Cherry-picking of changed from emscripten-libs-18 onto llvmorg-19.1.4 (#13)

Author: aheejin

compiler-rt/lib/asan/asan_errors.cpp

@@ -500,6 +500,17 @@ ErrorGeneric::ErrorGeneric(u32 tid, uptr pc_, uptr bp_, uptr sp_, uptr addr,
       scariness.Scare(bug_type_score + read_after_free_bonus, bug_descr);
       if (far_from_bounds) scariness.Scare(10, "far-from-bounds");
     }
+#if SANITIZER_EMSCRIPTEN
+    // If address is in the first page (64 KB), then it is likely that the
+    // access is a result of a null pointer dereference.
+    else if (addr < 65536) {
+      bug_descr = "null-pointer-dereference";
+      scariness.Scare(25, bug_descr);
+    } else if (AddrIsInShadow(addr)) {
+      bug_descr = "shadow-access";
+      scariness.Scare(25, bug_descr);
+    }
+#endif
   }
 }
 

compiler-rt/lib/asan/asan_flags.cpp

@@ -22,6 +22,12 @@
 #include "ubsan/ubsan_flags.h"
 #include "ubsan/ubsan_platform.h"
 
+#if SANITIZER_EMSCRIPTEN
+#include <emscripten/heap.h>
+#include "emscripten_internal.h"
+#endif
+
+
 namespace __asan {
 
 Flags asan_flags_dont_use_directly;  // use via flags().
@@ -54,7 +60,11 @@ void InitializeFlags() {
     CommonFlags cf;
     cf.CopyFrom(*common_flags());
     cf.detect_leaks = cf.detect_leaks && CAN_SANITIZE_LEAKS;
+#if !SANITIZER_EMSCRIPTEN
+    // getenv on emscripten uses malloc, which we can't when using LSan.
+    // You can't run external symbolizer executables anyway.
     cf.external_symbolizer_path = GetEnv("ASAN_SYMBOLIZER_PATH");
+#endif
     cf.malloc_context_size = kDefaultMallocContextSize;
     cf.intercept_tls_get_addr = true;
     cf.exitcode = 1;
@@ -115,6 +125,23 @@ void InitializeFlags() {
   lsan_parser.ParseString(lsan_default_options);
 #endif
 
+#if SANITIZER_EMSCRIPTEN
+  char *options;
+  // Override from Emscripten Module.
+  // TODO: add EM_ASM_I64 and avoid using a double for a 64-bit pointer.
+#define MAKE_OPTION_LOAD(parser, name) \
+    options = _emscripten_sanitizer_get_option(name); \
+    parser.ParseString(options); \
+    emscripten_builtin_free(options);
+
+  MAKE_OPTION_LOAD(asan_parser, "ASAN_OPTIONS");
+#if CAN_SANITIZE_LEAKS
+  MAKE_OPTION_LOAD(lsan_parser, "LSAN_OPTIONS");
+#endif
+#if CAN_SANITIZE_UB
+  MAKE_OPTION_LOAD(ubsan_parser, "UBSAN_OPTIONS");
+#endif
+#else
   // Override from command line.
   asan_parser.ParseStringFromEnv("ASAN_OPTIONS");
 #if CAN_SANITIZE_LEAKS
@@ -123,12 +150,18 @@ void InitializeFlags() {
 #if CAN_SANITIZE_UB
   ubsan_parser.ParseStringFromEnv("UBSAN_OPTIONS");
 #endif
+#endif // SANITIZER_EMSCRIPTEN
 
   InitializeCommonFlags();
 
   // TODO(eugenis): dump all flags at verbosity>=2?
   if (Verbosity()) ReportUnrecognizedFlags();
 
+#if SANITIZER_EMSCRIPTEN
+  if (common_flags()->malloc_context_size <= 1)
+    StackTrace::snapshot_stack = false;
+#endif // SANITIZER_EMSCRIPTEN
+
   if (common_flags()->help) {
     // TODO(samsonov): print all of the flags (ASan, LSan, common).
     asan_parser.PrintFlagDescriptions();

compiler-rt/lib/asan/asan_globals.cpp

@@ -325,6 +325,7 @@ void PrintGlobalLocation(InternalScopedString *str, const __asan_global &g,
 // ---------------------- Interface ---------------- {{{1
 using namespace __asan;
 
+#if !SANITIZER_EMSCRIPTEN
 // Apply __asan_register_globals to all globals found in the same loaded
 // executable or shared library as `flag'. The flag tracks whether globals have
 // already been registered or not for this image.
@@ -362,6 +363,7 @@ void __asan_unregister_elf_globals(uptr *flag, void *start, void *stop) {
   __asan_unregister_globals(globals_start, globals_stop - globals_start);
   *flag = 0;
 }
+#endif
 
 // Register an array of globals.
 void __asan_register_globals(__asan_global *globals, uptr n) {

compiler-rt/lib/asan/asan_interceptors.cpp

@@ -27,10 +27,10 @@
 #include "sanitizer_common/sanitizer_internal_defs.h"
 #include "sanitizer_common/sanitizer_libc.h"
 
-// There is no general interception at all on Fuchsia.
+// There is no general interception at all on Fuchsia or Emscripten.
 // Only the functions in asan_interceptors_memintrinsics.cpp are
 // really defined to replace libc functions.
-#if !SANITIZER_FUCHSIA
+#if !SANITIZER_FUCHSIA && !SANITIZER_EMSCRIPTEN
 
 #  if SANITIZER_POSIX
 #    include "sanitizer_common/sanitizer_posix.h"
@@ -883,4 +883,4 @@ void InitializeAsanInterceptors() {
 
 } // namespace __asan
 
-#endif  // !SANITIZER_FUCHSIA
+#endif  // !SANITIZER_FUCHSIA && !SANITIZER_RTEMS && !SANITIZER_EMSCRIPTEN

compiler-rt/lib/asan/asan_interceptors_memintrinsics.cpp

@@ -71,7 +71,7 @@ void *__asan_memmove(void *to, const void *from, uptr size) {
   ASAN_MEMMOVE_IMPL(nullptr, to, from, size);
 }
 
-#if SANITIZER_FUCHSIA
+#if SANITIZER_FUCHSIA || SANITIZER_EMSCRIPTEN
 
 // Fuchsia doesn't use sanitizer_common_interceptors.inc, but
 // the only things there it wants are these three.  Just define them
@@ -81,7 +81,7 @@ extern "C" decltype(__asan_memcpy) memcpy[[gnu::alias("__asan_memcpy")]];
 extern "C" decltype(__asan_memmove) memmove[[gnu::alias("__asan_memmove")]];
 extern "C" decltype(__asan_memset) memset[[gnu::alias("__asan_memset")]];
 
-#else  // SANITIZER_FUCHSIA
+#else  // SANITIZER_FUCHSIA || SANITIZER_EMSCRIPTEN
 
 #define COMMON_INTERCEPTOR_MEMMOVE_IMPL(ctx, to, from, size) \
   do {                                                       \
@@ -103,4 +103,4 @@ extern "C" decltype(__asan_memset) memset[[gnu::alias("__asan_memset")]];
 
 #include "sanitizer_common/sanitizer_common_interceptors_memintrinsics.inc"
 
-#endif  // SANITIZER_FUCHSIA
+#endif  // SANITIZER_FUCHSIA || SANITIZER_EMSCRIPTEN

compiler-rt/lib/asan/asan_malloc_linux.cpp

@@ -15,7 +15,7 @@
 
 #include "sanitizer_common/sanitizer_platform.h"
 #if SANITIZER_FREEBSD || SANITIZER_FUCHSIA || SANITIZER_LINUX || \
-    SANITIZER_NETBSD || SANITIZER_SOLARIS
+    SANITIZER_NETBSD || SANITIZER_SOLARIS || SANITIZER_EMSCRIPTEN
 
 #  include "asan_allocator.h"
 #  include "asan_interceptors.h"

compiler-rt/lib/asan/asan_mapping.h

@@ -272,6 +272,8 @@ extern uptr kHighMemEnd, kMidMemBeg, kMidMemEnd;  // Initialized in __asan_init.
 
 #  if defined(__sparc__) && SANITIZER_WORDSIZE == 64
 #    include "asan_mapping_sparc64.h"
+#  elif SANITIZER_EMSCRIPTEN
+#    include "asan_mapping_emscripten.h"
 #  else
 #    define MEM_TO_SHADOW(mem) \
       (((mem) >> ASAN_SHADOW_SCALE) + (ASAN_SHADOW_OFFSET))

compiler-rt/lib/asan/asan_poisoning.cpp

@@ -174,6 +174,15 @@ uptr __asan_region_is_poisoned(uptr beg, uptr size) {
   if (!size)
     return 0;
   uptr end = beg + size;
+#if SANITIZER_EMSCRIPTEN
+  // XXX Emscripten hack XXX
+  // Null pointer handling, since Emscripten does not crash on null pointer,
+  // ASan must catch null pointer dereference by itself.
+  // Unfortunately, this function returns 0 to mean the region is not
+  // poisoned, so we must return 1 instead if we receive a region
+  // starting at 0.
+  if (!beg) return 1;
+#endif
   if (!AddrIsInMem(beg))
     return beg;
   if (!AddrIsInMem(end))

compiler-rt/lib/asan/asan_poisoning.h

@@ -51,6 +51,10 @@ ALWAYS_INLINE void FastPoisonShadow(uptr aligned_beg, uptr aligned_size,
   // probably provide higher-level interface for these operations.
   // For now, just memset on Windows.
   if (value || SANITIZER_WINDOWS == 1 ||
+      // Emscripten doesn't have a nice way to zero whole pages.
+      // The bulk memory proposal will allow memset to be optimized, but
+      // even then, we still must use memset.
+      SANITIZER_EMSCRIPTEN == 1 ||
       shadow_end - shadow_beg < common_flags()->clear_shadow_mmap_threshold) {
     REAL(memset)((void*)shadow_beg, value, shadow_end - shadow_beg);
   } else {

compiler-rt/lib/asan/asan_posix.cpp

@@ -41,6 +41,9 @@ void AsanOnDeadlySignal(int signo, void *siginfo, void *context) {
 }
 
 bool PlatformUnpoisonStacks() {
+#if SANITIZER_EMSCRIPTEN
+  return false;
+#else
   stack_t signal_stack;
   CHECK_EQ(0, sigaltstack(nullptr, &signal_stack));
   uptr sigalt_bottom = (uptr)signal_stack.ss_sp;
@@ -64,6 +67,7 @@ bool PlatformUnpoisonStacks() {
                        &tls_size);
   UnpoisonStack(default_bottom, default_bottom + stack_size, "default");
   return true;
+#endif
 }
 
 // ---------------------- TSD ---------------- {{{1

compiler-rt/lib/asan/asan_rtl.cpp

@@ -54,6 +54,7 @@ static void AsanDie() {
 
   WaitForDebugger(flags()->sleep_before_dying, "before dying");
 
+#if !SANITIZER_EMSCRIPTEN
   if (flags()->unmap_shadow_on_exit) {
     if (kMidMemBeg) {
       UnmapOrDie((void*)kLowShadowBeg, kMidMemBeg - kLowShadowBeg);
@@ -63,6 +64,7 @@ static void AsanDie() {
         UnmapOrDie((void*)kLowShadowBeg, kHighShadowEnd - kLowShadowBeg);
     }
   }
+#endif
 }
 
 static void CheckUnwind() {
@@ -324,13 +326,15 @@ static void asan_atexit() {
 }
 
 static void InitializeHighMemEnd() {
+#if !SANITIZER_EMSCRIPTEN
 #if !ASAN_FIXED_MAPPING
   kHighMemEnd = GetMaxUserVirtualAddress();
   // Increase kHighMemEnd to make sure it's properly
   // aligned together with kHighMemBeg:
   kHighMemEnd |= (GetMmapGranularity() << ASAN_SHADOW_SCALE) - 1;
 #endif  // !ASAN_FIXED_MAPPING
   CHECK_EQ((kHighMemBeg % GetMmapGranularity()), 0);
+#endif  // !SANITIZER_EMSCRIPTEN
 }
 
 void PrintAddressSpaceLayout() {
@@ -449,12 +453,16 @@ static bool AsanInitInternal() {
 
   ReplaceSystemMalloc();
 
+#if !SANITIZER_EMSCRIPTEN
   DisableCoreDumperIfNecessary();
+#endif
 
   InitializeShadowMemory();
 
   AsanTSDInit(PlatformTSDDtor);
+#if !SANITIZER_EMSCRIPTEN
   InstallDeadlySignalHandlers(AsanOnDeadlySignal);
+#endif
 
   AllocatorOptions allocator_options;
   allocator_options.SetFrom(flags(), common_flags());

compiler-rt/lib/asan/asan_shadow_setup.cpp

@@ -13,8 +13,9 @@
 
 #include "sanitizer_common/sanitizer_platform.h"
 
-// asan_fuchsia.cpp has their own InitializeShadowMemory implementation.
-#if !SANITIZER_FUCHSIA
+// asan_fuchsia.cpp and asan_emscripten.cc have have their own
+// InitializeShadowMemory implementation.
+#if !SANITIZER_FUCHSIA && !SANITIZER_EMSCRIPTEN
 
 #  include "asan_internal.h"
 #  include "asan_mapping.h"

compiler-rt/lib/asan/asan_thread.cpp

@@ -133,8 +133,10 @@ void AsanThread::Destroy() {
     if (AsanThread *thread = GetCurrentThread())
       CHECK_EQ(this, thread);
     malloc_storage().CommitBack();
+#if !SANITIZER_EMSCRIPTEN
     if (common_flags()->use_sigaltstack)
       UnsetAlternateSignalStack();
+#endif
     FlushToDeadThreadStats(&stats_);
     // We also clear the shadow on thread destruction because
     // some code may still be executing in later TSD destructors
@@ -288,8 +290,10 @@ void AsanThread::ThreadStart(tid_t os_id) {
   Init();
   asanThreadRegistry().StartThread(tid(), os_id, ThreadType::Regular, nullptr);
 
+#if !SANITIZER_EMSCRIPTEN
   if (common_flags()->use_sigaltstack)
     SetAlternateSignalStack();
+#endif
 }
 
 AsanThread *CreateMainThread() {
@@ -563,6 +567,12 @@ void GetRunningThreadsLocked(InternalMmapVector<tid_t> *threads) {
 
 }  // namespace __lsan
 
+namespace __sanitizer {
+ThreadRegistry *GetThreadRegistryLocked() {
+  return __lsan::GetAsanThreadRegistryLocked();
+}
+}  // namespace __sanitizer
+
 // ---------------------- Interface ---------------- {{{1
 using namespace __asan;
 

compiler-rt/lib/builtins/fp_compare_impl.inc

@@ -12,7 +12,7 @@
 // functions. We need to ensure that the return value is sign-extended in the
 // same way as GCC expects (since otherwise GCC-generated __builtin_isinf
 // returns true for finite 128-bit floating-point numbers).
-#ifdef __aarch64__
+#if defined(__aarch64__) || defined(__wasm__)
 // AArch64 GCC overrides libgcc_cmp_return to use int instead of long.
 typedef int CMP_RESULT;
 #elif __SIZEOF_POINTER__ == 8 && __SIZEOF_LONG__ == 4

compiler-rt/lib/interception/interception.h

@@ -19,7 +19,7 @@
 
 #if !SANITIZER_LINUX && !SANITIZER_FREEBSD && !SANITIZER_APPLE &&    \
     !SANITIZER_NETBSD && !SANITIZER_WINDOWS && !SANITIZER_FUCHSIA && \
-    !SANITIZER_SOLARIS
+    !SANITIZER_SOLARIS && !SANITIZER_EMSCRIPTEN
 #  error "Interception doesn't work on this operating system."
 #endif
 
@@ -157,6 +157,10 @@ const interpose_substitution substitution_##func_name[]             \
     extern "C" ret_type func(__VA_ARGS__);
 # define DECLARE_WRAPPER_WINAPI(ret_type, func, ...)  \
     extern "C" __declspec(dllimport) ret_type __stdcall func(__VA_ARGS__);
+#elif SANITIZER_EMSCRIPTEN
+# define WRAP(x) x
+# define INTERCEPTOR_ATTRIBUTE
+# define DECLARE_WRAPPER(ret_type, func, ...)
 #elif !SANITIZER_FUCHSIA  // LINUX, FREEBSD, NETBSD, SOLARIS
 # define INTERCEPTOR_ATTRIBUTE __attribute__((visibility("default")))
 # if ASM_INTERCEPTOR_TRAMPOLINE_SUPPORT
@@ -240,6 +244,13 @@ const interpose_substitution substitution_##func_name[]             \
 # define INTERCEPTOR_ATTRIBUTE __attribute__((visibility("default")))
 # define REAL(x) __unsanitized_##x
 # define DECLARE_REAL(ret_type, func, ...)
+#elif SANITIZER_EMSCRIPTEN
+// Sanitizer runtimes on Emscripten just define functions directly to override
+// the libc functions. If the real version is really needed, they can be defined
+// with the emscripten_builtin_ prefix.
+# define REAL(x) emscripten_builtin_##x
+# define DECLARE_REAL(ret_type, func, ...) \
+    extern "C" ret_type REAL(func)(__VA_ARGS__);
 #elif !SANITIZER_APPLE
 # define PTR_TO_REAL(x) real_##x
 # define REAL(x) __interception::PTR_TO_REAL(x)
@@ -278,7 +289,7 @@ const interpose_substitution substitution_##func_name[]             \
 // macros does its job. In exceptional cases you may need to call REAL(foo)
 // without defining INTERCEPTOR(..., foo, ...). For example, if you override
 // foo with an interceptor for other function.
-#if !SANITIZER_APPLE && !SANITIZER_FUCHSIA
+#if !SANITIZER_APPLE && !SANITIZER_FUCHSIA && !SANITIZER_EMSCRIPTEN
 #  define DEFINE_REAL(ret_type, func, ...)            \
     typedef ret_type (*FUNC_TYPE(func))(__VA_ARGS__); \
     namespace __interception {                        \
@@ -344,9 +355,9 @@ const interpose_substitution substitution_##func_name[]             \
 // INTERCEPT_FUNCTION macro, only its name.
 namespace __interception {
 #if defined(_WIN64)
-typedef unsigned long long uptr;
+typedef unsigned long long uptr;  // NOLINT
 #else
-typedef unsigned long uptr;
+typedef unsigned long uptr;  // NOLINT
 #endif  // _WIN64
 
 #if defined(__ELF__) && !SANITIZER_FUCHSIA
@@ -365,7 +376,7 @@ inline void DoesNotSupportStaticLinking() {}
 #define INCLUDED_FROM_INTERCEPTION_LIB
 
 #if SANITIZER_LINUX || SANITIZER_FREEBSD || SANITIZER_NETBSD || \
-    SANITIZER_SOLARIS
+    SANITIZER_SOLARIS || SANITIZER_EMSCRIPTEN
 
 # include "interception_linux.h"
 # define INTERCEPT_FUNCTION(func) INTERCEPT_FUNCTION_LINUX_OR_FREEBSD(func)

compiler-rt/lib/interception/interception_linux.h

@@ -12,7 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #if SANITIZER_LINUX || SANITIZER_FREEBSD || SANITIZER_NETBSD || \
-    SANITIZER_SOLARIS
+    SANITIZER_SOLARIS || SANITIZER_EMSCRIPTEN
 
 #if !defined(INCLUDED_FROM_INTERCEPTION_LIB)
 # error interception_linux.h should be included from interception library only