[sanitizer_common] Add darwin-specific MemoryRangeIsAvailable (#167797)

The fixes a TOCTOU bug in the code that initializes shadow memory in
ASAN:


https://github.com/llvm/llvm-project/blob/4b05581bae0e3432cfa514788418fb2fc2144904/compiler-rt/lib/asan/asan_shadow_setup.cpp#L66-L91

1. During initialization, we call `FindDynamicShadowStart` to search the
memory mapping for enough space to dynamically allocate shadow memory.
2. We call `MemoryRangeIsAvailable(shadow_start, kHighShadowEnd);`,
which goes into `MemoryMappingLayout`.
3. We actually map the shadow with `ReserveShadowMemoryRange`.

In step 2, `MemoryMappingLayout` makes various allocations using the
internal allocator. This can cause the allocator to map more memory! In
some cases, this can actually allocate memory that overlaps with the
shadow region returned by` FindDynamicShadowStart` in step 1. This is
not actually fatal, but it memory corruption; MAP_FIXED is allowed to
overlap other regions, and the effect is any overlapping memory is
zeroed.

------

To address this, this PR implements `MemoryRangeIsAvailable` on Darwin
without any heap allocations:

- Move `IntervalsAreSeparate` into sanitizer_common.h
- Guard existing sanitizer_posix implementation of
`MemoryRangeIsAvailable` behind !SANITIZER_APPLE
- `IsAddressInMappedRegion` in sanitizer_mac becomes
`MemoryRangeIsAvailable`, which also checks for overlap with the DYLD
shared cache.

After this fix, it should be possible to re-land #166005, which
triggered this issue on the x86 iOS simulators.

rdar://164208439

Author: ndrewh

compiler-rt/lib/sanitizer_common/sanitizer_common.h

@@ -487,6 +487,13 @@ inline uptr Log2(uptr x) {
   return LeastSignificantSetBitIndex(x);
 }
 
+inline bool IntervalsAreSeparate(uptr start1, uptr end1, uptr start2,
+                                 uptr end2) {
+  CHECK_LE(start1, end1);
+  CHECK_LE(start2, end2);
+  return (end1 < start2) || (end2 < start1);
+}
+
 // Don't use std::min, std::max or std::swap, to minimize dependency
 // on libstdc++.
 template <class T>

compiler-rt/lib/sanitizer_common/sanitizer_mac.cpp

@@ -103,6 +103,8 @@ extern "C" {
     natural_t *nesting_depth,
     vm_region_recurse_info_t info,
     mach_msg_type_number_t *infoCnt);
+
+  extern const void* _dyld_get_shared_cache_range(size_t* length);
 }
 
 #  if !SANITIZER_GO
@@ -1403,15 +1405,27 @@ uptr FindAvailableMemoryRange(uptr size, uptr alignment, uptr left_padding,
   return 0;
 }
 
-// Returns true if the address is definitely mapped, and false if it is not
-// mapped or could not be determined.
-bool IsAddressInMappedRegion(uptr addr) {
+// This function (when used during initialization when there is
+// only a single thread), can be used to verify that a range
+// of memory hasn't already been mapped, and won't be mapped
+// later in the shared cache.
+//
+// If the syscall mach_vm_region_recurse fails (due to sandbox),
+// we assume that the memory is not mapped so that execution can continue.
+//
+// NOTE: range_end is inclusive
+//
+// WARNING: This function must NOT allocate memory, since it is
+// used in InitializeShadowMemory between where we search for
+// space for shadow and where we actually allocate it.
+bool MemoryRangeIsAvailable(uptr range_start, uptr range_end) {
   mach_vm_size_t vmsize = 0;
   natural_t depth = 0;
   vm_region_submap_short_info_data_64_t vminfo;
   mach_msg_type_number_t count = VM_REGION_SUBMAP_SHORT_INFO_COUNT_64;
-  mach_vm_address_t address = addr;
+  mach_vm_address_t address = range_start;
 
+  // First, check if the range is already mapped.
   kern_return_t kr =
       mach_vm_region_recurse(mach_task_self(), &address, &vmsize, &depth,
                              (vm_region_info_t)&vminfo, &count);
@@ -1423,7 +1437,24 @@ bool IsAddressInMappedRegion(uptr addr) {
     Report("HINT: Is mach_vm_region_recurse allowed by sandbox?\n");
   }
 
-  return (kr == KERN_SUCCESS && addr >= address && addr < address + vmsize);
+  if (kr == KERN_SUCCESS && !IntervalsAreSeparate(address, address + vmsize - 1,
+                                                  range_start, range_end)) {
+    // Overlaps with already-mapped memory
+    return false;
+  }
+
+  size_t cacheLength;
+  uptr cacheStart = (uptr)_dyld_get_shared_cache_range(&cacheLength);
+
+  if (cacheStart &&
+      !IntervalsAreSeparate(cacheStart, cacheStart + cacheLength - 1,
+                            range_start, range_end)) {
+    // Overlaps with shared cache region
+    return false;
+  }
+
+  // We believe this address is available.
+  return true;
 }
 
 // FIXME implement on this platform.

compiler-rt/lib/sanitizer_common/sanitizer_mac.h

@@ -76,8 +76,6 @@ struct ThreadEventCallbacks {
 
 void InstallPthreadIntrospectionHook(const ThreadEventCallbacks &callbacks);
 
-bool IsAddressInMappedRegion(uptr addr);
-
 }  // namespace __sanitizer
 
 #endif  // SANITIZER_APPLE

compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp

@@ -225,17 +225,9 @@ void *MapWritableFileToMemory(void *addr, uptr size, fd_t fd, OFF_T offset) {
   return (void *)p;
 }
 
-static inline bool IntervalsAreSeparate(uptr start1, uptr end1,
-                                        uptr start2, uptr end2) {
-  CHECK(start1 <= end1);
-  CHECK(start2 <= end2);
-  return (end1 < start2) || (end2 < start1);
-}
-
+#  if !SANITIZER_APPLE
 // FIXME: this is thread-unsafe, but should not cause problems most of the time.
-// When the shadow is mapped only a single thread usually exists (plus maybe
-// several worker threads on Mac, which aren't expected to map big chunks of
-// memory).
+// When the shadow is mapped only a single thread usually exists
 bool MemoryRangeIsAvailable(uptr range_start, uptr range_end) {
   MemoryMappingLayout proc_maps(/*cache_enabled*/true);
   if (proc_maps.Error())
@@ -251,7 +243,6 @@ bool MemoryRangeIsAvailable(uptr range_start, uptr range_end) {
   return true;
 }
 
-#if !SANITIZER_APPLE
 void DumpProcessMap() {
   MemoryMappingLayout proc_maps(/*cache_enabled*/true);
   const sptr kBufSize = 4095;
@@ -265,7 +256,7 @@ void DumpProcessMap() {
   Report("End of process memory map.\n");
   UnmapOrDie(filename, kBufSize);
 }
-#endif
+#  endif
 
 const char *GetPwd() {
   return GetEnv("PWD");

compiler-rt/lib/tsan/rtl/tsan_platform_mac.cpp

@@ -235,7 +235,7 @@ void InitializePlatformEarly() {
   }
   // In some configurations, the max_vm is expanded, but much of this space is
   // already mapped. TSAN will not work in this configuration.
-  if (IsAddressInMappedRegion(HiAppMemEnd() - 1)) {
+  if (!MemoryRangeIsAvailable(HiAppMemEnd() - 1, HiAppMemEnd())) {
     Report(
         "ThreadSanitizer: Unsupported virtual memory layout: Address %p is "
         "already mapped.\n",