesoftqueue; + spinlock_t queuelock; + void *shadow[SEC_QUEUE_LEN]; +}; + +/** + * struct sec_hw_sge: Track each of the 64 element SEC HW SGL entries + * @buf: The IOV dma address for this entry. + * @len: Length of this IOV. + * @pad: Reserved space. + */ +struct sec_hw_sge { + dma_addr_t buf; + unsigned int len; + unsigned int pad; +}; + +/** + * struct sec_hw_sgl: One hardware SGL entry. + * @next_sgl: The next entry if we need to chain dma address. Null if last. + * @entry_sum_in_chain: The full count of SGEs - only matters for first SGL. + * @entry_sum_in_sgl: The number of SGEs in this SGL element. + * @flag: Unused in skciphers. + * @serial_num: Unsued in skciphers. + * @cpuid: Currently unused. + * @data_bytes_in_sgl: Count of bytes from all SGEs in this SGL. + * @next: Virtual address used to stash the next sgl - useful in completion. + * @reserved: A reserved field not currently used. + * @sge_entries: The (up to) 64 Scatter Gather Entries, representing IOVs. + * @node: Currently unused. + */ +struct sec_hw_sgl { + dma_addr_t next_sgl; + u16 entry_sum_in_chain; + u16 entry_sum_in_sgl; + u32 flag; + u64 serial_num; + u32 cpuid; + u32 data_bytes_in_sgl; + struct sec_hw_sgl *next; + u64 reserved; + struct sec_hw_sge sge_entries[SEC_MAX_SGE_NUM]; + u8 node[16]; +}; + +struct dma_pool; + +/** + * struct sec_dev_info: The full SEC unit comprising queues and processors. + * @sec_id: Index used to track which SEC this is when more than one is present. + * @num_saas: The number of backed processors enabled. + * @regs: iomapped register regions shared by whole SEC unit. + * @dev_lock: Protects concurrent queue allocation / freeing for the SEC. + * @queues: The 16 queues that this SEC instance provides. + * @dev: Device pointer. + * @hw_sgl_pool: DMA pool used to mimise mapping for the scatter gather lists. + */ +struct sec_dev_info { + int sec_id; + int num_saas; + void __iomem *regs[SEC_NUM_ADDR_REGIONS]; + struct mutex dev_lock; + int queues_in_use; + struct sec_queue queues[SEC_Q_NUM]; + struct device *dev; + struct dma_pool *hw_sgl_pool; +}; + +int sec_queue_send(struct sec_queue *queue, struct sec_bd_info *msg, void *ctx); +bool sec_queue_can_enqueue(struct sec_queue *queue, int num); +int sec_queue_stop_release(struct sec_queue *queue); +struct sec_queue *sec_queue_alloc_start_safe(void); +bool sec_queue_empty(struct sec_queue *queue); + +/* Algorithm specific elements from sec_algs.c */ +void sec_alg_callback(struct sec_bd_info *resp, void *ctx); +int sec_algs_register(void); +void sec_algs_unregister(void); + +#endif /* _SEC_DRV_H_ */ diff --git a/kae_driver/hisilicon/sec2/Makefile b/kae_driver/hisilicon/sec2/Makefile new file mode 100755 index 0000000..b4f6cf1 --- /dev/null +++ b/kae_driver/hisilicon/sec2/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_SEC2) += hisi_sec2.o +hisi_sec2-objs = sec_main.o sec_crypto.o diff --git a/kae_driver/hisilicon/sec2/sec.h b/kae_driver/hisilicon/sec2/sec.h new file mode 100755 index 0000000..5059f4b --- /dev/null +++ b/kae_driver/hisilicon/sec2/sec.h @@ -0,0 +1,228 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ + +#ifndef __HISI_SEC_V2_H +#define __HISI_SEC_V2_H + +#include "../hisi_acc_qm.h" +#include "sec_crypto.h" + +/* Algorithm resource per hardware SEC queue */ +struct sec_alg_res { + u8 *pbuf; + dma_addr_t pbuf_dma; + u8 *c_ivin; + dma_addr_t c_ivin_dma; + u8 *a_ivin; + dma_addr_t a_ivin_dma; + u8 *out_mac; + dma_addr_t out_mac_dma; + u16 depth; +}; + +/* Cipher request of SEC private */ +struct sec_cipher_req { + struct hisi_acc_hw_sgl *c_out; + dma_addr_t c_out_dma; + u8 *c_ivin; + dma_addr_t c_ivin_dma; + struct skcipher_request *sk_req; + u32 c_len; + bool encrypt; +}; + +struct sec_aead_req { + u8 *out_mac; + dma_addr_t out_mac_dma; + u8 *a_ivin; + dma_addr_t a_ivin_dma; + struct aead_request *aead_req; +}; + +/* SEC request of Crypto */ +struct sec_req { + union { + struct sec_sqe sec_sqe; + struct sec_sqe3 sec_sqe3; + }; + struct sec_ctx *ctx; + struct sec_qp_ctx *qp_ctx; + + /** + * Common parameter of the SEC request. + */ + struct hisi_acc_hw_sgl *in; + dma_addr_t in_dma; + struct sec_cipher_req c_req; + struct sec_aead_req aead_req; + struct list_head backlog_head; + + int err_type; + int req_id; + u32 flag; + + /* Status of the SEC request */ + bool fake_busy; + bool use_pbuf; +}; + +/** + * struct sec_req_op - Operations for SEC request + * @buf_map: DMA map the SGL buffers of the request + * @buf_unmap: DMA unmap the SGL buffers of the request + * @bd_fill: Fill the SEC queue BD + * @bd_send: Send the SEC BD into the hardware queue + * @callback: Call back for the request + * @process: Main processing logic of Skcipher + */ +struct sec_req_op { + int (*buf_map)(struct sec_ctx *ctx, struct sec_req *req); + void (*buf_unmap)(struct sec_ctx *ctx, struct sec_req *req); + void (*do_transfer)(struct sec_ctx *ctx, struct sec_req *req); + int (*bd_fill)(struct sec_ctx *ctx, struct sec_req *req); + int (*bd_send)(struct sec_ctx *ctx, struct sec_req *req); + void (*callback)(struct sec_ctx *ctx, struct sec_req *req, int err); + int (*process)(struct sec_ctx *ctx, struct sec_req *req); +}; + +/* SEC auth context */ +struct sec_auth_ctx { + dma_addr_t a_key_dma; + u8 *a_key; + u8 a_key_len; + u8 mac_len; + u8 a_alg; + bool fallback; + struct crypto_shash *hash_tfm; + struct crypto_aead *fallback_aead_tfm; +}; + +/* SEC cipher context which cipher's relatives */ +struct sec_cipher_ctx { + u8 *c_key; + dma_addr_t c_key_dma; + sector_t iv_offset; + u32 c_gran_size; + u32 ivsize; + u8 c_mode; + u8 c_alg; + u8 c_key_len; + + /* add software support */ + bool fallback; + struct crypto_sync_skcipher *fbtfm; +}; + +/* SEC queue context which defines queue's relatives */ +struct sec_qp_ctx { + struct hisi_qp *qp; + struct sec_req **req_list; + struct idr req_idr; + struct sec_alg_res *res; + struct sec_ctx *ctx; + spinlock_t req_lock; + struct list_head backlog; + struct hisi_acc_sgl_pool *c_in_pool; + struct hisi_acc_sgl_pool *c_out_pool; +}; + +enum sec_alg_type { + SEC_SKCIPHER, + SEC_AEAD +}; + +/* SEC Crypto TFM context which defines queue and cipher .etc relatives */ +struct sec_ctx { + struct sec_qp_ctx *qp_ctx; + struct sec_dev *sec; + const struct sec_req_op *req_op; + struct hisi_qp **qps; + + /* Half queues for encipher, and half for decipher */ + u32 hlf_q_num; + + /* Threshold for fake busy, trigger to return -EBUSY to user */ + u32 fake_req_limit; + + /* Current cyclic index to select a queue for encipher */ + atomic_t enc_qcyclic; + + /* Current cyclic index to select a queue for decipher */ + atomic_t dec_qcyclic; + + enum sec_alg_type alg_type; + bool pbuf_supported; + struct sec_cipher_ctx c_ctx; + struct sec_auth_ctx a_ctx; + u8 type_supported; + struct device *dev; +}; + + +enum sec_debug_file_index { + SEC_CLEAR_ENABLE, + SEC_DEBUG_FILE_NUM, +}; + +struct sec_debug_file { + enum sec_debug_file_index index; + spinlock_t lock; + struct hisi_qm *qm; +}; + +struct sec_dfx { + atomic64_t send_cnt; + atomic64_t recv_cnt; + atomic64_t send_busy_cnt; + atomic64_t recv_busy_cnt; + atomic64_t err_bd_cnt; + atomic64_t invalid_req_cnt; + atomic64_t done_flag_cnt; +}; + +struct sec_debug { + struct sec_dfx dfx; + struct sec_debug_file files[SEC_DEBUG_FILE_NUM]; +}; + +struct sec_dev { + struct hisi_qm qm; + struct sec_debug debug; + u32 ctx_q_num; + bool iommu_used; +}; + +enum sec_cap_type { + SEC_QM_NFE_MASK_CAP = 0x0, + SEC_QM_RESET_MASK_CAP, + SEC_QM_OOO_SHUTDOWN_MASK_CAP, + SEC_QM_CE_MASK_CAP, + SEC_NFE_MASK_CAP, + SEC_RESET_MASK_CAP, + SEC_OOO_SHUTDOWN_MASK_CAP, + SEC_CE_MASK_CAP, + SEC_CLUSTER_NUM_CAP, + SEC_CORE_TYPE_NUM_CAP, + SEC_CORE_NUM_CAP, + SEC_CORES_PER_CLUSTER_NUM_CAP, + SEC_CORE_ENABLE_BITMAP, + SEC_DRV_ALG_BITMAP_LOW, + SEC_DRV_ALG_BITMAP_HIGH, + SEC_DEV_ALG_BITMAP_LOW, + SEC_DEV_ALG_BITMAP_HIGH, + SEC_CORE1_ALG_BITMAP_LOW, + SEC_CORE1_ALG_BITMAP_HIGH, + SEC_CORE2_ALG_BITMAP_LOW, + SEC_CORE2_ALG_BITMAP_HIGH, + SEC_CORE3_ALG_BITMAP_LOW, + SEC_CORE3_ALG_BITMAP_HIGH, + SEC_CORE4_ALG_BITMAP_LOW, + SEC_CORE4_ALG_BITMAP_HIGH, +}; + +void sec_destroy_qps(struct hisi_qp **qps, int qp_num); +struct hisi_qp **sec_create_qps(void); +int sec_register_to_crypto(struct hisi_qm *qm); +void sec_unregister_from_crypto(struct hisi_qm *qm); +u64 sec_get_alg_bitmap(struct hisi_qm *qm, u32 high, u32 low); +#endif diff --git a/kae_driver/hisilicon/sec2/sec_crypto.c b/kae_driver/hisilicon/sec2/sec_crypto.c new file mode 100755 index 0000000..84ae8dd --- /dev/null +++ b/kae_driver/hisilicon/sec2/sec_crypto.c @@ -0,0 +1,2569 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "sec.h" +#include "sec_crypto.h" + +#define SEC_PRIORITY 4001 +#define SEC_XTS_MIN_KEY_SIZE (2 * AES_MIN_KEY_SIZE) +#define SEC_XTS_MID_KEY_SIZE (3 * AES_MIN_KEY_SIZE) +#define SEC_XTS_MAX_KEY_SIZE (2 * AES_MAX_KEY_SIZE) +#define SEC_DES3_2KEY_SIZE (2 * DES_KEY_SIZE) +#define SEC_DES3_3KEY_SIZE (3 * DES_KEY_SIZE) + +/* SEC sqe(bd) bit operational relative MACRO */ +#define SEC_DE_OFFSET 1 +#define SEC_CIPHER_OFFSET 4 +#define SEC_SCENE_OFFSET 3 +#define SEC_DST_SGL_OFFSET 2 +#define SEC_SRC_SGL_OFFSET 7 +#define SEC_CKEY_OFFSET 9 +#define SEC_CMODE_OFFSET 12 +#define SEC_AKEY_OFFSET 5 +#define SEC_AEAD_ALG_OFFSET 11 +#define SEC_AUTH_OFFSET 6 + +#define SEC_DE_OFFSET_V3 9 +#define SEC_SCENE_OFFSET_V3 5 +#define SEC_CKEY_OFFSET_V3 13 +#define SEC_CTR_CNT_OFFSET 25 +#define SEC_CTR_CNT_ROLLOVER 2 +#define SEC_SRC_SGL_OFFSET_V3 11 +#define SEC_DST_SGL_OFFSET_V3 14 +#define SEC_CALG_OFFSET_V3 4 +#define SEC_AKEY_OFFSET_V3 9 +#define SEC_MAC_OFFSET_V3 4 +#define SEC_AUTH_ALG_OFFSET_V3 15 +#define SEC_CIPHER_AUTH_V3 0xbf +#define SEC_AUTH_CIPHER_V3 0x40 +#define SEC_FLAG_OFFSET 7 +#define SEC_FLAG_MASK 0x0780 +#define SEC_TYPE_MASK 0x0F +#define SEC_DONE_MASK 0x0001 +#define SEC_ICV_MASK 0x000E +#define SEC_SQE_LEN_RATE_MASK 0x3 + +#define SEC_TOTAL_IV_SZ(depth) (SEC_IV_SIZE * (depth)) +#define SEC_SGL_SGE_NR 128 +#define SEC_CIPHER_AUTH 0xfe +#define SEC_AUTH_CIPHER 0x1 +#define SEC_MAX_MAC_LEN 64 +#define SEC_MAX_AAD_LEN 65535 +#define SEC_MAX_CCM_AAD_LEN 65279 +#define SEC_TOTAL_MAC_SZ(depth) (SEC_MAX_MAC_LEN * (depth)) + +#define SEC_PBUF_SZ 512 +#define SEC_PBUF_IV_OFFSET SEC_PBUF_SZ +#define SEC_PBUF_MAC_OFFSET (SEC_PBUF_SZ + SEC_IV_SIZE) +#define SEC_PBUF_PKG (SEC_PBUF_SZ + SEC_IV_SIZE + \ + SEC_MAX_MAC_LEN * 2) +#define SEC_PBUF_NUM (PAGE_SIZE / SEC_PBUF_PKG) +#define SEC_PBUF_PAGE_NUM(depth) ((depth) / SEC_PBUF_NUM) +#define SEC_PBUF_LEFT_SZ(depth) (SEC_PBUF_PKG * ((depth) - \ + SEC_PBUF_PAGE_NUM(depth) * SEC_PBUF_NUM)) +#define SEC_TOTAL_PBUF_SZ(depth) (PAGE_SIZE * SEC_PBUF_PAGE_NUM(depth) + \ + SEC_PBUF_LEFT_SZ(depth)) + +#define SEC_SQE_LEN_RATE 4 +#define SEC_SQE_CFLAG 2 +#define SEC_SQE_AEAD_FLAG 3 +#define SEC_SQE_DONE 0x1 +#define SEC_ICV_ERR 0x2 +#define MIN_MAC_LEN 4 +#define MAC_LEN_MASK 0x1U +#define MAX_INPUT_DATA_LEN 0xFFFE00 +#define BITS_MASK 0xFF +#define BYTE_BITS 0x8 +#define SEC_XTS_NAME_SZ 0x3 +#define IV_CM_CAL_NUM 2 +#define IV_CL_MASK 0x7 +#define IV_CL_MIN 2 +#define IV_CL_MID 4 +#define IV_CL_MAX 8 +#define IV_FLAGS_OFFSET 0x6 +#define IV_CM_OFFSET 0x3 +#define IV_LAST_BYTE1 1 +#define IV_LAST_BYTE2 2 +#define IV_LAST_BYTE_MASK 0xFF +#define IV_CTR_INIT 0x1 +#define IV_BYTE_OFFSET 0x8 + +struct sec_skcipher { + u64 alg_msk; + struct skcipher_alg alg; +}; + +struct sec_aead { + u64 alg_msk; + struct aead_alg alg; +}; + +/* Get an en/de-cipher queue cyclically to balance load over queues of TFM */ +static inline int sec_alloc_queue_id(struct sec_ctx *ctx, struct sec_req *req) +{ + if (req->c_req.encrypt) + return (u32)atomic_inc_return(&ctx->enc_qcyclic) % + ctx->hlf_q_num; + + return (u32)atomic_inc_return(&ctx->dec_qcyclic) % ctx->hlf_q_num + + ctx->hlf_q_num; +} + +static inline void sec_free_queue_id(struct sec_ctx *ctx, struct sec_req *req) +{ + if (req->c_req.encrypt) + atomic_dec(&ctx->enc_qcyclic); + else + atomic_dec(&ctx->dec_qcyclic); +} + +static int sec_alloc_req_id(struct sec_req *req, struct sec_qp_ctx *qp_ctx) +{ + int req_id; + + spin_lock_bh(&qp_ctx->req_lock); + req_id = idr_alloc_cyclic(&qp_ctx->req_idr, NULL, 0, qp_ctx->qp->sq_depth, GFP_ATOMIC); + spin_unlock_bh(&qp_ctx->req_lock); + if (unlikely(req_id < 0)) { + dev_err(req->ctx->dev, "alloc req id fail!\n"); + return req_id; + } + + req->qp_ctx = qp_ctx; + qp_ctx->req_list[req_id] = req; + + return req_id; +} + +static void sec_free_req_id(struct sec_req *req) +{ + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + int req_id = req->req_id; + + if (unlikely(req_id < 0 || req_id >= qp_ctx->qp->sq_depth)) { + dev_err(req->ctx->dev, "free request id invalid!\n"); + return; + } + + qp_ctx->req_list[req_id] = NULL; + req->qp_ctx = NULL; + + spin_lock_bh(&qp_ctx->req_lock); + idr_remove(&qp_ctx->req_idr, req_id); + spin_unlock_bh(&qp_ctx->req_lock); +} + +static u8 pre_parse_finished_bd(struct bd_status *status, void *resp) +{ + struct sec_sqe *bd = resp; + + status->done = le16_to_cpu(bd->type2.done_flag) & SEC_DONE_MASK; + status->icv = (le16_to_cpu(bd->type2.done_flag) & SEC_ICV_MASK) >> 1; + status->flag = (le16_to_cpu(bd->type2.done_flag) & + SEC_FLAG_MASK) >> SEC_FLAG_OFFSET; + status->tag = le16_to_cpu(bd->type2.tag); + status->err_type = bd->type2.error_type; + + return bd->type_cipher_auth & SEC_TYPE_MASK; +} + +static u8 pre_parse_finished_bd3(struct bd_status *status, void *resp) +{ + struct sec_sqe3 *bd3 = resp; + + status->done = le16_to_cpu(bd3->done_flag) & SEC_DONE_MASK; + status->icv = (le16_to_cpu(bd3->done_flag) & SEC_ICV_MASK) >> 1; + status->flag = (le16_to_cpu(bd3->done_flag) & + SEC_FLAG_MASK) >> SEC_FLAG_OFFSET; + status->tag = le64_to_cpu(bd3->tag); + status->err_type = bd3->error_type; + + return le32_to_cpu(bd3->bd_param) & SEC_TYPE_MASK; +} + +static int sec_cb_status_check(struct sec_req *req, + struct bd_status *status) +{ + struct sec_ctx *ctx = req->ctx; + + if (unlikely(req->err_type || status->done != SEC_SQE_DONE)) { + dev_err_ratelimited(ctx->dev, "err_type[%d], done[%u]\n", + req->err_type, status->done); + return -EIO; + } + + if (unlikely(ctx->alg_type == SEC_SKCIPHER)) { + if (unlikely(status->flag != SEC_SQE_CFLAG)) { + dev_err_ratelimited(ctx->dev, "flag[%u]\n", + status->flag); + return -EIO; + } + } else if (unlikely(ctx->alg_type == SEC_AEAD)) { + if (unlikely(status->flag != SEC_SQE_AEAD_FLAG || + status->icv == SEC_ICV_ERR)) { + dev_err_ratelimited(ctx->dev, + "flag[%u], icv[%u]\n", + status->flag, status->icv); + return -EBADMSG; + } + } + + return 0; +} + +static void sec_req_cb(struct hisi_qp *qp, void *resp) +{ + struct sec_qp_ctx *qp_ctx = qp->qp_ctx; + struct sec_dfx *dfx = &qp_ctx->ctx->sec->debug.dfx; + u8 type_supported = qp_ctx->ctx->type_supported; + struct bd_status status; + struct sec_ctx *ctx; + struct sec_req *req; + int err; + u8 type; + + if (type_supported == SEC_BD_TYPE2) { + type = pre_parse_finished_bd(&status, resp); + req = qp_ctx->req_list[status.tag]; + } else { + type = pre_parse_finished_bd3(&status, resp); + req = (void *)(uintptr_t)status.tag; + } + + if (unlikely(type != type_supported)) { + atomic64_inc(&dfx->err_bd_cnt); + pr_err("err bd type [%u]\n", type); + return; + } + + if (unlikely(!req)) { + atomic64_inc(&dfx->invalid_req_cnt); + atomic_inc(&qp->qp_status.used); + return; + } + + req->err_type = status.err_type; + ctx = req->ctx; + err = sec_cb_status_check(req, &status); + if (err) + atomic64_inc(&dfx->done_flag_cnt); + + atomic64_inc(&dfx->recv_cnt); + + ctx->req_op->buf_unmap(ctx, req); + + ctx->req_op->callback(ctx, req, err); +} + +static int sec_bd_send(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + int ret; + + if (ctx->fake_req_limit <= + atomic_read(&qp_ctx->qp->qp_status.used) && + !(req->flag & CRYPTO_TFM_REQ_MAY_BACKLOG)) + return -EBUSY; + + spin_lock_bh(&qp_ctx->req_lock); + ret = hisi_qp_send(qp_ctx->qp, &req->sec_sqe); + + if (ctx->fake_req_limit <= + atomic_read(&qp_ctx->qp->qp_status.used) && !ret) { + list_add_tail(&req->backlog_head, &qp_ctx->backlog); + atomic64_inc(&ctx->sec->debug.dfx.send_cnt); + atomic64_inc(&ctx->sec->debug.dfx.send_busy_cnt); + spin_unlock_bh(&qp_ctx->req_lock); + return -EBUSY; + } + spin_unlock_bh(&qp_ctx->req_lock); + + if (unlikely(ret == -EBUSY)) + return -ENOBUFS; + + if (likely(!ret)) { + ret = -EINPROGRESS; + atomic64_inc(&ctx->sec->debug.dfx.send_cnt); + } + + return ret; +} + +/* Get DMA memory resources */ +static int sec_alloc_civ_resource(struct device *dev, struct sec_alg_res *res) +{ + u16 q_depth = res->depth; + int i; + + res->c_ivin = dma_alloc_coherent(dev, SEC_TOTAL_IV_SZ(q_depth), + &res->c_ivin_dma, GFP_KERNEL); + if (!res->c_ivin) + return -ENOMEM; + + for (i = 1; i < q_depth; i++) { + res[i].c_ivin_dma = res->c_ivin_dma + i * SEC_IV_SIZE; + res[i].c_ivin = res->c_ivin + i * SEC_IV_SIZE; + } + + return 0; +} + +static void sec_free_civ_resource(struct device *dev, struct sec_alg_res *res) +{ + if (res->c_ivin) + dma_free_coherent(dev, SEC_TOTAL_IV_SZ(res->depth), + res->c_ivin, res->c_ivin_dma); +} + +static int sec_alloc_aiv_resource(struct device *dev, struct sec_alg_res *res) +{ + u16 q_depth = res->depth; + int i; + + res->a_ivin = dma_alloc_coherent(dev, SEC_TOTAL_IV_SZ(q_depth), + &res->a_ivin_dma, GFP_KERNEL); + if (!res->a_ivin) + return -ENOMEM; + + for (i = 1; i < q_depth; i++) { + res[i].a_ivin_dma = res->a_ivin_dma + i * SEC_IV_SIZE; + res[i].a_ivin = res->a_ivin + i * SEC_IV_SIZE; + } + + return 0; +} + +static void sec_free_aiv_resource(struct device *dev, struct sec_alg_res *res) +{ + if (res->a_ivin) + dma_free_coherent(dev, SEC_TOTAL_IV_SZ(res->depth), + res->a_ivin, res->a_ivin_dma); +} + +static int sec_alloc_mac_resource(struct device *dev, struct sec_alg_res *res) +{ + u16 q_depth = res->depth; + int i; + + res->out_mac = dma_alloc_coherent(dev, SEC_TOTAL_MAC_SZ(q_depth) << 1, + &res->out_mac_dma, GFP_KERNEL); + if (!res->out_mac) + return -ENOMEM; + + for (i = 1; i < q_depth; i++) { + res[i].out_mac_dma = res->out_mac_dma + + i * (SEC_MAX_MAC_LEN << 1); + res[i].out_mac = res->out_mac + i * (SEC_MAX_MAC_LEN << 1); + } + + return 0; +} + +static void sec_free_mac_resource(struct device *dev, struct sec_alg_res *res) +{ + if (res->out_mac) + dma_free_coherent(dev, SEC_TOTAL_MAC_SZ(res->depth) << 1, + res->out_mac, res->out_mac_dma); +} + +static void sec_free_pbuf_resource(struct device *dev, struct sec_alg_res *res) +{ + if (res->pbuf) + dma_free_coherent(dev, SEC_TOTAL_PBUF_SZ(res->depth), + res->pbuf, res->pbuf_dma); +} + +/* + * To improve performance, pbuffer is used for + * small packets (< 512Bytes) as IOMMU translation using. + */ +static int sec_alloc_pbuf_resource(struct device *dev, struct sec_alg_res *res) +{ + u16 q_depth = res->depth; + int size = SEC_PBUF_PAGE_NUM(q_depth); + int pbuf_page_offset; + int i, j, k; + + res->pbuf = dma_alloc_coherent(dev, SEC_TOTAL_PBUF_SZ(q_depth), + &res->pbuf_dma, GFP_KERNEL); + if (!res->pbuf) + return -ENOMEM; + + /* + * SEC_PBUF_PKG contains data pbuf, iv and + * out_mac : + * Every PAGE contains six SEC_PBUF_PKG + * The sec_qp_ctx contains QM_Q_DEPTH numbers of SEC_PBUF_PKG + * So we need SEC_PBUF_PAGE_NUM numbers of PAGE + * for the SEC_TOTAL_PBUF_SZ + */ + for (i = 0; i <= size; i++) { + pbuf_page_offset = PAGE_SIZE * i; + for (j = 0; j < SEC_PBUF_NUM; j++) { + k = i * SEC_PBUF_NUM + j; + if (k == q_depth) + break; + res[k].pbuf = res->pbuf + + j * SEC_PBUF_PKG + pbuf_page_offset; + res[k].pbuf_dma = res->pbuf_dma + + j * SEC_PBUF_PKG + pbuf_page_offset; + } + } + + return 0; +} + +static int sec_alg_resource_alloc(struct sec_ctx *ctx, + struct sec_qp_ctx *qp_ctx) +{ + struct sec_alg_res *res = qp_ctx->res; + struct device *dev = ctx->dev; + int ret; + + ret = sec_alloc_civ_resource(dev, res); + if (ret) + return ret; + + if (ctx->alg_type == SEC_AEAD) { + ret = sec_alloc_aiv_resource(dev, res); + if (ret) + goto alloc_aiv_fail; + + ret = sec_alloc_mac_resource(dev, res); + if (ret) + goto alloc_mac_fail; + } + if (ctx->pbuf_supported) { + ret = sec_alloc_pbuf_resource(dev, res); + if (ret) { + dev_err(dev, "fail to alloc pbuf dma resource!\n"); + goto alloc_pbuf_fail; + } + } + + return 0; + +alloc_pbuf_fail: + if (ctx->alg_type == SEC_AEAD) + sec_free_mac_resource(dev, qp_ctx->res); +alloc_mac_fail: + if (ctx->alg_type == SEC_AEAD) + sec_free_aiv_resource(dev, res); +alloc_aiv_fail: + sec_free_civ_resource(dev, res); + return ret; +} + +static void sec_alg_resource_free(struct sec_ctx *ctx, + struct sec_qp_ctx *qp_ctx) +{ + struct device *dev = ctx->dev; + + sec_free_civ_resource(dev, qp_ctx->res); + + if (ctx->pbuf_supported) + sec_free_pbuf_resource(dev, qp_ctx->res); + if (ctx->alg_type == SEC_AEAD) + sec_free_mac_resource(dev, qp_ctx->res); +} + +static int sec_alloc_qp_ctx_resource(struct hisi_qm *qm, struct sec_ctx *ctx, + struct sec_qp_ctx *qp_ctx) +{ + u16 q_depth = qp_ctx->qp->sq_depth; + struct device *dev = ctx->dev; + int ret = -ENOMEM; + + qp_ctx->req_list = kcalloc(q_depth, sizeof(struct sec_req *), GFP_KERNEL); + if (!qp_ctx->req_list) + return ret; + + qp_ctx->res = kcalloc(q_depth, sizeof(struct sec_alg_res), GFP_KERNEL); + if (!qp_ctx->res) + goto err_free_req_list; + qp_ctx->res->depth = q_depth; + + qp_ctx->c_in_pool = hisi_acc_create_sgl_pool(dev, q_depth, SEC_SGL_SGE_NR); + if (IS_ERR(qp_ctx->c_in_pool)) { + dev_err(dev, "fail to create sgl pool for input!\n"); + goto err_free_res; + } + + qp_ctx->c_out_pool = hisi_acc_create_sgl_pool(dev, q_depth, SEC_SGL_SGE_NR); + if (IS_ERR(qp_ctx->c_out_pool)) { + dev_err(dev, "fail to create sgl pool for output!\n"); + goto err_free_c_in_pool; + } + + ret = sec_alg_resource_alloc(ctx, qp_ctx); + if (ret) + goto err_free_c_out_pool; + + return 0; + +err_free_c_out_pool: + hisi_acc_free_sgl_pool(dev, qp_ctx->c_out_pool); +err_free_c_in_pool: + hisi_acc_free_sgl_pool(dev, qp_ctx->c_in_pool); +err_free_res: + kfree(qp_ctx->res); +err_free_req_list: + kfree(qp_ctx->req_list); + return ret; +} + +static void sec_free_qp_ctx_resource(struct sec_ctx *ctx, struct sec_qp_ctx *qp_ctx) +{ + struct device *dev = ctx->dev; + + sec_alg_resource_free(ctx, qp_ctx); + hisi_acc_free_sgl_pool(dev, qp_ctx->c_out_pool); + hisi_acc_free_sgl_pool(dev, qp_ctx->c_in_pool); + kfree(qp_ctx->res); + kfree(qp_ctx->req_list); +} + +static int sec_create_qp_ctx(struct hisi_qm *qm, struct sec_ctx *ctx, + int qp_ctx_id, int alg_type) +{ + struct sec_qp_ctx *qp_ctx; + struct hisi_qp *qp; + int ret; + + qp_ctx = &ctx->qp_ctx[qp_ctx_id]; + qp = ctx->qps[qp_ctx_id]; + qp->req_type = 0; + qp->qp_ctx = qp_ctx; + qp_ctx->qp = qp; + qp_ctx->ctx = ctx; + + qp->req_cb = sec_req_cb; + + spin_lock_init(&qp_ctx->req_lock); + idr_init(&qp_ctx->req_idr); + INIT_LIST_HEAD(&qp_ctx->backlog); + + ret = sec_alloc_qp_ctx_resource(qm, ctx, qp_ctx); + if (ret) + goto err_destroy_idr; + + ret = hisi_qm_start_qp(qp, 0); + if (ret < 0) + goto err_resource_free; + + return 0; + +err_resource_free: + sec_free_qp_ctx_resource(ctx, qp_ctx); +err_destroy_idr: + idr_destroy(&qp_ctx->req_idr); + return ret; +} + +static void sec_release_qp_ctx(struct sec_ctx *ctx, + struct sec_qp_ctx *qp_ctx) +{ + hisi_qm_stop_qp(qp_ctx->qp); + sec_free_qp_ctx_resource(ctx, qp_ctx); + idr_destroy(&qp_ctx->req_idr); +} + +static int sec_ctx_base_init(struct sec_ctx *ctx) +{ + struct sec_dev *sec; + int i, ret; + + ctx->qps = sec_create_qps(); + if (!ctx->qps) { + pr_err("Can not create sec qps!\n"); + return -ENODEV; + } + + sec = container_of(ctx->qps[0]->qm, struct sec_dev, qm); + ctx->sec = sec; + ctx->dev = &sec->qm.pdev->dev; + ctx->hlf_q_num = sec->ctx_q_num >> 1; + + ctx->pbuf_supported = ctx->sec->iommu_used; + + /* Half of queue depth is taken as fake requests limit in the queue. */ + ctx->fake_req_limit = ctx->qps[0]->sq_depth >> 1; + ctx->qp_ctx = kcalloc(sec->ctx_q_num, sizeof(struct sec_qp_ctx), + GFP_KERNEL); + if (!ctx->qp_ctx) { + ret = -ENOMEM; + goto err_destroy_qps; + } + + for (i = 0; i < sec->ctx_q_num; i++) { + ret = sec_create_qp_ctx(&sec->qm, ctx, i, 0); + if (ret) + goto err_sec_release_qp_ctx; + } + + return 0; + +err_sec_release_qp_ctx: + for (i = i - 1; i >= 0; i--) + sec_release_qp_ctx(ctx, &ctx->qp_ctx[i]); + kfree(ctx->qp_ctx); +err_destroy_qps: + sec_destroy_qps(ctx->qps, sec->ctx_q_num); + return ret; +} + +static void sec_ctx_base_uninit(struct sec_ctx *ctx) +{ + int i; + + for (i = 0; i < ctx->sec->ctx_q_num; i++) + sec_release_qp_ctx(ctx, &ctx->qp_ctx[i]); + + sec_destroy_qps(ctx->qps, ctx->sec->ctx_q_num); + kfree(ctx->qp_ctx); +} + +static int sec_cipher_init(struct sec_ctx *ctx) +{ + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + + c_ctx->c_key = dma_alloc_coherent(ctx->dev, SEC_MAX_KEY_SIZE, + &c_ctx->c_key_dma, GFP_KERNEL); + if (!c_ctx->c_key) + return -ENOMEM; + + return 0; +} + +static void sec_cipher_uninit(struct sec_ctx *ctx) +{ + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + + memzero_explicit(c_ctx->c_key, SEC_MAX_KEY_SIZE); + dma_free_coherent(ctx->dev, SEC_MAX_KEY_SIZE, + c_ctx->c_key, c_ctx->c_key_dma); +} + +static int sec_auth_init(struct sec_ctx *ctx) +{ + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + + a_ctx->a_key = dma_alloc_coherent(ctx->dev, SEC_MAX_AKEY_SIZE, + &a_ctx->a_key_dma, GFP_KERNEL); + if (!a_ctx->a_key) + return -ENOMEM; + + return 0; +} + +static void sec_auth_uninit(struct sec_ctx *ctx) +{ + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + + memzero_explicit(a_ctx->a_key, SEC_MAX_AKEY_SIZE); + dma_free_coherent(ctx->dev, SEC_MAX_AKEY_SIZE, + a_ctx->a_key, a_ctx->a_key_dma); +} + +static int sec_skcipher_fbtfm_init(struct crypto_skcipher *tfm) +{ + const char *alg = crypto_tfm_alg_name(&tfm->base); + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + + c_ctx->fallback = false; + + /* Currently, only XTS mode need fallback tfm when using 192bit key */ + if (likely(strncmp(alg, "xts", SEC_XTS_NAME_SZ))) + return 0; + + c_ctx->fbtfm = crypto_alloc_sync_skcipher(alg, 0, + CRYPTO_ALG_NEED_FALLBACK); + if (IS_ERR(c_ctx->fbtfm)) { + pr_err("failed to alloc xts mode fallback tfm!\n"); + return PTR_ERR(c_ctx->fbtfm); + } + + return 0; +} + +static int sec_skcipher_init(struct crypto_skcipher *tfm) +{ + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + int ret; + + ctx->alg_type = SEC_SKCIPHER; + crypto_skcipher_set_reqsize(tfm, sizeof(struct sec_req)); + ctx->c_ctx.ivsize = crypto_skcipher_ivsize(tfm); + if (ctx->c_ctx.ivsize > SEC_IV_SIZE) { + pr_err("get error skcipher iv size!\n"); + return -EINVAL; + } + + ret = sec_ctx_base_init(ctx); + if (ret) + return ret; + + ret = sec_cipher_init(ctx); + if (ret) + goto err_cipher_init; + + ret = sec_skcipher_fbtfm_init(tfm); + if (ret) + goto err_fbtfm_init; + + return 0; + +err_fbtfm_init: + sec_cipher_uninit(ctx); +err_cipher_init: + sec_ctx_base_uninit(ctx); + return ret; +} + +static void sec_skcipher_uninit(struct crypto_skcipher *tfm) +{ + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + + if (ctx->c_ctx.fbtfm) + crypto_free_sync_skcipher(ctx->c_ctx.fbtfm); + + sec_cipher_uninit(ctx); + sec_ctx_base_uninit(ctx); +} + +static int sec_skcipher_3des_setkey(struct crypto_skcipher *tfm, const u8 *key, + const u32 keylen, + const enum sec_cmode c_mode) +{ + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + int ret; + + ret = verify_skcipher_des3_key(tfm, key); + if (ret) + return ret; + + switch (keylen) { + case SEC_DES3_2KEY_SIZE: + c_ctx->c_key_len = SEC_CKEY_3DES_2KEY; + break; + case SEC_DES3_3KEY_SIZE: + c_ctx->c_key_len = SEC_CKEY_3DES_3KEY; + break; + default: + return -EINVAL; + } + + return 0; +} + +static int sec_skcipher_aes_sm4_setkey(struct sec_cipher_ctx *c_ctx, + const u32 keylen, + const enum sec_cmode c_mode) +{ + if (c_mode == SEC_CMODE_XTS) { + switch (keylen) { + case SEC_XTS_MIN_KEY_SIZE: + c_ctx->c_key_len = SEC_CKEY_128BIT; + break; + case SEC_XTS_MID_KEY_SIZE: + c_ctx->fallback = true; + break; + case SEC_XTS_MAX_KEY_SIZE: + c_ctx->c_key_len = SEC_CKEY_256BIT; + break; + default: + pr_err("hisi_sec2: xts mode key error!\n"); + return -EINVAL; + } + } else { + if (c_ctx->c_alg == SEC_CALG_SM4 && + keylen != AES_KEYSIZE_128) { + pr_err("hisi_sec2: sm4 key error!\n"); + return -EINVAL; + } else { + switch (keylen) { + case AES_KEYSIZE_128: + c_ctx->c_key_len = SEC_CKEY_128BIT; + break; + case AES_KEYSIZE_192: + c_ctx->c_key_len = SEC_CKEY_192BIT; + break; + case AES_KEYSIZE_256: + c_ctx->c_key_len = SEC_CKEY_256BIT; + break; + default: + pr_err("hisi_sec2: aes key error!\n"); + return -EINVAL; + } + } + } + + return 0; +} + +static int sec_skcipher_setkey(struct crypto_skcipher *tfm, const u8 *key, + const u32 keylen, const enum sec_calg c_alg, + const enum sec_cmode c_mode) +{ + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + struct device *dev = ctx->dev; + int ret; + + if (c_mode == SEC_CMODE_XTS) { + ret = xts_verify_key(tfm, key, keylen); + if (ret) { + dev_err(dev, "xts mode key err!\n"); + return ret; + } + } + + c_ctx->c_alg = c_alg; + c_ctx->c_mode = c_mode; + + switch (c_alg) { + case SEC_CALG_3DES: + ret = sec_skcipher_3des_setkey(tfm, key, keylen, c_mode); + break; + case SEC_CALG_AES: + case SEC_CALG_SM4: + ret = sec_skcipher_aes_sm4_setkey(c_ctx, keylen, c_mode); + break; + default: + return -EINVAL; + } + + if (ret) { + dev_err(dev, "set sec key err!\n"); + return ret; + } + + memcpy(c_ctx->c_key, key, keylen); + if (c_ctx->fallback && c_ctx->fbtfm) { + ret = crypto_sync_skcipher_setkey(c_ctx->fbtfm, key, keylen); + if (ret) { + dev_err(dev, "failed to set fallback skcipher key!\n"); + return ret; + } + } + return 0; +} + +#define GEN_SEC_SETKEY_FUNC(name, c_alg, c_mode) \ +static int sec_setkey_##name(struct crypto_skcipher *tfm, const u8 *key,\ + u32 keylen) \ +{ \ + return sec_skcipher_setkey(tfm, key, keylen, c_alg, c_mode); \ +} + +GEN_SEC_SETKEY_FUNC(aes_ecb, SEC_CALG_AES, SEC_CMODE_ECB) +GEN_SEC_SETKEY_FUNC(aes_cbc, SEC_CALG_AES, SEC_CMODE_CBC) +GEN_SEC_SETKEY_FUNC(aes_xts, SEC_CALG_AES, SEC_CMODE_XTS) +GEN_SEC_SETKEY_FUNC(aes_ofb, SEC_CALG_AES, SEC_CMODE_OFB) +GEN_SEC_SETKEY_FUNC(aes_cfb, SEC_CALG_AES, SEC_CMODE_CFB) +GEN_SEC_SETKEY_FUNC(aes_ctr, SEC_CALG_AES, SEC_CMODE_CTR) +GEN_SEC_SETKEY_FUNC(3des_ecb, SEC_CALG_3DES, SEC_CMODE_ECB) +GEN_SEC_SETKEY_FUNC(3des_cbc, SEC_CALG_3DES, SEC_CMODE_CBC) +GEN_SEC_SETKEY_FUNC(sm4_xts, SEC_CALG_SM4, SEC_CMODE_XTS) +GEN_SEC_SETKEY_FUNC(sm4_cbc, SEC_CALG_SM4, SEC_CMODE_CBC) +GEN_SEC_SETKEY_FUNC(sm4_ofb, SEC_CALG_SM4, SEC_CMODE_OFB) +GEN_SEC_SETKEY_FUNC(sm4_cfb, SEC_CALG_SM4, SEC_CMODE_CFB) +GEN_SEC_SETKEY_FUNC(sm4_ctr, SEC_CALG_SM4, SEC_CMODE_CTR) + +static int sec_cipher_pbuf_map(struct sec_ctx *ctx, struct sec_req *req, + struct scatterlist *src) +{ + struct sec_aead_req *a_req = &req->aead_req; + struct aead_request *aead_req = a_req->aead_req; + struct sec_cipher_req *c_req = &req->c_req; + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + struct device *dev = ctx->dev; + int copy_size, pbuf_length; + int req_id = req->req_id; + struct crypto_aead *tfm; + size_t authsize; + u8 *mac_offset; + + if (ctx->alg_type == SEC_AEAD) + copy_size = aead_req->cryptlen + aead_req->assoclen; + else + copy_size = c_req->c_len; + + pbuf_length = sg_copy_to_buffer(src, sg_nents(src), + qp_ctx->res[req_id].pbuf, copy_size); + if (unlikely(pbuf_length != copy_size)) { + dev_err(dev, "copy src data to pbuf error!\n"); + return -EINVAL; + } + if (!c_req->encrypt && ctx->alg_type == SEC_AEAD) { + tfm = crypto_aead_reqtfm(aead_req); + authsize = crypto_aead_authsize(tfm); + mac_offset = qp_ctx->res[req_id].pbuf + copy_size - authsize; + memcpy(a_req->out_mac, mac_offset, authsize); + } + + req->in_dma = qp_ctx->res[req_id].pbuf_dma; + c_req->c_out_dma = req->in_dma; + + return 0; +} + +static void sec_cipher_pbuf_unmap(struct sec_ctx *ctx, struct sec_req *req, + struct scatterlist *dst) +{ + struct aead_request *aead_req = req->aead_req.aead_req; + struct sec_cipher_req *c_req = &req->c_req; + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + int copy_size, pbuf_length; + int req_id = req->req_id; + + if (ctx->alg_type == SEC_AEAD) + copy_size = c_req->c_len + aead_req->assoclen; + else + copy_size = c_req->c_len; + + pbuf_length = sg_copy_from_buffer(dst, sg_nents(dst), + qp_ctx->res[req_id].pbuf, copy_size); + if (unlikely(pbuf_length != copy_size)) + dev_err(ctx->dev, "copy pbuf data to dst error!\n"); +} + +static int sec_aead_mac_init(struct sec_aead_req *req) +{ + struct aead_request *aead_req = req->aead_req; + struct crypto_aead *tfm = crypto_aead_reqtfm(aead_req); + size_t authsize = crypto_aead_authsize(tfm); + u8 *mac_out = req->out_mac; + struct scatterlist *sgl = aead_req->src; + size_t copy_size; + off_t skip_size; + + /* Copy input mac */ + skip_size = aead_req->assoclen + aead_req->cryptlen - authsize; + copy_size = sg_pcopy_to_buffer(sgl, sg_nents(sgl), mac_out, + authsize, skip_size); + if (unlikely(copy_size != authsize)) + return -EINVAL; + + return 0; +} + +static int sec_cipher_map(struct sec_ctx *ctx, struct sec_req *req, + struct scatterlist *src, struct scatterlist *dst) +{ + struct sec_cipher_req *c_req = &req->c_req; + struct sec_aead_req *a_req = &req->aead_req; + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + struct sec_alg_res *res = &qp_ctx->res[req->req_id]; + struct device *dev = ctx->dev; + int ret; + + if (req->use_pbuf) { + c_req->c_ivin = res->pbuf + SEC_PBUF_IV_OFFSET; + c_req->c_ivin_dma = res->pbuf_dma + SEC_PBUF_IV_OFFSET; + if (ctx->alg_type == SEC_AEAD) { + a_req->a_ivin = res->a_ivin; + a_req->a_ivin_dma = res->a_ivin_dma; + a_req->out_mac = res->pbuf + SEC_PBUF_MAC_OFFSET; + a_req->out_mac_dma = res->pbuf_dma + + SEC_PBUF_MAC_OFFSET; + } + ret = sec_cipher_pbuf_map(ctx, req, src); + + return ret; + } + c_req->c_ivin = res->c_ivin; + c_req->c_ivin_dma = res->c_ivin_dma; + if (ctx->alg_type == SEC_AEAD) { + a_req->a_ivin = res->a_ivin; + a_req->a_ivin_dma = res->a_ivin_dma; + a_req->out_mac = res->out_mac; + a_req->out_mac_dma = res->out_mac_dma; + } + + req->in = hisi_acc_sg_buf_map_to_hw_sgl(dev, src, + qp_ctx->c_in_pool, + req->req_id, + &req->in_dma); + if (IS_ERR(req->in)) { + dev_err(dev, "fail to dma map input sgl buffers!\n"); + return PTR_ERR(req->in); + } + + if (!c_req->encrypt && ctx->alg_type == SEC_AEAD) { + ret = sec_aead_mac_init(a_req); + if (unlikely(ret)) { + dev_err(dev, "fail to init mac data for ICV!\n"); + return ret; + } + } + + if (dst == src) { + c_req->c_out = req->in; + c_req->c_out_dma = req->in_dma; + } else { + c_req->c_out = hisi_acc_sg_buf_map_to_hw_sgl(dev, dst, + qp_ctx->c_out_pool, + req->req_id, + &c_req->c_out_dma); + + if (IS_ERR(c_req->c_out)) { + dev_err(dev, "fail to dma map output sgl buffers!\n"); + hisi_acc_sg_buf_unmap(dev, src, req->in); + return PTR_ERR(c_req->c_out); + } + } + + return 0; +} + +static void sec_cipher_unmap(struct sec_ctx *ctx, struct sec_req *req, + struct scatterlist *src, struct scatterlist *dst) +{ + struct sec_cipher_req *c_req = &req->c_req; + struct device *dev = ctx->dev; + + if (req->use_pbuf) { + sec_cipher_pbuf_unmap(ctx, req, dst); + } else { + if (dst != src) + hisi_acc_sg_buf_unmap(dev, src, req->in); + + hisi_acc_sg_buf_unmap(dev, dst, c_req->c_out); + } +} + +static int sec_skcipher_sgl_map(struct sec_ctx *ctx, struct sec_req *req) +{ + struct skcipher_request *sq = req->c_req.sk_req; + + return sec_cipher_map(ctx, req, sq->src, sq->dst); +} + +static void sec_skcipher_sgl_unmap(struct sec_ctx *ctx, struct sec_req *req) +{ + struct skcipher_request *sq = req->c_req.sk_req; + + sec_cipher_unmap(ctx, req, sq->src, sq->dst); +} + +static int sec_aead_aes_set_key(struct sec_cipher_ctx *c_ctx, + struct crypto_authenc_keys *keys) +{ + switch (keys->enckeylen) { + case AES_KEYSIZE_128: + c_ctx->c_key_len = SEC_CKEY_128BIT; + break; + case AES_KEYSIZE_192: + c_ctx->c_key_len = SEC_CKEY_192BIT; + break; + case AES_KEYSIZE_256: + c_ctx->c_key_len = SEC_CKEY_256BIT; + break; + default: + pr_err("hisi_sec2: aead aes key error!\n"); + return -EINVAL; + } + memcpy(c_ctx->c_key, keys->enckey, keys->enckeylen); + + return 0; +} + +static int sec_aead_auth_set_key(struct sec_auth_ctx *ctx, + struct crypto_authenc_keys *keys) +{ + struct crypto_shash *hash_tfm = ctx->hash_tfm; + int blocksize, digestsize, ret; + + if (!keys->authkeylen) { + pr_err("hisi_sec2: aead auth key error!\n"); + return -EINVAL; + } + + blocksize = crypto_shash_blocksize(hash_tfm); + digestsize = crypto_shash_digestsize(hash_tfm); + if (keys->authkeylen > blocksize) { + ret = crypto_shash_tfm_digest(hash_tfm, keys->authkey, + keys->authkeylen, ctx->a_key); + if (ret) { + pr_err("hisi_sec2: aead auth digest error!\n"); + return -EINVAL; + } + ctx->a_key_len = digestsize; + } else { + memcpy(ctx->a_key, keys->authkey, keys->authkeylen); + ctx->a_key_len = keys->authkeylen; + } + + return 0; +} + +static int sec_aead_setauthsize(struct crypto_aead *aead, unsigned int authsize) +{ + struct crypto_tfm *tfm = crypto_aead_tfm(aead); + struct sec_ctx *ctx = crypto_tfm_ctx(tfm); + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + + if (unlikely(a_ctx->fallback_aead_tfm)) + return crypto_aead_setauthsize(a_ctx->fallback_aead_tfm, authsize); + + return 0; +} + +static int sec_aead_fallback_setkey(struct sec_auth_ctx *a_ctx, + struct crypto_aead *tfm, const u8 *key, + unsigned int keylen) +{ + crypto_aead_clear_flags(a_ctx->fallback_aead_tfm, CRYPTO_TFM_REQ_MASK); + crypto_aead_set_flags(a_ctx->fallback_aead_tfm, + crypto_aead_get_flags(tfm) & CRYPTO_TFM_REQ_MASK); + return crypto_aead_setkey(a_ctx->fallback_aead_tfm, key, keylen); +} + +static int sec_aead_setkey(struct crypto_aead *tfm, const u8 *key, + const u32 keylen, const enum sec_hash_alg a_alg, + const enum sec_calg c_alg, + const enum sec_mac_len mac_len, + const enum sec_cmode c_mode) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + struct device *dev = ctx->dev; + struct crypto_authenc_keys keys; + int ret; + + ctx->a_ctx.a_alg = a_alg; + ctx->c_ctx.c_alg = c_alg; + ctx->a_ctx.mac_len = mac_len; + c_ctx->c_mode = c_mode; + + if (c_mode == SEC_CMODE_CCM || c_mode == SEC_CMODE_GCM) { + ret = sec_skcipher_aes_sm4_setkey(c_ctx, keylen, c_mode); + if (ret) { + dev_err(dev, "set sec aes ccm cipher key err!\n"); + return ret; + } + memcpy(c_ctx->c_key, key, keylen); + + if (unlikely(a_ctx->fallback_aead_tfm)) { + ret = sec_aead_fallback_setkey(a_ctx, tfm, key, keylen); + if (ret) + return ret; + } + + return 0; + } + + if (crypto_authenc_extractkeys(&keys, key, keylen)) + goto bad_key; + + ret = sec_aead_aes_set_key(c_ctx, &keys); + if (ret) { + dev_err(dev, "set sec cipher key err!\n"); + goto bad_key; + } + + ret = sec_aead_auth_set_key(&ctx->a_ctx, &keys); + if (ret) { + dev_err(dev, "set sec auth key err!\n"); + goto bad_key; + } + + if ((ctx->a_ctx.mac_len & SEC_SQE_LEN_RATE_MASK) || + (ctx->a_ctx.a_key_len & SEC_SQE_LEN_RATE_MASK)) { + dev_err(dev, "MAC or AUTH key length error!\n"); + goto bad_key; + } + + return 0; + +bad_key: + memzero_explicit(&keys, sizeof(struct crypto_authenc_keys)); + return -EINVAL; +} + + +#define GEN_SEC_AEAD_SETKEY_FUNC(name, aalg, calg, maclen, cmode) \ +static int sec_setkey_##name(struct crypto_aead *tfm, const u8 *key, \ + u32 keylen) \ +{ \ + return sec_aead_setkey(tfm, key, keylen, aalg, calg, maclen, cmode);\ +} + +GEN_SEC_AEAD_SETKEY_FUNC(aes_cbc_sha1, SEC_A_HMAC_SHA1, + SEC_CALG_AES, SEC_HMAC_SHA1_MAC, SEC_CMODE_CBC) +GEN_SEC_AEAD_SETKEY_FUNC(aes_cbc_sha256, SEC_A_HMAC_SHA256, + SEC_CALG_AES, SEC_HMAC_SHA256_MAC, SEC_CMODE_CBC) +GEN_SEC_AEAD_SETKEY_FUNC(aes_cbc_sha512, SEC_A_HMAC_SHA512, + SEC_CALG_AES, SEC_HMAC_SHA512_MAC, SEC_CMODE_CBC) +GEN_SEC_AEAD_SETKEY_FUNC(aes_ccm, 0, SEC_CALG_AES, + SEC_HMAC_CCM_MAC, SEC_CMODE_CCM) +GEN_SEC_AEAD_SETKEY_FUNC(aes_gcm, 0, SEC_CALG_AES, + SEC_HMAC_GCM_MAC, SEC_CMODE_GCM) +GEN_SEC_AEAD_SETKEY_FUNC(sm4_ccm, 0, SEC_CALG_SM4, + SEC_HMAC_CCM_MAC, SEC_CMODE_CCM) +GEN_SEC_AEAD_SETKEY_FUNC(sm4_gcm, 0, SEC_CALG_SM4, + SEC_HMAC_GCM_MAC, SEC_CMODE_GCM) + +static int sec_aead_sgl_map(struct sec_ctx *ctx, struct sec_req *req) +{ + struct aead_request *aq = req->aead_req.aead_req; + + return sec_cipher_map(ctx, req, aq->src, aq->dst); +} + +static void sec_aead_sgl_unmap(struct sec_ctx *ctx, struct sec_req *req) +{ + struct aead_request *aq = req->aead_req.aead_req; + + sec_cipher_unmap(ctx, req, aq->src, aq->dst); +} + +static int sec_request_transfer(struct sec_ctx *ctx, struct sec_req *req) +{ + int ret; + + ret = ctx->req_op->buf_map(ctx, req); + if (unlikely(ret)) + return ret; + + ctx->req_op->do_transfer(ctx, req); + + ret = ctx->req_op->bd_fill(ctx, req); + if (unlikely(ret)) + goto unmap_req_buf; + + return ret; + +unmap_req_buf: + ctx->req_op->buf_unmap(ctx, req); + return ret; +} + +static void sec_request_untransfer(struct sec_ctx *ctx, struct sec_req *req) +{ + ctx->req_op->buf_unmap(ctx, req); +} + +static void sec_skcipher_copy_iv(struct sec_ctx *ctx, struct sec_req *req) +{ + struct skcipher_request *sk_req = req->c_req.sk_req; + struct sec_cipher_req *c_req = &req->c_req; + + memcpy(c_req->c_ivin, sk_req->iv, ctx->c_ctx.ivsize); +} + +static int sec_skcipher_bd_fill(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + struct sec_cipher_req *c_req = &req->c_req; + struct sec_sqe *sec_sqe = &req->sec_sqe; + u8 scene, sa_type, da_type; + u8 bd_type, cipher; + u8 de = 0; + + memset(sec_sqe, 0, sizeof(struct sec_sqe)); + + sec_sqe->type2.c_key_addr = cpu_to_le64(c_ctx->c_key_dma); + sec_sqe->type2.c_ivin_addr = cpu_to_le64(c_req->c_ivin_dma); + sec_sqe->type2.data_src_addr = cpu_to_le64(req->in_dma); + sec_sqe->type2.data_dst_addr = cpu_to_le64(c_req->c_out_dma); + + sec_sqe->type2.icvw_kmode |= cpu_to_le16(((u16)c_ctx->c_mode) << + SEC_CMODE_OFFSET); + sec_sqe->type2.c_alg = c_ctx->c_alg; + sec_sqe->type2.icvw_kmode |= cpu_to_le16(((u16)c_ctx->c_key_len) << + SEC_CKEY_OFFSET); + + bd_type = SEC_BD_TYPE2; + if (c_req->encrypt) + cipher = SEC_CIPHER_ENC << SEC_CIPHER_OFFSET; + else + cipher = SEC_CIPHER_DEC << SEC_CIPHER_OFFSET; + sec_sqe->type_cipher_auth = bd_type | cipher; + + /* Set destination and source address type */ + if (req->use_pbuf) { + sa_type = SEC_PBUF << SEC_SRC_SGL_OFFSET; + da_type = SEC_PBUF << SEC_DST_SGL_OFFSET; + } else { + sa_type = SEC_SGL << SEC_SRC_SGL_OFFSET; + da_type = SEC_SGL << SEC_DST_SGL_OFFSET; + } + + sec_sqe->sdm_addr_type |= da_type; + scene = SEC_COMM_SCENE << SEC_SCENE_OFFSET; + if (req->in_dma != c_req->c_out_dma) + de = 0x1 << SEC_DE_OFFSET; + + sec_sqe->sds_sa_type = (de | scene | sa_type); + + sec_sqe->type2.clen_ivhlen |= cpu_to_le32(c_req->c_len); + sec_sqe->type2.tag = cpu_to_le16((u16)req->req_id); + + return 0; +} + +static int sec_skcipher_bd_fill_v3(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_sqe3 *sec_sqe3 = &req->sec_sqe3; + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + struct sec_cipher_req *c_req = &req->c_req; + u32 bd_param = 0; + u16 cipher; + + memset(sec_sqe3, 0, sizeof(struct sec_sqe3)); + + sec_sqe3->c_key_addr = cpu_to_le64(c_ctx->c_key_dma); + sec_sqe3->no_scene.c_ivin_addr = cpu_to_le64(c_req->c_ivin_dma); + sec_sqe3->data_src_addr = cpu_to_le64(req->in_dma); + sec_sqe3->data_dst_addr = cpu_to_le64(c_req->c_out_dma); + + sec_sqe3->c_mode_alg = ((u8)c_ctx->c_alg << SEC_CALG_OFFSET_V3) | + c_ctx->c_mode; + sec_sqe3->c_icv_key |= cpu_to_le16(((u16)c_ctx->c_key_len) << + SEC_CKEY_OFFSET_V3); + + if (c_req->encrypt) + cipher = SEC_CIPHER_ENC; + else + cipher = SEC_CIPHER_DEC; + sec_sqe3->c_icv_key |= cpu_to_le16(cipher); + + /* Set the CTR counter mode is 128bit rollover */ + sec_sqe3->auth_mac_key = cpu_to_le32((u32)SEC_CTR_CNT_ROLLOVER << + SEC_CTR_CNT_OFFSET); + + if (req->use_pbuf) { + bd_param |= SEC_PBUF << SEC_SRC_SGL_OFFSET_V3; + bd_param |= SEC_PBUF << SEC_DST_SGL_OFFSET_V3; + } else { + bd_param |= SEC_SGL << SEC_SRC_SGL_OFFSET_V3; + bd_param |= SEC_SGL << SEC_DST_SGL_OFFSET_V3; + } + + bd_param |= SEC_COMM_SCENE << SEC_SCENE_OFFSET_V3; + if (req->in_dma != c_req->c_out_dma) + bd_param |= 0x1 << SEC_DE_OFFSET_V3; + + bd_param |= SEC_BD_TYPE3; + sec_sqe3->bd_param = cpu_to_le32(bd_param); + + sec_sqe3->c_len_ivin |= cpu_to_le32(c_req->c_len); + sec_sqe3->tag = cpu_to_le64(req); + + return 0; +} + +/* increment counter (128-bit int) */ +static void ctr_iv_inc(__u8 *counter, __u8 bits, __u32 nums) +{ + do { + --bits; + nums += counter[bits]; + counter[bits] = nums & BITS_MASK; + nums >>= BYTE_BITS; + } while (bits && nums); +} + +static void sec_update_iv(struct sec_req *req, enum sec_alg_type alg_type) +{ + struct aead_request *aead_req = req->aead_req.aead_req; + struct skcipher_request *sk_req = req->c_req.sk_req; + u32 iv_size = req->ctx->c_ctx.ivsize; + struct scatterlist *sgl; + unsigned int cryptlen; + size_t sz; + u8 *iv; + + if (req->c_req.encrypt) + sgl = alg_type == SEC_SKCIPHER ? sk_req->dst : aead_req->dst; + else + sgl = alg_type == SEC_SKCIPHER ? sk_req->src : aead_req->src; + + if (alg_type == SEC_SKCIPHER) { + iv = sk_req->iv; + cryptlen = sk_req->cryptlen; + } else { + iv = aead_req->iv; + cryptlen = aead_req->cryptlen; + } + + if (req->ctx->c_ctx.c_mode == SEC_CMODE_CBC) { + sz = sg_pcopy_to_buffer(sgl, sg_nents(sgl), iv, iv_size, + cryptlen - iv_size); + if (unlikely(sz != iv_size)) + dev_err(req->ctx->dev, "copy output iv error!\n"); + } else { + sz = cryptlen / iv_size; + if (cryptlen % iv_size) + sz += 1; + ctr_iv_inc(iv, iv_size, sz); + } +} + +static struct sec_req *sec_back_req_clear(struct sec_ctx *ctx, + struct sec_qp_ctx *qp_ctx) +{ + struct sec_req *backlog_req = NULL; + + spin_lock_bh(&qp_ctx->req_lock); + if (ctx->fake_req_limit >= + atomic_read(&qp_ctx->qp->qp_status.used) && + !list_empty(&qp_ctx->backlog)) { + backlog_req = list_first_entry(&qp_ctx->backlog, + typeof(*backlog_req), backlog_head); + list_del(&backlog_req->backlog_head); + } + spin_unlock_bh(&qp_ctx->req_lock); + + return backlog_req; +} + +static void sec_skcipher_callback(struct sec_ctx *ctx, struct sec_req *req, + int err) +{ + struct skcipher_request *sk_req = req->c_req.sk_req; + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + struct skcipher_request *backlog_sk_req; + struct sec_req *backlog_req; + + sec_free_req_id(req); + + /* IV output at encrypto of CBC/CTR mode */ + if (!err && (ctx->c_ctx.c_mode == SEC_CMODE_CBC || + ctx->c_ctx.c_mode == SEC_CMODE_CTR) && req->c_req.encrypt) + sec_update_iv(req, SEC_SKCIPHER); + + while (1) { + backlog_req = sec_back_req_clear(ctx, qp_ctx); + if (!backlog_req) + break; + + backlog_sk_req = backlog_req->c_req.sk_req; + backlog_sk_req->base.complete(&backlog_sk_req->base, + -EINPROGRESS); + atomic64_inc(&ctx->sec->debug.dfx.recv_busy_cnt); + } + + sk_req->base.complete(&sk_req->base, err); +} + +static void set_aead_auth_iv(struct sec_ctx *ctx, struct sec_req *req) +{ + struct aead_request *aead_req = req->aead_req.aead_req; + struct sec_cipher_req *c_req = &req->c_req; + struct sec_aead_req *a_req = &req->aead_req; + size_t authsize = ctx->a_ctx.mac_len; + u32 data_size = aead_req->cryptlen; + u8 flage = 0; + u8 cm, cl; + + /* the specification has been checked in aead_iv_demension_check() */ + cl = c_req->c_ivin[0] + 1; + c_req->c_ivin[ctx->c_ctx.ivsize - cl] = 0x00; + memset(&c_req->c_ivin[ctx->c_ctx.ivsize - cl], 0, cl); + c_req->c_ivin[ctx->c_ctx.ivsize - IV_LAST_BYTE1] = IV_CTR_INIT; + + /* the last 3bit is L' */ + flage |= c_req->c_ivin[0] & IV_CL_MASK; + + /* the M' is bit3~bit5, the Flags is bit6 */ + cm = (authsize - IV_CM_CAL_NUM) / IV_CM_CAL_NUM; + flage |= cm << IV_CM_OFFSET; + if (aead_req->assoclen) + flage |= 0x01 << IV_FLAGS_OFFSET; + + memcpy(a_req->a_ivin, c_req->c_ivin, ctx->c_ctx.ivsize); + a_req->a_ivin[0] = flage; + + /* + * the last 32bit is counter's initial number, + * but the nonce uses the first 16bit + * the tail 16bit fill with the cipher length + */ + if (!c_req->encrypt) + data_size = aead_req->cryptlen - authsize; + + a_req->a_ivin[ctx->c_ctx.ivsize - IV_LAST_BYTE1] = + data_size & IV_LAST_BYTE_MASK; + data_size >>= IV_BYTE_OFFSET; + a_req->a_ivin[ctx->c_ctx.ivsize - IV_LAST_BYTE2] = + data_size & IV_LAST_BYTE_MASK; +} + +static void sec_aead_set_iv(struct sec_ctx *ctx, struct sec_req *req) +{ + struct aead_request *aead_req = req->aead_req.aead_req; + struct crypto_aead *tfm = crypto_aead_reqtfm(aead_req); + size_t authsize = crypto_aead_authsize(tfm); + struct sec_cipher_req *c_req = &req->c_req; + struct sec_aead_req *a_req = &req->aead_req; + + memcpy(c_req->c_ivin, aead_req->iv, ctx->c_ctx.ivsize); + + if (ctx->c_ctx.c_mode == SEC_CMODE_CCM) { + /* + * CCM 16Byte Cipher_IV: {1B_Flage,13B_IV,2B_counter}, + * the counter must set to 0x01 + */ + ctx->a_ctx.mac_len = authsize; + /* CCM 16Byte Auth_IV: {1B_AFlage,13B_IV,2B_Ptext_length} */ + set_aead_auth_iv(ctx, req); + } + + /* GCM 12Byte Cipher_IV == Auth_IV */ + if (ctx->c_ctx.c_mode == SEC_CMODE_GCM) { + ctx->a_ctx.mac_len = authsize; + memcpy(a_req->a_ivin, c_req->c_ivin, SEC_AIV_SIZE); + } +} + +static void sec_auth_bd_fill_xcm(struct sec_auth_ctx *ctx, int dir, + struct sec_req *req, struct sec_sqe *sec_sqe) +{ + struct sec_aead_req *a_req = &req->aead_req; + struct aead_request *aq = a_req->aead_req; + + /* C_ICV_Len is MAC size, 0x4 ~ 0x10 */ + sec_sqe->type2.icvw_kmode |= cpu_to_le16((u16)ctx->mac_len); + + /* mode set to CCM/GCM, don't set {A_Alg, AKey_Len, MAC_Len} */ + sec_sqe->type2.a_key_addr = sec_sqe->type2.c_key_addr; + sec_sqe->type2.a_ivin_addr = cpu_to_le64(a_req->a_ivin_dma); + sec_sqe->type_cipher_auth |= SEC_NO_AUTH << SEC_AUTH_OFFSET; + + if (dir) + sec_sqe->sds_sa_type &= SEC_CIPHER_AUTH; + else + sec_sqe->sds_sa_type |= SEC_AUTH_CIPHER; + + sec_sqe->type2.alen_ivllen = cpu_to_le32(aq->assoclen); + sec_sqe->type2.auth_src_offset = cpu_to_le16(0x0); + sec_sqe->type2.cipher_src_offset = cpu_to_le16((u16)aq->assoclen); + + sec_sqe->type2.mac_addr = cpu_to_le64(a_req->out_mac_dma); +} + +static void sec_auth_bd_fill_xcm_v3(struct sec_auth_ctx *ctx, int dir, + struct sec_req *req, struct sec_sqe3 *sqe3) +{ + struct sec_aead_req *a_req = &req->aead_req; + struct aead_request *aq = a_req->aead_req; + + /* C_ICV_Len is MAC size, 0x4 ~ 0x10 */ + sqe3->c_icv_key |= cpu_to_le16((u16)ctx->mac_len << SEC_MAC_OFFSET_V3); + + /* mode set to CCM/GCM, don't set {A_Alg, AKey_Len, MAC_Len} */ + sqe3->a_key_addr = sqe3->c_key_addr; + sqe3->auth_ivin.a_ivin_addr = cpu_to_le64(a_req->a_ivin_dma); + sqe3->auth_mac_key |= SEC_NO_AUTH; + + if (dir) + sqe3->huk_iv_seq &= SEC_CIPHER_AUTH_V3; + else + sqe3->huk_iv_seq |= SEC_AUTH_CIPHER_V3; + + sqe3->a_len_key = cpu_to_le32(aq->assoclen); + sqe3->auth_src_offset = cpu_to_le16(0x0); + sqe3->cipher_src_offset = cpu_to_le16((u16)aq->assoclen); + sqe3->mac_addr = cpu_to_le64(a_req->out_mac_dma); +} + +static void sec_auth_bd_fill_ex(struct sec_auth_ctx *ctx, int dir, + struct sec_req *req, struct sec_sqe *sec_sqe) +{ + struct sec_aead_req *a_req = &req->aead_req; + struct sec_cipher_req *c_req = &req->c_req; + struct aead_request *aq = a_req->aead_req; + + sec_sqe->type2.a_key_addr = cpu_to_le64(ctx->a_key_dma); + + sec_sqe->type2.mac_key_alg = + cpu_to_le32(ctx->mac_len / SEC_SQE_LEN_RATE); + + sec_sqe->type2.mac_key_alg |= + cpu_to_le32((u32)((ctx->a_key_len) / + SEC_SQE_LEN_RATE) << SEC_AKEY_OFFSET); + + sec_sqe->type2.mac_key_alg |= + cpu_to_le32((u32)(ctx->a_alg) << SEC_AEAD_ALG_OFFSET); + + if (dir) { + sec_sqe->type_cipher_auth |= SEC_AUTH_TYPE1 << SEC_AUTH_OFFSET; + sec_sqe->sds_sa_type &= SEC_CIPHER_AUTH; + } else { + sec_sqe->type_cipher_auth |= SEC_AUTH_TYPE2 << SEC_AUTH_OFFSET; + sec_sqe->sds_sa_type |= SEC_AUTH_CIPHER; + } + sec_sqe->type2.alen_ivllen = cpu_to_le32(c_req->c_len + aq->assoclen); + + sec_sqe->type2.cipher_src_offset = cpu_to_le16((u16)aq->assoclen); + + sec_sqe->type2.mac_addr = cpu_to_le64(a_req->out_mac_dma); +} + +static int sec_aead_bd_fill(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_auth_ctx *auth_ctx = &ctx->a_ctx; + struct sec_sqe *sec_sqe = &req->sec_sqe; + int ret; + + ret = sec_skcipher_bd_fill(ctx, req); + if (unlikely(ret)) { + dev_err(ctx->dev, "skcipher bd fill is error!\n"); + return ret; + } + + if (ctx->c_ctx.c_mode == SEC_CMODE_CCM || + ctx->c_ctx.c_mode == SEC_CMODE_GCM) + sec_auth_bd_fill_xcm(auth_ctx, req->c_req.encrypt, req, sec_sqe); + else + sec_auth_bd_fill_ex(auth_ctx, req->c_req.encrypt, req, sec_sqe); + + return 0; +} + +static void sec_auth_bd_fill_ex_v3(struct sec_auth_ctx *ctx, int dir, + struct sec_req *req, struct sec_sqe3 *sqe3) +{ + struct sec_aead_req *a_req = &req->aead_req; + struct sec_cipher_req *c_req = &req->c_req; + struct aead_request *aq = a_req->aead_req; + + sqe3->a_key_addr = cpu_to_le64(ctx->a_key_dma); + + sqe3->auth_mac_key |= + cpu_to_le32((u32)(ctx->mac_len / + SEC_SQE_LEN_RATE) << SEC_MAC_OFFSET_V3); + + sqe3->auth_mac_key |= + cpu_to_le32((u32)(ctx->a_key_len / + SEC_SQE_LEN_RATE) << SEC_AKEY_OFFSET_V3); + + sqe3->auth_mac_key |= + cpu_to_le32((u32)(ctx->a_alg) << SEC_AUTH_ALG_OFFSET_V3); + + if (dir) { + sqe3->auth_mac_key |= cpu_to_le32((u32)SEC_AUTH_TYPE1); + sqe3->huk_iv_seq &= SEC_CIPHER_AUTH_V3; + } else { + sqe3->auth_mac_key |= cpu_to_le32((u32)SEC_AUTH_TYPE2); + sqe3->huk_iv_seq |= SEC_AUTH_CIPHER_V3; + } + sqe3->a_len_key = cpu_to_le32(c_req->c_len + aq->assoclen); + + sqe3->cipher_src_offset = cpu_to_le16((u16)aq->assoclen); + + sqe3->mac_addr = cpu_to_le64(a_req->out_mac_dma); +} + +static int sec_aead_bd_fill_v3(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_auth_ctx *auth_ctx = &ctx->a_ctx; + struct sec_sqe3 *sec_sqe3 = &req->sec_sqe3; + int ret; + + ret = sec_skcipher_bd_fill_v3(ctx, req); + if (unlikely(ret)) { + dev_err(ctx->dev, "skcipher bd3 fill is error!\n"); + return ret; + } + + if (ctx->c_ctx.c_mode == SEC_CMODE_CCM || + ctx->c_ctx.c_mode == SEC_CMODE_GCM) + sec_auth_bd_fill_xcm_v3(auth_ctx, req->c_req.encrypt, + req, sec_sqe3); + else + sec_auth_bd_fill_ex_v3(auth_ctx, req->c_req.encrypt, + req, sec_sqe3); + + return 0; +} + +static void sec_aead_callback(struct sec_ctx *c, struct sec_req *req, int err) +{ + struct aead_request *a_req = req->aead_req.aead_req; + struct crypto_aead *tfm = crypto_aead_reqtfm(a_req); + struct sec_aead_req *aead_req = &req->aead_req; + struct sec_cipher_req *c_req = &req->c_req; + size_t authsize = crypto_aead_authsize(tfm); + struct sec_qp_ctx *qp_ctx = req->qp_ctx; + struct aead_request *backlog_aead_req; + struct sec_req *backlog_req; + size_t sz; + + if (!err && c->c_ctx.c_mode == SEC_CMODE_CBC && c_req->encrypt) + sec_update_iv(req, SEC_AEAD); + + /* Copy output mac */ + if (!err && c_req->encrypt) { + struct scatterlist *sgl = a_req->dst; + + sz = sg_pcopy_from_buffer(sgl, sg_nents(sgl), + aead_req->out_mac, + authsize, a_req->cryptlen + + a_req->assoclen); + if (unlikely(sz != authsize)) { + dev_err(c->dev, "copy out mac err!\n"); + err = -EINVAL; + } + } + + sec_free_req_id(req); + + while (1) { + backlog_req = sec_back_req_clear(c, qp_ctx); + if (!backlog_req) + break; + + backlog_aead_req = backlog_req->aead_req.aead_req; + backlog_aead_req->base.complete(&backlog_aead_req->base, + -EINPROGRESS); + atomic64_inc(&c->sec->debug.dfx.recv_busy_cnt); + } + + a_req->base.complete(&a_req->base, err); +} + +static void sec_request_uninit(struct sec_ctx *ctx, struct sec_req *req) +{ + sec_free_req_id(req); + sec_free_queue_id(ctx, req); +} + +static int sec_request_init(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_qp_ctx *qp_ctx; + int queue_id; + + /* To load balance */ + queue_id = sec_alloc_queue_id(ctx, req); + qp_ctx = &ctx->qp_ctx[queue_id]; + + req->req_id = sec_alloc_req_id(req, qp_ctx); + if (unlikely(req->req_id < 0)) { + sec_free_queue_id(ctx, req); + return req->req_id; + } + + return 0; +} + +static int sec_process(struct sec_ctx *ctx, struct sec_req *req) +{ + struct sec_cipher_req *c_req = &req->c_req; + int ret; + + ret = sec_request_init(ctx, req); + if (unlikely(ret)) + return ret; + + ret = sec_request_transfer(ctx, req); + if (unlikely(ret)) + goto err_uninit_req; + + /* Output IV as decrypto */ + if (!req->c_req.encrypt && (ctx->c_ctx.c_mode == SEC_CMODE_CBC || + ctx->c_ctx.c_mode == SEC_CMODE_CTR)) + sec_update_iv(req, ctx->alg_type); + + ret = ctx->req_op->bd_send(ctx, req); + if (unlikely((ret != -EBUSY && ret != -EINPROGRESS) || + (ret == -EBUSY && !(req->flag & CRYPTO_TFM_REQ_MAY_BACKLOG)))) { + dev_err_ratelimited(ctx->dev, "send sec request failed!\n"); + goto err_send_req; + } + + return ret; + +err_send_req: + /* As failing, restore the IV from user */ + if (ctx->c_ctx.c_mode == SEC_CMODE_CBC && !req->c_req.encrypt) { + if (ctx->alg_type == SEC_SKCIPHER) + memcpy(req->c_req.sk_req->iv, c_req->c_ivin, + ctx->c_ctx.ivsize); + else + memcpy(req->aead_req.aead_req->iv, c_req->c_ivin, + ctx->c_ctx.ivsize); + } + + sec_request_untransfer(ctx, req); +err_uninit_req: + sec_request_uninit(ctx, req); + return ret; +} + +static const struct sec_req_op sec_skcipher_req_ops = { + .buf_map = sec_skcipher_sgl_map, + .buf_unmap = sec_skcipher_sgl_unmap, + .do_transfer = sec_skcipher_copy_iv, + .bd_fill = sec_skcipher_bd_fill, + .bd_send = sec_bd_send, + .callback = sec_skcipher_callback, + .process = sec_process, +}; + +static const struct sec_req_op sec_aead_req_ops = { + .buf_map = sec_aead_sgl_map, + .buf_unmap = sec_aead_sgl_unmap, + .do_transfer = sec_aead_set_iv, + .bd_fill = sec_aead_bd_fill, + .bd_send = sec_bd_send, + .callback = sec_aead_callback, + .process = sec_process, +}; + +static const struct sec_req_op sec_skcipher_req_ops_v3 = { + .buf_map = sec_skcipher_sgl_map, + .buf_unmap = sec_skcipher_sgl_unmap, + .do_transfer = sec_skcipher_copy_iv, + .bd_fill = sec_skcipher_bd_fill_v3, + .bd_send = sec_bd_send, + .callback = sec_skcipher_callback, + .process = sec_process, +}; + +static const struct sec_req_op sec_aead_req_ops_v3 = { + .buf_map = sec_aead_sgl_map, + .buf_unmap = sec_aead_sgl_unmap, + .do_transfer = sec_aead_set_iv, + .bd_fill = sec_aead_bd_fill_v3, + .bd_send = sec_bd_send, + .callback = sec_aead_callback, + .process = sec_process, +}; + +static int sec_skcipher_ctx_init(struct crypto_skcipher *tfm) +{ + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + int ret; + + ret = sec_skcipher_init(tfm); + if (ret) + return ret; + + if (ctx->sec->qm.ver < QM_HW_V3) { + ctx->type_supported = SEC_BD_TYPE2; + ctx->req_op = &sec_skcipher_req_ops; + } else { + ctx->type_supported = SEC_BD_TYPE3; + ctx->req_op = &sec_skcipher_req_ops_v3; + } + + return ret; +} + +static void sec_skcipher_ctx_exit(struct crypto_skcipher *tfm) +{ + sec_skcipher_uninit(tfm); +} + +static int sec_aead_init(struct crypto_aead *tfm) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + int ret; + + crypto_aead_set_reqsize(tfm, sizeof(struct sec_req)); + ctx->alg_type = SEC_AEAD; + ctx->c_ctx.ivsize = crypto_aead_ivsize(tfm); + if (ctx->c_ctx.ivsize < SEC_AIV_SIZE || + ctx->c_ctx.ivsize > SEC_IV_SIZE) { + pr_err("get error aead iv size!\n"); + return -EINVAL; + } + + ret = sec_ctx_base_init(ctx); + if (ret) + return ret; + if (ctx->sec->qm.ver < QM_HW_V3) { + ctx->type_supported = SEC_BD_TYPE2; + ctx->req_op = &sec_aead_req_ops; + } else { + ctx->type_supported = SEC_BD_TYPE3; + ctx->req_op = &sec_aead_req_ops_v3; + } + + ret = sec_auth_init(ctx); + if (ret) + goto err_auth_init; + + ret = sec_cipher_init(ctx); + if (ret) + goto err_cipher_init; + + return ret; + +err_cipher_init: + sec_auth_uninit(ctx); +err_auth_init: + sec_ctx_base_uninit(ctx); + return ret; +} + +static void sec_aead_exit(struct crypto_aead *tfm) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + + sec_cipher_uninit(ctx); + sec_auth_uninit(ctx); + sec_ctx_base_uninit(ctx); +} + +static int sec_aead_ctx_init(struct crypto_aead *tfm, const char *hash_name) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + struct sec_auth_ctx *auth_ctx = &ctx->a_ctx; + int ret; + + ret = sec_aead_init(tfm); + if (ret) { + pr_err("hisi_sec2: aead init error!\n"); + return ret; + } + + auth_ctx->hash_tfm = crypto_alloc_shash(hash_name, 0, 0); + if (IS_ERR(auth_ctx->hash_tfm)) { + dev_err(ctx->dev, "aead alloc shash error!\n"); + sec_aead_exit(tfm); + return PTR_ERR(auth_ctx->hash_tfm); + } + + return 0; +} + +static void sec_aead_ctx_exit(struct crypto_aead *tfm) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + + crypto_free_shash(ctx->a_ctx.hash_tfm); + sec_aead_exit(tfm); +} + +static int sec_aead_xcm_ctx_init(struct crypto_aead *tfm) +{ + struct aead_alg *alg = crypto_aead_alg(tfm); + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + const char *aead_name = alg->base.cra_name; + int ret; + + ret = sec_aead_init(tfm); + if (ret) { + dev_err(ctx->dev, "hisi_sec2: aead xcm init error!\n"); + return ret; + } + + a_ctx->fallback_aead_tfm = crypto_alloc_aead(aead_name, 0, + CRYPTO_ALG_NEED_FALLBACK | + CRYPTO_ALG_ASYNC); + if (IS_ERR(a_ctx->fallback_aead_tfm)) { + dev_err(ctx->dev, "aead driver alloc fallback tfm error!\n"); + sec_aead_exit(tfm); + return PTR_ERR(a_ctx->fallback_aead_tfm); + } + a_ctx->fallback = false; + + return 0; +} + +static void sec_aead_xcm_ctx_exit(struct crypto_aead *tfm) +{ + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + + crypto_free_aead(ctx->a_ctx.fallback_aead_tfm); + sec_aead_exit(tfm); +} + +static int sec_aead_sha1_ctx_init(struct crypto_aead *tfm) +{ + return sec_aead_ctx_init(tfm, "sha1"); +} + +static int sec_aead_sha256_ctx_init(struct crypto_aead *tfm) +{ + return sec_aead_ctx_init(tfm, "sha256"); +} + +static int sec_aead_sha512_ctx_init(struct crypto_aead *tfm) +{ + return sec_aead_ctx_init(tfm, "sha512"); +} + +static int sec_skcipher_cryptlen_ckeck(struct sec_ctx *ctx, + struct sec_req *sreq) +{ + u32 cryptlen = sreq->c_req.sk_req->cryptlen; + struct device *dev = ctx->dev; + u8 c_mode = ctx->c_ctx.c_mode; + int ret = 0; + + switch (c_mode) { + case SEC_CMODE_XTS: + if (unlikely(cryptlen < AES_BLOCK_SIZE)) { + dev_err(dev, "skcipher XTS mode input length error!\n"); + ret = -EINVAL; + } + break; + case SEC_CMODE_ECB: + case SEC_CMODE_CBC: + if (unlikely(cryptlen & (AES_BLOCK_SIZE - 1))) { + dev_err(dev, "skcipher AES input length error!\n"); + ret = -EINVAL; + } + break; + case SEC_CMODE_CFB: + case SEC_CMODE_OFB: + case SEC_CMODE_CTR: + if (unlikely(ctx->sec->qm.ver < QM_HW_V3)) { + dev_err(dev, "skcipher HW version error!\n"); + ret = -EINVAL; + } + break; + default: + ret = -EINVAL; + } + + return ret; +} + +static int sec_skcipher_param_check(struct sec_ctx *ctx, struct sec_req *sreq) +{ + struct skcipher_request *sk_req = sreq->c_req.sk_req; + struct device *dev = ctx->dev; + u8 c_alg = ctx->c_ctx.c_alg; + + if (unlikely(!sk_req->src || !sk_req->dst || + sk_req->cryptlen > MAX_INPUT_DATA_LEN)) { + dev_err(dev, "skcipher input param error!\n"); + return -EINVAL; + } + sreq->c_req.c_len = sk_req->cryptlen; + + if (ctx->pbuf_supported && sk_req->cryptlen <= SEC_PBUF_SZ) + sreq->use_pbuf = true; + else + sreq->use_pbuf = false; + + if (c_alg == SEC_CALG_3DES) { + if (unlikely(sk_req->cryptlen & (DES3_EDE_BLOCK_SIZE - 1))) { + dev_err(dev, "skcipher 3des input length error!\n"); + return -EINVAL; + } + return 0; + } else if (c_alg == SEC_CALG_AES || c_alg == SEC_CALG_SM4) { + return sec_skcipher_cryptlen_ckeck(ctx, sreq); + } + + dev_err(dev, "skcipher algorithm error!\n"); + + return -EINVAL; +} + +static int sec_skcipher_soft_crypto(struct sec_ctx *ctx, + struct skcipher_request *sreq, bool encrypt) +{ + struct sec_cipher_ctx *c_ctx = &ctx->c_ctx; + SYNC_SKCIPHER_REQUEST_ON_STACK(subreq, c_ctx->fbtfm); + struct device *dev = ctx->dev; + int ret; + + if (!c_ctx->fbtfm) { + dev_err_ratelimited(dev, "the soft tfm isn't supported in the current system.\n"); + return -EINVAL; + } + + skcipher_request_set_sync_tfm(subreq, c_ctx->fbtfm); + + /* software need sync mode to do crypto */ + skcipher_request_set_callback(subreq, sreq->base.flags, + NULL, NULL); + skcipher_request_set_crypt(subreq, sreq->src, sreq->dst, + sreq->cryptlen, sreq->iv); + if (encrypt) + ret = crypto_skcipher_encrypt(subreq); + else + ret = crypto_skcipher_decrypt(subreq); + + skcipher_request_zero(subreq); + + return ret; +} + +static int sec_skcipher_crypto(struct skcipher_request *sk_req, bool encrypt) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(sk_req); + struct sec_req *req = skcipher_request_ctx(sk_req); + struct sec_ctx *ctx = crypto_skcipher_ctx(tfm); + int ret; + + if (!sk_req->cryptlen) { + if (ctx->c_ctx.c_mode == SEC_CMODE_XTS) + return -EINVAL; + return 0; + } + + req->flag = sk_req->base.flags; + req->c_req.sk_req = sk_req; + req->c_req.encrypt = encrypt; + req->ctx = ctx; + + ret = sec_skcipher_param_check(ctx, req); + if (unlikely(ret)) + return -EINVAL; + + if (unlikely(ctx->c_ctx.fallback)) + return sec_skcipher_soft_crypto(ctx, sk_req, encrypt); + + return ctx->req_op->process(ctx, req); +} + +static int sec_skcipher_encrypt(struct skcipher_request *sk_req) +{ + return sec_skcipher_crypto(sk_req, true); +} + +static int sec_skcipher_decrypt(struct skcipher_request *sk_req) +{ + return sec_skcipher_crypto(sk_req, false); +} + +#define SEC_SKCIPHER_GEN_ALG(sec_cra_name, sec_set_key, sec_min_key_size, \ + sec_max_key_size, ctx_init, ctx_exit, blk_size, iv_size)\ +{\ + .base = {\ + .cra_name = sec_cra_name,\ + .cra_driver_name = "hisi_sec_"sec_cra_name,\ + .cra_priority = SEC_PRIORITY,\ + .cra_flags = CRYPTO_ALG_ASYNC |\ + CRYPTO_ALG_NEED_FALLBACK,\ + .cra_blocksize = blk_size,\ + .cra_ctxsize = sizeof(struct sec_ctx),\ + .cra_module = THIS_MODULE,\ + },\ + .init = ctx_init,\ + .exit = ctx_exit,\ + .setkey = sec_set_key,\ + .decrypt = sec_skcipher_decrypt,\ + .encrypt = sec_skcipher_encrypt,\ + .min_keysize = sec_min_key_size,\ + .max_keysize = sec_max_key_size,\ + .ivsize = iv_size,\ +} + +#define SEC_SKCIPHER_ALG(name, key_func, min_key_size, \ + max_key_size, blk_size, iv_size) \ + SEC_SKCIPHER_GEN_ALG(name, key_func, min_key_size, max_key_size, \ + sec_skcipher_ctx_init, sec_skcipher_ctx_exit, blk_size, iv_size) + +static struct sec_skcipher sec_skciphers[] = { + { + .alg_msk = BIT(0), + .alg = SEC_SKCIPHER_ALG("ecb(aes)", sec_setkey_aes_ecb, AES_MIN_KEY_SIZE, + AES_MAX_KEY_SIZE, AES_BLOCK_SIZE, 0), + }, + { + .alg_msk = BIT(1), + .alg = SEC_SKCIPHER_ALG("cbc(aes)", sec_setkey_aes_cbc, AES_MIN_KEY_SIZE, + AES_MAX_KEY_SIZE, AES_BLOCK_SIZE, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(2), + .alg = SEC_SKCIPHER_ALG("ctr(aes)", sec_setkey_aes_ctr, AES_MIN_KEY_SIZE, + AES_MAX_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(3), + .alg = SEC_SKCIPHER_ALG("xts(aes)", sec_setkey_aes_xts, SEC_XTS_MIN_KEY_SIZE, + SEC_XTS_MAX_KEY_SIZE, AES_BLOCK_SIZE, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(4), + .alg = SEC_SKCIPHER_ALG("ofb(aes)", sec_setkey_aes_ofb, AES_MIN_KEY_SIZE, + AES_MAX_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(5), + .alg = SEC_SKCIPHER_ALG("cfb(aes)", sec_setkey_aes_cfb, AES_MIN_KEY_SIZE, + AES_MAX_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(12), + .alg = SEC_SKCIPHER_ALG("cbc(sm4)", sec_setkey_sm4_cbc, AES_MIN_KEY_SIZE, + AES_MIN_KEY_SIZE, AES_BLOCK_SIZE, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(13), + .alg = SEC_SKCIPHER_ALG("ctr(sm4)", sec_setkey_sm4_ctr, AES_MIN_KEY_SIZE, + AES_MIN_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(14), + .alg = SEC_SKCIPHER_ALG("xts(sm4)", sec_setkey_sm4_xts, SEC_XTS_MIN_KEY_SIZE, + SEC_XTS_MIN_KEY_SIZE, AES_BLOCK_SIZE, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(15), + .alg = SEC_SKCIPHER_ALG("ofb(sm4)", sec_setkey_sm4_ofb, AES_MIN_KEY_SIZE, + AES_MIN_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(16), + .alg = SEC_SKCIPHER_ALG("cfb(sm4)", sec_setkey_sm4_cfb, AES_MIN_KEY_SIZE, + AES_MIN_KEY_SIZE, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(23), + .alg = SEC_SKCIPHER_ALG("ecb(des3_ede)", sec_setkey_3des_ecb, SEC_DES3_3KEY_SIZE, + SEC_DES3_3KEY_SIZE, DES3_EDE_BLOCK_SIZE, 0), + }, + { + .alg_msk = BIT(24), + .alg = SEC_SKCIPHER_ALG("cbc(des3_ede)", sec_setkey_3des_cbc, SEC_DES3_3KEY_SIZE, + SEC_DES3_3KEY_SIZE, DES3_EDE_BLOCK_SIZE, + DES3_EDE_BLOCK_SIZE), + }, +}; + +static int aead_iv_demension_check(struct aead_request *aead_req) +{ + u8 cl; + + cl = aead_req->iv[0] + 1; + if (cl < IV_CL_MIN || cl > IV_CL_MAX) + return -EINVAL; + + if (cl < IV_CL_MID && aead_req->cryptlen >> (BYTE_BITS * cl)) + return -EOVERFLOW; + + return 0; +} + +static int sec_aead_spec_check(struct sec_ctx *ctx, struct sec_req *sreq) +{ + struct aead_request *req = sreq->aead_req.aead_req; + struct crypto_aead *tfm = crypto_aead_reqtfm(req); + size_t authsize = crypto_aead_authsize(tfm); + u8 c_mode = ctx->c_ctx.c_mode; + struct device *dev = ctx->dev; + int ret; + + if (unlikely(req->cryptlen + req->assoclen > MAX_INPUT_DATA_LEN || + req->assoclen > SEC_MAX_AAD_LEN)) { + dev_err(dev, "aead input spec error!\n"); + return -EINVAL; + } + + if (unlikely((c_mode == SEC_CMODE_GCM && authsize < DES_BLOCK_SIZE) || + (c_mode == SEC_CMODE_CCM && (authsize < MIN_MAC_LEN || + authsize & MAC_LEN_MASK)))) { + dev_err(dev, "aead input mac length error!\n"); + return -EINVAL; + } + + if (c_mode == SEC_CMODE_CCM) { + if (unlikely(req->assoclen > SEC_MAX_CCM_AAD_LEN)) { + dev_err_ratelimited(dev, "CCM input aad parameter is too long!\n"); + return -EINVAL; + } + ret = aead_iv_demension_check(req); + if (ret) { + dev_err(dev, "aead input iv param error!\n"); + return ret; + } + } + + if (sreq->c_req.encrypt) + sreq->c_req.c_len = req->cryptlen; + else + sreq->c_req.c_len = req->cryptlen - authsize; + if (c_mode == SEC_CMODE_CBC) { + if (unlikely(sreq->c_req.c_len & (AES_BLOCK_SIZE - 1))) { + dev_err(dev, "aead crypto length error!\n"); + return -EINVAL; + } + } + + return 0; +} + +static int sec_aead_param_check(struct sec_ctx *ctx, struct sec_req *sreq) +{ + struct aead_request *req = sreq->aead_req.aead_req; + struct crypto_aead *tfm = crypto_aead_reqtfm(req); + size_t authsize = crypto_aead_authsize(tfm); + struct device *dev = ctx->dev; + u8 c_alg = ctx->c_ctx.c_alg; + + if (unlikely(!req->src || !req->dst)) { + dev_err(dev, "aead input param error!\n"); + return -EINVAL; + } + + if (ctx->sec->qm.ver == QM_HW_V2) { + if (unlikely(!req->cryptlen || (!sreq->c_req.encrypt && + req->cryptlen <= authsize))) { + ctx->a_ctx.fallback = true; + return -EINVAL; + } + } + + /* Support AES or SM4 */ + if (unlikely(c_alg != SEC_CALG_AES && c_alg != SEC_CALG_SM4)) { + dev_err(dev, "aead crypto alg error!\n"); + return -EINVAL; + } + + if (unlikely(sec_aead_spec_check(ctx, sreq))) + return -EINVAL; + + if (ctx->pbuf_supported && (req->cryptlen + req->assoclen) <= + SEC_PBUF_SZ) + sreq->use_pbuf = true; + else + sreq->use_pbuf = false; + + return 0; +} + +static int sec_aead_soft_crypto(struct sec_ctx *ctx, + struct aead_request *aead_req, + bool encrypt) +{ + struct sec_auth_ctx *a_ctx = &ctx->a_ctx; + struct device *dev = ctx->dev; + struct aead_request *subreq; + int ret; + + /* Kunpeng920 aead mode not support input 0 size */ + if (!a_ctx->fallback_aead_tfm) { + dev_err(dev, "aead fallback tfm is NULL!\n"); + return -EINVAL; + } + + subreq = aead_request_alloc(a_ctx->fallback_aead_tfm, GFP_KERNEL); + if (!subreq) + return -ENOMEM; + + aead_request_set_tfm(subreq, a_ctx->fallback_aead_tfm); + aead_request_set_callback(subreq, aead_req->base.flags, + aead_req->base.complete, aead_req->base.data); + aead_request_set_crypt(subreq, aead_req->src, aead_req->dst, + aead_req->cryptlen, aead_req->iv); + aead_request_set_ad(subreq, aead_req->assoclen); + + if (encrypt) + ret = crypto_aead_encrypt(subreq); + else + ret = crypto_aead_decrypt(subreq); + aead_request_free(subreq); + + return ret; +} + +static int sec_aead_crypto(struct aead_request *a_req, bool encrypt) +{ + struct crypto_aead *tfm = crypto_aead_reqtfm(a_req); + struct sec_req *req = aead_request_ctx(a_req); + struct sec_ctx *ctx = crypto_aead_ctx(tfm); + int ret; + + req->flag = a_req->base.flags; + req->aead_req.aead_req = a_req; + req->c_req.encrypt = encrypt; + req->ctx = ctx; + + ret = sec_aead_param_check(ctx, req); + if (unlikely(ret)) { + if (ctx->a_ctx.fallback) + return sec_aead_soft_crypto(ctx, a_req, encrypt); + return -EINVAL; + } + + return ctx->req_op->process(ctx, req); +} + +static int sec_aead_encrypt(struct aead_request *a_req) +{ + return sec_aead_crypto(a_req, true); +} + +static int sec_aead_decrypt(struct aead_request *a_req) +{ + return sec_aead_crypto(a_req, false); +} + +#define SEC_AEAD_ALG(sec_cra_name, sec_set_key, ctx_init,\ + ctx_exit, blk_size, iv_size, max_authsize)\ +{\ + .base = {\ + .cra_name = sec_cra_name,\ + .cra_driver_name = "hisi_sec_"sec_cra_name,\ + .cra_priority = SEC_PRIORITY,\ + .cra_flags = CRYPTO_ALG_ASYNC |\ + CRYPTO_ALG_NEED_FALLBACK,\ + .cra_blocksize = blk_size,\ + .cra_ctxsize = sizeof(struct sec_ctx),\ + .cra_module = THIS_MODULE,\ + },\ + .init = ctx_init,\ + .exit = ctx_exit,\ + .setkey = sec_set_key,\ + .setauthsize = sec_aead_setauthsize,\ + .decrypt = sec_aead_decrypt,\ + .encrypt = sec_aead_encrypt,\ + .ivsize = iv_size,\ + .maxauthsize = max_authsize,\ +} + +static struct sec_aead sec_aeads[] = { + { + .alg_msk = BIT(6), + .alg = SEC_AEAD_ALG("ccm(aes)", sec_setkey_aes_ccm, sec_aead_xcm_ctx_init, + sec_aead_xcm_ctx_exit, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE, + AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(7), + .alg = SEC_AEAD_ALG("gcm(aes)", sec_setkey_aes_gcm, sec_aead_xcm_ctx_init, + sec_aead_xcm_ctx_exit, SEC_MIN_BLOCK_SZ, SEC_AIV_SIZE, + AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(17), + .alg = SEC_AEAD_ALG("ccm(sm4)", sec_setkey_sm4_ccm, sec_aead_xcm_ctx_init, + sec_aead_xcm_ctx_exit, SEC_MIN_BLOCK_SZ, AES_BLOCK_SIZE, + AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(18), + .alg = SEC_AEAD_ALG("gcm(sm4)", sec_setkey_sm4_gcm, sec_aead_xcm_ctx_init, + sec_aead_xcm_ctx_exit, SEC_MIN_BLOCK_SZ, SEC_AIV_SIZE, + AES_BLOCK_SIZE), + }, + { + .alg_msk = BIT(43), + .alg = SEC_AEAD_ALG("authenc(hmac(sha1),cbc(aes))", sec_setkey_aes_cbc_sha1, + sec_aead_sha1_ctx_init, sec_aead_ctx_exit, AES_BLOCK_SIZE, + AES_BLOCK_SIZE, SHA1_DIGEST_SIZE), + }, + { + .alg_msk = BIT(44), + .alg = SEC_AEAD_ALG("authenc(hmac(sha256),cbc(aes))", sec_setkey_aes_cbc_sha256, + sec_aead_sha256_ctx_init, sec_aead_ctx_exit, AES_BLOCK_SIZE, + AES_BLOCK_SIZE, SHA256_DIGEST_SIZE), + }, + { + .alg_msk = BIT(45), + .alg = SEC_AEAD_ALG("authenc(hmac(sha512),cbc(aes))", sec_setkey_aes_cbc_sha512, + sec_aead_sha512_ctx_init, sec_aead_ctx_exit, AES_BLOCK_SIZE, + AES_BLOCK_SIZE, SHA512_DIGEST_SIZE), + }, +}; + +static void sec_unregister_skcipher(u64 alg_mask, int end) +{ + int i; + + for (i = 0; i < end; i++) + if (sec_skciphers[i].alg_msk & alg_mask) + crypto_unregister_skcipher(&sec_skciphers[i].alg); +} + +static int sec_register_skcipher(u64 alg_mask) +{ + int i, ret, count; + + count = ARRAY_SIZE(sec_skciphers); + + for (i = 0; i < count; i++) { + if (!(sec_skciphers[i].alg_msk & alg_mask)) + continue; + + ret = crypto_register_skcipher(&sec_skciphers[i].alg); + if (ret) + goto err; + } + + return 0; + +err: + sec_unregister_skcipher(alg_mask, i); + + return ret; +} + +static void sec_unregister_aead(u64 alg_mask, int end) +{ + int i; + + for (i = 0; i < end; i++) + if (sec_aeads[i].alg_msk & alg_mask) + crypto_unregister_aead(&sec_aeads[i].alg); +} + +static int sec_register_aead(u64 alg_mask) +{ + int i, ret, count; + + count = ARRAY_SIZE(sec_aeads); + + for (i = 0; i < count; i++) { + if (!(sec_aeads[i].alg_msk & alg_mask)) + continue; + + ret = crypto_register_aead(&sec_aeads[i].alg); + if (ret) + goto err; + } + + return 0; + +err: + sec_unregister_aead(alg_mask, i); + + return ret; +} + +int sec_register_to_crypto(struct hisi_qm *qm) +{ + u64 alg_mask = sec_get_alg_bitmap(qm, SEC_DRV_ALG_BITMAP_HIGH, SEC_DRV_ALG_BITMAP_LOW); + int ret; + + ret = sec_register_skcipher(alg_mask); + if (ret) + return ret; + + ret = sec_register_aead(alg_mask); + if (ret) + sec_unregister_skcipher(alg_mask, ARRAY_SIZE(sec_skciphers)); + + return ret; +} + +void sec_unregister_from_crypto(struct hisi_qm *qm) +{ + u64 alg_mask = sec_get_alg_bitmap(qm, SEC_DRV_ALG_BITMAP_HIGH, SEC_DRV_ALG_BITMAP_LOW); + + sec_unregister_aead(alg_mask, ARRAY_SIZE(sec_aeads)); + sec_unregister_skcipher(alg_mask, ARRAY_SIZE(sec_skciphers)); +} diff --git a/kae_driver/hisilicon/sec2/sec_crypto.h b/kae_driver/hisilicon/sec2/sec_crypto.h new file mode 100755 index 0000000..d033f63 --- /dev/null +++ b/kae_driver/hisilicon/sec2/sec_crypto.h @@ -0,0 +1,410 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ + +#ifndef __HISI_SEC_V2_CRYPTO_H +#define __HISI_SEC_V2_CRYPTO_H + +#define SEC_AIV_SIZE 12 +#define SEC_IV_SIZE 24 +#define SEC_MAX_KEY_SIZE 64 +#define SEC_MAX_AKEY_SIZE 128 +#define SEC_COMM_SCENE 0 +#define SEC_MIN_BLOCK_SZ 1 + +enum sec_calg { + SEC_CALG_3DES = 0x1, + SEC_CALG_AES = 0x2, + SEC_CALG_SM4 = 0x3, +}; + +enum sec_hash_alg { + SEC_A_HMAC_SHA1 = 0x10, + SEC_A_HMAC_SHA256 = 0x11, + SEC_A_HMAC_SHA512 = 0x15, +}; + +enum sec_mac_len { + SEC_HMAC_CCM_MAC = 16, + SEC_HMAC_GCM_MAC = 16, + SEC_SM3_MAC = 32, + SEC_HMAC_SM3_MAC = 32, + SEC_HMAC_MD5_MAC = 16, + SEC_HMAC_SHA1_MAC = 20, + SEC_HMAC_SHA256_MAC = 32, + SEC_HMAC_SHA512_MAC = 64, +}; + +enum sec_cmode { + SEC_CMODE_ECB = 0x0, + SEC_CMODE_CBC = 0x1, + SEC_CMODE_CFB = 0x2, + SEC_CMODE_OFB = 0x3, + SEC_CMODE_CTR = 0x4, + SEC_CMODE_CCM = 0x5, + SEC_CMODE_GCM = 0x6, + SEC_CMODE_XTS = 0x7, +}; + +enum sec_ckey_type { + SEC_CKEY_128BIT = 0x0, + SEC_CKEY_192BIT = 0x1, + SEC_CKEY_256BIT = 0x2, + SEC_CKEY_3DES_3KEY = 0x1, + SEC_CKEY_3DES_2KEY = 0x3, +}; + +enum sec_bd_type { + SEC_BD_TYPE1 = 0x1, + SEC_BD_TYPE2 = 0x2, + SEC_BD_TYPE3 = 0x3, +}; + +enum sec_auth { + SEC_NO_AUTH = 0x0, + SEC_AUTH_TYPE1 = 0x1, + SEC_AUTH_TYPE2 = 0x2, +}; + +enum sec_cipher_dir { + SEC_CIPHER_ENC = 0x1, + SEC_CIPHER_DEC = 0x2, +}; + +enum sec_addr_type { + SEC_PBUF = 0x0, + SEC_SGL = 0x1, + SEC_PRP = 0x2, +}; + +struct bd_status { + u64 tag; + u8 done; + u8 err_type; + u16 flag; + u16 icv; +}; + +enum { + AUTHPAD_PAD, + AUTHPAD_NOPAD, +}; + +enum { + AIGEN_GEN, + AIGEN_NOGEN, +}; + +struct sec_sqe_type2 { + /* + * mac_len: 0~4 bits + * a_key_len: 5~10 bits + * a_alg: 11~16 bits + */ + __le32 mac_key_alg; + + /* + * c_icv_len: 0~5 bits + * c_width: 6~8 bits + * c_key_len: 9~11 bits + * c_mode: 12~15 bits + */ + __le16 icvw_kmode; + + /* c_alg: 0~3 bits */ + __u8 c_alg; + __u8 rsvd4; + + /* + * a_len: 0~23 bits + * iv_offset_l: 24~31 bits + */ + __le32 alen_ivllen; + + /* + * c_len: 0~23 bits + * iv_offset_h: 24~31 bits + */ + __le32 clen_ivhlen; + + __le16 auth_src_offset; + __le16 cipher_src_offset; + __le16 cs_ip_header_offset; + __le16 cs_udp_header_offset; + __le16 pass_word_len; + __le16 dk_len; + __u8 salt3; + __u8 salt2; + __u8 salt1; + __u8 salt0; + + __le16 tag; + __le16 rsvd5; + + /* + * c_pad_type: 0~3 bits + * c_pad_len: 4~11 bits + * c_pad_data_type: 12~15 bits + */ + __le16 cph_pad; + + /* c_pad_len_field: 0~1 bits */ + __le16 c_pad_len_field; + + __le64 long_a_data_len; + __le64 a_ivin_addr; + __le64 a_key_addr; + __le64 mac_addr; + __le64 c_ivin_addr; + __le64 c_key_addr; + + __le64 data_src_addr; + __le64 data_dst_addr; + + /* + * done: 0 bit + * icv: 1~3 bits + * csc: 4~6 bits + * flag: 7-10 bits + * dif_check: 11~13 bits + */ + __le16 done_flag; + + __u8 error_type; + __u8 warning_type; + __u8 mac_i3; + __u8 mac_i2; + __u8 mac_i1; + __u8 mac_i0; + __le16 check_sum_i; + __u8 tls_pad_len_i; + __u8 rsvd12; + __le32 counter; +}; + +struct sec_sqe { + /* + * type: 0~3 bits + * cipher: 4~5 bits + * auth: 6~7 bit s + */ + __u8 type_cipher_auth; + + /* + * seq: 0 bit + * de: 1~2 bits + * scene: 3~6 bits + * src_addr_type: ~7 bit, with sdm_addr_type 0-1 bits + */ + __u8 sds_sa_type; + + /* + * src_addr_type: 0~1 bits, not used now, + * if support PRP, set this field, or set zero. + * dst_addr_type: 2~4 bits + * mac_addr_type: 5~7 bits + */ + __u8 sdm_addr_type; + __u8 rsvd0; + + /* + * nonce_len(type2): 0~3 bits + * huk(type2): 4 bit + * key_s(type2): 5 bit + * ci_gen: 6~7 bits + */ + __u8 huk_key_ci; + + /* + * ai_gen: 0~1 bits + * a_pad(type2): 2~3 bits + * c_s(type2): 4~5 bits + */ + __u8 ai_apd_cs; + + /* + * rhf(type2): 0 bit + * c_key_type: 1~2 bits + * a_key_type: 3~4 bits + * write_frame_len(type2): 5~7 bits + */ + __u8 rca_key_frm; + + /* + * cal_iv_addr_en(type2): 0 bit + * tls_up(type2): 1 bit + * inveld: 7 bit + */ + __u8 iv_tls_ld; + + /* Just using type2 BD now */ + struct sec_sqe_type2 type2; +}; + +struct bd3_auth_ivin { + __le64 a_ivin_addr; + __le32 rsvd0; + __le32 rsvd1; +} __packed __aligned(4); + +struct bd3_skip_data { + __le32 rsvd0; + + /* + * gran_num: 0~15 bits + * reserved: 16~31 bits + */ + __le32 gran_num; + + /* + * src_skip_data_len: 0~24 bits + * reserved: 25~31 bits + */ + __le32 src_skip_data_len; + + /* + * dst_skip_data_len: 0~24 bits + * reserved: 25~31 bits + */ + __le32 dst_skip_data_len; +}; + +struct bd3_stream_scene { + __le64 c_ivin_addr; + __le64 long_a_data_len; + + /* + * auth_pad: 0~1 bits + * stream_protocol: 2~4 bits + * reserved: 5~7 bits + */ + __u8 stream_auth_pad; + __u8 plaintext_type; + __le16 pad_len_1p3; +} __packed __aligned(4); + +struct bd3_no_scene { + __le64 c_ivin_addr; + __le32 rsvd0; + __le32 rsvd1; + __le32 rsvd2; +} __packed __aligned(4); + +struct bd3_check_sum { + __u8 rsvd0; + __u8 hac_sva_status; + __le16 check_sum_i; +}; + +struct bd3_tls_type_back { + __u8 tls_1p3_type_back; + __u8 hac_sva_status; + __le16 pad_len_1p3_back; +}; + +struct sec_sqe3 { + /* + * type: 0~3 bit + * bd_invalid: 4 bit + * scene: 5~8 bit + * de: 9~10 bit + * src_addr_type: 11~13 bit + * dst_addr_type: 14~16 bit + * mac_addr_type: 17~19 bit + * reserved: 20~31 bits + */ + __le32 bd_param; + + /* + * cipher: 0~1 bits + * ci_gen: 2~3 bit + * c_icv_len: 4~9 bit + * c_width: 10~12 bits + * c_key_len: 13~15 bits + */ + __le16 c_icv_key; + + /* + * c_mode : 0~3 bits + * c_alg : 4~7 bits + */ + __u8 c_mode_alg; + + /* + * nonce_len : 0~3 bits + * huk : 4 bits + * cal_iv_addr_en : 5 bits + * seq : 6 bits + * reserved : 7 bits + */ + __u8 huk_iv_seq; + + __le64 tag; + __le64 data_src_addr; + __le64 a_key_addr; + union { + struct bd3_auth_ivin auth_ivin; + struct bd3_skip_data skip_data; + }; + + __le64 c_key_addr; + + /* + * auth: 0~1 bits + * ai_gen: 2~3 bits + * mac_len: 4~8 bits + * akey_len: 9~14 bits + * a_alg: 15~20 bits + * key_sel: 21~24 bits + * ctr_count_mode/sm4_xts: 25~26 bits + * sva_prefetch: 27 bits + * key_wrap_num: 28~30 bits + * update_key: 31 bits + */ + __le32 auth_mac_key; + __le32 salt; + __le16 auth_src_offset; + __le16 cipher_src_offset; + + /* + * auth_len: 0~23 bit + * auth_key_offset: 24~31 bits + */ + __le32 a_len_key; + + /* + * cipher_len: 0~23 bit + * auth_ivin_offset: 24~31 bits + */ + __le32 c_len_ivin; + __le64 data_dst_addr; + __le64 mac_addr; + union { + struct bd3_stream_scene stream_scene; + struct bd3_no_scene no_scene; + }; + + /* + * done: 0 bit + * icv: 1~3 bit + * csc: 4~6 bit + * flag: 7~10 bit + * reserved: 11~15 bit + */ + __le16 done_flag; + __u8 error_type; + __u8 warning_type; + union { + __le32 mac_i; + __le32 kek_key_addr_l; + }; + union { + __le32 kek_key_addr_h; + struct bd3_check_sum check_sum; + struct bd3_tls_type_back tls_type_back; + }; + __le32 counter; +} __packed __aligned(4); + +int sec_register_to_crypto(struct hisi_qm *qm); +void sec_unregister_from_crypto(struct hisi_qm *qm); +#endif diff --git a/kae_driver/hisilicon/sec2/sec_main.c b/kae_driver/hisilicon/sec2/sec_main.c new file mode 100755 index 0000000..e28b501 --- /dev/null +++ b/kae_driver/hisilicon/sec2/sec_main.c @@ -0,0 +1,1370 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "sec.h" + +#define SEC_VF_NUM 63 +#define SEC_QUEUE_NUM_V1 4096 +#define PCI_DEVICE_ID_HUAWEI_SEC_PF 0xa255 +#define PCI_DEVICE_ID_HUAWEI_SEC_VF 0xa256 + +#define SEC_BD_ERR_CHK_EN0 0xEFFFFFFF +#define SEC_BD_ERR_CHK_EN1 0x7ffff7fd +#define SEC_BD_ERR_CHK_EN3 0xffffbfff + +#define SEC_SQE_SIZE 128 +#define SEC_PF_DEF_Q_NUM 256 +#define SEC_PF_DEF_Q_BASE 0 +#define SEC_CTX_Q_NUM_DEF 2 +#define SEC_CTX_Q_NUM_MAX 32 + +#define SEC_CTRL_CNT_CLR_CE 0x301120 +#define SEC_CTRL_CNT_CLR_CE_BIT BIT(0) +#define SEC_CORE_INT_SOURCE 0x301010 +#define SEC_CORE_INT_MASK 0x301000 +#define SEC_CORE_INT_STATUS 0x301008 +#define SEC_CORE_SRAM_ECC_ERR_INFO 0x301C14 +#define SEC_ECC_NUM 16 +#define SEC_ECC_MASH 0xFF +#define SEC_CORE_INT_DISABLE 0x0 + +#define SEC_RAS_CE_REG 0x301050 +#define SEC_RAS_FE_REG 0x301054 +#define SEC_RAS_NFE_REG 0x301058 +#define SEC_RAS_FE_ENB_MSK 0x0 +#define SEC_OOO_SHUTDOWN_SEL 0x301014 +#define SEC_RAS_DISABLE 0x0 +#define SEC_MEM_START_INIT_REG 0x301100 +#define SEC_MEM_INIT_DONE_REG 0x301104 + +/* clock gating */ +#define SEC_CONTROL_REG 0x301200 +#define SEC_DYNAMIC_GATE_REG 0x30121c +#define SEC_CORE_AUTO_GATE 0x30212c +#define SEC_DYNAMIC_GATE_EN 0x7fff +#define SEC_CORE_AUTO_GATE_EN GENMASK(3, 0) +#define SEC_CLK_GATE_ENABLE BIT(3) +#define SEC_CLK_GATE_DISABLE (~BIT(3)) + +#define SEC_TRNG_EN_SHIFT 8 +#define SEC_AXI_SHUTDOWN_ENABLE BIT(12) +#define SEC_AXI_SHUTDOWN_DISABLE 0xFFFFEFFF + +#define SEC_INTERFACE_USER_CTRL0_REG 0x301220 +#define SEC_INTERFACE_USER_CTRL1_REG 0x301224 +#define SEC_SAA_EN_REG 0x301270 +#define SEC_BD_ERR_CHK_EN_REG0 0x301380 +#define SEC_BD_ERR_CHK_EN_REG1 0x301384 +#define SEC_BD_ERR_CHK_EN_REG3 0x30138c + +#define SEC_USER0_SMMU_NORMAL (BIT(23) | BIT(15)) +#define SEC_USER1_SMMU_NORMAL (BIT(31) | BIT(23) | BIT(15) | BIT(7)) +#define SEC_USER1_ENABLE_CONTEXT_SSV BIT(24) +#define SEC_USER1_ENABLE_DATA_SSV BIT(16) +#define SEC_USER1_WB_CONTEXT_SSV BIT(8) +#define SEC_USER1_WB_DATA_SSV BIT(0) +#define SEC_USER1_SVA_SET (SEC_USER1_ENABLE_CONTEXT_SSV | \ + SEC_USER1_ENABLE_DATA_SSV | \ + SEC_USER1_WB_CONTEXT_SSV | \ + SEC_USER1_WB_DATA_SSV) +#define SEC_USER1_SMMU_SVA (SEC_USER1_SMMU_NORMAL | SEC_USER1_SVA_SET) +#define SEC_USER1_SMMU_MASK (~SEC_USER1_SVA_SET) +#define SEC_INTERFACE_USER_CTRL0_REG_V3 0x302220 +#define SEC_INTERFACE_USER_CTRL1_REG_V3 0x302224 +#define SEC_USER1_SMMU_NORMAL_V3 (BIT(23) | BIT(17) | BIT(11) | BIT(5)) +#define SEC_USER1_SMMU_MASK_V3 0xFF79E79E +#define SEC_CORE_INT_STATUS_M_ECC BIT(2) + +#define SEC_PREFETCH_CFG 0x301130 +#define SEC_SVA_TRANS 0x301EC4 +#define SEC_PREFETCH_ENABLE (~(BIT(0) | BIT(1) | BIT(11))) +#define SEC_PREFETCH_DISABLE BIT(1) +#define SEC_SVA_DISABLE_READY (BIT(7) | BIT(11)) + +#define SEC_DELAY_10_US 10 +#define SEC_POLL_TIMEOUT_US 1000 +#define SEC_DBGFS_VAL_MAX_LEN 20 +#define SEC_SINGLE_PORT_MAX_TRANS 0x2060 + +#define SEC_SQE_MASK_OFFSET 64 +#define SEC_SQE_MASK_LEN 48 +#define SEC_SHAPER_TYPE_RATE 400 + +#define SEC_DFX_BASE 0x301000 +#define SEC_DFX_CORE 0x302100 +#define SEC_DFX_COMMON1 0x301600 +#define SEC_DFX_COMMON2 0x301C00 +#define SEC_DFX_BASE_LEN 0x9D +#define SEC_DFX_CORE_LEN 0x32B +#define SEC_DFX_COMMON1_LEN 0x45 +#define SEC_DFX_COMMON2_LEN 0xBA + +#define SEC_ALG_BITMAP_SHIFT 32 + +#define SEC_CIPHER_BITMAP (GENMASK_ULL(5, 0) | GENMASK_ULL(16, 12) | \ + GENMASK(24, 21)) +#define SEC_DIGEST_BITMAP (GENMASK_ULL(11, 8) | GENMASK_ULL(20, 19) | \ + GENMASK_ULL(42, 25)) +#define SEC_AEAD_BITMAP (GENMASK_ULL(7, 6) | GENMASK_ULL(18, 17) | \ + GENMASK_ULL(45, 43)) +#define SEC_DEV_ALG_MAX_LEN 256 + +struct sec_hw_error { + u32 int_msk; + const char *msg; +}; + +struct sec_dfx_item { + const char *name; + u32 offset; +}; + +struct sec_dev_alg { + u64 alg_msk; + const char *algs; +}; + +static const char sec_name[] = "hisi_sec2"; +static struct dentry *sec_debugfs_root; + +static struct hisi_qm_list sec_devices = { + .register_to_crypto = sec_register_to_crypto, + .unregister_from_crypto = sec_unregister_from_crypto, +}; + +static const struct hisi_qm_cap_info sec_basic_info[] = { + {SEC_QM_NFE_MASK_CAP, 0x3124, 0, GENMASK(31, 0), 0x0, 0x1C77, 0x7C77}, + {SEC_QM_RESET_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0xC77, 0x6C77}, + {SEC_QM_OOO_SHUTDOWN_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0x4, 0x6C77}, + {SEC_QM_CE_MASK_CAP, 0x312C, 0, GENMASK(31, 0), 0x0, 0x8, 0x8}, + {SEC_NFE_MASK_CAP, 0x3130, 0, GENMASK(31, 0), 0x0, 0x177, 0x60177}, + {SEC_RESET_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x177, 0x177}, + {SEC_OOO_SHUTDOWN_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x4, 0x177}, + {SEC_CE_MASK_CAP, 0x3138, 0, GENMASK(31, 0), 0x0, 0x88, 0xC088}, + {SEC_CLUSTER_NUM_CAP, 0x313c, 20, GENMASK(3, 0), 0x1, 0x1, 0x1}, + {SEC_CORE_TYPE_NUM_CAP, 0x313c, 16, GENMASK(3, 0), 0x1, 0x1, 0x1}, + {SEC_CORE_NUM_CAP, 0x313c, 8, GENMASK(7, 0), 0x4, 0x4, 0x4}, + {SEC_CORES_PER_CLUSTER_NUM_CAP, 0x313c, 0, GENMASK(7, 0), 0x4, 0x4, 0x4}, + {SEC_CORE_ENABLE_BITMAP, 0x3140, 32, GENMASK(31, 0), 0x17F, 0x17F, 0xF}, + {SEC_DRV_ALG_BITMAP_LOW, 0x3144, 0, GENMASK(31, 0), 0x18050CB, 0x18050CB, 0x187F0FF}, + {SEC_DRV_ALG_BITMAP_HIGH, 0x3148, 0, GENMASK(31, 0), 0x395C, 0x395C, 0x395C}, + {SEC_DEV_ALG_BITMAP_LOW, 0x314c, 0, GENMASK(31, 0), 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}, + {SEC_DEV_ALG_BITMAP_HIGH, 0x3150, 0, GENMASK(31, 0), 0x3FFF, 0x3FFF, 0x3FFF}, + {SEC_CORE1_ALG_BITMAP_LOW, 0x3154, 0, GENMASK(31, 0), 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}, + {SEC_CORE1_ALG_BITMAP_HIGH, 0x3158, 0, GENMASK(31, 0), 0x3FFF, 0x3FFF, 0x3FFF}, + {SEC_CORE2_ALG_BITMAP_LOW, 0x315c, 0, GENMASK(31, 0), 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}, + {SEC_CORE2_ALG_BITMAP_HIGH, 0x3160, 0, GENMASK(31, 0), 0x3FFF, 0x3FFF, 0x3FFF}, + {SEC_CORE3_ALG_BITMAP_LOW, 0x3164, 0, GENMASK(31, 0), 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}, + {SEC_CORE3_ALG_BITMAP_HIGH, 0x3168, 0, GENMASK(31, 0), 0x3FFF, 0x3FFF, 0x3FFF}, + {SEC_CORE4_ALG_BITMAP_LOW, 0x316c, 0, GENMASK(31, 0), 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}, + {SEC_CORE4_ALG_BITMAP_HIGH, 0x3170, 0, GENMASK(31, 0), 0x3FFF, 0x3FFF, 0x3FFF}, +}; + +static const struct sec_dev_alg sec_dev_algs[] = { { + .alg_msk = SEC_CIPHER_BITMAP, + .algs = "cipher\n", + }, { + .alg_msk = SEC_DIGEST_BITMAP, + .algs = "digest\n", + }, { + .alg_msk = SEC_AEAD_BITMAP, + .algs = "aead\n", + }, +}; + +static const struct sec_hw_error sec_hw_errors[] = { + { + .int_msk = BIT(0), + .msg = "sec_axi_rresp_err_rint" + }, + { + .int_msk = BIT(1), + .msg = "sec_axi_bresp_err_rint" + }, + { + .int_msk = BIT(2), + .msg = "sec_ecc_2bit_err_rint" + }, + { + .int_msk = BIT(3), + .msg = "sec_ecc_1bit_err_rint" + }, + { + .int_msk = BIT(4), + .msg = "sec_req_trng_timeout_rint" + }, + { + .int_msk = BIT(5), + .msg = "sec_fsm_hbeat_rint" + }, + { + .int_msk = BIT(6), + .msg = "sec_channel_req_rng_timeout_rint" + }, + { + .int_msk = BIT(7), + .msg = "sec_bd_err_rint" + }, + { + .int_msk = BIT(8), + .msg = "sec_chain_buff_err_rint" + }, + { + .int_msk = BIT(14), + .msg = "sec_no_secure_access" + }, + { + .int_msk = BIT(15), + .msg = "sec_wrapping_key_auth_err" + }, + { + .int_msk = BIT(16), + .msg = "sec_km_key_crc_fail" + }, + { + .int_msk = BIT(17), + .msg = "sec_axi_poison_err" + }, + { + .int_msk = BIT(18), + .msg = "sec_sva_err" + }, + {} +}; + +static const char * const sec_dbg_file_name[] = { + [SEC_CLEAR_ENABLE] = "clear_enable", +}; + +static struct sec_dfx_item sec_dfx_labels[] = { + {"send_cnt", offsetof(struct sec_dfx, send_cnt)}, + {"recv_cnt", offsetof(struct sec_dfx, recv_cnt)}, + {"send_busy_cnt", offsetof(struct sec_dfx, send_busy_cnt)}, + {"recv_busy_cnt", offsetof(struct sec_dfx, recv_busy_cnt)}, + {"err_bd_cnt", offsetof(struct sec_dfx, err_bd_cnt)}, + {"invalid_req_cnt", offsetof(struct sec_dfx, invalid_req_cnt)}, + {"done_flag_cnt", offsetof(struct sec_dfx, done_flag_cnt)}, +}; + +static const struct debugfs_reg32 sec_dfx_regs[] = { + {"SEC_PF_ABNORMAL_INT_SOURCE ", 0x301010}, + {"SEC_SAA_EN ", 0x301270}, + {"SEC_BD_LATENCY_MIN ", 0x301600}, + {"SEC_BD_LATENCY_MAX ", 0x301608}, + {"SEC_BD_LATENCY_AVG ", 0x30160C}, + {"SEC_BD_NUM_IN_SAA0 ", 0x301670}, + {"SEC_BD_NUM_IN_SAA1 ", 0x301674}, + {"SEC_BD_NUM_IN_SEC ", 0x301680}, + {"SEC_ECC_1BIT_CNT ", 0x301C00}, + {"SEC_ECC_1BIT_INFO ", 0x301C04}, + {"SEC_ECC_2BIT_CNT ", 0x301C10}, + {"SEC_ECC_2BIT_INFO ", 0x301C14}, + {"SEC_BD_SAA0 ", 0x301C20}, + {"SEC_BD_SAA1 ", 0x301C24}, + {"SEC_BD_SAA2 ", 0x301C28}, + {"SEC_BD_SAA3 ", 0x301C2C}, + {"SEC_BD_SAA4 ", 0x301C30}, + {"SEC_BD_SAA5 ", 0x301C34}, + {"SEC_BD_SAA6 ", 0x301C38}, + {"SEC_BD_SAA7 ", 0x301C3C}, + {"SEC_BD_SAA8 ", 0x301C40}, +}; + +/* define the SEC's dfx regs region and region length */ +static struct dfx_diff_registers sec_diff_regs[] = { + { + .reg_offset = SEC_DFX_BASE, + .reg_len = SEC_DFX_BASE_LEN, + }, { + .reg_offset = SEC_DFX_COMMON1, + .reg_len = SEC_DFX_COMMON1_LEN, + }, { + .reg_offset = SEC_DFX_COMMON2, + .reg_len = SEC_DFX_COMMON2_LEN, + }, { + .reg_offset = SEC_DFX_CORE, + .reg_len = SEC_DFX_CORE_LEN, + }, +}; + +static int sec_diff_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + + hisi_qm_acc_diff_regs_dump(qm, s, qm->debug.acc_diff_regs, + ARRAY_SIZE(sec_diff_regs)); + + return 0; +} +DEFINE_SHOW_ATTRIBUTE(sec_diff_regs); + +static int sec_pf_q_num_set(const char *val, const struct kernel_param *kp) +{ + return q_num_set(val, kp, PCI_DEVICE_ID_HUAWEI_SEC_PF); +} + +static const struct kernel_param_ops sec_pf_q_num_ops = { + .set = sec_pf_q_num_set, + .get = param_get_int, +}; + +static u32 pf_q_num = SEC_PF_DEF_Q_NUM; +module_param_cb(pf_q_num, &sec_pf_q_num_ops, &pf_q_num, 0444); +MODULE_PARM_DESC(pf_q_num, "Number of queues in PF(v1 2-4096, v2 2-1024)"); + +static int sec_ctx_q_num_set(const char *val, const struct kernel_param *kp) +{ + u32 ctx_q_num; + int ret; + + if (!val) + return -EINVAL; + + ret = kstrtou32(val, 10, &ctx_q_num); + if (ret) + return -EINVAL; + + if (!ctx_q_num || ctx_q_num > SEC_CTX_Q_NUM_MAX || ctx_q_num & 0x1) { + pr_err("ctx queue num[%u] is invalid!\n", ctx_q_num); + return -EINVAL; + } + + return param_set_int(val, kp); +} + +static const struct kernel_param_ops sec_ctx_q_num_ops = { + .set = sec_ctx_q_num_set, + .get = param_get_int, +}; +static u32 ctx_q_num = SEC_CTX_Q_NUM_DEF; +module_param_cb(ctx_q_num, &sec_ctx_q_num_ops, &ctx_q_num, 0444); +MODULE_PARM_DESC(ctx_q_num, "Queue num in ctx (2 default, 2, 4, ..., 32)"); + +static const struct kernel_param_ops vfs_num_ops = { + .set = vfs_num_set, + .get = param_get_int, +}; + +static u32 vfs_num; +module_param_cb(vfs_num, &vfs_num_ops, &vfs_num, 0444); +MODULE_PARM_DESC(vfs_num, "Number of VFs to enable(1-63), 0(default)"); + +void sec_destroy_qps(struct hisi_qp **qps, int qp_num) +{ + hisi_qm_free_qps(qps, qp_num); + kfree(qps); +} + +struct hisi_qp **sec_create_qps(void) +{ + int node = cpu_to_node(smp_processor_id()); + u32 ctx_num = ctx_q_num; + struct hisi_qp **qps; + int ret; + + qps = kcalloc(ctx_num, sizeof(struct hisi_qp *), GFP_KERNEL); + if (!qps) + return NULL; + + ret = hisi_qm_alloc_qps_node(&sec_devices, ctx_num, 0, node, qps); + if (!ret) + return qps; + + kfree(qps); + return NULL; +} + +u64 sec_get_alg_bitmap(struct hisi_qm *qm, u32 high, u32 low) +{ + u32 cap_val_h, cap_val_l; + + cap_val_h = hisi_qm_get_hw_info(qm, sec_basic_info, high, qm->cap_ver); + cap_val_l = hisi_qm_get_hw_info(qm, sec_basic_info, low, qm->cap_ver); + + return ((u64)cap_val_h << SEC_ALG_BITMAP_SHIFT) | (u64)cap_val_l; +} + +static const struct kernel_param_ops sec_uacce_mode_ops = { + .set = uacce_mode_set, + .get = param_get_int, +}; + +/* + * uacce_mode = 0 means sec only register to crypto, + * uacce_mode = 1 means sec both register to crypto and uacce. + */ +static u32 uacce_mode = UACCE_MODE_NOUACCE; +module_param_cb(uacce_mode, &sec_uacce_mode_ops, &uacce_mode, 0444); +MODULE_PARM_DESC(uacce_mode, UACCE_MODE_DESC); + +static const struct pci_device_id sec_dev_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_SEC_PF) }, + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_SEC_VF) }, + { 0, } +}; +MODULE_DEVICE_TABLE(pci, sec_dev_ids); + +static void sec_set_endian(struct hisi_qm *qm) +{ + u32 reg; + + reg = readl_relaxed(qm->io_base + SEC_CONTROL_REG); + reg &= ~(BIT(1) | BIT(0)); + if (!IS_ENABLED(CONFIG_64BIT)) + reg |= BIT(1); + + + if (!IS_ENABLED(CONFIG_CPU_LITTLE_ENDIAN)) + reg |= BIT(0); + + writel_relaxed(reg, qm->io_base + SEC_CONTROL_REG); +} + +static void sec_engine_sva_config(struct hisi_qm *qm) +{ + u32 reg; + + if (qm->ver > QM_HW_V2) { + reg = readl_relaxed(qm->io_base + + SEC_INTERFACE_USER_CTRL0_REG_V3); + reg |= SEC_USER0_SMMU_NORMAL; + writel_relaxed(reg, qm->io_base + + SEC_INTERFACE_USER_CTRL0_REG_V3); + + reg = readl_relaxed(qm->io_base + + SEC_INTERFACE_USER_CTRL1_REG_V3); + reg &= SEC_USER1_SMMU_MASK_V3; + reg |= SEC_USER1_SMMU_NORMAL_V3; + writel_relaxed(reg, qm->io_base + + SEC_INTERFACE_USER_CTRL1_REG_V3); + } else { + reg = readl_relaxed(qm->io_base + + SEC_INTERFACE_USER_CTRL0_REG); + reg |= SEC_USER0_SMMU_NORMAL; + writel_relaxed(reg, qm->io_base + + SEC_INTERFACE_USER_CTRL0_REG); + reg = readl_relaxed(qm->io_base + + SEC_INTERFACE_USER_CTRL1_REG); + reg &= SEC_USER1_SMMU_MASK; + if (qm->use_sva) + reg |= SEC_USER1_SMMU_SVA; + else + reg |= SEC_USER1_SMMU_NORMAL; + writel_relaxed(reg, qm->io_base + + SEC_INTERFACE_USER_CTRL1_REG); + } +} + +static void sec_open_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + /* Enable prefetch */ + val = readl_relaxed(qm->io_base + SEC_PREFETCH_CFG); + val &= SEC_PREFETCH_ENABLE; + writel(val, qm->io_base + SEC_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + SEC_PREFETCH_CFG, + val, !(val & SEC_PREFETCH_DISABLE), + SEC_DELAY_10_US, SEC_POLL_TIMEOUT_US); + if (ret) + pci_err(qm->pdev, "failed to open sva prefetch\n"); +} + +static void sec_close_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + val = readl_relaxed(qm->io_base + SEC_PREFETCH_CFG); + val |= SEC_PREFETCH_DISABLE; + writel(val, qm->io_base + SEC_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + SEC_SVA_TRANS, + val, !(val & SEC_SVA_DISABLE_READY), + SEC_DELAY_10_US, SEC_POLL_TIMEOUT_US); + if (ret) + pci_err(qm->pdev, "failed to close sva prefetch\n"); +} + +static void sec_enable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl_relaxed(qm->io_base + SEC_CONTROL_REG); + val |= SEC_CLK_GATE_ENABLE; + writel_relaxed(val, qm->io_base + SEC_CONTROL_REG); + + val = readl(qm->io_base + SEC_DYNAMIC_GATE_REG); + val |= SEC_DYNAMIC_GATE_EN; + writel(val, qm->io_base + SEC_DYNAMIC_GATE_REG); + + val = readl(qm->io_base + SEC_CORE_AUTO_GATE); + val |= SEC_CORE_AUTO_GATE_EN; + writel(val, qm->io_base + SEC_CORE_AUTO_GATE); +} + +static void sec_disable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + /* Kunpeng920 needs to close clock gating */ + val = readl_relaxed(qm->io_base + SEC_CONTROL_REG); + val &= SEC_CLK_GATE_DISABLE; + writel_relaxed(val, qm->io_base + SEC_CONTROL_REG); +} + +static int sec_engine_init(struct hisi_qm *qm) +{ + int ret; + u32 reg; + + /* disable clock gate control before mem init */ + sec_disable_clock_gate(qm); + + writel_relaxed(0x1, qm->io_base + SEC_MEM_START_INIT_REG); + + ret = readl_relaxed_poll_timeout(qm->io_base + SEC_MEM_INIT_DONE_REG, + reg, reg & 0x1, SEC_DELAY_10_US, + SEC_POLL_TIMEOUT_US); + if (ret) { + pci_err(qm->pdev, "fail to init sec mem\n"); + return ret; + } + + reg = readl_relaxed(qm->io_base + SEC_CONTROL_REG); + reg |= (0x1 << SEC_TRNG_EN_SHIFT); + writel_relaxed(reg, qm->io_base + SEC_CONTROL_REG); + + sec_engine_sva_config(qm); + + writel(SEC_SINGLE_PORT_MAX_TRANS, + qm->io_base + AM_CFG_SINGLE_PORT_MAX_TRANS); + + reg = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_CORE_ENABLE_BITMAP, qm->cap_ver); + writel(reg, qm->io_base + SEC_SAA_EN_REG); + + if (qm->ver < QM_HW_V3) { + /* HW V2 enable sm4 extra mode, as ctr/ecb */ + writel_relaxed(SEC_BD_ERR_CHK_EN0, + qm->io_base + SEC_BD_ERR_CHK_EN_REG0); + + /* HW V2 enable sm4 xts mode multiple iv */ + writel_relaxed(SEC_BD_ERR_CHK_EN1, + qm->io_base + SEC_BD_ERR_CHK_EN_REG1); + writel_relaxed(SEC_BD_ERR_CHK_EN3, + qm->io_base + SEC_BD_ERR_CHK_EN_REG3); + } + + /* config endian */ + sec_set_endian(qm); + + sec_enable_clock_gate(qm); + + return 0; +} + +static int sec_set_user_domain_and_cache(struct hisi_qm *qm) +{ + /* qm user domain */ + writel(AXUSER_BASE, qm->io_base + QM_ARUSER_M_CFG_1); + writel(ARUSER_M_CFG_ENABLE, qm->io_base + QM_ARUSER_M_CFG_ENABLE); + writel(AXUSER_BASE, qm->io_base + QM_AWUSER_M_CFG_1); + writel(AWUSER_M_CFG_ENABLE, qm->io_base + QM_AWUSER_M_CFG_ENABLE); + writel(WUSER_M_CFG_ENABLE, qm->io_base + QM_WUSER_M_CFG_ENABLE); + + /* qm cache */ + writel(AXI_M_CFG, qm->io_base + QM_AXI_M_CFG); + writel(AXI_M_CFG_ENABLE, qm->io_base + QM_AXI_M_CFG_ENABLE); + + /* disable FLR triggered by BME(bus master enable) */ + writel(PEH_AXUSER_CFG, qm->io_base + QM_PEH_AXUSER_CFG); + writel(PEH_AXUSER_CFG_ENABLE, qm->io_base + QM_PEH_AXUSER_CFG_ENABLE); + + /* enable sqc,cqc writeback */ + writel(SQC_CACHE_ENABLE | CQC_CACHE_ENABLE | SQC_CACHE_WB_ENABLE | + CQC_CACHE_WB_ENABLE | FIELD_PREP(SQC_CACHE_WB_THRD, 1) | + FIELD_PREP(CQC_CACHE_WB_THRD, 1), qm->io_base + QM_CACHE_CTL); + + return sec_engine_init(qm); +} + +/* sec_debug_regs_clear() - clear the sec debug regs */ +static void sec_debug_regs_clear(struct hisi_qm *qm) +{ + int i; + + /* clear sec dfx regs */ + writel(0x1, qm->io_base + SEC_CTRL_CNT_CLR_CE); + for (i = 0; i < ARRAY_SIZE(sec_dfx_regs); i++) + readl(qm->io_base + sec_dfx_regs[i].offset); + + /* clear rdclr_en */ + writel(0x0, qm->io_base + SEC_CTRL_CNT_CLR_CE); + + hisi_qm_debug_regs_clear(qm); +} + +static void sec_master_ooo_ctrl(struct hisi_qm *qm, bool enable) +{ + u32 val1, val2; + + val1 = readl(qm->io_base + SEC_CONTROL_REG); + if (enable) { + val1 |= SEC_AXI_SHUTDOWN_ENABLE; + val2 = hisi_qm_get_hw_info(qm, sec_basic_info, + SEC_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + } else { + val1 &= SEC_AXI_SHUTDOWN_DISABLE; + val2 = 0x0; + } + + if (qm->ver > QM_HW_V2) + writel(val2, qm->io_base + SEC_OOO_SHUTDOWN_SEL); + + writel(val1, qm->io_base + SEC_CONTROL_REG); +} + +static void sec_hw_error_enable(struct hisi_qm *qm) +{ + u32 ce, nfe; + + if (qm->ver == QM_HW_V1) { + writel(SEC_CORE_INT_DISABLE, qm->io_base + SEC_CORE_INT_MASK); + pci_info(qm->pdev, "V1 not support hw error handle\n"); + return; + } + + ce = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_CE_MASK_CAP, qm->cap_ver); + nfe = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_NFE_MASK_CAP, qm->cap_ver); + + /* clear SEC hw error source if having */ + writel(ce | nfe | SEC_RAS_FE_ENB_MSK, qm->io_base + SEC_CORE_INT_SOURCE); + + /* enable RAS int */ + writel(ce, qm->io_base + SEC_RAS_CE_REG); + writel(SEC_RAS_FE_ENB_MSK, qm->io_base + SEC_RAS_FE_REG); + writel(nfe, qm->io_base + SEC_RAS_NFE_REG); + + /* enable SEC block master OOO when nfe occurs on Kunpeng930 */ + sec_master_ooo_ctrl(qm, true); + + /* enable SEC hw error interrupts */ + writel(ce | nfe | SEC_RAS_FE_ENB_MSK, qm->io_base + SEC_CORE_INT_MASK); +} + +static void sec_hw_error_disable(struct hisi_qm *qm) +{ + /* disable SEC hw error interrupts */ + writel(SEC_CORE_INT_DISABLE, qm->io_base + SEC_CORE_INT_MASK); + + /* disable SEC block master OOO when nfe occurs on Kunpeng930 */ + sec_master_ooo_ctrl(qm, false); + + /* disable RAS int */ + writel(SEC_RAS_DISABLE, qm->io_base + SEC_RAS_CE_REG); + writel(SEC_RAS_DISABLE, qm->io_base + SEC_RAS_FE_REG); + writel(SEC_RAS_DISABLE, qm->io_base + SEC_RAS_NFE_REG); +} + +static u32 sec_clear_enable_read(struct hisi_qm *qm) +{ + return readl(qm->io_base + SEC_CTRL_CNT_CLR_CE) & + SEC_CTRL_CNT_CLR_CE_BIT; +} + +static int sec_clear_enable_write(struct hisi_qm *qm, u32 val) +{ + u32 tmp; + + if (val != 1 && val) + return -EINVAL; + + tmp = (readl(qm->io_base + SEC_CTRL_CNT_CLR_CE) & + ~SEC_CTRL_CNT_CLR_CE_BIT) | val; + writel(tmp, qm->io_base + SEC_CTRL_CNT_CLR_CE); + + return 0; +} + +static ssize_t sec_debug_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct sec_debug_file *file = filp->private_data; + char tbuf[SEC_DBGFS_VAL_MAX_LEN]; + struct hisi_qm *qm = file->qm; + u32 val; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + + switch (file->index) { + case SEC_CLEAR_ENABLE: + val = sec_clear_enable_read(qm); + break; + default: + goto err_input; + } + + spin_unlock_irq(&file->lock); + + hisi_qm_put_dfx_access(qm); + ret = snprintf(tbuf, SEC_DBGFS_VAL_MAX_LEN, "%u\n", val); + return simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return -EINVAL; +} + +static ssize_t sec_debug_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct sec_debug_file *file = filp->private_data; + char tbuf[SEC_DBGFS_VAL_MAX_LEN]; + struct hisi_qm *qm = file->qm; + unsigned long val; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= SEC_DBGFS_VAL_MAX_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, SEC_DBGFS_VAL_MAX_LEN - 1, + pos, buf, count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + if (kstrtoul(tbuf, 0, &val)) + return -EFAULT; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + + switch (file->index) { + case SEC_CLEAR_ENABLE: + ret = sec_clear_enable_write(qm, val); + if (ret) + goto err_input; + break; + default: + ret = -EINVAL; + goto err_input; + } + + ret = count; + + err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return ret; +} + +static const struct file_operations sec_dbg_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = sec_debug_read, + .write = sec_debug_write, +}; + +static int sec_debugfs_atomic64_get(void *data, u64 *val) +{ + *val = atomic64_read((atomic64_t *)data); + + return 0; +} + +static int sec_debugfs_atomic64_set(void *data, u64 val) +{ + if (val) + return -EINVAL; + + atomic64_set((atomic64_t *)data, 0); + + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(sec_atomic64_ops, sec_debugfs_atomic64_get, + sec_debugfs_atomic64_set, "%lld\n"); + +static int sec_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(sec_regs); + +static int sec_core_debug_init(struct hisi_qm *qm) +{ + struct dfx_diff_registers *sec_regs = qm->debug.acc_diff_regs; + struct sec_dev *sec = container_of(qm, struct sec_dev, qm); + struct device *dev = &qm->pdev->dev; + struct sec_dfx *dfx = &sec->debug.dfx; + struct debugfs_regset32 *regset; + struct dentry *tmp_d; + int i; + + tmp_d = debugfs_create_dir("sec_dfx", qm->debug.debug_root); + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOMEM; + + regset->regs = sec_dfx_regs; + regset->nregs = ARRAY_SIZE(sec_dfx_regs); + regset->base = qm->io_base; + regset->dev = dev; + + if (qm->pdev->device == PCI_DEVICE_ID_HUAWEI_SEC_PF) + debugfs_create_file("regs", 0444, tmp_d, regset, &sec_regs_fops); + if (qm->fun_type == QM_HW_PF && sec_regs) + debugfs_create_file("diff_regs", 0444, tmp_d, + qm, &sec_diff_regs_fops); + + for (i = 0; i < ARRAY_SIZE(sec_dfx_labels); i++) { + atomic64_t *data = (atomic64_t *)((uintptr_t)dfx + + sec_dfx_labels[i].offset); + debugfs_create_file(sec_dfx_labels[i].name, 0644, + tmp_d, data, &sec_atomic64_ops); + } + + return 0; +} + +static int sec_debug_init(struct hisi_qm *qm) +{ + struct sec_dev *sec = container_of(qm, struct sec_dev, qm); + int i; + + if (qm->pdev->device == PCI_DEVICE_ID_HUAWEI_SEC_PF) { + for (i = SEC_CLEAR_ENABLE; i < SEC_DEBUG_FILE_NUM; i++) { + spin_lock_init(&sec->debug.files[i].lock); + sec->debug.files[i].index = i; + sec->debug.files[i].qm = qm; + + debugfs_create_file(sec_dbg_file_name[i], 0600, + qm->debug.debug_root, + sec->debug.files + i, + &sec_dbg_fops); + } + } + + return sec_core_debug_init(qm); +} + +static int sec_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + qm->debug.debug_root = debugfs_create_dir(dev_name(dev), + sec_debugfs_root); + qm->debug.sqe_mask_offset = SEC_SQE_MASK_OFFSET; + qm->debug.sqe_mask_len = SEC_SQE_MASK_LEN; + + ret = hisi_qm_diff_regs_init(qm, sec_diff_regs, + ARRAY_SIZE(sec_diff_regs)); + if (ret) { + dev_warn(dev, "Failed to init SEC diff regs!\n"); + goto debugfs_remove; + } + + hisi_qm_debug_init(qm); + + ret = sec_debug_init(qm); + if (ret) + goto failed_to_create; + + return 0; + +failed_to_create: + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(sec_diff_regs)); +debugfs_remove: + debugfs_remove_recursive(sec_debugfs_root); + return ret; +} + +static void sec_debugfs_exit(struct hisi_qm *qm) +{ + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(sec_diff_regs)); + + debugfs_remove_recursive(qm->debug.debug_root); +} + +static int sec_show_last_regs_init(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + int i; + + debug->last_words = kcalloc(ARRAY_SIZE(sec_dfx_regs), + sizeof(unsigned int), GFP_KERNEL); + if (!debug->last_words) + return -ENOMEM; + + for (i = 0; i < ARRAY_SIZE(sec_dfx_regs); i++) + debug->last_words[i] = readl_relaxed(qm->io_base + + sec_dfx_regs[i].offset); + + return 0; +} + +static void sec_show_last_regs_uninit(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + kfree(debug->last_words); + debug->last_words = NULL; +} + +static void sec_show_last_dfx_regs(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + struct pci_dev *pdev = qm->pdev; + u32 val; + int i; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + /* dumps last word of the debugging registers during controller reset */ + for (i = 0; i < ARRAY_SIZE(sec_dfx_regs); i++) { + val = readl_relaxed(qm->io_base + sec_dfx_regs[i].offset); + if (val != debug->last_words[i]) + pci_info(pdev, "%s \t= 0x%08x => 0x%08x\n", + sec_dfx_regs[i].name, debug->last_words[i], val); + } +} + +static void sec_log_hw_error(struct hisi_qm *qm, u32 err_sts) +{ + const struct sec_hw_error *errs = sec_hw_errors; + struct device *dev = &qm->pdev->dev; + u32 err_val; + + while (errs->msg) { + if (errs->int_msk & err_sts) { + dev_err(dev, "%s [error status=0x%x] found\n", + errs->msg, errs->int_msk); + + if (SEC_CORE_INT_STATUS_M_ECC & errs->int_msk) { + err_val = readl(qm->io_base + + SEC_CORE_SRAM_ECC_ERR_INFO); + dev_err(dev, "multi ecc sram num=0x%x\n", + ((err_val) >> SEC_ECC_NUM) & + SEC_ECC_MASH); + } + } + errs++; + } +} + +static u32 sec_get_hw_err_status(struct hisi_qm *qm) +{ + return readl(qm->io_base + SEC_CORE_INT_STATUS); +} + +static void sec_clear_hw_err_status(struct hisi_qm *qm, u32 err_sts) +{ + u32 nfe; + + writel(err_sts, qm->io_base + SEC_CORE_INT_SOURCE); + nfe = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_NFE_MASK_CAP, qm->cap_ver); + writel(nfe, qm->io_base + SEC_RAS_NFE_REG); +} + +static void sec_open_axi_master_ooo(struct hisi_qm *qm) +{ + u32 val; + + val = readl(qm->io_base + SEC_CONTROL_REG); + writel(val & SEC_AXI_SHUTDOWN_DISABLE, qm->io_base + SEC_CONTROL_REG); + writel(val | SEC_AXI_SHUTDOWN_ENABLE, qm->io_base + SEC_CONTROL_REG); +} + +static void sec_err_info_init(struct hisi_qm *qm) +{ + struct hisi_qm_err_info *err_info = &qm->err_info; + + err_info->fe = SEC_RAS_FE_ENB_MSK; + err_info->ce = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_QM_CE_MASK_CAP, qm->cap_ver); + err_info->nfe = hisi_qm_get_hw_info(qm, sec_basic_info, SEC_QM_NFE_MASK_CAP, qm->cap_ver); + err_info->ecc_2bits_mask = SEC_CORE_INT_STATUS_M_ECC; + err_info->qm_shutdown_mask = hisi_qm_get_hw_info(qm, sec_basic_info, + SEC_QM_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->dev_shutdown_mask = hisi_qm_get_hw_info(qm, sec_basic_info, + SEC_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_reset_mask = hisi_qm_get_hw_info(qm, sec_basic_info, + SEC_QM_RESET_MASK_CAP, qm->cap_ver); + err_info->dev_reset_mask = hisi_qm_get_hw_info(qm, sec_basic_info, + SEC_RESET_MASK_CAP, qm->cap_ver); + err_info->msi_wr_port = BIT(0); + err_info->acpi_rst = "SRST"; +} + +static const struct hisi_qm_err_ini sec_err_ini = { + .hw_init = sec_set_user_domain_and_cache, + .hw_err_enable = sec_hw_error_enable, + .hw_err_disable = sec_hw_error_disable, + .get_dev_hw_err_status = sec_get_hw_err_status, + .clear_dev_hw_err_status = sec_clear_hw_err_status, + .log_dev_hw_err = sec_log_hw_error, + .open_axi_master_ooo = sec_open_axi_master_ooo, + .open_sva_prefetch = sec_open_sva_prefetch, + .close_sva_prefetch = sec_close_sva_prefetch, + .show_last_dfx_regs = sec_show_last_dfx_regs, + .err_info_init = sec_err_info_init, +}; + +static int sec_pf_probe_init(struct sec_dev *sec) +{ + struct hisi_qm *qm = &sec->qm; + int ret; + + qm->err_ini = &sec_err_ini; + qm->err_ini->err_info_init(qm); + + ret = sec_set_user_domain_and_cache(qm); + if (ret) + return ret; + + sec_open_sva_prefetch(qm); + hisi_qm_dev_err_init(qm); + sec_debug_regs_clear(qm); + ret = sec_show_last_regs_init(qm); + if (ret) + pci_err(qm->pdev, "Failed to init last word regs!\n"); + + return ret; +} + +static int sec_set_qm_algs(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + char *algs, *ptr; + u64 alg_mask; + int i; + + if (!qm->use_sva) + return 0; + + algs = devm_kzalloc(dev, SEC_DEV_ALG_MAX_LEN * sizeof(char), GFP_KERNEL); + if (!algs) + return -ENOMEM; + + alg_mask = sec_get_alg_bitmap(qm, SEC_DEV_ALG_BITMAP_HIGH, SEC_DEV_ALG_BITMAP_LOW); + + for (i = 0; i < ARRAY_SIZE(sec_dev_algs); i++) + if (alg_mask & sec_dev_algs[i].alg_msk) + strcat(algs, sec_dev_algs[i].algs); + + ptr = strrchr(algs, '\n'); + if (ptr) + *ptr = '\0'; + + qm->uacce->algs = algs; + + return 0; +} + +static int sec_qm_init(struct hisi_qm *qm, struct pci_dev *pdev) +{ + int ret; + + qm->pdev = pdev; + qm->ver = pdev->revision; + qm->mode = uacce_mode; + qm->sqe_size = SEC_SQE_SIZE; + qm->dev_name = sec_name; + + qm->fun_type = (pdev->device == PCI_DEVICE_ID_HUAWEI_SEC_PF) ? + QM_HW_PF : QM_HW_VF; + if (qm->fun_type == QM_HW_PF) { + qm->qp_base = SEC_PF_DEF_Q_BASE; + qm->qp_num = pf_q_num; + qm->debug.curr_qm_qp_num = pf_q_num; + qm->qm_list = &sec_devices; + } else if (qm->fun_type == QM_HW_VF && qm->ver == QM_HW_V1) { + /* + * have no way to get qm configure in VM in v1 hardware, + * so currently force PF to uses SEC_PF_DEF_Q_NUM, and force + * to trigger only one VF in v1 hardware. + * v2 hardware has no such problem. + */ + qm->qp_base = SEC_PF_DEF_Q_NUM; + qm->qp_num = SEC_QUEUE_NUM_V1 - SEC_PF_DEF_Q_NUM; + } + + ret = hisi_qm_init(qm); + if (ret) { + pci_err(qm->pdev, "Failed to init sec qm configures!\n"); + return ret; + } + + ret = sec_set_qm_algs(qm); + if (ret) { + pci_err(qm->pdev, "Failed to set sec algs!\n"); + hisi_qm_uninit(qm); + } + + return ret; +} + +static void sec_qm_uninit(struct hisi_qm *qm) +{ + hisi_qm_uninit(qm); +} + +static int sec_probe_init(struct sec_dev *sec) +{ + u32 type_rate = SEC_SHAPER_TYPE_RATE; + struct hisi_qm *qm = &sec->qm; + int ret; + + if (qm->fun_type == QM_HW_PF) { + ret = sec_pf_probe_init(sec); + if (ret) + return ret; + /* enable shaper type 0 */ + if (qm->ver >= QM_HW_V3) { + type_rate |= QM_SHAPER_ENABLE; + qm->type_rate = type_rate; + } + } + + return 0; +} + +static void sec_probe_uninit(struct hisi_qm *qm) +{ + hisi_qm_dev_err_uninit(qm); +} + +static void sec_iommu_used_check(struct sec_dev *sec) +{ + struct iommu_domain *domain; + struct device *dev = &sec->qm.pdev->dev; + + domain = iommu_get_domain_for_dev(dev); + + /* Check if iommu is used */ + sec->iommu_used = false; + if (domain) { + if (domain->type & __IOMMU_DOMAIN_PAGING) + sec->iommu_used = true; + dev_info(dev, "SMMU Opened, the iommu type = %u\n", + domain->type); + } +} + +static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct sec_dev *sec; + struct hisi_qm *qm; + int ret; + + sec = devm_kzalloc(&pdev->dev, sizeof(*sec), GFP_KERNEL); + if (!sec) + return -ENOMEM; + + qm = &sec->qm; + ret = sec_qm_init(qm, pdev); + if (ret) { + pci_err(pdev, "Failed to init SEC QM (%d)!\n", ret); + return ret; + } + + sec->ctx_q_num = ctx_q_num; + sec_iommu_used_check(sec); + + ret = sec_probe_init(sec); + if (ret) { + pci_err(pdev, "Failed to probe!\n"); + goto err_qm_uninit; + } + + ret = hisi_qm_start(qm); + if (ret) { + pci_err(pdev, "Failed to start sec qm!\n"); + goto err_probe_uninit; + } + + ret = sec_debugfs_init(qm); + if (ret) + pci_warn(pdev, "Failed to init debugfs!\n"); + + if (qm->qp_num >= ctx_q_num) { + ret = hisi_qm_alg_register(qm, &sec_devices); + if (ret < 0) { + pr_err("Failed to register driver to crypto.\n"); + goto err_qm_stop; + } + } else { + pci_warn(qm->pdev, + "Failed to use kernel mode, qp not enough!\n"); + } + + if (qm->uacce) { + ret = uacce_register(qm->uacce); + if (ret) { + pci_err(pdev, "failed to register uacce (%d)!\n", ret); + goto err_alg_unregister; + } + } + + if (qm->fun_type == QM_HW_PF && vfs_num) { + ret = hisi_qm_sriov_enable(pdev, vfs_num); + if (ret < 0) + goto err_alg_unregister; + } + + hisi_qm_pm_init(qm); + + return 0; + +err_alg_unregister: + if (qm->qp_num >= ctx_q_num) + hisi_qm_alg_unregister(qm, &sec_devices); +err_qm_stop: + sec_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); +err_probe_uninit: + sec_show_last_regs_uninit(qm); + sec_probe_uninit(qm); +err_qm_uninit: + sec_qm_uninit(qm); + return ret; +} + +static void sec_remove(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + + hisi_qm_pm_uninit(qm); + hisi_qm_wait_task_finish(qm, &sec_devices); + if (qm->qp_num >= ctx_q_num) + hisi_qm_alg_unregister(qm, &sec_devices); + + if (qm->fun_type == QM_HW_PF && qm->vfs_num) + hisi_qm_sriov_disable(pdev, true); + + sec_debugfs_exit(qm); + + (void)hisi_qm_stop(qm, QM_NORMAL); + + if (qm->fun_type == QM_HW_PF) + sec_debug_regs_clear(qm); + sec_show_last_regs_uninit(qm); + + sec_probe_uninit(qm); + + sec_qm_uninit(qm); +} + +static const struct dev_pm_ops sec_pm_ops = { + SET_RUNTIME_PM_OPS(hisi_qm_suspend, hisi_qm_resume, NULL) +}; + +static const struct pci_error_handlers sec_err_handler = { + .error_detected = hisi_qm_dev_err_detected, + .slot_reset = hisi_qm_dev_slot_reset, + .reset_prepare = hisi_qm_reset_prepare, + .reset_done = hisi_qm_reset_done, +}; + +static struct pci_driver sec_pci_driver = { + .name = "hisi_sec2", + .id_table = sec_dev_ids, + .probe = sec_probe, + .remove = sec_remove, + .err_handler = &sec_err_handler, + .sriov_configure = hisi_qm_sriov_configure, + .shutdown = hisi_qm_dev_shutdown, + .driver.pm = &sec_pm_ops, +}; + +static void sec_register_debugfs(void) +{ + if (!debugfs_initialized()) + return; + + sec_debugfs_root = debugfs_create_dir("hisi_sec2", NULL); +} + +static void sec_unregister_debugfs(void) +{ + debugfs_remove_recursive(sec_debugfs_root); +} + +static int __init sec_init(void) +{ + int ret; + + hisi_qm_init_list(&sec_devices); + sec_register_debugfs(); + + ret = pci_register_driver(&sec_pci_driver); + if (ret < 0) { + sec_unregister_debugfs(); + pr_err("Failed to register pci driver.\n"); + return ret; + } + + return 0; +} + +static void __exit sec_exit(void) +{ + pci_unregister_driver(&sec_pci_driver); + sec_unregister_debugfs(); +} + +module_init(sec_init); +module_exit(sec_exit); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Zaibo Xu "); +MODULE_AUTHOR("Longfang Liu "); +MODULE_AUTHOR("Kai Ye "); +MODULE_AUTHOR("Wei Zhang "); +MODULE_DESCRIPTION("Driver for HiSilicon SEC accelerator"); diff --git a/kae_driver/hisilicon/sgl.c b/kae_driver/hisilicon/sgl.c new file mode 100755 index 0000000..dc58206 --- /dev/null +++ b/kae_driver/hisilicon/sgl.c @@ -0,0 +1,290 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include "hisi_acc_qm.h" +#include +#include + +#define HISI_ACC_SGL_SGE_NR_MIN 1 +#define HISI_ACC_SGL_NR_MAX 256 +#define HISI_ACC_SGL_ALIGN_SIZE 64 +#define HISI_ACC_MEM_BLOCK_NR 5 + +struct acc_hw_sge { + dma_addr_t buf; + void *page_ctrl; + __le32 len; + __le32 pad; + __le32 pad0; + __le32 pad1; +}; + +/* use default sgl head size 64B */ +struct hisi_acc_hw_sgl { + dma_addr_t next_dma; + __le16 entry_sum_in_chain; + __le16 entry_sum_in_sgl; + __le16 entry_length_in_sgl; + __le16 pad0; + __le64 pad1[5]; + struct hisi_acc_hw_sgl *next; + struct acc_hw_sge sge_entries[]; +} __aligned(1); + +struct hisi_acc_sgl_pool { + struct mem_block { + struct hisi_acc_hw_sgl *sgl; + dma_addr_t sgl_dma; + size_t size; + } mem_block[HISI_ACC_MEM_BLOCK_NR]; + u32 sgl_num_per_block; + u32 block_num; + u32 count; + u32 sge_nr; + size_t sgl_size; +}; + +/** + * hisi_acc_create_sgl_pool() - Create a hw sgl pool. + * @dev: The device which hw sgl pool belongs to. + * @count: Count of hisi_acc_hw_sgl in pool. + * @sge_nr: The count of sge in hw_sgl + * + * This function creates a hw sgl pool, after this user can get hw sgl memory + * from it. + */ +struct hisi_acc_sgl_pool *hisi_acc_create_sgl_pool(struct device *dev, + u32 count, u32 sge_nr) +{ + u32 sgl_size, block_size, sgl_num_per_block, block_num, remain_sgl; + struct hisi_acc_sgl_pool *pool; + struct mem_block *block; + u32 i, j; + + if (!dev || !count || !sge_nr || sge_nr > HISI_ACC_SGL_SGE_NR_MAX) + return ERR_PTR(-EINVAL); + + sgl_size = ALIGN(sizeof(struct acc_hw_sge) * sge_nr + + sizeof(struct hisi_acc_hw_sgl), + HISI_ACC_SGL_ALIGN_SIZE); + + /* + * the pool may allocate a block of memory of size PAGE_SIZE * 2^(MAX_ORDER - 1), + * block size may exceed 2^31 on ia64, so the max of block size is 2^31 + */ + block_size = 1 << (PAGE_SHIFT + MAX_ORDER <= 32 ? + PAGE_SHIFT + MAX_ORDER - 1 : 31); + sgl_num_per_block = block_size / sgl_size; + block_num = count / sgl_num_per_block; + remain_sgl = count % sgl_num_per_block; + + if ((!remain_sgl && block_num > HISI_ACC_MEM_BLOCK_NR) || + (remain_sgl > 0 && block_num > HISI_ACC_MEM_BLOCK_NR - 1)) + return ERR_PTR(-EINVAL); + + pool = kzalloc(sizeof(*pool), GFP_KERNEL); + if (!pool) + return ERR_PTR(-ENOMEM); + block = pool->mem_block; + + for (i = 0; i < block_num; i++) { + block[i].sgl = dma_alloc_coherent(dev, block_size, + &block[i].sgl_dma, + GFP_KERNEL); + if (!block[i].sgl) { + dev_err(dev, "Fail to allocate hw SG buffer!\n"); + goto err_free_mem; + } + + block[i].size = block_size; + } + + if (remain_sgl > 0) { + block[i].sgl = dma_alloc_coherent(dev, remain_sgl * sgl_size, + &block[i].sgl_dma, + GFP_KERNEL); + if (!block[i].sgl) { + dev_err(dev, "Fail to allocate remained hw SG buffer!\n"); + goto err_free_mem; + } + + block[i].size = remain_sgl * sgl_size; + } + + pool->sgl_num_per_block = sgl_num_per_block; + pool->block_num = remain_sgl ? block_num + 1 : block_num; + pool->count = count; + pool->sgl_size = sgl_size; + pool->sge_nr = sge_nr; + + return pool; + +err_free_mem: + for (j = 0; j < i; j++) { + dma_free_coherent(dev, block_size, block[j].sgl, + block[j].sgl_dma); + memset(block + j, 0, sizeof(*block)); + } + kfree(pool); + return ERR_PTR(-ENOMEM); +} +EXPORT_SYMBOL_GPL(hisi_acc_create_sgl_pool); + +/** + * hisi_acc_free_sgl_pool() - Free a hw sgl pool. + * @dev: The device which hw sgl pool belongs to. + * @pool: Pointer of pool. + * + * This function frees memory of a hw sgl pool. + */ +void hisi_acc_free_sgl_pool(struct device *dev, struct hisi_acc_sgl_pool *pool) +{ + struct mem_block *block; + int i; + + if (!dev || !pool) + return; + + block = pool->mem_block; + + for (i = 0; i < pool->block_num; i++) + dma_free_coherent(dev, block[i].size, block[i].sgl, + block[i].sgl_dma); + + kfree(pool); +} +EXPORT_SYMBOL_GPL(hisi_acc_free_sgl_pool); + +static struct hisi_acc_hw_sgl *acc_get_sgl(struct hisi_acc_sgl_pool *pool, + u32 index, dma_addr_t *hw_sgl_dma) +{ + struct mem_block *block; + u32 block_index, offset; + + if (!pool || !hw_sgl_dma || index >= pool->count) + return ERR_PTR(-EINVAL); + + block = pool->mem_block; + block_index = index / pool->sgl_num_per_block; + offset = index % pool->sgl_num_per_block; + + *hw_sgl_dma = block[block_index].sgl_dma + pool->sgl_size * offset; + return (void *)block[block_index].sgl + pool->sgl_size * offset; +} + +static void sg_map_to_hw_sg(struct scatterlist *sgl, + struct acc_hw_sge *hw_sge) +{ + hw_sge->buf = sg_dma_address(sgl); + hw_sge->len = cpu_to_le32(sg_dma_len(sgl)); + hw_sge->page_ctrl = sg_virt(sgl); +} + +static void inc_hw_sgl_sge(struct hisi_acc_hw_sgl *hw_sgl) +{ + u16 var = le16_to_cpu(hw_sgl->entry_sum_in_sgl); + + var++; + hw_sgl->entry_sum_in_sgl = cpu_to_le16(var); +} + +static void update_hw_sgl_sum_sge(struct hisi_acc_hw_sgl *hw_sgl, u16 sum) +{ + hw_sgl->entry_sum_in_chain = cpu_to_le16(sum); +} + +static void clear_hw_sgl_sge(struct hisi_acc_hw_sgl *hw_sgl) +{ + struct acc_hw_sge *hw_sge = hw_sgl->sge_entries; + int i; + + for (i = 0; i < le16_to_cpu(hw_sgl->entry_sum_in_sgl); i++) { + hw_sge[i].page_ctrl = NULL; + hw_sge[i].buf = 0; + hw_sge[i].len = 0; + } +} + +/** + * hisi_acc_sg_buf_map_to_hw_sgl - Map a scatterlist to a hw sgl. + * @dev: The device which hw sgl belongs to. + * @sgl: Scatterlist which will be mapped to hw sgl. + * @pool: Pool which hw sgl memory will be allocated in. + * @index: Index of hisi_acc_hw_sgl in pool. + * @hw_sgl_dma: The dma address of allocated hw sgl. + * + * This function builds hw sgl according input sgl, user can use hw_sgl_dma + * as src/dst in its BD. Only support single hw sgl currently. + */ +struct hisi_acc_hw_sgl * +hisi_acc_sg_buf_map_to_hw_sgl(struct device *dev, + struct scatterlist *sgl, + struct hisi_acc_sgl_pool *pool, + u32 index, dma_addr_t *hw_sgl_dma) +{ + struct hisi_acc_hw_sgl *curr_hw_sgl; + dma_addr_t curr_sgl_dma = 0; + struct acc_hw_sge *curr_hw_sge; + struct scatterlist *sg; + int i, sg_n, sg_n_mapped; + + if (!dev || !sgl || !pool || !hw_sgl_dma) + return ERR_PTR(-EINVAL); + + sg_n = sg_nents(sgl); + + sg_n_mapped = dma_map_sg(dev, sgl, sg_n, DMA_BIDIRECTIONAL); + if (!sg_n_mapped) { + dev_err(dev, "DMA mapping for SG error!\n"); + return ERR_PTR(-EINVAL); + } + + if (sg_n_mapped > pool->sge_nr) { + dev_err(dev, "the number of entries in input scatterlist is bigger than SGL pool setting.\n"); + return ERR_PTR(-EINVAL); + } + + curr_hw_sgl = acc_get_sgl(pool, index, &curr_sgl_dma); + if (IS_ERR(curr_hw_sgl)) { + dev_err(dev, "Get SGL error!\n"); + dma_unmap_sg(dev, sgl, sg_n, DMA_BIDIRECTIONAL); + return ERR_PTR(-ENOMEM); + + } + curr_hw_sgl->entry_length_in_sgl = cpu_to_le16(pool->sge_nr); + curr_hw_sge = curr_hw_sgl->sge_entries; + + for_each_sg(sgl, sg, sg_n_mapped, i) { + sg_map_to_hw_sg(sg, curr_hw_sge); + inc_hw_sgl_sge(curr_hw_sgl); + curr_hw_sge++; + } + + update_hw_sgl_sum_sge(curr_hw_sgl, pool->sge_nr); + *hw_sgl_dma = curr_sgl_dma; + + return curr_hw_sgl; +} +EXPORT_SYMBOL_GPL(hisi_acc_sg_buf_map_to_hw_sgl); + +/** + * hisi_acc_sg_buf_unmap() - Unmap allocated hw sgl. + * @dev: The device which hw sgl belongs to. + * @sgl: Related scatterlist. + * @hw_sgl: Virtual address of hw sgl. + * + * This function unmaps allocated hw sgl. + */ +void hisi_acc_sg_buf_unmap(struct device *dev, struct scatterlist *sgl, + struct hisi_acc_hw_sgl *hw_sgl) +{ + if (!dev || !sgl || !hw_sgl) + return; + + dma_unmap_sg(dev, sgl, sg_nents(sgl), DMA_BIDIRECTIONAL); + clear_hw_sgl_sge(hw_sgl); + hw_sgl->entry_sum_in_chain = 0; + hw_sgl->entry_sum_in_sgl = 0; + hw_sgl->entry_length_in_sgl = 0; +} +EXPORT_SYMBOL_GPL(hisi_acc_sg_buf_unmap); diff --git a/kae_driver/hisilicon/trng/Makefile b/kae_driver/hisilicon/trng/Makefile new file mode 100755 index 0000000..d909079 --- /dev/null +++ b/kae_driver/hisilicon/trng/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_TRNG) += hisi-trng-v2.o +hisi-trng-v2-objs = trng.o diff --git a/kae_driver/hisilicon/trng/trng.c b/kae_driver/hisilicon/trng/trng.c new file mode 100755 index 0000000..97e500d --- /dev/null +++ b/kae_driver/hisilicon/trng/trng.c @@ -0,0 +1,341 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define HISI_TRNG_REG 0x00F0 +#define HISI_TRNG_BYTES 4 +#define HISI_TRNG_QUALITY 512 +#define HISI_TRNG_VERSION 0x01B8 +#define HISI_TRNG_VER_V1 GENMASK(31, 0) +#define SLEEP_US 10 +#define TIMEOUT_US 10000 +#define SW_DRBG_NUM_SHIFT 2 +#define SW_DRBG_KEY_BASE 0x082C +#define SW_DRBG_SEED(n) (SW_DRBG_KEY_BASE - ((n) << SW_DRBG_NUM_SHIFT)) +#define SW_DRBG_SEED_REGS_NUM 12 +#define SW_DRBG_SEED_SIZE 48 +#define SW_DRBG_BLOCKS 0x0830 +#define SW_DRBG_INIT 0x0834 +#define SW_DRBG_GEN 0x083c +#define SW_DRBG_STATUS 0x0840 +#define SW_DRBG_BLOCKS_NUM 4095 +#define SW_DRBG_DATA_BASE 0x0850 +#define SW_DRBG_DATA_NUM 4 +#define SW_DRBG_DATA(n) (SW_DRBG_DATA_BASE - ((n) << SW_DRBG_NUM_SHIFT)) +#define SW_DRBG_BYTES 16 +#define SW_DRBG_ENABLE_SHIFT 12 +#define SEED_SHIFT_24 24 +#define SEED_SHIFT_16 16 +#define SEED_SHIFT_8 8 + +struct hisi_trng_list { + struct mutex lock; + struct list_head list; + bool is_init; +}; + +struct hisi_trng { + void __iomem *base; + struct hisi_trng_list *trng_list; + struct list_head list; + struct hwrng rng; + u32 ver; + bool is_used; + struct mutex mutex; +}; + +struct hisi_trng_ctx { + struct hisi_trng *trng; +}; + +static atomic_t trng_active_devs; +static struct hisi_trng_list trng_devices; + +static void hisi_trng_set_seed(struct hisi_trng *trng, const u8 *seed) +{ + u32 val, seed_reg, i; + + for (i = 0; i < SW_DRBG_SEED_SIZE; + i += SW_DRBG_SEED_SIZE / SW_DRBG_SEED_REGS_NUM) { + val = seed[i] << SEED_SHIFT_24; + val |= seed[i + 1UL] << SEED_SHIFT_16; + val |= seed[i + 2UL] << SEED_SHIFT_8; + val |= seed[i + 3UL]; + + seed_reg = (i >> SW_DRBG_NUM_SHIFT) % SW_DRBG_SEED_REGS_NUM; + writel(val, trng->base + SW_DRBG_SEED(seed_reg)); + } +} + +static int hisi_trng_seed(struct crypto_rng *tfm, const u8 *seed, + unsigned int slen) +{ + struct hisi_trng_ctx *ctx = crypto_rng_ctx(tfm); + struct hisi_trng *trng = ctx->trng; + u32 val = 0; + int ret = 0; + + if (slen < SW_DRBG_SEED_SIZE) { + pr_err("slen(%u) is not matched with trng(%d)\n", slen, + SW_DRBG_SEED_SIZE); + return -EINVAL; + } + + writel(0x0, trng->base + SW_DRBG_BLOCKS); + hisi_trng_set_seed(trng, seed); + + writel(SW_DRBG_BLOCKS_NUM | (0x1 << SW_DRBG_ENABLE_SHIFT), + trng->base + SW_DRBG_BLOCKS); + writel(0x1, trng->base + SW_DRBG_INIT); + + ret = readl_relaxed_poll_timeout(trng->base + SW_DRBG_STATUS, + val, val & BIT(0), SLEEP_US, TIMEOUT_US); + if (ret) + pr_err("fail to init trng(%d)\n", ret); + + return ret; +} + +static int hisi_trng_generate(struct crypto_rng *tfm, const u8 *src, + unsigned int slen, u8 *dstn, unsigned int dlen) +{ + struct hisi_trng_ctx *ctx = crypto_rng_ctx(tfm); + struct hisi_trng *trng = ctx->trng; + u32 data[SW_DRBG_DATA_NUM]; + u32 currsize = 0; + u32 val = 0; + int ret; + u32 i; + + if (dlen > SW_DRBG_BLOCKS_NUM * SW_DRBG_BYTES || dlen == 0) { + pr_err("dlen(%d) exceeds limit(%d)!\n", dlen, + SW_DRBG_BLOCKS_NUM * SW_DRBG_BYTES); + return -EINVAL; + } + + do { + ret = readl_relaxed_poll_timeout(trng->base + SW_DRBG_STATUS, + val, val & BIT(1), SLEEP_US, TIMEOUT_US); + if (ret) { + pr_err("fail to generate random number(%d)!\n", ret); + break; + } + + for (i = 0; i < SW_DRBG_DATA_NUM; i++) + data[i] = readl(trng->base + SW_DRBG_DATA(i)); + + if (dlen - currsize >= SW_DRBG_BYTES) { + memcpy(dstn + currsize, data, SW_DRBG_BYTES); + currsize += SW_DRBG_BYTES; + } else { + memcpy(dstn + currsize, data, dlen - currsize); + currsize = dlen; + } + + writel(0x1, trng->base + SW_DRBG_GEN); + } while (currsize < dlen); + + return ret; +} + +static int hisi_trng_init(struct crypto_tfm *tfm) +{ + struct hisi_trng_ctx *ctx = crypto_tfm_ctx(tfm); + struct hisi_trng *trng; + int ret = -EBUSY; + + mutex_lock(&trng_devices.lock); + list_for_each_entry(trng, &trng_devices.list, list) { + if (!trng->is_used) { + trng->is_used = true; + ctx->trng = trng; + ret = 0; + break; + } + } + mutex_unlock(&trng_devices.lock); + + return ret; +} + +static void hisi_trng_exit(struct crypto_tfm *tfm) +{ + struct hisi_trng_ctx *ctx = crypto_tfm_ctx(tfm); + + mutex_lock(&trng_devices.lock); + ctx->trng->is_used = false; + mutex_unlock(&trng_devices.lock); +} + +static int hisi_trng_read(struct hwrng *rng, void *buf, size_t max, bool wait) +{ + struct hisi_trng *trng; + int currsize = 0; + u32 val = 0; + int ret; + + trng = container_of(rng, struct hisi_trng, rng); + + do { + ret = readl_poll_timeout(trng->base + HISI_TRNG_REG, val, + val, SLEEP_US, TIMEOUT_US); + if (ret) + return currsize; + + if (max - currsize >= HISI_TRNG_BYTES) { + memcpy(buf + currsize, &val, HISI_TRNG_BYTES); + currsize += HISI_TRNG_BYTES; + if (currsize == max) + return currsize; + continue; + } + + /* copy remaining bytes */ + memcpy(buf + currsize, &val, max - currsize); + currsize = max; + } while (currsize < max); + + return currsize; +} + +static struct rng_alg hisi_trng_alg = { + .generate = hisi_trng_generate, + .seed = hisi_trng_seed, + .seedsize = SW_DRBG_SEED_SIZE, + .base = { + .cra_name = "stdrng", + .cra_driver_name = "hisi_stdrng", + .cra_priority = 300, + .cra_ctxsize = sizeof(struct hisi_trng_ctx), + .cra_module = THIS_MODULE, + .cra_init = hisi_trng_init, + .cra_exit = hisi_trng_exit, + }, +}; + +static void hisi_trng_add_to_list(struct hisi_trng *trng) +{ + mutex_lock(&trng_devices.lock); + list_add_tail(&trng->list, &trng_devices.list); + mutex_unlock(&trng_devices.lock); +} + +static int hisi_trng_del_from_list(struct hisi_trng *trng) +{ + int ret = -EBUSY; + + mutex_lock(&trng_devices.lock); + if (!trng->is_used) { + list_del(&trng->list); + ret = 0; + } + mutex_unlock(&trng_devices.lock); + + return ret; +} + +static int hisi_trng_probe(struct platform_device *pdev) +{ + struct hisi_trng *trng; + int ret; + + trng = devm_kzalloc(&pdev->dev, sizeof(*trng), GFP_KERNEL); + if (!trng) + return -ENOMEM; + + platform_set_drvdata(pdev, trng); + + trng->base = devm_platform_ioremap_resource(pdev, 0); + if (IS_ERR(trng->base)) + return PTR_ERR(trng->base); + + trng->is_used = false; + trng->ver = readl(trng->base + HISI_TRNG_VERSION); + if (!trng_devices.is_init) { + INIT_LIST_HEAD(&trng_devices.list); + mutex_init(&trng_devices.lock); + trng_devices.is_init = true; + } + + hisi_trng_add_to_list(trng); + if (trng->ver != HISI_TRNG_VER_V1 && + atomic_inc_return(&trng_active_devs) == 1) { + ret = crypto_register_rng(&hisi_trng_alg); + if (ret) { + dev_err(&pdev->dev, + "failed to register crypto(%d)\n", ret); + atomic_dec_return(&trng_active_devs); + goto err_remove_from_list; + } + } + + trng->rng.name = pdev->name; + trng->rng.read = hisi_trng_read; + trng->rng.quality = HISI_TRNG_QUALITY; + ret = devm_hwrng_register(&pdev->dev, &trng->rng); + if (ret) { + dev_err(&pdev->dev, "failed to register hwrng: %d!\n", ret); + goto err_crypto_unregister; + } + + return ret; + +err_crypto_unregister: + if (trng->ver != HISI_TRNG_VER_V1 && + atomic_dec_return(&trng_active_devs) == 0) + crypto_unregister_rng(&hisi_trng_alg); + +err_remove_from_list: + hisi_trng_del_from_list(trng); + return ret; +} + +static int hisi_trng_remove(struct platform_device *pdev) +{ + struct hisi_trng *trng = platform_get_drvdata(pdev); + + /* Wait until the task is finished */ + while (hisi_trng_del_from_list(trng)) + ; + + if (trng->ver != HISI_TRNG_VER_V1 && + atomic_dec_return(&trng_active_devs) == 0) + crypto_unregister_rng(&hisi_trng_alg); + + return 0; +} + +static const struct acpi_device_id hisi_trng_acpi_match[] = { + { "HISI02B3", 0 }, + { } +}; +MODULE_DEVICE_TABLE(acpi, hisi_trng_acpi_match); + +static struct platform_driver hisi_trng_driver = { + .probe = hisi_trng_probe, + .remove = hisi_trng_remove, + .driver = { + .name = "hisi-trng-v2", + .acpi_match_table = ACPI_PTR(hisi_trng_acpi_match), + }, +}; + +module_platform_driver(hisi_trng_driver); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Weili Qian "); +MODULE_AUTHOR("Zaibo Xu "); +MODULE_DESCRIPTION("HiSilicon true random number generator V2 driver"); diff --git a/kae_driver/hisilicon/zip/Makefile b/kae_driver/hisilicon/zip/Makefile new file mode 100755 index 0000000..a936f09 --- /dev/null +++ b/kae_driver/hisilicon/zip/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_ZIP) += hisi_zip.o +hisi_zip-objs = zip_main.o zip_crypto.o diff --git a/kae_driver/hisilicon/zip/zip.h b/kae_driver/hisilicon/zip/zip.h new file mode 100755 index 0000000..3c9d8e1 --- /dev/null +++ b/kae_driver/hisilicon/zip/zip.h @@ -0,0 +1,88 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ +#ifndef HISI_ZIP_H +#define HISI_ZIP_H + +#undef pr_fmt +#define pr_fmt(fmt) "hisi_zip: " fmt + +#include +#include "../hisi_acc_qm.h" + +enum hisi_zip_error_type { + /* negative compression */ + HZIP_NC_ERR = 0x0d, +}; + +struct hisi_zip_dfx { + atomic64_t send_cnt; + atomic64_t recv_cnt; + atomic64_t send_busy_cnt; + atomic64_t err_bd_cnt; +}; + +struct hisi_zip_ctrl; + +struct hisi_zip { + struct hisi_qm qm; + struct hisi_zip_ctrl *ctrl; + struct hisi_zip_dfx dfx; +}; + +struct hisi_zip_sqe { + u32 consumed; + u32 produced; + u32 comp_data_length; + /* + * status: 0~7 bits + * rsvd: 8~31 bits + */ + u32 dw3; + u32 input_data_length; + u32 dw5; + u32 dw6; + /* + * in_sge_data_offset: 0~23 bits + * rsvd: 24~27 bits + * sqe_type: 29~31 bits + */ + u32 dw7; + /* + * out_sge_data_offset: 0~23 bits + * rsvd: 24~31 bits + */ + u32 dw8; + /* + * request_type: 0~7 bits + * buffer_type: 8~11 bits + * rsvd: 13~31 bits + */ + u32 dw9; + u32 dw10; + u32 dw11; + u32 dw12; + /* tag: in sqe type 0 */ + u32 dw13; + u32 dest_avail_out; + u32 dw15; + u32 dw16; + u32 dw17; + u32 source_addr_l; + u32 source_addr_h; + u32 dest_addr_l; + u32 dest_addr_h; + u32 dw22; + u32 dw23; + u32 dw24; + u32 dw25; + /* tag: in sqe type 3 */ + u32 dw26; + u32 dw27; + u32 rsvd1[4]; +}; + +int zip_create_qps(struct hisi_qp **qps, int qp_num, int node); +int hisi_zip_register_to_crypto(struct hisi_qm *qm); +void hisi_zip_unregister_from_crypto(struct hisi_qm *qm); +bool hisi_zip_alg_support(struct hisi_qm *qm, u32 alg); +#endif diff --git a/kae_driver/hisilicon/zip/zip_crypto.c b/kae_driver/hisilicon/zip/zip_crypto.c new file mode 100755 index 0000000..6608971 --- /dev/null +++ b/kae_driver/hisilicon/zip/zip_crypto.c @@ -0,0 +1,839 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include "zip.h" + +/* hisi_zip_sqe dw3 */ +#define HZIP_BD_STATUS_M GENMASK(7, 0) +/* hisi_zip_sqe dw7 */ +#define HZIP_IN_SGE_DATA_OFFSET_M GENMASK(23, 0) +#define HZIP_SQE_TYPE_M GENMASK(31, 28) +/* hisi_zip_sqe dw8 */ +#define HZIP_OUT_SGE_DATA_OFFSET_M GENMASK(23, 0) +/* hisi_zip_sqe dw9 */ +#define HZIP_REQ_TYPE_M GENMASK(7, 0) +#define HZIP_ALG_TYPE_ZLIB 0x02 +#define HZIP_ALG_TYPE_GZIP 0x03 +#define HZIP_BUF_TYPE_M GENMASK(11, 8) +#define HZIP_PBUFFER 0x0 +#define HZIP_SGL 0x1 + +#define HZIP_ZLIB_HEAD_SIZE 2 +#define HZIP_GZIP_HEAD_SIZE 10 + +#define GZIP_HEAD_FHCRC_BIT BIT(1) +#define GZIP_HEAD_FEXTRA_BIT BIT(2) +#define GZIP_HEAD_FNAME_BIT BIT(3) +#define GZIP_HEAD_FCOMMENT_BIT BIT(4) + +#define GZIP_HEAD_FLG_SHIFT 3 +#define GZIP_HEAD_FEXTRA_SHIFT 10 +#define GZIP_HEAD_FEXTRA_XLEN 2UL +#define GZIP_HEAD_FHCRC_SIZE 2 + +#define HZIP_GZIP_HEAD_BUF 256 +#define HZIP_ALG_PRIORITY 300 +#define HZIP_SGL_SGE_NR 10 + +#define HZIP_ALG_ZLIB GENMASK(1, 0) +#define HZIP_ALG_GZIP GENMASK(3, 2) + +static const u8 zlib_head[HZIP_ZLIB_HEAD_SIZE] = {0x78, 0x9c}; +static const u8 gzip_head[HZIP_GZIP_HEAD_SIZE] = { + 0x1f, 0x8b, 0x08, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x03 +}; + +enum hisi_zip_alg_type { + HZIP_ALG_TYPE_COMP = 0, + HZIP_ALG_TYPE_DECOMP = 1, +}; + +enum { + HZIP_QPC_COMP, + HZIP_QPC_DECOMP, + HZIP_CTX_Q_NUM +}; + +#define COMP_NAME_TO_TYPE(alg_name) \ + (!strcmp((alg_name), "zlib-deflate") ? HZIP_ALG_TYPE_ZLIB : \ + !strcmp((alg_name), "gzip") ? HZIP_ALG_TYPE_GZIP : 0) \ + +#define TO_HEAD_SIZE(req_type) \ + (((req_type) == HZIP_ALG_TYPE_ZLIB) ? sizeof(zlib_head) : \ + ((req_type) == HZIP_ALG_TYPE_GZIP) ? sizeof(gzip_head) : 0) \ + +#define TO_HEAD(req_type) \ + (((req_type) == HZIP_ALG_TYPE_ZLIB) ? zlib_head : \ + ((req_type) == HZIP_ALG_TYPE_GZIP) ? gzip_head : NULL) \ + +struct hisi_zip_req { + struct acomp_req *req; + u32 sskip; + u32 dskip; + struct hisi_acc_hw_sgl *hw_src; + struct hisi_acc_hw_sgl *hw_dst; + dma_addr_t dma_src; + dma_addr_t dma_dst; + u16 req_id; +}; + +struct hisi_zip_req_q { + struct hisi_zip_req *q; + unsigned long *req_bitmap; + rwlock_t req_lock; + u16 size; +}; + +struct hisi_zip_qp_ctx { + struct hisi_qp *qp; + struct hisi_zip_req_q req_q; + struct hisi_acc_sgl_pool *sgl_pool; + struct hisi_zip *zip_dev; + struct hisi_zip_ctx *ctx; +}; + +struct hisi_zip_sqe_ops { + u8 sqe_type; + void (*fill_addr)(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req); + void (*fill_buf_size)(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req); + void (*fill_buf_type)(struct hisi_zip_sqe *sqe, u8 buf_type); + void (*fill_req_type)(struct hisi_zip_sqe *sqe, u8 req_type); + void (*fill_tag)(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req); + void (*fill_sqe_type)(struct hisi_zip_sqe *sqe, u8 sqe_type); + u32 (*get_tag)(struct hisi_zip_sqe *sqe); + u32 (*get_status)(struct hisi_zip_sqe *sqe); + u32 (*get_dstlen)(struct hisi_zip_sqe *sqe); +}; + +struct hisi_zip_ctx { + struct hisi_zip_qp_ctx qp_ctx[HZIP_CTX_Q_NUM]; + const struct hisi_zip_sqe_ops *ops; +}; + +static int sgl_sge_nr_set(const char *val, const struct kernel_param *kp) +{ + int ret; + u16 n; + + if (!val) + return -EINVAL; + + ret = kstrtou16(val, 10, &n); + if (ret || n == 0 || n > HISI_ACC_SGL_SGE_NR_MAX) + return -EINVAL; + + return param_set_ushort(val, kp); +} + +static const struct kernel_param_ops sgl_sge_nr_ops = { + .set = sgl_sge_nr_set, + .get = param_get_ushort, +}; + +static u16 sgl_sge_nr = HZIP_SGL_SGE_NR; +module_param_cb(sgl_sge_nr, &sgl_sge_nr_ops, &sgl_sge_nr, 0444); +MODULE_PARM_DESC(sgl_sge_nr, "Number of sge in sgl(1-255)"); + +static u32 get_extra_field_size(const u8 *start) +{ + return *((u16 *)start) + GZIP_HEAD_FEXTRA_XLEN; +} + +static u32 get_name_field_size(const u8 *start) +{ + return strlen(start) + 1; +} + +static u32 get_comment_field_size(const u8 *start) +{ + return strlen(start) + 1; +} + +static u32 __get_gzip_head_size(const u8 *src) +{ + u8 head_flg = *(src + GZIP_HEAD_FLG_SHIFT); + u32 size = GZIP_HEAD_FEXTRA_SHIFT; + + if (head_flg & GZIP_HEAD_FEXTRA_BIT) + size += get_extra_field_size(src + size); + if (head_flg & GZIP_HEAD_FNAME_BIT) + size += get_name_field_size(src + size); + if (head_flg & GZIP_HEAD_FCOMMENT_BIT) + size += get_comment_field_size(src + size); + if (head_flg & GZIP_HEAD_FHCRC_BIT) + size += GZIP_HEAD_FHCRC_SIZE; + + return size; +} + +static u32 __maybe_unused get_gzip_head_size(struct scatterlist *sgl) +{ + char buf[HZIP_GZIP_HEAD_BUF]; + + sg_copy_to_buffer(sgl, sg_nents(sgl), buf, sizeof(buf)); + + return __get_gzip_head_size(buf); +} + +static int add_comp_head(struct scatterlist *dst, u8 req_type) +{ + int head_size = TO_HEAD_SIZE(req_type); + const u8 *head = TO_HEAD(req_type); + int ret; + + ret = sg_copy_from_buffer(dst, sg_nents(dst), head, head_size); + if (unlikely(ret != head_size)) { + pr_err("the head size of buffer is wrong (%d)!\n", ret); + return -ENOMEM; + } + + return head_size; +} + +static int get_comp_head_size(struct acomp_req *acomp_req, u8 req_type) +{ + if (unlikely(!acomp_req->src || !acomp_req->slen)) + return -EINVAL; + + if (unlikely(req_type == HZIP_ALG_TYPE_GZIP && + acomp_req->slen < GZIP_HEAD_FEXTRA_SHIFT)) + return -EINVAL; + + switch (req_type) { + case HZIP_ALG_TYPE_ZLIB: + return TO_HEAD_SIZE(HZIP_ALG_TYPE_ZLIB); + case HZIP_ALG_TYPE_GZIP: + return TO_HEAD_SIZE(HZIP_ALG_TYPE_GZIP); + default: + pr_err("request type does not support!\n"); + return -EINVAL; + } +} + +static struct hisi_zip_req *hisi_zip_create_req(struct acomp_req *req, + struct hisi_zip_qp_ctx *qp_ctx, + size_t head_size, bool is_comp) +{ + struct hisi_zip_req_q *req_q = &qp_ctx->req_q; + struct hisi_zip_req *q = req_q->q; + struct hisi_zip_req *req_cache; + int req_id; + + write_lock(&req_q->req_lock); + + req_id = find_first_zero_bit(req_q->req_bitmap, req_q->size); + if (req_id >= req_q->size) { + write_unlock(&req_q->req_lock); + dev_dbg(&qp_ctx->qp->qm->pdev->dev, "req cache is full!\n"); + return ERR_PTR(-EAGAIN); + } + set_bit(req_id, req_q->req_bitmap); + + write_unlock(&req_q->req_lock); + + req_cache = q + req_id; + req_cache->req_id = req_id; + req_cache->req = req; + + if (is_comp) { + req_cache->sskip = 0; + req_cache->dskip = head_size; + } else { + req_cache->sskip = head_size; + req_cache->dskip = 0; + } + + return req_cache; +} + +static void hisi_zip_remove_req(struct hisi_zip_qp_ctx *qp_ctx, + struct hisi_zip_req *req) +{ + struct hisi_zip_req_q *req_q = &qp_ctx->req_q; + + write_lock(&req_q->req_lock); + clear_bit(req->req_id, req_q->req_bitmap); + write_unlock(&req_q->req_lock); +} + +static void hisi_zip_fill_addr(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req) +{ + sqe->source_addr_l = lower_32_bits(req->dma_src); + sqe->source_addr_h = upper_32_bits(req->dma_src); + sqe->dest_addr_l = lower_32_bits(req->dma_dst); + sqe->dest_addr_h = upper_32_bits(req->dma_dst); +} + +static void hisi_zip_fill_buf_size(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req) +{ + struct acomp_req *a_req = req->req; + + sqe->input_data_length = a_req->slen - req->sskip; + sqe->dest_avail_out = a_req->dlen - req->dskip; + sqe->dw7 = FIELD_PREP(HZIP_IN_SGE_DATA_OFFSET_M, req->sskip); + sqe->dw8 = FIELD_PREP(HZIP_OUT_SGE_DATA_OFFSET_M, req->dskip); +} + +static void hisi_zip_fill_buf_type(struct hisi_zip_sqe *sqe, u8 buf_type) +{ + u32 val; + + val = sqe->dw9 & ~HZIP_BUF_TYPE_M; + val |= FIELD_PREP(HZIP_BUF_TYPE_M, buf_type); + sqe->dw9 = val; +} + +static void hisi_zip_fill_req_type(struct hisi_zip_sqe *sqe, u8 req_type) +{ + u32 val; + + val = sqe->dw9 & ~HZIP_REQ_TYPE_M; + val |= FIELD_PREP(HZIP_REQ_TYPE_M, req_type); + sqe->dw9 = val; +} + +static void hisi_zip_fill_tag_v1(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req) +{ + sqe->dw13 = req->req_id; +} + +static void hisi_zip_fill_tag_v2(struct hisi_zip_sqe *sqe, struct hisi_zip_req *req) +{ + sqe->dw26 = req->req_id; +} + +static void hisi_zip_fill_sqe_type(struct hisi_zip_sqe *sqe, u8 sqe_type) +{ + u32 val; + + val = sqe->dw7 & ~HZIP_SQE_TYPE_M; + val |= FIELD_PREP(HZIP_SQE_TYPE_M, sqe_type); + sqe->dw7 = val; +} + +static void hisi_zip_fill_sqe(struct hisi_zip_ctx *ctx, struct hisi_zip_sqe *sqe, + u8 req_type, struct hisi_zip_req *req) +{ + const struct hisi_zip_sqe_ops *ops = ctx->ops; + + memset(sqe, 0, sizeof(struct hisi_zip_sqe)); + + ops->fill_addr(sqe, req); + ops->fill_buf_size(sqe, req); + ops->fill_buf_type(sqe, HZIP_SGL); + ops->fill_req_type(sqe, req_type); + ops->fill_tag(sqe, req); + ops->fill_sqe_type(sqe, ops->sqe_type); +} + +static int hisi_zip_do_work(struct hisi_zip_req *req, + struct hisi_zip_qp_ctx *qp_ctx) +{ + struct hisi_acc_sgl_pool *pool = qp_ctx->sgl_pool; + struct hisi_zip_dfx *dfx = &qp_ctx->zip_dev->dfx; + struct acomp_req *a_req = req->req; + struct hisi_qp *qp = qp_ctx->qp; + struct device *dev = &qp->qm->pdev->dev; + struct hisi_zip_sqe zip_sqe; + int ret; + + if (unlikely(!a_req->src || !a_req->slen || !a_req->dst || !a_req->dlen)) + return -EINVAL; + + req->hw_src = hisi_acc_sg_buf_map_to_hw_sgl(dev, a_req->src, pool, + req->req_id << 1, &req->dma_src); + if (IS_ERR(req->hw_src)) { + dev_err(dev, "failed to map the src buffer to hw sgl (%ld)!\n", + PTR_ERR(req->hw_src)); + return PTR_ERR(req->hw_src); + } + + req->hw_dst = hisi_acc_sg_buf_map_to_hw_sgl(dev, a_req->dst, pool, + (req->req_id << 1) + 1, + &req->dma_dst); + if (IS_ERR(req->hw_dst)) { + ret = PTR_ERR(req->hw_dst); + dev_err(dev, "failed to map the dst buffer to hw slg (%d)!\n", + ret); + goto err_unmap_input; + } + + hisi_zip_fill_sqe(qp_ctx->ctx, &zip_sqe, qp->req_type, req); + + /* send command to start a task */ + atomic64_inc(&dfx->send_cnt); + ret = hisi_qp_send(qp, &zip_sqe); + if (unlikely(ret < 0)) { + atomic64_inc(&dfx->send_busy_cnt); + ret = -EAGAIN; + dev_dbg_ratelimited(dev, "failed to send request!\n"); + goto err_unmap_output; + } + + return -EINPROGRESS; + +err_unmap_output: + hisi_acc_sg_buf_unmap(dev, a_req->dst, req->hw_dst); +err_unmap_input: + hisi_acc_sg_buf_unmap(dev, a_req->src, req->hw_src); + return ret; +} + +static u32 hisi_zip_get_tag_v1(struct hisi_zip_sqe *sqe) +{ + return sqe->dw13; +} + +static u32 hisi_zip_get_tag_v2(struct hisi_zip_sqe *sqe) +{ + return sqe->dw26; +} + +static u32 hisi_zip_get_status(struct hisi_zip_sqe *sqe) +{ + return sqe->dw3 & HZIP_BD_STATUS_M; +} + +static u32 hisi_zip_get_dstlen(struct hisi_zip_sqe *sqe) +{ + return sqe->produced; +} + +static void hisi_zip_acomp_cb(struct hisi_qp *qp, void *data) +{ + struct hisi_zip_qp_ctx *qp_ctx = qp->qp_ctx; + const struct hisi_zip_sqe_ops *ops = qp_ctx->ctx->ops; + struct hisi_zip_dfx *dfx = &qp_ctx->zip_dev->dfx; + struct hisi_zip_req_q *req_q = &qp_ctx->req_q; + struct device *dev = &qp->qm->pdev->dev; + struct hisi_zip_sqe *sqe = data; + u32 tag = ops->get_tag(sqe); + struct hisi_zip_req *req = req_q->q + tag; + struct acomp_req *acomp_req = req->req; + u32 status, dlen, head_size; + int err = 0; + + atomic64_inc(&dfx->recv_cnt); + status = ops->get_status(sqe); + if (unlikely(status != 0 && status != HZIP_NC_ERR)) { + dev_err(dev, "%scompress fail in qp%u: %u, output: %u\n", + (qp->alg_type == 0) ? "" : "de", qp->qp_id, status, + sqe->produced); + atomic64_inc(&dfx->err_bd_cnt); + err = -EIO; + } + + dlen = ops->get_dstlen(sqe); + + hisi_acc_sg_buf_unmap(dev, acomp_req->src, req->hw_src); + hisi_acc_sg_buf_unmap(dev, acomp_req->dst, req->hw_dst); + + head_size = (qp->alg_type == 0) ? TO_HEAD_SIZE(qp->req_type) : 0; + acomp_req->dlen = dlen + head_size; + + if (acomp_req->base.complete) + acomp_request_complete(acomp_req, err); + + hisi_zip_remove_req(qp_ctx, req); +} + +static int hisi_zip_acompress(struct acomp_req *acomp_req) +{ + struct hisi_zip_ctx *ctx = crypto_tfm_ctx(acomp_req->base.tfm); + struct hisi_zip_qp_ctx *qp_ctx = &ctx->qp_ctx[HZIP_QPC_COMP]; + struct device *dev = &qp_ctx->qp->qm->pdev->dev; + struct hisi_zip_req *req; + int head_size; + int ret; + + /* let's output compression head now */ + head_size = add_comp_head(acomp_req->dst, qp_ctx->qp->req_type); + if (unlikely(head_size < 0)) { + dev_err_ratelimited(dev, "failed to add comp head (%d)!\n", + head_size); + return head_size; + } + + req = hisi_zip_create_req(acomp_req, qp_ctx, head_size, true); + if (IS_ERR(req)) + return PTR_ERR(req); + + ret = hisi_zip_do_work(req, qp_ctx); + if (unlikely(ret != -EINPROGRESS)) { + dev_info_ratelimited(dev, "failed to do compress (%d)!\n", ret); + hisi_zip_remove_req(qp_ctx, req); + } + + return ret; +} + +static int hisi_zip_adecompress(struct acomp_req *acomp_req) +{ + struct hisi_zip_ctx *ctx = crypto_tfm_ctx(acomp_req->base.tfm); + struct hisi_zip_qp_ctx *qp_ctx = &ctx->qp_ctx[HZIP_QPC_DECOMP]; + struct device *dev = &qp_ctx->qp->qm->pdev->dev; + struct hisi_zip_req *req; + int head_size, ret; + + head_size = get_comp_head_size(acomp_req, qp_ctx->qp->req_type); + if (unlikely(head_size < 0)) { + dev_err_ratelimited(dev, "failed to get comp head size (%d)!\n", + head_size); + return head_size; + } + + req = hisi_zip_create_req(acomp_req, qp_ctx, head_size, false); + if (IS_ERR(req)) + return PTR_ERR(req); + + ret = hisi_zip_do_work(req, qp_ctx); + if (unlikely(ret != -EINPROGRESS)) { + dev_info_ratelimited(dev, "failed to do decompress (%d)!\n", + ret); + hisi_zip_remove_req(qp_ctx, req); + } + + return ret; +} + +static int hisi_zip_start_qp(struct hisi_qp *qp, struct hisi_zip_qp_ctx *qp_ctx, + int alg_type, int req_type) +{ + struct device *dev = &qp->qm->pdev->dev; + int ret; + + qp->req_type = req_type; + qp->alg_type = alg_type; + qp->qp_ctx = qp_ctx; + + ret = hisi_qm_start_qp(qp, 0); + if (ret < 0) { + dev_err(dev, "failed to start qp (%d)!\n", ret); + return ret; + } + + qp_ctx->qp = qp; + + return 0; +} + +static void hisi_zip_release_qp(struct hisi_zip_qp_ctx *qp_ctx) +{ + hisi_qm_stop_qp(qp_ctx->qp); + hisi_qm_free_qps(&qp_ctx->qp, 1); +} + +static const struct hisi_zip_sqe_ops hisi_zip_ops_v1 = { + .sqe_type = 0, + .fill_addr = hisi_zip_fill_addr, + .fill_buf_size = hisi_zip_fill_buf_size, + .fill_buf_type = hisi_zip_fill_buf_type, + .fill_req_type = hisi_zip_fill_req_type, + .fill_tag = hisi_zip_fill_tag_v1, + .fill_sqe_type = hisi_zip_fill_sqe_type, + .get_tag = hisi_zip_get_tag_v1, + .get_status = hisi_zip_get_status, + .get_dstlen = hisi_zip_get_dstlen, +}; + +static const struct hisi_zip_sqe_ops hisi_zip_ops_v2 = { + .sqe_type = 0x3, + .fill_addr = hisi_zip_fill_addr, + .fill_buf_size = hisi_zip_fill_buf_size, + .fill_buf_type = hisi_zip_fill_buf_type, + .fill_req_type = hisi_zip_fill_req_type, + .fill_tag = hisi_zip_fill_tag_v2, + .fill_sqe_type = hisi_zip_fill_sqe_type, + .get_tag = hisi_zip_get_tag_v2, + .get_status = hisi_zip_get_status, + .get_dstlen = hisi_zip_get_dstlen, +}; + +static int hisi_zip_ctx_init(struct hisi_zip_ctx *hisi_zip_ctx, u8 req_type, int node) +{ + struct hisi_qp *qps[HZIP_CTX_Q_NUM] = { NULL }; + struct hisi_zip_qp_ctx *qp_ctx; + struct hisi_zip *hisi_zip; + int ret, i, j; + + ret = zip_create_qps(qps, HZIP_CTX_Q_NUM, node); + if (ret) { + pr_err("failed to create zip qps (%d)!\n", ret); + return -ENODEV; + } + + hisi_zip = container_of(qps[0]->qm, struct hisi_zip, qm); + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) { + /* alg_type = 0 for compress, 1 for decompress in hw sqe */ + qp_ctx = &hisi_zip_ctx->qp_ctx[i]; + qp_ctx->ctx = hisi_zip_ctx; + ret = hisi_zip_start_qp(qps[i], qp_ctx, i, req_type); + if (ret) { + for (j = i - 1; j >= 0; j--) + hisi_qm_stop_qp(hisi_zip_ctx->qp_ctx[j].qp); + + hisi_qm_free_qps(qps, HZIP_CTX_Q_NUM); + return ret; + } + + qp_ctx->zip_dev = hisi_zip; + } + + if (hisi_zip->qm.ver < QM_HW_V3) + hisi_zip_ctx->ops = &hisi_zip_ops_v1; + else + hisi_zip_ctx->ops = &hisi_zip_ops_v2; + + return 0; +} + +static void hisi_zip_ctx_exit(struct hisi_zip_ctx *hisi_zip_ctx) +{ + int i; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) + hisi_zip_release_qp(&hisi_zip_ctx->qp_ctx[i]); +} + +static int hisi_zip_create_req_q(struct hisi_zip_ctx *ctx) +{ + u16 q_depth = ctx->qp_ctx[0].qp->sq_depth; + struct hisi_zip_req_q *req_q; + int i, ret; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) { + req_q = &ctx->qp_ctx[i].req_q; + req_q->size = q_depth; + + req_q->req_bitmap = bitmap_zalloc(req_q->size, GFP_KERNEL); + if (!req_q->req_bitmap) { + ret = -ENOMEM; + if (i == 0) + return ret; + + goto err_free_comp_q; + } + rwlock_init(&req_q->req_lock); + + req_q->q = kcalloc(req_q->size, sizeof(struct hisi_zip_req), + GFP_KERNEL); + if (!req_q->q) { + ret = -ENOMEM; + if (i == 0) + goto err_free_comp_bitmap; + else + goto err_free_decomp_bitmap; + } + } + + return 0; + +err_free_decomp_bitmap: + bitmap_free(ctx->qp_ctx[HZIP_QPC_DECOMP].req_q.req_bitmap); +err_free_comp_q: + kfree(ctx->qp_ctx[HZIP_QPC_COMP].req_q.q); +err_free_comp_bitmap: + bitmap_free(ctx->qp_ctx[HZIP_QPC_COMP].req_q.req_bitmap); + return ret; +} + +static void hisi_zip_release_req_q(struct hisi_zip_ctx *ctx) +{ + int i; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) { + kfree(ctx->qp_ctx[i].req_q.q); + bitmap_free(ctx->qp_ctx[i].req_q.req_bitmap); + } +} + +static int hisi_zip_create_sgl_pool(struct hisi_zip_ctx *ctx) +{ + u16 q_depth = ctx->qp_ctx[0].qp->sq_depth; + struct hisi_zip_qp_ctx *tmp; + struct device *dev; + int i; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) { + tmp = &ctx->qp_ctx[i]; + dev = &tmp->qp->qm->pdev->dev; + tmp->sgl_pool = hisi_acc_create_sgl_pool(dev, q_depth << 1, + sgl_sge_nr); + if (IS_ERR(tmp->sgl_pool)) { + if (i == 1) + goto err_free_sgl_pool0; + return -ENOMEM; + } + } + + return 0; + +err_free_sgl_pool0: + hisi_acc_free_sgl_pool(&ctx->qp_ctx[HZIP_QPC_COMP].qp->qm->pdev->dev, + ctx->qp_ctx[HZIP_QPC_COMP].sgl_pool); + return -ENOMEM; +} + +static void hisi_zip_release_sgl_pool(struct hisi_zip_ctx *ctx) +{ + int i; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) + hisi_acc_free_sgl_pool(&ctx->qp_ctx[i].qp->qm->pdev->dev, + ctx->qp_ctx[i].sgl_pool); +} + +static void hisi_zip_set_acomp_cb(struct hisi_zip_ctx *ctx, + void (*fn)(struct hisi_qp *, void *)) +{ + int i; + + for (i = 0; i < HZIP_CTX_Q_NUM; i++) + ctx->qp_ctx[i].qp->req_cb = fn; +} + +static int hisi_zip_acomp_init(struct crypto_acomp *tfm) +{ + const char *alg_name = crypto_tfm_alg_name(&tfm->base); + struct hisi_zip_ctx *ctx = crypto_tfm_ctx(&tfm->base); + struct device *dev; + int ret; + + ret = hisi_zip_ctx_init(ctx, COMP_NAME_TO_TYPE(alg_name), tfm->base.node); + if (ret) { + pr_err("failed to init ctx (%d)!\n", ret); + return ret; + } + + dev = &ctx->qp_ctx[0].qp->qm->pdev->dev; + + ret = hisi_zip_create_req_q(ctx); + if (ret) { + dev_err(dev, "failed to create request queue (%d)!\n", ret); + goto err_ctx_exit; + } + + ret = hisi_zip_create_sgl_pool(ctx); + if (ret) { + dev_err(dev, "failed to create sgl pool (%d)!\n", ret); + goto err_release_req_q; + } + + hisi_zip_set_acomp_cb(ctx, hisi_zip_acomp_cb); + + return 0; + +err_release_req_q: + hisi_zip_release_req_q(ctx); +err_ctx_exit: + hisi_zip_ctx_exit(ctx); + return ret; +} + +static void hisi_zip_acomp_exit(struct crypto_acomp *tfm) +{ + struct hisi_zip_ctx *ctx = crypto_tfm_ctx(&tfm->base); + + hisi_zip_set_acomp_cb(ctx, NULL); + hisi_zip_release_sgl_pool(ctx); + hisi_zip_release_req_q(ctx); + hisi_zip_ctx_exit(ctx); +} + +static struct acomp_alg hisi_zip_acomp_zlib = { + .init = hisi_zip_acomp_init, + .exit = hisi_zip_acomp_exit, + .compress = hisi_zip_acompress, + .decompress = hisi_zip_adecompress, + .base = { + .cra_name = "zlib-deflate", + .cra_driver_name = "hisi-zlib-acomp", + .cra_module = THIS_MODULE, + .cra_priority = HZIP_ALG_PRIORITY, + .cra_ctxsize = sizeof(struct hisi_zip_ctx), + } +}; + +static int hisi_zip_register_zlib(struct hisi_qm *qm) +{ + int ret; + + if (!hisi_zip_alg_support(qm, HZIP_ALG_ZLIB)) + return 0; + + ret = crypto_register_acomp(&hisi_zip_acomp_zlib); + if (ret) + dev_err(&qm->pdev->dev, "failed to register to zlib (%d)!\n", ret); + + return ret; +} + +static void hisi_zip_unregister_zlib(struct hisi_qm *qm) +{ + if (!hisi_zip_alg_support(qm, HZIP_ALG_ZLIB)) + return; + + crypto_unregister_acomp(&hisi_zip_acomp_zlib); +} + +static struct acomp_alg hisi_zip_acomp_gzip = { + .init = hisi_zip_acomp_init, + .exit = hisi_zip_acomp_exit, + .compress = hisi_zip_acompress, + .decompress = hisi_zip_adecompress, + .base = { + .cra_name = "gzip", + .cra_driver_name = "hisi-gzip-acomp", + .cra_module = THIS_MODULE, + .cra_priority = HZIP_ALG_PRIORITY, + .cra_ctxsize = sizeof(struct hisi_zip_ctx), + } +}; + +static int hisi_zip_register_gzip(struct hisi_qm *qm) +{ + int ret; + + if (!hisi_zip_alg_support(qm, HZIP_ALG_GZIP)) + return 0; + + ret = crypto_register_acomp(&hisi_zip_acomp_gzip); + if (ret) + dev_err(&qm->pdev->dev, "failed to register to gzip (%d)!\n", ret); + + return ret; +} + +static void hisi_zip_unregister_gzip(struct hisi_qm *qm) +{ + if (!hisi_zip_alg_support(qm, HZIP_ALG_GZIP)) + return; + + crypto_unregister_acomp(&hisi_zip_acomp_gzip); +} + +int hisi_zip_register_to_crypto(struct hisi_qm *qm) +{ + int ret = 0; + + ret = hisi_zip_register_zlib(qm); + if (ret) + return ret; + + ret = hisi_zip_register_gzip(qm); + if (ret) + hisi_zip_unregister_zlib(qm); + + return ret; +} + +void hisi_zip_unregister_from_crypto(struct hisi_qm *qm) +{ + hisi_zip_unregister_zlib(qm); + hisi_zip_unregister_gzip(qm); +} diff --git a/kae_driver/hisilicon/zip/zip_main.c b/kae_driver/hisilicon/zip/zip_main.c new file mode 100755 index 0000000..a61a176 --- /dev/null +++ b/kae_driver/hisilicon/zip/zip_main.c @@ -0,0 +1,1354 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "zip.h" + +#define PCI_DEVICE_ID_HUAWEI_ZIP_PF 0xa250 +#define PCI_DEVICE_ID_HUAWEI_ZIP_VF 0xa251 + +#define HZIP_QUEUE_NUM_V1 4096 + +#define HZIP_CLOCK_GATE_CTRL 0x301004 +#define HZIP_DECOMP_CHECK_ENABLE BIT(16) +#define HZIP_FSM_MAX_CNT 0x301008 + +#define HZIP_PORT_ARCA_CHE_0 0x301040 +#define HZIP_PORT_ARCA_CHE_1 0x301044 +#define HZIP_PORT_AWCA_CHE_0 0x301060 +#define HZIP_PORT_AWCA_CHE_1 0x301064 +#define HZIP_CACHE_ALL_EN 0xffffffff + +#define HZIP_BD_RUSER_32_63 0x301110 +#define HZIP_SGL_RUSER_32_63 0x30111c +#define HZIP_DATA_RUSER_32_63 0x301128 +#define HZIP_DATA_WUSER_32_63 0x301134 +#define HZIP_BD_WUSER_32_63 0x301140 + +#define HZIP_QM_IDEL_STATUS 0x3040e4 + +#define HZIP_CORE_DFX_BASE 0x301000 +#define HZIP_CLOCK_GATED_CONTL 0X301004 +#define HZIP_CORE_DFX_COMP_0 0x302000 +#define HZIP_CORE_DFX_COMP_1 0x303000 +#define HZIP_CORE_DFX_DECOMP_0 0x304000 +#define HZIP_CORE_DFX_DECOMP_1 0x305000 +#define HZIP_CORE_DFX_DECOMP_2 0x306000 +#define HZIP_CORE_DFX_DECOMP_3 0x307000 +#define HZIP_CORE_DFX_DECOMP_4 0x308000 +#define HZIP_CORE_DFX_DECOMP_5 0x309000 +#define HZIP_CORE_REGS_BASE_LEN 0xB0 +#define HZIP_CORE_REGS_DFX_LEN 0x28 + +#define HZIP_CORE_INT_SOURCE 0x3010A0 +#define HZIP_CORE_INT_MASK_REG 0x3010A4 +#define HZIP_CORE_INT_SET 0x3010A8 +#define HZIP_CORE_INT_STATUS 0x3010AC +#define HZIP_CORE_INT_STATUS_M_ECC BIT(1) +#define HZIP_CORE_SRAM_ECC_ERR_INFO 0x301148 +#define HZIP_CORE_INT_RAS_CE_ENB 0x301160 +#define HZIP_CORE_INT_RAS_NFE_ENB 0x301164 +#define HZIP_CORE_INT_RAS_FE_ENB 0x301168 +#define HZIP_CORE_INT_RAS_FE_ENB_MASK 0x0 +#define HZIP_OOO_SHUTDOWN_SEL 0x30120C +#define HZIP_SRAM_ECC_ERR_NUM_SHIFT 16 +#define HZIP_SRAM_ECC_ERR_ADDR_SHIFT 24 +#define HZIP_CORE_INT_MASK_ALL GENMASK(12, 0) +#define HZIP_SQE_SIZE 128 +#define HZIP_PF_DEF_Q_NUM 64 +#define HZIP_PF_DEF_Q_BASE 0 + +#define HZIP_SOFT_CTRL_CNT_CLR_CE 0x301000 +#define HZIP_SOFT_CTRL_CNT_CLR_CE_BIT BIT(0) +#define HZIP_SOFT_CTRL_ZIP_CONTROL 0x30100C +#define HZIP_AXI_SHUTDOWN_ENABLE BIT(14) +#define HZIP_WR_PORT BIT(11) + +#define HZIP_DEV_ALG_MAX_LEN 256 +#define HZIP_ALG_ZLIB_BIT GENMASK(1, 0) +#define HZIP_ALG_GZIP_BIT GENMASK(3, 2) +#define HZIP_ALG_DEFLATE_BIT GENMASK(5, 4) +#define HZIP_ALG_LZ77_BIT GENMASK(7, 6) + +#define HZIP_BUF_SIZE 22 +#define HZIP_SQE_MASK_OFFSET 64 +#define HZIP_SQE_MASK_LEN 48 + +#define HZIP_CNT_CLR_CE_EN BIT(0) +#define HZIP_RO_CNT_CLR_CE_EN BIT(2) +#define HZIP_RD_CNT_CLR_CE_EN (HZIP_CNT_CLR_CE_EN | \ + HZIP_RO_CNT_CLR_CE_EN) + +#define HZIP_PREFETCH_CFG 0x3011B0 +#define HZIP_SVA_TRANS 0x3011C4 +#define HZIP_PREFETCH_ENABLE (~(BIT(26) | BIT(17) | BIT(0))) +#define HZIP_SVA_PREFETCH_DISABLE BIT(26) +#define HZIP_SVA_DISABLE_READY (BIT(26) | BIT(30)) +#define HZIP_SHAPER_RATE_COMPRESS 750 +#define HZIP_SHAPER_RATE_DECOMPRESS 140 +#define HZIP_DELAY_1_US 1 +#define HZIP_POLL_TIMEOUT_US 1000 + +/* clock gating */ +#define HZIP_PEH_CFG_AUTO_GATE 0x3011A8 +#define HZIP_PEH_CFG_AUTO_GATE_EN BIT(0) +#define HZIP_CORE_GATED_EN GENMASK(15, 8) +#define HZIP_CORE_GATED_OOO_EN BIT(29) +#define HZIP_CLOCK_GATED_EN (HZIP_CORE_GATED_EN | \ + HZIP_CORE_GATED_OOO_EN) + +static const char hisi_zip_name[] = "hisi_zip"; +static struct dentry *hzip_debugfs_root; + +struct hisi_zip_hw_error { + u32 int_msk; + const char *msg; +}; + +struct zip_dfx_item { + const char *name; + u32 offset; +}; + +struct zip_dev_alg { + u32 alg_msk; + const char *algs; +}; + +static const struct zip_dev_alg zip_dev_algs[] = { { + .alg_msk = HZIP_ALG_ZLIB_BIT, + .algs = "zlib\n", + }, { + .alg_msk = HZIP_ALG_GZIP_BIT, + .algs = "gzip\n", + }, { + .alg_msk = HZIP_ALG_DEFLATE_BIT, + .algs = "deflate\n", + }, { + .alg_msk = HZIP_ALG_LZ77_BIT, + .algs = "lz77_zstd\n", + }, +}; + +static struct hisi_qm_list zip_devices = { + .register_to_crypto = hisi_zip_register_to_crypto, + .unregister_from_crypto = hisi_zip_unregister_from_crypto, +}; + +static struct zip_dfx_item zip_dfx_files[] = { + {"send_cnt", offsetof(struct hisi_zip_dfx, send_cnt)}, + {"recv_cnt", offsetof(struct hisi_zip_dfx, recv_cnt)}, + {"send_busy_cnt", offsetof(struct hisi_zip_dfx, send_busy_cnt)}, + {"err_bd_cnt", offsetof(struct hisi_zip_dfx, err_bd_cnt)}, +}; + +static const struct hisi_zip_hw_error zip_hw_error[] = { + { .int_msk = BIT(0), .msg = "zip_ecc_1bitt_err" }, + { .int_msk = BIT(1), .msg = "zip_ecc_2bit_err" }, + { .int_msk = BIT(2), .msg = "zip_axi_rresp_err" }, + { .int_msk = BIT(3), .msg = "zip_axi_bresp_err" }, + { .int_msk = BIT(4), .msg = "zip_src_addr_parse_err" }, + { .int_msk = BIT(5), .msg = "zip_dst_addr_parse_err" }, + { .int_msk = BIT(6), .msg = "zip_pre_in_addr_err" }, + { .int_msk = BIT(7), .msg = "zip_pre_in_data_err" }, + { .int_msk = BIT(8), .msg = "zip_com_inf_err" }, + { .int_msk = BIT(9), .msg = "zip_enc_inf_err" }, + { .int_msk = BIT(10), .msg = "zip_pre_out_err" }, + { .int_msk = BIT(11), .msg = "zip_axi_poison_err" }, + { .int_msk = BIT(12), .msg = "zip_sva_err" }, + { /* sentinel */ } +}; + +enum ctrl_debug_file_index { + HZIP_CLEAR_ENABLE, + HZIP_DEBUG_FILE_NUM, +}; + +static const char * const ctrl_debug_file_name[] = { + [HZIP_CLEAR_ENABLE] = "clear_enable", +}; + +struct ctrl_debug_file { + enum ctrl_debug_file_index index; + spinlock_t lock; + struct hisi_zip_ctrl *ctrl; +}; + +/* + * One ZIP controller has one PF and multiple VFs, some global configurations + * which PF has need this structure. + * + * Just relevant for PF. + */ +struct hisi_zip_ctrl { + struct hisi_zip *hisi_zip; + struct ctrl_debug_file files[HZIP_DEBUG_FILE_NUM]; +}; + +enum zip_cap_type { + ZIP_QM_NFE_MASK_CAP = 0x0, + ZIP_QM_RESET_MASK_CAP, + ZIP_QM_OOO_SHUTDOWN_MASK_CAP, + ZIP_QM_CE_MASK_CAP, + ZIP_NFE_MASK_CAP, + ZIP_RESET_MASK_CAP, + ZIP_OOO_SHUTDOWN_MASK_CAP, + ZIP_CE_MASK_CAP, + ZIP_CLUSTER_NUM_CAP, + ZIP_CORE_TYPE_NUM_CAP, + ZIP_CORE_NUM_CAP, + ZIP_CLUSTER_COMP_NUM_CAP, + ZIP_CLUSTER_DECOMP_NUM_CAP, + ZIP_DECOMP_ENABLE_BITMAP, + ZIP_COMP_ENABLE_BITMAP, + ZIP_DRV_ALG_BITMAP, + ZIP_DEV_ALG_BITMAP, + ZIP_CORE1_ALG_BITMAP, + ZIP_CORE2_ALG_BITMAP, + ZIP_CORE3_ALG_BITMAP, + ZIP_CORE4_ALG_BITMAP, + ZIP_CORE5_ALG_BITMAP, + ZIP_CAP_MAX +}; + +static struct hisi_qm_cap_info zip_basic_cap_info[] = { + {ZIP_QM_NFE_MASK_CAP, 0x3124, 0, GENMASK(31, 0), 0x0, 0x1C57, 0x7C77}, + {ZIP_QM_RESET_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0xC57, 0x6C77}, + {ZIP_QM_OOO_SHUTDOWN_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0x4, 0x6C77}, + {ZIP_QM_CE_MASK_CAP, 0x312C, 0, GENMASK(31, 0), 0x0, 0x8, 0x8}, + {ZIP_NFE_MASK_CAP, 0x3130, 0, GENMASK(31, 0), 0x0, 0x7FE, 0x1FFE}, + {ZIP_RESET_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x7FE, 0x7FE}, + {ZIP_OOO_SHUTDOWN_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x2, 0x7FE}, + {ZIP_CE_MASK_CAP, 0x3138, 0, GENMASK(31, 0), 0x0, 0x1, 0x1}, + {ZIP_CLUSTER_NUM_CAP, 0x313C, 28, GENMASK(3, 0), 0x1, 0x1, 0x1}, + {ZIP_CORE_TYPE_NUM_CAP, 0x313C, 24, GENMASK(3, 0), 0x2, 0x2, 0x2}, + {ZIP_CORE_NUM_CAP, 0x313C, 16, GENMASK(7, 0), 0x8, 0x8, 0x5}, + {ZIP_CLUSTER_COMP_NUM_CAP, 0x313C, 8, GENMASK(7, 0), 0x2, 0x2, 0x2}, + {ZIP_CLUSTER_DECOMP_NUM_CAP, 0x313C, 0, GENMASK(7, 0), 0x6, 0x6, 0x3}, + {ZIP_DECOMP_ENABLE_BITMAP, 0x3140, 16, GENMASK(15, 0), 0xFC, 0xFC, 0x1C}, + {ZIP_COMP_ENABLE_BITMAP, 0x3140, 0, GENMASK(15, 0), 0x3, 0x3, 0x3}, + {ZIP_DRV_ALG_BITMAP, 0x3144, 0, GENMASK(31, 0), 0xF, 0xF, 0xF}, + {ZIP_DEV_ALG_BITMAP, 0x3148, 0, GENMASK(31, 0), 0xF, 0xF, 0xFF}, + {ZIP_CORE1_ALG_BITMAP, 0x314C, 0, GENMASK(31, 0), 0x5, 0x5, 0xD5}, + {ZIP_CORE2_ALG_BITMAP, 0x3150, 0, GENMASK(31, 0), 0x5, 0x5, 0xD5}, + {ZIP_CORE3_ALG_BITMAP, 0x3154, 0, GENMASK(31, 0), 0xA, 0xA, 0x2A}, + {ZIP_CORE4_ALG_BITMAP, 0x3158, 0, GENMASK(31, 0), 0xA, 0xA, 0x2A}, + {ZIP_CORE5_ALG_BITMAP, 0x315C, 0, GENMASK(31, 0), 0xA, 0xA, 0x2A}, + {ZIP_CAP_MAX, 0x317c, 0, GENMASK(0, 0), 0x0, 0x0, 0x0} +}; + +enum { + HZIP_COMP_CORE0, + HZIP_COMP_CORE1, + HZIP_DECOMP_CORE0, + HZIP_DECOMP_CORE1, + HZIP_DECOMP_CORE2, + HZIP_DECOMP_CORE3, + HZIP_DECOMP_CORE4, + HZIP_DECOMP_CORE5, +}; + +static const u64 core_offsets[] = { + [HZIP_COMP_CORE0] = 0x302000, + [HZIP_COMP_CORE1] = 0x303000, + [HZIP_DECOMP_CORE0] = 0x304000, + [HZIP_DECOMP_CORE1] = 0x305000, + [HZIP_DECOMP_CORE2] = 0x306000, + [HZIP_DECOMP_CORE3] = 0x307000, + [HZIP_DECOMP_CORE4] = 0x308000, + [HZIP_DECOMP_CORE5] = 0x309000, +}; + +static const struct debugfs_reg32 hzip_dfx_regs[] = { + {"HZIP_GET_BD_NUM ", 0x00ull}, + {"HZIP_GET_RIGHT_BD ", 0x04ull}, + {"HZIP_GET_ERROR_BD ", 0x08ull}, + {"HZIP_DONE_BD_NUM ", 0x0cull}, + {"HZIP_WORK_CYCLE ", 0x10ull}, + {"HZIP_IDLE_CYCLE ", 0x18ull}, + {"HZIP_MAX_DELAY ", 0x20ull}, + {"HZIP_MIN_DELAY ", 0x24ull}, + {"HZIP_AVG_DELAY ", 0x28ull}, + {"HZIP_MEM_VISIBLE_DATA ", 0x30ull}, + {"HZIP_MEM_VISIBLE_ADDR ", 0x34ull}, + {"HZIP_CONSUMED_BYTE ", 0x38ull}, + {"HZIP_PRODUCED_BYTE ", 0x40ull}, + {"HZIP_COMP_INF ", 0x70ull}, + {"HZIP_PRE_OUT ", 0x78ull}, + {"HZIP_BD_RD ", 0x7cull}, + {"HZIP_BD_WR ", 0x80ull}, + {"HZIP_GET_BD_AXI_ERR_NUM ", 0x84ull}, + {"HZIP_GET_BD_PARSE_ERR_NUM ", 0x88ull}, + {"HZIP_ADD_BD_AXI_ERR_NUM ", 0x8cull}, + {"HZIP_DECOMP_STF_RELOAD_CURR_ST ", 0x94ull}, + {"HZIP_DECOMP_LZ77_CURR_ST ", 0x9cull}, +}; + +static const struct debugfs_reg32 hzip_com_dfx_regs[] = { + {"HZIP_CLOCK_GATE_CTRL ", 0x301004}, + {"HZIP_CORE_INT_RAS_CE_ENB ", 0x301160}, + {"HZIP_CORE_INT_RAS_NFE_ENB ", 0x301164}, + {"HZIP_CORE_INT_RAS_FE_ENB ", 0x301168}, + {"HZIP_UNCOM_ERR_RAS_CTRL ", 0x30116C}, +}; + +static const struct debugfs_reg32 hzip_dump_dfx_regs[] = { + {"HZIP_GET_BD_NUM ", 0x00ull}, + {"HZIP_GET_RIGHT_BD ", 0x04ull}, + {"HZIP_GET_ERROR_BD ", 0x08ull}, + {"HZIP_DONE_BD_NUM ", 0x0cull}, + {"HZIP_MAX_DELAY ", 0x20ull}, +}; + +/* define the ZIP's dfx regs region and region length */ +static struct dfx_diff_registers hzip_diff_regs[] = { + { + .reg_offset = HZIP_CORE_DFX_BASE, + .reg_len = HZIP_CORE_REGS_BASE_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_COMP_0, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_COMP_1, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_0, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_1, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_2, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_3, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_4, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, { + .reg_offset = HZIP_CORE_DFX_DECOMP_5, + .reg_len = HZIP_CORE_REGS_DFX_LEN, + }, +}; + +static int hzip_diff_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + + hisi_qm_acc_diff_regs_dump(qm, s, qm->debug.acc_diff_regs, + ARRAY_SIZE(hzip_diff_regs)); + + return 0; +} +DEFINE_SHOW_ATTRIBUTE(hzip_diff_regs); +static const struct kernel_param_ops zip_uacce_mode_ops = { + .set = uacce_mode_set, + .get = param_get_int, +}; + +/* + * uacce_mode = 0 means zip only register to crypto, + * uacce_mode = 1 means zip both register to crypto and uacce. + */ +static u32 uacce_mode = UACCE_MODE_NOUACCE; +module_param_cb(uacce_mode, &zip_uacce_mode_ops, &uacce_mode, 0444); +MODULE_PARM_DESC(uacce_mode, UACCE_MODE_DESC); + +static int pf_q_num_set(const char *val, const struct kernel_param *kp) +{ + return q_num_set(val, kp, PCI_DEVICE_ID_HUAWEI_ZIP_PF); +} + +static const struct kernel_param_ops pf_q_num_ops = { + .set = pf_q_num_set, + .get = param_get_int, +}; + +static u32 pf_q_num = HZIP_PF_DEF_Q_NUM; +module_param_cb(pf_q_num, &pf_q_num_ops, &pf_q_num, 0444); +MODULE_PARM_DESC(pf_q_num, "Number of queues in PF(v1 2-4096, v2 2-1024)"); + +static const struct kernel_param_ops vfs_num_ops = { + .set = vfs_num_set, + .get = param_get_int, +}; + +static u32 vfs_num; +module_param_cb(vfs_num, &vfs_num_ops, &vfs_num, 0444); +MODULE_PARM_DESC(vfs_num, "Number of VFs to enable(1-63), 0(default)"); + +static const struct pci_device_id hisi_zip_dev_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_ZIP_PF) }, + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_ZIP_VF) }, + { 0, } +}; +MODULE_DEVICE_TABLE(pci, hisi_zip_dev_ids); + +int zip_create_qps(struct hisi_qp **qps, int qp_num, int node) +{ + if (node == NUMA_NO_NODE) + node = cpu_to_node(smp_processor_id()); + + return hisi_qm_alloc_qps_node(&zip_devices, qp_num, 0, node, qps); +} + +bool hisi_zip_alg_support(struct hisi_qm *qm, u32 alg) +{ + u32 cap_val; + + cap_val = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_DRV_ALG_BITMAP, qm->cap_ver); + if ((alg & cap_val) == alg) + return true; + + return false; +} + +static int hisi_zip_set_qm_algs(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + char *algs, *ptr; + u32 alg_mask; + int i; + + if (!qm->use_sva) + return 0; + + algs = devm_kzalloc(dev, HZIP_DEV_ALG_MAX_LEN * sizeof(char), GFP_KERNEL); + if (!algs) + return -ENOMEM; + + alg_mask = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_DEV_ALG_BITMAP, qm->cap_ver); + + for (i = 0; i < ARRAY_SIZE(zip_dev_algs); i++) + if (alg_mask & zip_dev_algs[i].alg_msk) + strcat(algs, zip_dev_algs[i].algs); + + ptr = strrchr(algs, '\n'); + if (ptr) + *ptr = '\0'; + + qm->uacce->algs = algs; + + return 0; +} + +static void hisi_zip_open_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + /* Enable prefetch */ + val = readl_relaxed(qm->io_base + HZIP_PREFETCH_CFG); + val &= HZIP_PREFETCH_ENABLE; + writel(val, qm->io_base + HZIP_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HZIP_PREFETCH_CFG, + val, !(val & HZIP_SVA_PREFETCH_DISABLE), + HZIP_DELAY_1_US, HZIP_POLL_TIMEOUT_US); + if (ret) + pci_err(qm->pdev, "failed to open sva prefetch\n"); +} + +static void hisi_zip_close_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + val = readl_relaxed(qm->io_base + HZIP_PREFETCH_CFG); + val |= HZIP_SVA_PREFETCH_DISABLE; + writel(val, qm->io_base + HZIP_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HZIP_SVA_TRANS, + val, !(val & HZIP_SVA_DISABLE_READY), + HZIP_DELAY_1_US, HZIP_POLL_TIMEOUT_US); + if (ret) + pci_err(qm->pdev, "failed to close sva prefetch\n"); +} + +static void hisi_zip_enable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + HZIP_CLOCK_GATE_CTRL); + val |= HZIP_CLOCK_GATED_EN; + writel(val, qm->io_base + HZIP_CLOCK_GATE_CTRL); + + val = readl(qm->io_base + HZIP_PEH_CFG_AUTO_GATE); + val |= HZIP_PEH_CFG_AUTO_GATE_EN; + writel(val, qm->io_base + HZIP_PEH_CFG_AUTO_GATE); +} + +static int hisi_zip_set_user_domain_and_cache(struct hisi_qm *qm) +{ + void __iomem *base = qm->io_base; + u32 dcomp_bm, comp_bm; + + /* qm user domain */ + writel(AXUSER_BASE, base + QM_ARUSER_M_CFG_1); + writel(ARUSER_M_CFG_ENABLE, base + QM_ARUSER_M_CFG_ENABLE); + writel(AXUSER_BASE, base + QM_AWUSER_M_CFG_1); + writel(AWUSER_M_CFG_ENABLE, base + QM_AWUSER_M_CFG_ENABLE); + writel(WUSER_M_CFG_ENABLE, base + QM_WUSER_M_CFG_ENABLE); + + /* qm cache */ + writel(AXI_M_CFG, base + QM_AXI_M_CFG); + writel(AXI_M_CFG_ENABLE, base + QM_AXI_M_CFG_ENABLE); + + /* disable FLR triggered by BME(bus master enable) */ + writel(PEH_AXUSER_CFG, base + QM_PEH_AXUSER_CFG); + writel(PEH_AXUSER_CFG_ENABLE, base + QM_PEH_AXUSER_CFG_ENABLE); + + /* cache */ + writel(HZIP_CACHE_ALL_EN, base + HZIP_PORT_ARCA_CHE_0); + writel(HZIP_CACHE_ALL_EN, base + HZIP_PORT_ARCA_CHE_1); + writel(HZIP_CACHE_ALL_EN, base + HZIP_PORT_AWCA_CHE_0); + writel(HZIP_CACHE_ALL_EN, base + HZIP_PORT_AWCA_CHE_1); + + /* user domain configurations */ + writel(AXUSER_BASE, base + HZIP_BD_RUSER_32_63); + writel(AXUSER_BASE, base + HZIP_BD_WUSER_32_63); + + if (qm->use_sva && qm->ver == QM_HW_V2) { + writel(AXUSER_BASE | AXUSER_SSV, base + HZIP_DATA_RUSER_32_63); + writel(AXUSER_BASE | AXUSER_SSV, base + HZIP_DATA_WUSER_32_63); + writel(AXUSER_BASE | AXUSER_SSV, base + HZIP_SGL_RUSER_32_63); + } else { + writel(AXUSER_BASE, base + HZIP_DATA_RUSER_32_63); + writel(AXUSER_BASE, base + HZIP_DATA_WUSER_32_63); + writel(AXUSER_BASE, base + HZIP_SGL_RUSER_32_63); + } + + /* let's open all compression/decompression cores */ + dcomp_bm = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_DECOMP_ENABLE_BITMAP, qm->cap_ver); + comp_bm = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_COMP_ENABLE_BITMAP, qm->cap_ver); + writel(HZIP_DECOMP_CHECK_ENABLE | dcomp_bm | comp_bm, base + HZIP_CLOCK_GATE_CTRL); + + /* enable sqc,cqc writeback */ + writel(SQC_CACHE_ENABLE | CQC_CACHE_ENABLE | SQC_CACHE_WB_ENABLE | + CQC_CACHE_WB_ENABLE | FIELD_PREP(SQC_CACHE_WB_THRD, 1) | + FIELD_PREP(CQC_CACHE_WB_THRD, 1), base + QM_CACHE_CTL); + + hisi_zip_enable_clock_gate(qm); + + return 0; +} + +static void hisi_zip_master_ooo_ctrl(struct hisi_qm *qm, bool enable) +{ + u32 val1, val2; + + val1 = readl(qm->io_base + HZIP_SOFT_CTRL_ZIP_CONTROL); + if (enable) { + val1 |= HZIP_AXI_SHUTDOWN_ENABLE; + val2 = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + } else { + val1 &= ~HZIP_AXI_SHUTDOWN_ENABLE; + val2 = 0x0; + } + + if (qm->ver > QM_HW_V2) + writel(val2, qm->io_base + HZIP_OOO_SHUTDOWN_SEL); + + writel(val1, qm->io_base + HZIP_SOFT_CTRL_ZIP_CONTROL); +} + +static void hisi_zip_hw_error_enable(struct hisi_qm *qm) +{ + u32 nfe, ce; + + if (qm->ver == QM_HW_V1) { + writel(HZIP_CORE_INT_MASK_ALL, + qm->io_base + HZIP_CORE_INT_MASK_REG); + dev_info(&qm->pdev->dev, "Does not support hw error handle\n"); + return; + } + + nfe = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_NFE_MASK_CAP, qm->cap_ver); + ce = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CE_MASK_CAP, qm->cap_ver); + + /* clear ZIP hw error source if having */ + writel(ce | nfe | HZIP_CORE_INT_RAS_FE_ENB_MASK, qm->io_base + HZIP_CORE_INT_SOURCE); + + /* configure error type */ + writel(ce, qm->io_base + HZIP_CORE_INT_RAS_CE_ENB); + writel(HZIP_CORE_INT_RAS_FE_ENB_MASK, qm->io_base + HZIP_CORE_INT_RAS_FE_ENB); + writel(nfe, qm->io_base + HZIP_CORE_INT_RAS_NFE_ENB); + + hisi_zip_master_ooo_ctrl(qm, true); + + /* enable ZIP hw error interrupts */ + writel(0, qm->io_base + HZIP_CORE_INT_MASK_REG); +} + +static void hisi_zip_hw_error_disable(struct hisi_qm *qm) +{ + u32 nfe, ce; + + /* disable ZIP hw error interrupts */ + nfe = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_NFE_MASK_CAP, qm->cap_ver); + ce = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CE_MASK_CAP, qm->cap_ver); + writel(ce | nfe | HZIP_CORE_INT_RAS_FE_ENB_MASK, qm->io_base + HZIP_CORE_INT_MASK_REG); + + hisi_zip_master_ooo_ctrl(qm, false); +} + +static inline struct hisi_qm *file_to_qm(struct ctrl_debug_file *file) +{ + struct hisi_zip *hisi_zip = file->ctrl->hisi_zip; + + return &hisi_zip->qm; +} + +static u32 clear_enable_read(struct hisi_qm *qm) +{ + return readl(qm->io_base + HZIP_SOFT_CTRL_CNT_CLR_CE) & + HZIP_SOFT_CTRL_CNT_CLR_CE_BIT; +} + +static int clear_enable_write(struct hisi_qm *qm, u32 val) +{ + u32 tmp; + + if (val != 1 && val != 0) + return -EINVAL; + + tmp = (readl(qm->io_base + HZIP_SOFT_CTRL_CNT_CLR_CE) & + ~HZIP_SOFT_CTRL_CNT_CLR_CE_BIT) | val; + writel(tmp, qm->io_base + HZIP_SOFT_CTRL_CNT_CLR_CE); + + return 0; +} + +static ssize_t hisi_zip_ctrl_debug_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct ctrl_debug_file *file = filp->private_data; + struct hisi_qm *qm = file_to_qm(file); + char tbuf[HZIP_BUF_SIZE]; + u32 val; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->index) { + case HZIP_CLEAR_ENABLE: + val = clear_enable_read(qm); + break; + default: + goto err_input; + } + spin_unlock_irq(&file->lock); + + hisi_qm_put_dfx_access(qm); + ret = scnprintf(tbuf, sizeof(tbuf), "%u\n", val); + return simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return -EINVAL; +} + +static ssize_t hisi_zip_ctrl_debug_write(struct file *filp, + const char __user *buf, + size_t count, loff_t *pos) +{ + struct ctrl_debug_file *file = filp->private_data; + struct hisi_qm *qm = file_to_qm(file); + char tbuf[HZIP_BUF_SIZE]; + unsigned long val; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= HZIP_BUF_SIZE) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, HZIP_BUF_SIZE - 1, pos, buf, count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + ret = kstrtoul(tbuf, 0, &val); + if (ret) + return ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->index) { + case HZIP_CLEAR_ENABLE: + ret = clear_enable_write(qm, val); + if (ret) + goto err_input; + break; + default: + ret = -EINVAL; + goto err_input; + } + + ret = count; + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return ret; +} + +static const struct file_operations ctrl_debug_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = hisi_zip_ctrl_debug_read, + .write = hisi_zip_ctrl_debug_write, +}; + +static int zip_debugfs_atomic64_set(void *data, u64 val) +{ + if (val) + return -EINVAL; + + atomic64_set((atomic64_t *)data, 0); + + return 0; +} + +static int zip_debugfs_atomic64_get(void *data, u64 *val) +{ + *val = atomic64_read((atomic64_t *)data); + + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(zip_atomic64_ops, zip_debugfs_atomic64_get, + zip_debugfs_atomic64_set, "%llu\n"); + +static int hisi_zip_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hisi_zip_regs); + +static int hisi_zip_core_debug_init(struct hisi_qm *qm) +{ + u32 zip_core_num, zip_comp_core_num; + struct device *dev = &qm->pdev->dev; + struct debugfs_regset32 *regset; + struct dentry *tmp_d; + char buf[HZIP_BUF_SIZE]; + int i; + + zip_core_num = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CORE_NUM_CAP, qm->cap_ver); + zip_comp_core_num = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CLUSTER_COMP_NUM_CAP, + qm->cap_ver); + + for (i = 0; i < zip_core_num; i++) { + if (i < zip_comp_core_num) + scnprintf(buf, sizeof(buf), "comp_core%d", i); + else + scnprintf(buf, sizeof(buf), "decomp_core%d", + i - zip_comp_core_num); + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOENT; + + regset->regs = hzip_dfx_regs; + regset->nregs = ARRAY_SIZE(hzip_dfx_regs); + regset->base = qm->io_base + core_offsets[i]; + regset->dev = dev; + + tmp_d = debugfs_create_dir(buf, qm->debug.debug_root); + debugfs_create_file("regs", 0444, tmp_d, regset, + &hisi_zip_regs_fops); + } + + return 0; +} + +static void hisi_zip_dfx_debug_init(struct hisi_qm *qm) +{ + struct dfx_diff_registers *hzip_regs = qm->debug.acc_diff_regs; + struct hisi_zip *zip = container_of(qm, struct hisi_zip, qm); + struct hisi_zip_dfx *dfx = &zip->dfx; + struct dentry *tmp_dir; + void *data; + int i; + + tmp_dir = debugfs_create_dir("zip_dfx", qm->debug.debug_root); + for (i = 0; i < ARRAY_SIZE(zip_dfx_files); i++) { + data = (atomic64_t *)((uintptr_t)dfx + zip_dfx_files[i].offset); + debugfs_create_file(zip_dfx_files[i].name, + 0644, tmp_dir, data, + &zip_atomic64_ops); + } + + if (qm->fun_type == QM_HW_PF && hzip_regs) + debugfs_create_file("diff_regs", 0444, tmp_dir, + qm, &hzip_diff_regs_fops); +} + +static int hisi_zip_ctrl_debug_init(struct hisi_qm *qm) +{ + struct hisi_zip *zip = container_of(qm, struct hisi_zip, qm); + int i; + + for (i = HZIP_CLEAR_ENABLE; i < HZIP_DEBUG_FILE_NUM; i++) { + spin_lock_init(&zip->ctrl->files[i].lock); + zip->ctrl->files[i].ctrl = zip->ctrl; + zip->ctrl->files[i].index = i; + + debugfs_create_file(ctrl_debug_file_name[i], 0600, + qm->debug.debug_root, + zip->ctrl->files + i, + &ctrl_debug_fops); + } + + return hisi_zip_core_debug_init(qm); +} + +static int hisi_zip_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct dentry *dev_d; + int ret; + + dev_d = debugfs_create_dir(dev_name(dev), hzip_debugfs_root); + + qm->debug.sqe_mask_offset = HZIP_SQE_MASK_OFFSET; + qm->debug.sqe_mask_len = HZIP_SQE_MASK_LEN; + qm->debug.debug_root = dev_d; + ret = hisi_qm_diff_regs_init(qm, hzip_diff_regs, + ARRAY_SIZE(hzip_diff_regs)); + if (ret) { + dev_warn(dev, "Failed to init ZIP diff regs!\n"); + goto debugfs_remove; + } + + hisi_qm_debug_init(qm); + + if (qm->fun_type == QM_HW_PF) { + ret = hisi_zip_ctrl_debug_init(qm); + if (ret) + goto failed_to_create; + } + + hisi_zip_dfx_debug_init(qm); + + return 0; + +failed_to_create: + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(hzip_diff_regs)); +debugfs_remove: + debugfs_remove_recursive(hzip_debugfs_root); + return ret; +} + +/* hisi_zip_debug_regs_clear() - clear the zip debug regs */ +static void hisi_zip_debug_regs_clear(struct hisi_qm *qm) +{ + int i, j; + + /* enable register read_clear bit */ + writel(HZIP_RD_CNT_CLR_CE_EN, qm->io_base + HZIP_SOFT_CTRL_CNT_CLR_CE); + for (i = 0; i < ARRAY_SIZE(core_offsets); i++) + for (j = 0; j < ARRAY_SIZE(hzip_dfx_regs); j++) + readl(qm->io_base + core_offsets[i] + + hzip_dfx_regs[j].offset); + + /* disable register read_clear bit */ + writel(0x0, qm->io_base + HZIP_SOFT_CTRL_CNT_CLR_CE); + + hisi_qm_debug_regs_clear(qm); +} + +static void hisi_zip_debugfs_exit(struct hisi_qm *qm) +{ + hisi_qm_diff_regs_uninit(qm, ARRAY_SIZE(hzip_diff_regs)); + + debugfs_remove_recursive(qm->debug.debug_root); + + if (qm->fun_type == QM_HW_PF) { + hisi_zip_debug_regs_clear(qm); + qm->debug.curr_qm_qp_num = 0; + } +} + +static int hisi_zip_show_last_regs_init(struct hisi_qm *qm) +{ + int core_dfx_regs_num = ARRAY_SIZE(hzip_dump_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hzip_com_dfx_regs); + struct qm_debug *debug = &qm->debug; + void __iomem *io_base; + u32 zip_core_num; + int i, j, idx; + + zip_core_num = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CORE_NUM_CAP, qm->cap_ver); + + debug->last_words = kcalloc(core_dfx_regs_num * zip_core_num + com_dfx_regs_num, + sizeof(unsigned int), GFP_KERNEL); + if (!debug->last_words) + return -ENOMEM; + + for (i = 0; i < com_dfx_regs_num; i++) { + io_base = qm->io_base + hzip_com_dfx_regs[i].offset; + debug->last_words[i] = readl_relaxed(io_base); + } + + for (i = 0; i < zip_core_num; i++) { + io_base = qm->io_base + core_offsets[i]; + for (j = 0; j < core_dfx_regs_num; j++) { + idx = com_dfx_regs_num + i * core_dfx_regs_num + j; + debug->last_words[idx] = readl_relaxed( + io_base + hzip_dump_dfx_regs[j].offset); + } + } + + return 0; +} + +static void hisi_zip_show_last_regs_uninit(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + kfree(debug->last_words); + debug->last_words = NULL; +} + +static void hisi_zip_show_last_dfx_regs(struct hisi_qm *qm) +{ + int core_dfx_regs_num = ARRAY_SIZE(hzip_dump_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hzip_com_dfx_regs); + u32 zip_core_num, zip_comp_core_num; + struct qm_debug *debug = &qm->debug; + char buf[HZIP_BUF_SIZE]; + void __iomem *base; + int i, j, idx; + u32 val; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + for (i = 0; i < com_dfx_regs_num; i++) { + val = readl_relaxed(qm->io_base + hzip_com_dfx_regs[i].offset); + if (debug->last_words[i] != val) + pci_info(qm->pdev, "com_dfx: %s \t= 0x%08x => 0x%08x\n", + hzip_com_dfx_regs[i].name, debug->last_words[i], val); + } + + zip_core_num = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CORE_NUM_CAP, qm->cap_ver); + zip_comp_core_num = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_CLUSTER_COMP_NUM_CAP, + qm->cap_ver); + for (i = 0; i < zip_core_num; i++) { + if (i < zip_comp_core_num) + scnprintf(buf, sizeof(buf), "Comp_core-%d", i); + else + scnprintf(buf, sizeof(buf), "Decomp_core-%d", + i - zip_comp_core_num); + base = qm->io_base + core_offsets[i]; + + pci_info(qm->pdev, "==>%s:\n", buf); + /* dump last word for dfx regs during control resetting */ + for (j = 0; j < core_dfx_regs_num; j++) { + idx = com_dfx_regs_num + i * core_dfx_regs_num + j; + val = readl_relaxed(base + hzip_dump_dfx_regs[j].offset); + if (debug->last_words[idx] != val) + pci_info(qm->pdev, "%s \t= 0x%08x => 0x%08x\n", + hzip_dump_dfx_regs[j].name, + debug->last_words[idx], val); + } + } +} + +static void hisi_zip_log_hw_error(struct hisi_qm *qm, u32 err_sts) +{ + const struct hisi_zip_hw_error *err = zip_hw_error; + struct device *dev = &qm->pdev->dev; + u32 err_val; + + while (err->msg) { + if (err->int_msk & err_sts) { + dev_err(dev, "%s [error status=0x%x] found\n", + err->msg, err->int_msk); + + if (err->int_msk & HZIP_CORE_INT_STATUS_M_ECC) { + err_val = readl(qm->io_base + + HZIP_CORE_SRAM_ECC_ERR_INFO); + dev_err(dev, "hisi-zip multi ecc sram num=0x%x\n", + ((err_val >> + HZIP_SRAM_ECC_ERR_NUM_SHIFT) & 0xFF)); + } + } + err++; + } +} + +static u32 hisi_zip_get_hw_err_status(struct hisi_qm *qm) +{ + return readl(qm->io_base + HZIP_CORE_INT_STATUS); +} + +static void hisi_zip_clear_hw_err_status(struct hisi_qm *qm, u32 err_sts) +{ + u32 nfe; + + writel(err_sts, qm->io_base + HZIP_CORE_INT_SOURCE); + nfe = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_NFE_MASK_CAP, qm->cap_ver); + writel(nfe, qm->io_base + HZIP_CORE_INT_RAS_NFE_ENB); +} + +static void hisi_zip_open_axi_master_ooo(struct hisi_qm *qm) +{ + u32 val; + + val = readl(qm->io_base + HZIP_SOFT_CTRL_ZIP_CONTROL); + + writel(val & ~HZIP_AXI_SHUTDOWN_ENABLE, + qm->io_base + HZIP_SOFT_CTRL_ZIP_CONTROL); + + writel(val | HZIP_AXI_SHUTDOWN_ENABLE, + qm->io_base + HZIP_SOFT_CTRL_ZIP_CONTROL); +} + +static void hisi_zip_close_axi_master_ooo(struct hisi_qm *qm) +{ + u32 nfe_enb; + + /* Disable ECC Mbit error report. */ + nfe_enb = readl(qm->io_base + HZIP_CORE_INT_RAS_NFE_ENB); + writel(nfe_enb & ~HZIP_CORE_INT_STATUS_M_ECC, + qm->io_base + HZIP_CORE_INT_RAS_NFE_ENB); + + /* Inject zip ECC Mbit error to block master ooo. */ + writel(HZIP_CORE_INT_STATUS_M_ECC, + qm->io_base + HZIP_CORE_INT_SET); +} + +static void hisi_zip_err_info_init(struct hisi_qm *qm) +{ + struct hisi_qm_err_info *err_info = &qm->err_info; + + err_info->fe = HZIP_CORE_INT_RAS_FE_ENB_MASK; + err_info->ce = hisi_qm_get_hw_info(qm, zip_basic_cap_info, ZIP_QM_CE_MASK_CAP, qm->cap_ver); + err_info->nfe = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_QM_NFE_MASK_CAP, qm->cap_ver); + err_info->ecc_2bits_mask = HZIP_CORE_INT_STATUS_M_ECC; + err_info->qm_shutdown_mask = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_QM_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->dev_shutdown_mask = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_reset_mask = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_QM_RESET_MASK_CAP, qm->cap_ver); + err_info->dev_reset_mask = hisi_qm_get_hw_info(qm, zip_basic_cap_info, + ZIP_RESET_MASK_CAP, qm->cap_ver); + err_info->msi_wr_port = HZIP_WR_PORT; + err_info->acpi_rst = "ZRST"; +} + +static const struct hisi_qm_err_ini hisi_zip_err_ini = { + .hw_init = hisi_zip_set_user_domain_and_cache, + .hw_err_enable = hisi_zip_hw_error_enable, + .hw_err_disable = hisi_zip_hw_error_disable, + .get_dev_hw_err_status = hisi_zip_get_hw_err_status, + .clear_dev_hw_err_status = hisi_zip_clear_hw_err_status, + .log_dev_hw_err = hisi_zip_log_hw_error, + .open_axi_master_ooo = hisi_zip_open_axi_master_ooo, + .close_axi_master_ooo = hisi_zip_close_axi_master_ooo, + .open_sva_prefetch = hisi_zip_open_sva_prefetch, + .close_sva_prefetch = hisi_zip_close_sva_prefetch, + .show_last_dfx_regs = hisi_zip_show_last_dfx_regs, + .err_info_init = hisi_zip_err_info_init, +}; + +static int hisi_zip_pf_probe_init(struct hisi_zip *hisi_zip) +{ + struct hisi_qm *qm = &hisi_zip->qm; + struct hisi_zip_ctrl *ctrl; + int ret; + + ctrl = devm_kzalloc(&qm->pdev->dev, sizeof(*ctrl), GFP_KERNEL); + if (!ctrl) + return -ENOMEM; + + hisi_zip->ctrl = ctrl; + ctrl->hisi_zip = hisi_zip; + qm->err_ini = &hisi_zip_err_ini; + qm->err_ini->err_info_init(qm); + + ret = hisi_zip_set_user_domain_and_cache(qm); + if (ret) + return ret; + + hisi_zip_open_sva_prefetch(qm); + hisi_qm_dev_err_init(qm); + hisi_zip_debug_regs_clear(qm); + + ret = hisi_zip_show_last_regs_init(qm); + if (ret) + pci_err(qm->pdev, "Failed to init last word regs!\n"); + + return ret; +} + +static int hisi_zip_qm_init(struct hisi_qm *qm, struct pci_dev *pdev) +{ + int ret; + + qm->pdev = pdev; + qm->ver = pdev->revision; + qm->mode = uacce_mode; + qm->sqe_size = HZIP_SQE_SIZE; + qm->dev_name = hisi_zip_name; + + qm->fun_type = (pdev->device == PCI_DEVICE_ID_HUAWEI_ZIP_PF) ? + QM_HW_PF : QM_HW_VF; + if (qm->fun_type == QM_HW_PF) { + qm->qp_base = HZIP_PF_DEF_Q_BASE; + qm->qp_num = pf_q_num; + qm->debug.curr_qm_qp_num = pf_q_num; + qm->qm_list = &zip_devices; + } else if (qm->fun_type == QM_HW_VF && qm->ver == QM_HW_V1) { + /* + * have no way to get qm configure in VM in v1 hardware, + * so currently force PF to uses HZIP_PF_DEF_Q_NUM, and force + * to trigger only one VF in v1 hardware. + * + * v2 hardware has no such problem. + */ + qm->qp_base = HZIP_PF_DEF_Q_NUM; + qm->qp_num = HZIP_QUEUE_NUM_V1 - HZIP_PF_DEF_Q_NUM; + } + + ret = hisi_qm_init(qm); + if (ret) { + pci_err(qm->pdev, "Failed to init zip qm configures!\n"); + return ret; + } + + ret = hisi_zip_set_qm_algs(qm); + if (ret) { + pci_err(qm->pdev, "Failed to set zip algs!\n"); + hisi_qm_uninit(qm); + } + + return ret; +} + +static void hisi_zip_qm_uninit(struct hisi_qm *qm) +{ + hisi_qm_uninit(qm); +} + +static int hisi_zip_probe_init(struct hisi_zip *hisi_zip) +{ + u32 type_rate = HZIP_SHAPER_RATE_COMPRESS; + struct hisi_qm *qm = &hisi_zip->qm; + int ret; + + if (qm->fun_type == QM_HW_PF) { + ret = hisi_zip_pf_probe_init(hisi_zip); + if (ret) + return ret; + /* enable shaper type 0 */ + if (qm->ver >= QM_HW_V3) { + type_rate |= QM_SHAPER_ENABLE; + + /* ZIP need to enable shaper type 1 */ + type_rate |= HZIP_SHAPER_RATE_DECOMPRESS << QM_SHAPER_TYPE1_OFFSET; + qm->type_rate = type_rate; + } + } + + return 0; +} + +static int hisi_zip_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct hisi_zip *hisi_zip; + struct hisi_qm *qm; + int ret; + + hisi_zip = devm_kzalloc(&pdev->dev, sizeof(*hisi_zip), GFP_KERNEL); + if (!hisi_zip) + return -ENOMEM; + + qm = &hisi_zip->qm; + + ret = hisi_zip_qm_init(qm, pdev); + if (ret) { + pci_err(pdev, "Failed to init ZIP QM (%d)!\n", ret); + return ret; + } + + ret = hisi_zip_probe_init(hisi_zip); + if (ret) { + pci_err(pdev, "Failed to probe (%d)!\n", ret); + goto err_qm_uninit; + } + + ret = hisi_qm_start(qm); + if (ret) + goto err_dev_err_uninit; + + ret = hisi_zip_debugfs_init(qm); + if (ret) + pci_err(pdev, "failed to init debugfs (%d)!\n", ret); + + ret = hisi_qm_alg_register(qm, &zip_devices); + if (ret < 0) { + pci_err(pdev, "failed to register driver to crypto!\n"); + goto err_qm_stop; + } + + if (qm->uacce) { + ret = uacce_register(qm->uacce); + if (ret) { + pci_err(pdev, "failed to register uacce (%d)!\n", ret); + goto err_qm_alg_unregister; + } + } + + if (qm->fun_type == QM_HW_PF && vfs_num > 0) { + ret = hisi_qm_sriov_enable(pdev, vfs_num); + if (ret < 0) + goto err_qm_alg_unregister; + } + + hisi_qm_pm_init(qm); + + return 0; + +err_qm_alg_unregister: + hisi_qm_alg_unregister(qm, &zip_devices); + +err_qm_stop: + hisi_zip_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + +err_dev_err_uninit: + hisi_zip_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + +err_qm_uninit: + hisi_zip_qm_uninit(qm); + + return ret; +} + +static void hisi_zip_remove(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + + hisi_qm_pm_uninit(qm); + hisi_qm_wait_task_finish(qm, &zip_devices); + hisi_qm_alg_unregister(qm, &zip_devices); + + if (qm->fun_type == QM_HW_PF && qm->vfs_num) + hisi_qm_sriov_disable(pdev, true); + + hisi_zip_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + hisi_zip_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + hisi_zip_qm_uninit(qm); +} + +static const struct dev_pm_ops hisi_zip_pm_ops = { + SET_RUNTIME_PM_OPS(hisi_qm_suspend, hisi_qm_resume, NULL) +}; + +static const struct pci_error_handlers hisi_zip_err_handler = { + .error_detected = hisi_qm_dev_err_detected, + .slot_reset = hisi_qm_dev_slot_reset, + .reset_prepare = hisi_qm_reset_prepare, + .reset_done = hisi_qm_reset_done, +}; + +static struct pci_driver hisi_zip_pci_driver = { + .name = "hisi_zip", + .id_table = hisi_zip_dev_ids, + .probe = hisi_zip_probe, + .remove = hisi_zip_remove, + .sriov_configure = IS_ENABLED(CONFIG_PCI_IOV) ? + hisi_qm_sriov_configure : NULL, + .err_handler = &hisi_zip_err_handler, + .shutdown = hisi_qm_dev_shutdown, + .driver.pm = &hisi_zip_pm_ops, +}; + +static void hisi_zip_register_debugfs(void) +{ + if (!debugfs_initialized()) + return; + + hzip_debugfs_root = debugfs_create_dir("hisi_zip", NULL); +} + +static void hisi_zip_unregister_debugfs(void) +{ + debugfs_remove_recursive(hzip_debugfs_root); +} + +static int __init hisi_zip_init(void) +{ + int ret; + + hisi_qm_init_list(&zip_devices); + hisi_zip_register_debugfs(); + + ret = pci_register_driver(&hisi_zip_pci_driver); + if (ret < 0) { + hisi_zip_unregister_debugfs(); + pr_err("Failed to register pci driver.\n"); + } + + return ret; +} + +static void __exit hisi_zip_exit(void) +{ + pci_unregister_driver(&hisi_zip_pci_driver); + hisi_zip_unregister_debugfs(); +} + +module_init(hisi_zip_init); +module_exit(hisi_zip_exit); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Zhou Wang "); +MODULE_DESCRIPTION("Driver for HiSilicon ZIP accelerator"); diff --git a/kae_driver/include_linux/uacce.h b/kae_driver/include_linux/uacce.h new file mode 100755 index 0000000..9ce88c2 --- /dev/null +++ b/kae_driver/include_linux/uacce.h @@ -0,0 +1,149 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +#ifndef _LINUX_UACCE_H +#define _LINUX_UACCE_H + +#include +#include + +#define UACCE_NAME "uacce" +#define UACCE_MAX_REGION 2 +#define UACCE_MAX_NAME_SIZE 64 + +struct uacce_queue; +struct uacce_device; + +/** + * struct uacce_qfile_region - structure of queue file region + * @type: type of the region + */ +struct uacce_qfile_region { + enum uacce_qfrt type; +}; + +/** + * struct uacce_ops - uacce device operations + * @get_available_instances: get available instances left of the device + * @get_queue: get a queue from the device + * @put_queue: free a queue to the device + * @start_queue: make the queue start work after get_queue + * @stop_queue: make the queue stop work before put_queue + * @is_q_updated: check whether the task is finished + * @mmap: mmap addresses of queue to user space + * @ioctl: ioctl for user space users of the queue + */ +struct uacce_ops { + int (*get_available_instances)(struct uacce_device *uacce); + int (*get_queue)(struct uacce_device *uacce, unsigned long arg, + struct uacce_queue *q); + void (*put_queue)(struct uacce_queue *q); + int (*start_queue)(struct uacce_queue *q); + void (*stop_queue)(struct uacce_queue *q); + int (*is_q_updated)(struct uacce_queue *q); + int (*mmap)(struct uacce_queue *q, struct vm_area_struct *vma, + struct uacce_qfile_region *qfr); + long (*ioctl)(struct uacce_queue *q, unsigned int cmd, + unsigned long arg); +}; + +/** + * struct uacce_interface - interface required for uacce_register() + * @name: the uacce device name. Will show up in sysfs + * @flags: uacce device attributes + * @ops: pointer to the struct uacce_ops + */ +struct uacce_interface { + char name[UACCE_MAX_NAME_SIZE]; + unsigned int flags; + const struct uacce_ops *ops; +}; + +enum uacce_q_state { + UACCE_Q_ZOMBIE = 0, + UACCE_Q_INIT, + UACCE_Q_STARTED, +}; + +/** + * struct uacce_queue + * @uacce: pointer to uacce + * @priv: private pointer + * @wait: wait queue head + * @list: index into uacce queues list + * @qfrs: pointer of qfr regions + * @mutex: protects queue state + * @state: queue state machine + * @pasid: pasid associated to the mm + * @handle: iommu_sva handle returned by iommu_sva_bind_device() + */ +struct uacce_queue { + struct uacce_device *uacce; + void *priv; + wait_queue_head_t wait; + struct list_head list; + struct uacce_qfile_region *qfrs[UACCE_MAX_REGION]; + struct mutex mutex; + enum uacce_q_state state; + u32 pasid; + struct iommu_sva *handle; +}; + +/** + * struct uacce_device + * @algs: supported algorithms + * @api_ver: api version + * @ops: pointer to the struct uacce_ops + * @qf_pg_num: page numbers of the queue file regions + * @parent: pointer to the parent device + * @is_vf: whether virtual function + * @flags: uacce attributes + * @dev_id: id of the uacce device + * @cdev: cdev of the uacce + * @dev: dev of the uacce + * @mutex: protects uacce operation + * @priv: private pointer of the uacce + * @queues: list of queues + * @inode: core vfs + */ +struct uacce_device { + const char *algs; + const char *api_ver; + const struct uacce_ops *ops; + unsigned long qf_pg_num[UACCE_MAX_REGION]; + struct device *parent; + bool is_vf; + u32 flags; + u32 dev_id; + struct cdev *cdev; + struct device dev; + struct mutex mutex; + void *priv; + struct list_head queues; + struct inode *inode; +}; + +#if IS_ENABLED(CONFIG_UACCE) + +struct uacce_device *uacce_alloc(struct device *parent, + struct uacce_interface *interface); +int uacce_register(struct uacce_device *uacce); +void uacce_remove(struct uacce_device *uacce); + +#else /* CONFIG_UACCE */ + +static inline +struct uacce_device *uacce_alloc(struct device *parent, + struct uacce_interface *interface) +{ + return ERR_PTR(-ENODEV); +} + +static inline int uacce_register(struct uacce_device *uacce) +{ + return -EINVAL; +} + +static inline void uacce_remove(struct uacce_device *uacce) {} + +#endif /* CONFIG_UACCE */ + +#endif /* _LINUX_UACCE_H */ diff --git a/kae_driver/include_uapi_linux/uacce.h b/kae_driver/include_uapi_linux/uacce.h new file mode 100755 index 0000000..cc71856 --- /dev/null +++ b/kae_driver/include_uapi_linux/uacce.h @@ -0,0 +1,38 @@ +/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */ +#ifndef _UAPIUUACCE_H +#define _UAPIUUACCE_H + +#include +#include + +/* + * UACCE_CMD_START_Q: Start queue + */ +#define UACCE_CMD_START_Q _IO('W', 0) + +/* + * UACCE_CMD_PUT_Q: + * User actively stop queue and free queue resource immediately + * Optimization method since close fd may delay + */ +#define UACCE_CMD_PUT_Q _IO('W', 1) + +/* + * UACCE Device flags: + * UACCE_DEV_SVA: Shared Virtual Addresses + * Support PASID + * Support device page faults (PCI PRI or SMMU Stall) + */ +#define UACCE_DEV_SVA BIT(0) + +/** + * enum uacce_qfrt: queue file region type + * @UACCE_QFRT_MMIO: device mmio region + * @UACCE_QFRT_DUS: device user share region + */ +enum uacce_qfrt { + UACCE_QFRT_MMIO = 0, + UACCE_QFRT_DUS = 1, +}; + +#endif diff --git a/kae_driver/uacce/Kconfig b/kae_driver/uacce/Kconfig new file mode 100755 index 0000000..5e39b60 --- /dev/null +++ b/kae_driver/uacce/Kconfig @@ -0,0 +1,13 @@ +config UACCE + tristate "Accelerator Framework for User Land" + depends on IOMMU_API + help + UACCE provides interface for the user process to access the hardware + without interaction with the kernel space in data path. + + The user-space interface is described in + include/uapi/misc/uacce/uacce.h + + See Documentation/misc-devices/uacce.rst for more details. + + If you don't know what to do here, say N. diff --git a/kae_driver/uacce/Makefile b/kae_driver/uacce/Makefile new file mode 100755 index 0000000..5b4374e --- /dev/null +++ b/kae_driver/uacce/Makefile @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: GPL-2.0-or-later +obj-$(CONFIG_UACCE) += uacce.o diff --git a/kae_driver/uacce/uacce.c b/kae_driver/uacce/uacce.c new file mode 100755 index 0000000..faad393 --- /dev/null +++ b/kae_driver/uacce/uacce.c @@ -0,0 +1,595 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +#include +#include +#include +#include +#include +#include +#include "../include_linux/uacce.h" + +static struct class *uacce_class; +static dev_t uacce_devt; +static DEFINE_XARRAY_ALLOC(uacce_xa); + +/* + * If the parent driver or the device disappears, the queue state is invalid and + * ops are not usable anymore. + */ +static bool uacce_queue_is_valid(struct uacce_queue *q) +{ + return q->state == UACCE_Q_INIT || q->state == UACCE_Q_STARTED; +} + +static int uacce_start_queue(struct uacce_queue *q) +{ + int ret; + + if (q->state != UACCE_Q_INIT) + return -EINVAL; + + if (q->uacce->ops->start_queue) { + ret = q->uacce->ops->start_queue(q); + if (ret < 0) + return ret; + } + + q->state = UACCE_Q_STARTED; + return 0; +} + +static int uacce_put_queue(struct uacce_queue *q) +{ + struct uacce_device *uacce = q->uacce; + + if ((q->state == UACCE_Q_STARTED) && uacce->ops->stop_queue) + uacce->ops->stop_queue(q); + + if ((q->state == UACCE_Q_INIT || q->state == UACCE_Q_STARTED) && + uacce->ops->put_queue) + uacce->ops->put_queue(q); + + q->state = UACCE_Q_ZOMBIE; + + return 0; +} + +static long uacce_fops_unl_ioctl(struct file *filep, + unsigned int cmd, unsigned long arg) +{ + struct uacce_queue *q = filep->private_data; + struct uacce_device *uacce = q->uacce; + long ret = -ENXIO; + + /* + * uacce->ops->ioctl() may take the mmap_lock when copying arg to/from + * user. Avoid a circular lock dependency with uacce_fops_mmap(), which + * gets called with mmap_lock held, by taking uacce->mutex instead of + * q->mutex. Doing this in uacce_fops_mmap() is not possible because + * uacce_fops_open() calls iommu_sva_bind_device(), which takes + * mmap_lock, while holding uacce->mutex. + */ + mutex_lock(&uacce->mutex); + if (!uacce_queue_is_valid(q)) + goto out_unlock; + + switch (cmd) { + case UACCE_CMD_START_Q: + ret = uacce_start_queue(q); + break; + case UACCE_CMD_PUT_Q: + ret = uacce_put_queue(q); + break; + default: + if (uacce->ops->ioctl) + ret = uacce->ops->ioctl(q, cmd, arg); + else + ret = -EINVAL; + } +out_unlock: + mutex_unlock(&uacce->mutex); + return ret; +} + +#ifdef CONFIG_COMPAT +static long uacce_fops_compat_ioctl(struct file *filep, + unsigned int cmd, unsigned long arg) +{ + arg = (unsigned long)compat_ptr(arg); + + return uacce_fops_unl_ioctl(filep, cmd, arg); +} +#endif + +static int uacce_bind_queue(struct uacce_device *uacce, struct uacce_queue *q) +{ + u32 pasid; + struct iommu_sva *handle; + + if (!(uacce->flags & UACCE_DEV_SVA)) + return 0; + + handle = iommu_sva_bind_device(uacce->parent, current->mm, NULL); + if (IS_ERR(handle)) + return PTR_ERR(handle); + + pasid = iommu_sva_get_pasid(handle); + if (pasid == IOMMU_PASID_INVALID) { + iommu_sva_unbind_device(handle); + return -ENODEV; + } + + q->handle = handle; + q->pasid = pasid; + return 0; +} + +static void uacce_unbind_queue(struct uacce_queue *q) +{ + if (!q->handle) + return; + iommu_sva_unbind_device(q->handle); + q->handle = NULL; +} + +static int uacce_fops_open(struct inode *inode, struct file *filep) +{ + struct uacce_device *uacce; + struct uacce_queue *q; + int ret; + + uacce = xa_load(&uacce_xa, iminor(inode)); + if (!uacce) + return -ENODEV; + + q = kzalloc(sizeof(struct uacce_queue), GFP_KERNEL); + if (!q) + return -ENOMEM; + + mutex_lock(&uacce->mutex); + + if (!uacce->parent) { + ret = -EINVAL; + goto out_with_mem; + } + + ret = uacce_bind_queue(uacce, q); + if (ret) + goto out_with_mem; + + q->uacce = uacce; + + if (uacce->ops->get_queue) { + ret = uacce->ops->get_queue(uacce, q->pasid, q); + if (ret < 0) + goto out_with_bond; + } + + init_waitqueue_head(&q->wait); + filep->private_data = q; + uacce->inode = inode; + q->state = UACCE_Q_INIT; + mutex_init(&q->mutex); + list_add(&q->list, &uacce->queues); + mutex_unlock(&uacce->mutex); + + return 0; + +out_with_bond: + uacce_unbind_queue(q); +out_with_mem: + kfree(q); + mutex_unlock(&uacce->mutex); + return ret; +} + +static int uacce_fops_release(struct inode *inode, struct file *filep) +{ + struct uacce_queue *q = filep->private_data; + struct uacce_device *uacce = q->uacce; + + mutex_lock(&uacce->mutex); + uacce_put_queue(q); + uacce_unbind_queue(q); + list_del(&q->list); + mutex_unlock(&uacce->mutex); + kfree(q); + + return 0; +} + +static void uacce_vma_close(struct vm_area_struct *vma) +{ + struct uacce_queue *q = vma->vm_private_data; + struct uacce_qfile_region *qfr = NULL; + + if (vma->vm_pgoff < UACCE_MAX_REGION) + qfr = q->qfrs[vma->vm_pgoff]; + + kfree(qfr); +} + +static const struct vm_operations_struct uacce_vm_ops = { + .close = uacce_vma_close, +}; + +static int uacce_fops_mmap(struct file *filep, struct vm_area_struct *vma) +{ + struct uacce_queue *q = filep->private_data; + struct uacce_device *uacce = q->uacce; + struct uacce_qfile_region *qfr; + enum uacce_qfrt type = UACCE_MAX_REGION; + int ret = 0; + + if (vma->vm_pgoff < UACCE_MAX_REGION) + type = vma->vm_pgoff; + else + return -EINVAL; + + qfr = kzalloc(sizeof(*qfr), GFP_KERNEL); + if (!qfr) + return -ENOMEM; + + vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND | VM_WIPEONFORK; + vma->vm_ops = &uacce_vm_ops; + vma->vm_private_data = q; + qfr->type = type; + + mutex_lock(&q->mutex); + if (!uacce_queue_is_valid(q)) { + ret = -ENXIO; + goto out_with_lock; + } + + if (q->qfrs[type]) { + ret = -EEXIST; + goto out_with_lock; + } + + switch (type) { + case UACCE_QFRT_MMIO: + case UACCE_QFRT_DUS: + if (!uacce->ops->mmap) { + ret = -EINVAL; + goto out_with_lock; + } + + ret = uacce->ops->mmap(q, vma, qfr); + if (ret) + goto out_with_lock; + break; + + default: + ret = -EINVAL; + goto out_with_lock; + } + + q->qfrs[type] = qfr; + mutex_unlock(&q->mutex); + + return ret; + +out_with_lock: + mutex_unlock(&q->mutex); + kfree(qfr); + return ret; +} + +static __poll_t uacce_fops_poll(struct file *file, poll_table *wait) +{ + struct uacce_queue *q = file->private_data; + struct uacce_device *uacce = q->uacce; + __poll_t ret = 0; + + mutex_lock(&q->mutex); + if (!uacce_queue_is_valid(q)) + goto out_unlock; + + poll_wait(file, &q->wait, wait); + + if (uacce->ops->is_q_updated && uacce->ops->is_q_updated(q)) + ret = EPOLLIN | EPOLLRDNORM; + +out_unlock: + mutex_unlock(&q->mutex); + return ret; +} + +static const struct file_operations uacce_fops = { + .owner = THIS_MODULE, + .open = uacce_fops_open, + .release = uacce_fops_release, + .unlocked_ioctl = uacce_fops_unl_ioctl, +#ifdef CONFIG_COMPAT + .compat_ioctl = uacce_fops_compat_ioctl, +#endif + .mmap = uacce_fops_mmap, + .poll = uacce_fops_poll, +}; + +#define to_uacce_device(dev) container_of(dev, struct uacce_device, dev) + +static ssize_t api_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + return sprintf(buf, "%s\n", uacce->api_ver); +} + +static ssize_t flags_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + return sprintf(buf, "%u\n", uacce->flags); +} + +static ssize_t available_instances_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + if (!uacce->ops->get_available_instances) + return -ENODEV; + + return sprintf(buf, "%d\n", + uacce->ops->get_available_instances(uacce)); +} + +static ssize_t algorithms_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + return sprintf(buf, "%s\n", uacce->algs); +} + +static ssize_t region_mmio_size_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + return sprintf(buf, "%lu\n", + uacce->qf_pg_num[UACCE_QFRT_MMIO] << PAGE_SHIFT); +} + +static ssize_t region_dus_size_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + return sprintf(buf, "%lu\n", + uacce->qf_pg_num[UACCE_QFRT_DUS] << PAGE_SHIFT); +} + +static DEVICE_ATTR_RO(api); +static DEVICE_ATTR_RO(flags); +static DEVICE_ATTR_RO(available_instances); +static DEVICE_ATTR_RO(algorithms); +static DEVICE_ATTR_RO(region_mmio_size); +static DEVICE_ATTR_RO(region_dus_size); + +static struct attribute *uacce_dev_attrs[] = { + &dev_attr_api.attr, + &dev_attr_flags.attr, + &dev_attr_available_instances.attr, + &dev_attr_algorithms.attr, + &dev_attr_region_mmio_size.attr, + &dev_attr_region_dus_size.attr, + NULL, +}; + +static umode_t uacce_dev_is_visible(struct kobject *kobj, + struct attribute *attr, int n) +{ + struct device *dev = kobj_to_dev(kobj); + struct uacce_device *uacce = to_uacce_device(dev); + + if (((attr == &dev_attr_region_mmio_size.attr) && + (!uacce->qf_pg_num[UACCE_QFRT_MMIO])) || + ((attr == &dev_attr_region_dus_size.attr) && + (!uacce->qf_pg_num[UACCE_QFRT_DUS]))) + return 0; + + return attr->mode; +} + +static struct attribute_group uacce_dev_group = { + .is_visible = uacce_dev_is_visible, + .attrs = uacce_dev_attrs, +}; + +__ATTRIBUTE_GROUPS(uacce_dev); + +static void uacce_release(struct device *dev) +{ + struct uacce_device *uacce = to_uacce_device(dev); + + kfree(uacce); +} + +static unsigned int uacce_enable_sva(struct device *parent, unsigned int flags) +{ + int ret; + + if (!(flags & UACCE_DEV_SVA)) + return flags; + + flags &= ~UACCE_DEV_SVA; + + ret = iommu_dev_enable_feature(parent, IOMMU_DEV_FEAT_IOPF); + if (ret) { + dev_err(parent, "failed to enable IOPF feature! ret = %pe\n", ERR_PTR(ret)); + return flags; + } + + ret = iommu_dev_enable_feature(parent, IOMMU_DEV_FEAT_SVA); + if (ret) { + dev_err(parent, "failed to enable SVA feature! ret = %pe\n", ERR_PTR(ret)); + iommu_dev_disable_feature(parent, IOMMU_DEV_FEAT_IOPF); + return flags; + } + + return flags | UACCE_DEV_SVA; +} + +static void uacce_disable_sva(struct uacce_device *uacce) +{ + if (!(uacce->flags & UACCE_DEV_SVA)) + return; + + iommu_dev_disable_feature(uacce->parent, IOMMU_DEV_FEAT_SVA); + iommu_dev_disable_feature(uacce->parent, IOMMU_DEV_FEAT_IOPF); +} + +/** + * uacce_alloc() - alloc an accelerator + * @parent: pointer of uacce parent device + * @interface: pointer of uacce_interface for register + * + * Returns uacce pointer if success and ERR_PTR if not + * Need check returned negotiated uacce->flags + */ +struct uacce_device *uacce_alloc(struct device *parent, + struct uacce_interface *interface) +{ + unsigned int flags = interface->flags; + struct uacce_device *uacce; + int ret; + + uacce = kzalloc(sizeof(struct uacce_device), GFP_KERNEL); + if (!uacce) + return ERR_PTR(-ENOMEM); + + flags = uacce_enable_sva(parent, flags); + + uacce->parent = parent; + uacce->flags = flags; + uacce->ops = interface->ops; + + ret = xa_alloc(&uacce_xa, &uacce->dev_id, uacce, xa_limit_32b, + GFP_KERNEL); + if (ret < 0) + goto err_with_uacce; + + INIT_LIST_HEAD(&uacce->queues); + mutex_init(&uacce->mutex); + device_initialize(&uacce->dev); + uacce->dev.devt = MKDEV(MAJOR(uacce_devt), uacce->dev_id); + uacce->dev.class = uacce_class; + uacce->dev.groups = uacce_dev_groups; + uacce->dev.parent = uacce->parent; + uacce->dev.release = uacce_release; + dev_set_name(&uacce->dev, "%s-%d", interface->name, uacce->dev_id); + + return uacce; + +err_with_uacce: + uacce_disable_sva(uacce); + kfree(uacce); + return ERR_PTR(ret); +} +EXPORT_SYMBOL_GPL(uacce_alloc); + +/** + * uacce_register() - add the accelerator to cdev and export to user space + * @uacce: The initialized uacce device + * + * Return 0 if register succeeded, or an error. + */ +int uacce_register(struct uacce_device *uacce) +{ + if (!uacce) + return -ENODEV; + + uacce->cdev = cdev_alloc(); + if (!uacce->cdev) + return -ENOMEM; + + uacce->cdev->ops = &uacce_fops; + uacce->cdev->owner = THIS_MODULE; + + return cdev_device_add(uacce->cdev, &uacce->dev); +} +EXPORT_SYMBOL_GPL(uacce_register); + +/** + * uacce_remove() - remove the accelerator + * @uacce: the accelerator to remove + */ +void uacce_remove(struct uacce_device *uacce) +{ + struct uacce_queue *q, *next_q; + + if (!uacce) + return; + /* + * unmap remaining mapping from user space, preventing user still + * access the mmaped area while parent device is already removed + */ + if (uacce->inode) + unmap_mapping_range(uacce->inode->i_mapping, 0, 0, 1); + + /* + * uacce_fops_open() may be running concurrently, even after we remove + * the cdev. Holding uacce->mutex ensures that open() does not obtain a + * removed uacce device. + */ + mutex_lock(&uacce->mutex); + /* ensure no open queue remains */ + list_for_each_entry_safe(q, next_q, &uacce->queues, list) { + /* + * Taking q->mutex ensures that fops do not use the defunct + * uacce->ops after the queue is disabled. + */ + mutex_lock(&q->mutex); + uacce_put_queue(q); + mutex_unlock(&q->mutex); + uacce_unbind_queue(q); + } + + /* disable sva now since no opened queues */ + uacce_disable_sva(uacce); + + if (uacce->cdev) + cdev_device_del(uacce->cdev, &uacce->dev); + xa_erase(&uacce_xa, uacce->dev_id); + /* + * uacce exists as long as there are open fds, but ops will be freed + * now. Ensure that bugs cause NULL deref rather than use-after-free. + */ + uacce->ops = NULL; + uacce->parent = NULL; + mutex_unlock(&uacce->mutex); + put_device(&uacce->dev); +} +EXPORT_SYMBOL_GPL(uacce_remove); + +static int __init uacce_init(void) +{ + int ret; + + uacce_class = class_create(THIS_MODULE, UACCE_NAME); + if (IS_ERR(uacce_class)) + return PTR_ERR(uacce_class); + + ret = alloc_chrdev_region(&uacce_devt, 0, MINORMASK, UACCE_NAME); + if (ret) + class_destroy(uacce_class); + + return ret; +} + +static __exit void uacce_exit(void) +{ + unregister_chrdev_region(uacce_devt, MINORMASK); + class_destroy(uacce_class); +} + +subsys_initcall(uacce_init); +module_exit(uacce_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("HiSilicon Tech. Co., Ltd."); +MODULE_DESCRIPTION("Accelerator interface for Userland applications"); -- Gitee