[msan] Add 'MappingDesc::ALLOCATOR' type and check it is available (#85153)

MSan divides the virtual address space into APP, INVALID, SHADOW and
ORIGIN memory. The allocator usually just steals a bit of the APP
address space: typically the bottom portion of the PIE binaries section,
which works because the Linux kernel maps from the top of the PIE
binaries section. However, if ASLR is very aggressive, the binary may
end up mapped in the same location where the allocator wants to live;
this results in a segfault.

This patch adds in a MappingDesc::ALLOCATOR type and enforces that the
memory range for the allocator is not occupied by anything else.

Since the allocator range information is not readily available in
msan.h, we duplicate the information from msan_allocator.cpp.

Note: aggressive ASLR can also lead to a different type of failure,
where the PIE binaries/libraries are mapped entirely outside of the
APP/ALLOCATOR sections; that will be addressed in a separate patch
(https://github.com/llvm/llvm-project/pull/85142).

(cherry picked from commit af2bf86a372cacf5f536bae06e2f2d3886eefb7b)

3 files changed, 34 insertions, 13 deletions

diff --git a/compiler-rt/lib/msan/msan.h b/compiler-rt/lib/msan/msan.h
index 710447a..00f448f 100644
--- a/compiler-rt/lib/msan/msan.h
+++ b/compiler-rt/lib/msan/msan.h
@@ -33,12 +33,18 @@ struct MappingDesc {
   uptr start;
   uptr end;
   enum Type {
-    INVALID, APP, SHADOW, ORIGIN
+    INVALID = 1,
+    ALLOCATOR = 2,
+    APP = 4,
+    SHADOW = 8,
+    ORIGIN = 16,
   } type;
   const char *name;
 };
 
-
+// Note: MappingDesc::ALLOCATOR entries are only used to check for memory
+// layout compatibility. The actual allocation settings are in
+// msan_allocator.cpp, which need to be kept in sync.
 #if SANITIZER_LINUX && defined(__mips64)
 
 // MIPS64 maps:
@@ -84,7 +90,8 @@ const MappingDesc kMemoryLayout[] = {
     {0X0B00000000000, 0X0C00000000000, MappingDesc::SHADOW, "shadow-10-13"},
     {0X0C00000000000, 0X0D00000000000, MappingDesc::INVALID, "invalid"},
     {0X0D00000000000, 0X0E00000000000, MappingDesc::ORIGIN, "origin-10-13"},
-    {0X0E00000000000, 0X1000000000000, MappingDesc::APP, "app-15"},
+    {0x0E00000000000, 0x0E40000000000, MappingDesc::ALLOCATOR, "allocator"},
+    {0X0E40000000000, 0X1000000000000, MappingDesc::APP, "app-15"},
 };
 # define MEM_TO_SHADOW(mem) ((uptr)mem ^ 0xB00000000000ULL)
 # define SHADOW_TO_ORIGIN(shadow) (((uptr)(shadow)) + 0x200000000000ULL)
@@ -106,7 +113,8 @@ const MappingDesc kMemoryLayout[] = {
     {0x510000000000ULL, 0x600000000000ULL, MappingDesc::APP, "app-2"},
     {0x600000000000ULL, 0x610000000000ULL, MappingDesc::ORIGIN, "origin-1"},
     {0x610000000000ULL, 0x700000000000ULL, MappingDesc::INVALID, "invalid"},
-    {0x700000000000ULL, 0x800000000000ULL, MappingDesc::APP, "app-3"}};
+    {0x700000000000ULL, 0x740000000000ULL, MappingDesc::ALLOCATOR, "allocator"},
+    {0x740000000000ULL, 0x800000000000ULL, MappingDesc::APP, "app-3"}};
 #  define MEM_TO_SHADOW(mem) (((uptr)(mem)) ^ 0x500000000000ULL)
 #  define SHADOW_TO_ORIGIN(shadow) (((uptr)(shadow)) + 0x100000000000ULL)
 
@@ -118,7 +126,8 @@ const MappingDesc kMemoryLayout[] = {
     {0x180200000000ULL, 0x1C0000000000ULL, MappingDesc::INVALID, "invalid"},
     {0x1C0000000000ULL, 0x2C0200000000ULL, MappingDesc::ORIGIN, "origin"},
     {0x2C0200000000ULL, 0x300000000000ULL, MappingDesc::INVALID, "invalid"},
-    {0x300000000000ULL, 0x800000000000ULL, MappingDesc::APP, "high memory"}};
+    {0x300000000000ULL, 0x320000000000ULL, MappingDesc::ALLOCATOR, "allocator"},
+    {0x320000000000ULL, 0x800000000000ULL, MappingDesc::APP, "high memory"}};
 
 // Various kernels use different low end ranges but we can combine them into one
 // big range. They also use different high end ranges but we can map them all to
@@ -141,7 +150,8 @@ const MappingDesc kMemoryLayout[] = {
     {0x180000000000ULL, 0x1C0000000000ULL, MappingDesc::INVALID, "invalid"},
     {0x1C0000000000ULL, 0x2C0000000000ULL, MappingDesc::ORIGIN, "origin"},
     {0x2C0000000000ULL, 0x440000000000ULL, MappingDesc::INVALID, "invalid"},
-    {0x440000000000ULL, 0x500000000000ULL, MappingDesc::APP, "high memory"}};
+    {0x440000000000ULL, 0x460000000000ULL, MappingDesc::ALLOCATOR, "allocator"},
+    {0x460000000000ULL, 0x500000000000ULL, MappingDesc::APP, "high memory"}};
 
 #define MEM_TO_SHADOW(mem) \
   ((((uptr)(mem)) & ~0xC00000000000ULL) + 0x080000000000ULL)
@@ -208,7 +218,8 @@ const MappingDesc kMemoryLayout[] = {
     {0x510000000000ULL, 0x600000000000ULL, MappingDesc::APP, "app-2"},
     {0x600000000000ULL, 0x610000000000ULL, MappingDesc::ORIGIN, "origin-1"},
     {0x610000000000ULL, 0x700000000000ULL, MappingDesc::INVALID, "invalid"},
-    {0x700000000000ULL, 0x800000000000ULL, MappingDesc::APP, "app-3"}};
+    {0x700000000000ULL, 0x740000000000ULL, MappingDesc::ALLOCATOR, "allocator"},
+    {0x740000000000ULL, 0x800000000000ULL, MappingDesc::APP, "app-3"}};
 #define MEM_TO_SHADOW(mem) (((uptr)(mem)) ^ 0x500000000000ULL)
 #define SHADOW_TO_ORIGIN(mem) (((uptr)(mem)) + 0x100000000000ULL)
 
@@ -223,20 +234,22 @@ const uptr kMemoryLayoutSize = sizeof(kMemoryLayout) / sizeof(kMemoryLayout[0]);
 #ifndef __clang__
 __attribute__((optimize("unroll-loops")))
 #endif
-inline bool addr_is_type(uptr addr, MappingDesc::Type mapping_type) {
+inline bool
+addr_is_type(uptr addr, int mapping_types) {
 // It is critical for performance that this loop is unrolled (because then it is
 // simplified into just a few constant comparisons).
 #ifdef __clang__
 #pragma unroll
 #endif
   for (unsigned i = 0; i < kMemoryLayoutSize; ++i)
-    if (kMemoryLayout[i].type == mapping_type &&
+    if ((kMemoryLayout[i].type & mapping_types) &&
         addr >= kMemoryLayout[i].start && addr < kMemoryLayout[i].end)
       return true;
   return false;
 }
 
-#define MEM_IS_APP(mem) addr_is_type((uptr)(mem), MappingDesc::APP)
+#define MEM_IS_APP(mem) \
+  (addr_is_type((uptr)(mem), MappingDesc::APP | MappingDesc::ALLOCATOR))
 #define MEM_IS_SHADOW(mem) addr_is_type((uptr)(mem), MappingDesc::SHADOW)
 #define MEM_IS_ORIGIN(mem) addr_is_type((uptr)(mem), MappingDesc::ORIGIN)
 
diff --git a/compiler-rt/lib/msan/msan_allocator.cpp b/compiler-rt/lib/msan/msan_allocator.cpp
index 0b2dd2b..b1bc5b9 100644
--- a/compiler-rt/lib/msan/msan_allocator.cpp
+++ b/compiler-rt/lib/msan/msan_allocator.cpp
@@ -48,6 +48,9 @@ struct MsanMapUnmapCallback {
   }
 };
 
+// Note: to ensure that the allocator is compatible with the application memory
+// layout (especially with high-entropy ASLR), kSpaceBeg and kSpaceSize must be
+// duplicated as MappingDesc::ALLOCATOR in msan.h.
 #if defined(__mips64)
 static const uptr kMaxAllowedMallocSize = 2UL << 30;
 
diff --git a/compiler-rt/lib/msan/msan_linux.cpp b/compiler-rt/lib/msan/msan_linux.cpp
index c7ecb7c..b02994c 100644
--- a/compiler-rt/lib/msan/msan_linux.cpp
+++ b/compiler-rt/lib/msan/msan_linux.cpp
@@ -86,7 +86,7 @@ static void CheckMemoryLayoutSanity() {
     CHECK(addr_is_type(start, type));
     CHECK(addr_is_type((start + end) / 2, type));
     CHECK(addr_is_type(end - 1, type));
-    if (type == MappingDesc::APP) {
+    if (type == MappingDesc::APP || type == MappingDesc::ALLOCATOR) {
       uptr addr = start;
       CHECK(MEM_IS_SHADOW(MEM_TO_SHADOW(addr)));
       CHECK(MEM_IS_ORIGIN(MEM_TO_ORIGIN(addr)));
@@ -138,8 +138,13 @@ bool InitShadow(bool init_origins) {
     bool protect = type == MappingDesc::INVALID ||
                    (!init_origins && type == MappingDesc::ORIGIN);
     CHECK(!(map && protect));
-    if (!map && !protect)
-      CHECK(type == MappingDesc::APP);
+    if (!map && !protect) {
+      CHECK(type == MappingDesc::APP || type == MappingDesc::ALLOCATOR);
+
+      if (type == MappingDesc::ALLOCATOR &&
+          !CheckMemoryRangeAvailability(start, size))
+        return false;
+    }
     if (map) {
       if (!CheckMemoryRangeAvailability(start, size))
         return false;