[sanitizers] read/write page fault detection on mac.

[aizatsky]

nit

[eugenis]

LGTM

[vsk]

Hi Mike, some of our internal bots are having an issue with this commit:

error: no member named '__err' in '__darwin_arm_exception_state'
  return ucontext->uc_mcontext->__es.__err & 2 /*T_PF_WRITE*/ ? WRITE : READ;

I'd really appreciate it if you could shed some light on this. I see two definitions of this structure in the OS, and neither contain '__err'. Can you share which OS you built this code against?

[aizatsky]

Vedant,

I believe it is 10.11 on Intel. This could be CPU difference. I'll try to
add appropriate ifdefs tomorrow.

[mehdi_amini]

I'll revert in the meantime. Feel free to reapply with the fix.

For reference from what I see here: http://opensource.apple.com//source/cctools/cctools-836/include/mach/arm/_structs.h this was hitting:

#define _STRUCT_ARM_EXCEPTION_STATE	struct __darwin_arm_exception_state
_STRUCT_ARM_EXCEPTION_STATE
{
	__uint32_t	__exception; /* number of arm exception taken */
	__uint32_t	__fsr; /* Fault status */
	__uint32_t	__far; /* Virtual Fault Address */
};

[aizatsky]

Is there a kernel source as well for mac on aarch64? I'm trying to determine if illegal write or read has happened. Would be bad to leave aarch out of the implementation.

[vsk]

Do the xnu sources here help? The latest publicly-available one appears to be xnu-3248.20.55 (http://opensource.apple.com/source/xnu/).

[aizatsky]

New review with feature enabled only on intel: http://reviews.llvm.org/D19561

I can't find aarch64 support in published xnu sources.

compiler-rt/trunk/lib/sanitizer_common/sanitizer_mac.cc

#endif
   BlockingMutexLock l(&syslog_lock);
   if (common_flags()->log_to_syslog)
     WriteToSyslog(buffer);
 
   // The report is added to CrashLog as part of logging all of Printf output.
 }
 
 SignalContext::WriteFlag SignalContext::GetWriteFlag(void *context) {
-  return UNKNOWN;  // FIXME: implement this.
+  ucontext_t *ucontext = static_cast<ucontext_t*>(context);
+  return ucontext->uc_mcontext->__es.__err & 2 /*T_PF_WRITE*/ ? WRITE : READ;
 }
 
 void GetPcSpBp(void *context, uptr *pc, uptr *sp, uptr *bp) {
   ucontext_t *ucontext = (ucontext_t*)context;
 # if defined(__aarch64__)
   *pc = ucontext->uc_mcontext->__ss.__pc;
 #   if defined(__IPHONE_8_0) && __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_8_0
   *bp = ucontext->uc_mcontext->__ss.__fp;

compiler-rt/trunk/test/asan/TestCases/Linux/segv_read_write.c

-// RUN: %clangxx_asan -std=c++11 -O0 %s -o %t
-// RUN: not %run %t       2>&1 | FileCheck %s --check-prefix=READ
-// RUN: not %run %t write 2>&1 | FileCheck %s --check-prefix=WRITE
-// UNSUPPORTED: powerpc64,mips
-
-#include <sys/mman.h>
-
-static volatile int sink;
-__attribute__((noinline)) void Read(int *ptr) { sink = *ptr; }
-__attribute__((noinline)) void Write(int *ptr) { *ptr = 0; }
-int main(int argc, char **argv) {
-  // Writes to shadow are detected as reads from shadow gap (because of how the
-  // shadow mapping works). This is kinda hard to fix. Test a random address in
-  // the application part of the address space.
-  void *volatile p =
-      mmap(nullptr, 4096, PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
-  munmap(p, 4096);
-  if (argc == 1)
-    Read((int *)p);
-  else
-    Write((int *)p);
-}
-// READ: AddressSanitizer: SEGV on unknown address
-// READ: The signal is caused by a READ memory access.
-// WRITE: AddressSanitizer: SEGV on unknown address
-// WRITE: The signal is caused by a WRITE memory access.

compiler-rt/trunk/test/asan/TestCases/Posix/segv_read_write.c

+// RUN: %clangxx_asan -std=c++11 -O0 %s -o %t
+// RUN: not %run %t       2>&1 | FileCheck %s --check-prefix=READ
+// RUN: not %run %t write 2>&1 | FileCheck %s --check-prefix=WRITE
+// UNSUPPORTED: powerpc64,mips
+
+#include <sys/mman.h>
+
+static volatile int sink;
+__attribute__((noinline)) void Read(int *ptr) { sink = *ptr; }
+__attribute__((noinline)) void Write(int *ptr) { *ptr = 0; }
+int main(int argc, char **argv) {
+  // Writes to shadow are detected as reads from shadow gap (because of how the
+  // shadow mapping works). This is kinda hard to fix. Test a random address in
+  // the application part of the address space.
+  void *volatile p =
+      mmap(nullptr, 4096, PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+  munmap(p, 4096);
+  if (argc == 1)
+    Read((int *)p);
+  else
+    Write((int *)p);
+}
+// READ: AddressSanitizer: SEGV on unknown address
+// READ: The signal is caused by a READ memory access.
+// WRITE: AddressSanitizer: SEGV on unknown address
+// WRITE: The signal is caused by a WRITE memory access.