From dcbfcb916ca1a269b3feef86ee86835294758f84 Mon Sep 17 00:00:00 2001
From: Oran Agra <oran@redislabs.com>
Date: Tue, 28 Feb 2023 15:15:26 +0200
Subject: [PATCH] String pattern matching had exponential time complexity on
 pathological patterns (CVE-2022-36021) (#11858)

Authenticated users can use string matching commands with a
specially crafted pattern to trigger a denial-of-service attack on Redis,
causing it to hang and consume 100% CPU time.

Co-authored-by: Tom Levy <tomlevy93@gmail.com>
---
 src/util.c              | 27 +++++++++++++++++++++++----
 tests/unit/keyspace.tcl |  6 ++++++
 2 files changed, 29 insertions(+), 4 deletions(-)

diff --git a/src/util.c b/src/util.c
index d33f4522a507..26d92b92290e 100644
--- a/src/util.c
+++ b/src/util.c
@@ -44,8 +44,8 @@
 #include "sha1.h"

 /* Glob-style pattern matching. */
-int stringmatchlen(const char *pattern, int patternLen,
-        const char *string, int stringLen, int nocase)
+static int stringmatchlen_impl(const char *pattern, int patternLen,
+        const char *string, int stringLen, int nocase, int *skipLongerMatches)
 {
     while(patternLen && stringLen) {
         switch(pattern[0]) {
@@ -57,12 +57,24 @@
             if (patternLen == 1)
                 return 1; /* match */
             while(stringLen) {
-                if (stringmatchlen(pattern+1, patternLen-1,
-                            string, stringLen, nocase))
+               if (stringmatchlen_impl(pattern+1, patternLen-1,
+                            string, stringLen, nocase, skipLongerMatches))
                     return 1; /* match */
+               if (*skipLongerMatches)
+                    return 0; /* no match */
                 string++;
                 stringLen--;
             }
+           /* There was no match for the rest of the pattern starting
+             * from anywhere in the rest of the string. If there were
+             * any '*' earlier in the pattern, we can terminate the
+             * search early without trying to match them to longer
+             * substrings. This is because a longer match for the
+             * earlier part of the pattern would require the rest of the
+             * pattern to match starting later in the string, and we
+             * have just determined that there is no match for the rest
+             * of the pattern starting from anywhere in the current
+             * string. */
             return 0; /* no match */
             break;
         case '?':
@@ -166,10 +178,17 @@
     return 0;
 }

+int stringmatchlen(const char *pattern, int patternLen,
+        const char *string, int stringLen, int nocase) {
+    int skipLongerMatches = 0;
+    return stringmatchlen_impl(pattern,patternLen,string,stringLen,nocase,&skipLongerMatches);
+}
+
 int stringmatch(const char *pattern, const char *string, int nocase) {
     return stringmatchlen(pattern,strlen(pattern),string,strlen(string),nocase);
 }

+
 /* Convert a string representing an amount of memory into the number of
  * bytes, so for instance memtoll("1Gb") will return 1073741824 that is
  * (1024*1024*1024).

diff --git a/tests/unit/keyspace.tcl b/tests/unit/keyspace.tcl
index b173e0efcacc..43690d06b321 100644
--- a/tests/unit/keyspace.tcl
+++ b/tests/unit/keyspace.tcl
@@ -493,4 +493,10 @@ foreach {type large} [array get largevalue] {
         r keys *
         r keys *
     } {dlskeriewrioeuwqoirueioqwrueoqwrueqw}
+
+    test {Regression for pattern matching long nested loops} {
+        r flushdb
+        r SET aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa 1
+        r KEYS "a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*a*b"
+    } {}
 }