[hwasan] Runtime flags for malloc bisection.

[Eugene.Zelenko]

compiler-rt/lib/sanitizer_common/sanitizer_hash.h

12

Please insert empty line above.

[eugenis]

added an empty line

[eugenis]

revert unrelated changes

[pcc]

LGTM

compiler-rt/lib/hwasan/CMakeLists.txt

28

nit: alphabetical order

compiler-rt/test/hwasan/TestCases/malloc_bisect.c

5

Should this be 2^32 - 1 to prevent the test from being sensitive to the test environment?

[eugenis]

address comments

compiler-rt/lib/hwasan/CMakeLists.txt

 set(HWASAN_RTL_HEADERS
   hwasan.h
   hwasan_allocator.h
   hwasan_dynamic_shadow.h
   hwasan_flags.h
   hwasan_flags.inc
   hwasan_interface_internal.h
   hwasan_mapping.h
+  hwasan_malloc_bisect.h
   hwasan_poisoning.h
   hwasan_report.h
   hwasan_thread.h
   hwasan_thread_list.h
   )
 
 set(HWASAN_DEFINITIONS)
 append_list_if(COMPILER_RT_HWASAN_WITH_INTERCEPTORS HWASAN_WITH_INTERCEPTORS=1 HWASAN_DEFINITIONS)

compiler-rt/lib/hwasan/hwasan_allocator.cc

 //===----------------------------------------------------------------------===//
 
 #include "sanitizer_common/sanitizer_atomic.h"
 #include "sanitizer_common/sanitizer_errno.h"
 #include "sanitizer_common/sanitizer_stackdepot.h"
 #include "hwasan.h"
 #include "hwasan_allocator.h"
 #include "hwasan_mapping.h"
+#include "hwasan_malloc_bisect.h"
 #include "hwasan_thread.h"
 #include "hwasan_report.h"
 
 #if HWASAN_WITH_INTERCEPTORS
 DEFINE_REAL(void *, realloc, void *ptr, uptr size)
 DEFINE_REAL(void, free, void *ptr)
 #endif
 
if (zeroise) {
     uptr fill_size = Min(size, (uptr)flags()->max_malloc_fill_size);
     internal_memset(allocated, flags()->malloc_fill_byte, fill_size);
   }
   if (!right_align_mode)
     internal_memcpy(reinterpret_cast<u8 *>(allocated) + orig_size, tail_magic,
                     size - orig_size);
 
   void *user_ptr = allocated;
-  if (flags()->tag_in_malloc &&
-      atomic_load_relaxed(&hwasan_allocator_tagging_enabled))
-    user_ptr = (void *)TagMemoryAligned(
-        (uptr)user_ptr, size, t ? t->GenerateRandomTag() : kFallbackAllocTag);
+  // Tagging can only be skipped when both tag_in_malloc and tag_in_free are
+  // false. When tag_in_malloc = false and tag_in_free = true malloc needs to
+  // retag to 0.
+  if ((flags()->tag_in_malloc || flags()->tag_in_free) &&
+      atomic_load_relaxed(&hwasan_allocator_tagging_enabled)) {
+    tag_t tag = flags()->tag_in_malloc && malloc_bisect(stack, orig_size)
+                    ? (t ? t->GenerateRandomTag() : kFallbackAllocTag)
+                    : 0;
+    user_ptr = (void *)TagMemoryAligned((uptr)user_ptr, size, tag);
+  }
 
   if ((orig_size % kShadowAlignment) && (alignment <= kShadowAlignment) &&
       right_align_mode) {
     uptr as_uptr = reinterpret_cast<uptr>(user_ptr);
     if (right_align_mode == kRightAlignAlways ||
         GetTagFromPointer(as_uptr) & 1) {  // use a tag bit as a random bit.
       user_ptr = reinterpret_cast<void *>(AlignRight(as_uptr, orig_size));
       meta->right_aligned = 1;
static void HwasanDeallocate(StackTrace *stack, void *tagged_ptr) {
   // This memory will not be reused by anyone else, so we are free to keep it
   // poisoned.
   Thread *t = GetCurrentThread();
   if (flags()->max_free_fill_size > 0) {
     uptr fill_size =
         Min(TaggedSize(orig_size), (uptr)flags()->max_free_fill_size);
     internal_memset(aligned_ptr, flags()->free_fill_byte, fill_size);
   }
-  if (flags()->tag_in_free &&
+  if (flags()->tag_in_free && malloc_bisect(stack, 0) &&
       atomic_load_relaxed(&hwasan_allocator_tagging_enabled))
     TagMemoryAligned(reinterpret_cast<uptr>(aligned_ptr), TaggedSize(orig_size),
                      t ? t->GenerateRandomTag() : kFallbackFreeTag);
   if (t) {
     allocator.Deallocate(t->allocator_cache(), aligned_ptr);
     if (auto *ha = t->heap_allocations())
       ha->push({reinterpret_cast<uptr>(tagged_ptr), alloc_context_id,
                 free_context_id, static_cast<u32>(orig_size)});

compiler-rt/lib/hwasan/hwasan_flags.inc

HWASAN_FLAG(int, heap_history_size, 1023,
           "Affects the quality of heap-related reports, but not the ability "
           "to find bugs.")
 HWASAN_FLAG(bool, export_memory_stats, true,
             "Export up-to-date memory stats through /proc")
 HWASAN_FLAG(int, stack_history_size, 1024,
             "The number of stack frames remembered per thread. "
             "Affects the quality of stack-related reports, but not the ability "
             "to find bugs.")
+
+// Malloc / free bisection. Only tag malloc and free calls when a hash of
+// allocation size and stack trace is between malloc_bisect_left and
+// malloc_bisect_right (both inclusive). [0, 0] range is special and disables
+// bisection (i.e. everything is tagged). Once the range is narrowed down
+// enough, use malloc_bisect_dump to see interesting allocations.
+HWASAN_FLAG(uptr, malloc_bisect_left, 0,
+            "Left bound of malloc bisection, inclusive.")
+HWASAN_FLAG(uptr, malloc_bisect_right, 0,
+            "Right bound of malloc bisection, inclusive.")
+HWASAN_FLAG(bool, malloc_bisect_dump, false,
+            "Print all allocations within [malloc_bisect_left, "
+            "malloc_bisect_right] range ")

compiler-rt/lib/hwasan/hwasan_malloc_bisect.h

+//===-- hwasan_malloc_bisect.h ----------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file is a part of HWAddressSanitizer.
+//
+//===----------------------------------------------------------------------===//
+
+#include "sanitizer_common/sanitizer_hash.h"
+#include "hwasan.h"
+
+namespace __hwasan {
+
+static u32 malloc_hash(StackTrace *stack, uptr orig_size) {
+  uptr len = Min(stack->size, (unsigned)7);
+  MurMur2HashBuilder H(len);
+  H.add(orig_size);
+  // Start with frame #1 to skip __sanitizer_malloc frame, which is
+  // (a) almost always the same (well, could be operator new or new[])
+  // (b) can change hashes when compiler-rt is rebuilt, invalidating previous
+  // bisection results.
+  // Because of ASLR, use only offset inside the page.
+  for (uptr i = 1; i < len; ++i) H.add(((u32)stack->trace[i]) & 0xFFF);
+  return H.get();
+}
+
+static INLINE bool malloc_bisect(StackTrace *stack, uptr orig_size) {
+  uptr left = flags()->malloc_bisect_left;
+  uptr right = flags()->malloc_bisect_right;
+  if (LIKELY(left == 0 && right == 0))
+    return true;
+  if (!stack)
+    return true;
+  // Allow malloc_bisect_right > (u32)(-1) to avoid spelling the latter in
+  // decimal.
+  uptr h = (uptr)malloc_hash(stack, orig_size);
+  if (h < left || h > right)
+    return false;
+  if (flags()->malloc_bisect_dump) {
+    Printf("[alloc] %u %zu\n", h, orig_size);
+    stack->Print();
+  }
+  return true;
+}
+
+}  // namespace __hwasan

compiler-rt/lib/sanitizer_common/CMakeLists.txt

set(SANITIZER_IMPL_HEADERS
   sanitizer_errno_codes.h
   sanitizer_file.h
   sanitizer_flag_parser.h
   sanitizer_flags.h
   sanitizer_flags.inc
   sanitizer_freebsd.h
   sanitizer_fuchsia.h
   sanitizer_getauxval.h
+  sanitizer_hash.h
   sanitizer_interceptors_ioctl_netbsd.inc
   sanitizer_interface_internal.h
   sanitizer_internal_defs.h
   sanitizer_lfstack.h
   sanitizer_libc.h
   sanitizer_libignore.h
   sanitizer_linux.h
   sanitizer_list.h

compiler-rt/lib/sanitizer_common/sanitizer_hash.h

+//===-- sanitizer_common.h --------------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements a simple hash function.
+//===----------------------------------------------------------------------===//
+
+#ifndef SANITIZER_HASH_H
+#define SANITIZER_HASH_H
+
+#include "sanitizer_internal_defs.h"
+
+namespace __sanitizer {
+class MurMur2HashBuilder {
+  static const u32 m = 0x5bd1e995;
+  static const u32 seed = 0x9747b28c;
+  static const u32 r = 24;
+  u32 h;
+
+ public:
+  explicit MurMur2HashBuilder(u32 init = 0) { h = seed ^ init; }
+  void add(u32 k) {
+    k *= m;
+    k ^= k >> r;
+    k *= m;
+    h *= m;
+    h ^= k;
+  }
+  u32 get() {
+    u32 x = h;
+    x ^= x >> 13;
+    x *= m;
+    x ^= x >> 15;
+    return x;
+  }
+};
+}  //namespace __sanitizer
+
+#endif  // SANITIZER_HASH_H

compiler-rt/lib/sanitizer_common/sanitizer_stackdepot.cc

 //===-- sanitizer_stackdepot.cc -------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file is shared between AddressSanitizer and ThreadSanitizer
 // run-time libraries.
 //===----------------------------------------------------------------------===//
 
 #include "sanitizer_stackdepot.h"
 
 #include "sanitizer_common.h"
+#include "sanitizer_hash.h"
 #include "sanitizer_stackdepotbase.h"
 
 namespace __sanitizer {
 
 struct StackDepotNode {
   StackDepotNode *link;
   u32 id;
   atomic_uint32_t hash_and_use_count; // hash_bits : 12; use_count : 20;
for (; i < size; i++) {
       if (stack[i] != args.trace[i]) return false;
     }
     return true;
   }
   static uptr storage_size(const args_type &args) {
     return sizeof(StackDepotNode) + (args.size - 1) * sizeof(uptr);
   }
   static u32 hash(const args_type &args) {
-    // murmur2
-    const u32 m = 0x5bd1e995;
-    const u32 seed = 0x9747b28c;
-    const u32 r = 24;
-    u32 h = seed ^ (args.size * sizeof(uptr));
-    for (uptr i = 0; i < args.size; i++) {
-      u32 k = args.trace[i];
-      k *= m;
-      k ^= k >> r;
-      k *= m;
-      h *= m;
-      h ^= k;
-    }
-    h ^= h >> 13;
-    h *= m;
-    h ^= h >> 15;
-    return h;
+    MurMur2HashBuilder H(args.size * sizeof(uptr));
+    for (uptr i = 0; i < args.size; i++) H.add(args.trace[i]);
+    return H.get();
   }
   static bool is_valid(const args_type &args) {
     return args.size > 0 && args.trace;
   }
   void store(const args_type &args, u32 hash) {
     atomic_store(&hash_and_use_count, hash & kHashMask, memory_order_relaxed);
     size = args.size;
     tag = args.tag;

compiler-rt/test/hwasan/TestCases/malloc_bisect.c

+// RUN: %clang_hwasan -O0 %s -o %t
+// RUN: %env_hwasan_opts=malloc_bisect_left=0,malloc_bisect_right=0          not %run %t 2>&1 | \
+// RUN:     FileCheck %s --check-prefix=CRASH
+// RUN: %env_hwasan_opts=malloc_bisect_left=1000,malloc_bisect_right=999         %run %t 2>&1
+// RUN: %env_hwasan_opts=malloc_bisect_left=0,malloc_bisect_right=5000000000 not %run %t 2>&1 | \
+// RUN:     FileCheck %s --check-prefix=CRASH
+// RUN: %env_hwasan_opts=malloc_bisect_left=0,malloc_bisect_right=5000000000,malloc_bisect_dump=1 not %run %t 2>&1 | \
+// RUN:     FileCheck %s --check-prefixes=CRASH,DUMP
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <sanitizer/hwasan_interface.h>
+
+int main() {
+  __hwasan_enable_allocator_tagging();
+  // DUMP: [alloc] {{.*}} 10{{$}}
+  // DUMP: in main{{.*}}malloc_bisect.c
+  char * volatile p = (char*)malloc(10);
+  // CRASH: HWAddressSanitizer: tag-mismatch on address
+  // CRASH: in main{{.*}}malloc_bisect.c
+  char volatile x = p[16];
+  free(p);
+  __hwasan_disable_allocator_tagging();
+
+  return 0;
+}

compiler-rt/test/hwasan/TestCases/tag_in_free.c

+// RUN: %clang_hwasan -O0 %s -DMALLOC -DFREE -o %t.mf
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=1 not %run %t.mf 2>&1 | FileCheck %s --check-prefixes=FREE
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=1 not %run %t.mf 2>&1 | FileCheck %s --check-prefixes=MALLOC
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=0 not %run %t.mf 2>&1 | FileCheck %s --check-prefixes=MALLOC
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=0     %run %t.mf 2>&1
+
+// RUN: %clang_hwasan -O0 %s -DFREE -o %t.f
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=1 not %run %t.f 2>&1 | FileCheck %s --check-prefixes=FREE
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=1 not %run %t.f 2>&1 | FileCheck %s --check-prefixes=FREE
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=0     %run %t.f 2>&1
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=0     %run %t.f 2>&1
+
+// RUN: %clang_hwasan -O0 %s -DMALLOC -o %t.m
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=1     %run %t.m 2>&1
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=1 not %run %t.m 2>&1 | FileCheck %s --check-prefixes=MALLOC
+// RUN: %env_hwasan_opts=tag_in_malloc=1,tag_in_free=0 not %run %t.m 2>&1 | FileCheck %s --check-prefixes=MALLOC
+// RUN: %env_hwasan_opts=tag_in_malloc=0,tag_in_free=0     %run %t.m 2>&1
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <sanitizer/hwasan_interface.h>
+
+int main() {
+  __hwasan_enable_allocator_tagging();
+  // Loop for a while to make sure that the memory for the test below is reused after an earlier free(),
+  // and is potentially tagged (when tag_in_free == 1).
+  for (int i = 0; i < 100; ++i) {
+    char * volatile p = (char*)malloc(10);
+    free(p);
+  }
+
+  char * volatile p = (char*)malloc(10);
+#ifdef MALLOC
+  // MALLOC: READ of size 1 at
+  // MALLOC: is located 6 bytes to the right of 10-byte region
+  // MALLOC: allocated here:
+  char volatile x = p[16];
+#endif
+  free(p);
+#ifdef FREE
+  // FREE: READ of size 1 at
+  // FREE: is located 0 bytes inside of 10-byte region
+  // FREE: freed by thread T0 here:
+  // FREE: previously allocated here:
+  char volatile y = p[0];
+#endif
+
+  __hwasan_disable_allocator_tagging();
+
+  return 0;
+}