[msan] Add stat-family interceptors on Linux

[foobarrior]

Summary

Add following interceptors on Linux: stat, lstat, fstat, fstatat.

This fixes use-of-uninitialized value on platforms with GLIBC 2.33+.
In particular: Arch Linux, Ubuntu hirsute/impish.

The tests should have also been failing during the release on the mentioned platforms, but I cannot find any related discussion.

Most likely, the regression was introduced by glibc commit 8ed005daf0ab03e14250032:
all stat-family functions are now exported as shared functions.

Before, some of them (namely stat, lstat, fstat, fstatat) were provided as a part of libc_noshared.a and called their __xstat dopplegangers. This is still true for Debian Sid and earlier Ubuntu's. stat interceptors may be safely provided for them, no problem with that.

Closes https://github.com/google/sanitizers/issues/1452.
See also https://jira.mariadb.org/browse/MDEV-24841

[eugenis]

compiler-rt/lib/msan/msan_interceptors.cpp

711

Shouldn't this be "!SANITIZER_FREEBSD && !SANITIZER_NETBSD" ?

[foobarrior]

Fix debian build.

__fxstatat interceptor now resides together with fstatat (just like fstat/__fxstat)

INTERCEPT_FUNCTION -> MSAN_INTERCEPT_FUNC to report interception errors

[foobarrior]

MSAN_INTERCEPT_FUNC(__fxstatat64)

[foobarrior]

compiler-rt/lib/msan/msan_interceptors.cpp

711

yes, stupid mistake, sorry

[eugenis]

Please apply the clang-format lint changes in the comments.

compiler-rt/lib/msan/msan_interceptors.cpp

708

Please add _MAYBE_ to the macro name for consistency.

compiler-rt/test/msan/fstat.cpp

1 ↗

(On Diff #380690)

why is -fPIC necessary?

[foobarrior]

Add _MAYBE_ where necessary.
Remove accidentally added -fPIC

[foobarrior]

Should I fix lint checks from the review? They seem to be incorrect from my point of view

compiler-rt/test/msan/fstat.cpp

1 ↗

(On Diff #380690)

That snuck in accidentally, while i was testing. Removing

[foobarrior]

Reformat the code after eugenis comment is noticed

[foobarrior]

Reformat the code a bit less (no newlines after if's)

[eugenis]

In D111984#3074929, @foobarrior wrote:

Should I fix lint checks from the review? They seem to be incorrect from my point of view

Yes please. Otherwise they will simply pop up in someone else's patch. We do what clang-format tell us to do :)

LGTM. Do you need me to land this change for you? I can reformat before commit, too.

[foobarrior]

Yes, please. I have no commit rights.

[eugenis]

+llvm-commits

[vitalybuka]

compiler-rt/lib/msan/msan_interceptors.cpp

661

This condition is inconvenient
I have false msan reports on 2.33 because it's compiled on older glibc

[foobarrior]

compiler-rt/lib/msan/msan_interceptors.cpp

661

Hi Vitaly!
What's your setup?
And can you provide an exmple of false positive?

[vitalybuka]

compiler-rt/lib/msan/msan_interceptors.cpp

661

https://lab.llvm.org/staging/#/builders/72/builds/1685/steps/9/logs/stdio

[MaskRay]

compiler-rt/lib/msan/msan_interceptors.cpp

661

If __GLIBC_PREREQ is not defined (e.g. Linux musl), there will be a preprocessor error like:

• Clang: error: function-like macro '__GLIBC_PREREQ' is not defined
• GCC: error: missing binary operator before token "("

Looks like @thesamesam is trying to fix this for Linux musl.

compiler-rt/lib/msan/msan_interceptors.cpp

 
 INTERCEPTOR(int, putenv, char *string) {
   ENSURE_MSAN_INITED();
   int res = REAL(putenv)(string);
   if (!res) UnpoisonEnviron();
   return res;
 }
 
-#if SANITIZER_FREEBSD || SANITIZER_NETBSD
+#define SANITIZER_STAT_LINUX (SANITIZER_LINUX && __GLIBC_PREREQ(2, 33))
+#if SANITIZER_FREEBSD || SANITIZER_NETBSD || SANITIZER_STAT_LINUX
 INTERCEPTOR(int, fstat, int fd, void *buf) {
   ENSURE_MSAN_INITED();
   int res = REAL(fstat)(fd, buf);
   if (!res)
     __msan_unpoison(buf, __sanitizer::struct_stat_sz);
   return res;
 }
-#define MSAN_MAYBE_INTERCEPT_FSTAT INTERCEPT_FUNCTION(fstat)
+#  define MSAN_MAYBE_INTERCEPT_FSTAT MSAN_INTERCEPT_FUNC(fstat)
 #else
 #define MSAN_MAYBE_INTERCEPT_FSTAT
 #endif
 
 #if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
 INTERCEPTOR(int, __fxstat, int magic, int fd, void *buf) {
   ENSURE_MSAN_INITED();
   int res = REAL(__fxstat)(magic, fd, buf);
   if (!res)
     __msan_unpoison(buf, __sanitizer::struct_stat_sz);
   return res;
 }
-#define MSAN_MAYBE_INTERCEPT___FXSTAT INTERCEPT_FUNCTION(__fxstat)
+#  define MSAN_MAYBE_INTERCEPT___FXSTAT MSAN_INTERCEPT_FUNC(__fxstat)
 #else
 #define MSAN_MAYBE_INTERCEPT___FXSTAT
 #endif
 
 #if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
 INTERCEPTOR(int, __fxstat64, int magic, int fd, void *buf) {
   ENSURE_MSAN_INITED();
   int res = REAL(__fxstat64)(magic, fd, buf);
   if (!res)
     __msan_unpoison(buf, __sanitizer::struct_stat64_sz);
   return res;
 }
-#define MSAN_MAYBE_INTERCEPT___FXSTAT64 INTERCEPT_FUNCTION(__fxstat64)
+#  define MSAN_MAYBE_INTERCEPT___FXSTAT64 MSAN_INTERCEPT_FUNC(__fxstat64)
 #else
-#define MSAN_MAYBE_INTERCEPT___FXSTAT64
+#  define MSAN_MAYBE_INTERCEPT___FXSTAT64
 #endif
 
-#if SANITIZER_FREEBSD || SANITIZER_NETBSD
+#if SANITIZER_FREEBSD || SANITIZER_NETBSD || SANITIZER_STAT_LINUX
 INTERCEPTOR(int, fstatat, int fd, char *pathname, void *buf, int flags) {
   ENSURE_MSAN_INITED();
   int res = REAL(fstatat)(fd, pathname, buf, flags);
   if (!res) __msan_unpoison(buf, __sanitizer::struct_stat_sz);
   return res;
 }
-# define MSAN_INTERCEPT_FSTATAT INTERCEPT_FUNCTION(fstatat)
+#  define MSAN_MAYBE_INTERCEPT_FSTATAT MSAN_INTERCEPT_FUNC(fstatat)
 #else
+#  define MSAN_MAYBE_INTERCEPT_FSTATAT
+#endif
+
+#if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
 INTERCEPTOR(int, __fxstatat, int magic, int fd, char *pathname, void *buf,
             int flags) {
   ENSURE_MSAN_INITED();
   int res = REAL(__fxstatat)(magic, fd, pathname, buf, flags);
   if (!res) __msan_unpoison(buf, __sanitizer::struct_stat_sz);
   return res;
 }
-# define MSAN_INTERCEPT_FSTATAT INTERCEPT_FUNCTION(__fxstatat)
+#  define MSAN_MAYBE_INTERCEPT___FXSTATAT MSAN_INTERCEPT_FUNC(__fxstatat)
+#else
+#  define MSAN_MAYBE_INTERCEPT___FXSTATAT
 #endif
 
 #if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
 INTERCEPTOR(int, __fxstatat64, int magic, int fd, char *pathname, void *buf,
             int flags) {
   ENSURE_MSAN_INITED();
   int res = REAL(__fxstatat64)(magic, fd, pathname, buf, flags);
   if (!res) __msan_unpoison(buf, __sanitizer::struct_stat64_sz);
   return res;
 }
-#define MSAN_MAYBE_INTERCEPT___FXSTATAT64 INTERCEPT_FUNCTION(__fxstatat64)
+#  define MSAN_MAYBE_INTERCEPT___FXSTATAT64 MSAN_INTERCEPT_FUNC(__fxstatat64)
 #else
-#define MSAN_MAYBE_INTERCEPT___FXSTATAT64
+#  define MSAN_MAYBE_INTERCEPT___FXSTATAT64
 #endif
 
 INTERCEPTOR(int, pipe, int pipefd[2]) {
   if (msan_init_is_running)
     return REAL(pipe)(pipefd);
   ENSURE_MSAN_INITED();
   int res = REAL(pipe)(pipefd);
   if (!res)
#endif
   INTERCEPT_FUNCTION(wcscmp);
   INTERCEPT_FUNCTION(getenv);
   INTERCEPT_FUNCTION(setenv);
   INTERCEPT_FUNCTION(putenv);
   INTERCEPT_FUNCTION(gettimeofday);
   MSAN_MAYBE_INTERCEPT_FCVT;
   MSAN_MAYBE_INTERCEPT_FSTAT;
   MSAN_MAYBE_INTERCEPT___FXSTAT;
-  MSAN_INTERCEPT_FSTATAT;
+  MSAN_MAYBE_INTERCEPT_FSTATAT;
+  MSAN_MAYBE_INTERCEPT___FXSTATAT;
   MSAN_MAYBE_INTERCEPT___FXSTAT64;
   MSAN_MAYBE_INTERCEPT___FXSTATAT64;
   INTERCEPT_FUNCTION(pipe);
   INTERCEPT_FUNCTION(pipe2);
   INTERCEPT_FUNCTION(socketpair);
   MSAN_MAYBE_INTERCEPT_FGETS_UNLOCKED;
   INTERCEPT_FUNCTION(getrlimit);
   MSAN_MAYBE_INTERCEPT___GETRLIMIT;

compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h

 #define SANITIZER_INTERCEPT_CTERMID_R (SI_MAC || SI_FREEBSD || SI_SOLARIS)
 
 #define SANITIZER_INTERCEPTOR_HOOKS \
   (SI_LINUX || SI_MAC || SI_WINDOWS || SI_FREEBSD || SI_NETBSD || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_RECV_RECVFROM SI_POSIX
 #define SANITIZER_INTERCEPT_SEND_SENDTO SI_POSIX
 #define SANITIZER_INTERCEPT_EVENTFD_READ_WRITE SI_LINUX
 
-#define SANITIZER_INTERCEPT_STAT \
-  (SI_FREEBSD || SI_MAC || SI_ANDROID || SI_NETBSD || SI_SOLARIS)
+#define SI_STAT_LINUX (SI_LINUX && __GLIBC_PREREQ(2, 33))
+#define SANITIZER_INTERCEPT_STAT                                        \
+  (SI_FREEBSD || SI_MAC || SI_ANDROID || SI_NETBSD || SI_SOLARIS ||     \
+   SI_STAT_LINUX)
 #define SANITIZER_INTERCEPT_LSTAT (SI_NETBSD || SI_FREEBSD)
-#define SANITIZER_INTERCEPT___XSTAT (!SANITIZER_INTERCEPT_STAT && SI_POSIX)
+#define SANITIZER_INTERCEPT___XSTAT                             \
+  (!SANITIZER_INTERCEPT_STAT && SI_POSIX) || SI_STAT_LINUX
 #define SANITIZER_INTERCEPT___XSTAT64 SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT___LXSTAT SANITIZER_INTERCEPT___XSTAT
 #define SANITIZER_INTERCEPT___LXSTAT64 SI_LINUX_NOT_ANDROID
 
 #define SANITIZER_INTERCEPT_UTMP \
   (SI_POSIX && !SI_MAC && !SI_FREEBSD && !SI_NETBSD)
 #define SANITIZER_INTERCEPT_UTMPX \
   (SI_LINUX_NOT_ANDROID || SI_MAC || SI_FREEBSD || SI_NETBSD)