mozglue/build/TsanOptions.cpp

   1 /* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
   2 /* This Source Code Form is subject to the terms of the Mozilla Public
   3  * License, v. 2.0. If a copy of the MPL was not distributed with this file,
   4  * You can obtain one at http://mozilla.org/MPL/2.0/. */
   5
   6 #include "mozilla/Attributes.h"
   7 #include "mozilla/TsanOptions.h"
   8
   9 #ifndef _MSC_VER  // Not supported by clang-cl yet
  10
  11 //
  12 // When running with ThreadSanitizer, we sometimes need to suppress existing
  13 // races. However, in any case, it should be either because
  14 //
  15 //    1) a bug is on file. In this case, the bug number should always be
  16 //       included with the suppression.
  17 //
  18 // or 2) this is an intentional race. Please be very careful with judging
  19 //       races as intentional and benign. Races in C++ are undefined behavior
  20 //       and compilers increasingly rely on exploiting this for optimizations.
  21 //       Hence, many seemingly benign races cause harmful or unexpected
  22 //       side-effects.
  23 //
  24 //       See also:
  25 //       https://software.intel.com/en-us/blogs/2013/01/06/benign-data-races-what-could-possibly-go-wrong
  26 //
  27 //
  28 // Also, when adding any race suppressions here, make sure to always add
  29 // a signature for each of the two race stacks. Sometimes, TSan fails to
  30 // symbolize one of the two traces and this can cause suppressed races to
  31 // show up intermittently.
  32 //
  33 // clang-format off
  34 extern "C" const char* __tsan_default_suppressions() {
  35   return "# Add your suppressions below\n"
  36
  37          // External uninstrumented libraries
  38          MOZ_TSAN_DEFAULT_EXTLIB_SUPPRESSIONS
  39
  40          // TSan internals
  41          "race:__tsan::ProcessPendingSignals\n"
  42          "race:__tsan::CallUserSignalHandler\n"
  43
  44
  45
  46
  47
  48          // Uninstrumented code causing false positives
  49
  50          // These libraries are uninstrumented and cause mutex false positives.
  51          // However, they can be unloaded by GTK early which we cannot avoid.
  52          "mutex:libGL.so\n"
  53          "mutex:libGLdispatch\n"
  54          "mutex:libGLX\n"
  55          // Bug 1637707 - permanent
  56          "mutex:libEGL_mesa.so\n"
  57          // ~GLContextGLX unlocks a libGL mutex.
  58          "mutex:GLContextGLX::~GLContextGLX\n"
  59          // Bug 1825171
  60          "mutex:libffi.so\n"
  61          "mutex:wl_registry_destroy\n"
  62          // Bug 1824768
  63          "mutex:libdbus-1\n"
  64          "mutex:swrast_dri.so\n"
  65          // Bug 1651446 - permanent (ffmpeg)
  66          "race:libavcodec.so*\n"
  67          "race:libavutil.so*\n"
  68          // For some reason, the suppressions on libpulse.so
  69          // through `called_from_lib` only work partially.
  70          "race:libpulse.so\n"
  71          "race:pa_context_suspend_source_by_index\n"
  72          "race:pa_context_unref\n"
  73          "race:pa_format_info_set_prop_string_array\n"
  74          "race:pa_stream_get_index\n"
  75          "race:pa_stream_update_timing_info\n"
  76          "race:vorbis_synthesis_init\n"
  77          // This is a callback from libglib-2 that is apparently
  78          // not fully suppressed through `called_from_lib`.
  79          "race:g_main_context_dispatch\n"
  80          // This is likely a false positive involving a mutex from GTK.
  81          // See also bug 1642653 - permanent.
  82          "mutex:GetMaiAtkType\n"
  83          // Bug 1688716 - Failure due to fire_glxtest_process
  84          // calling into uninstrumented external graphics driver code.
  85          // For example: iris_dri.so and swrast_dri.so.
  86          "race:fire_glxtest_process\n"
  87          "race:iris_dri\n"
  88          // Bug 1824768
  89          "race:libLLVM-12\n"
  90          "race:radeonsi_dri\n"
  91          // Bug 1722721 - WebRender using uninstrumented Mesa drivers
  92          "race:swrast_dri.so\n"
  93          // Bug 1825171
  94          "race:libffi.so\n"
  95          "race:mozilla::widget::WaylandBuffer::BufferReleaseCallbackHandler\n"
  96
  97
  98
  99
 100          // Deadlock reports on single-threaded runtime.
 101          //
 102          // This is a known false positive from TSan where it reports
 103          // a potential deadlock even though all mutexes are only
 104          // taken by a single thread. For applications/tasks where we
 105          // are absolutely sure that no second thread will be involved
 106          // we should suppress these issues.
 107          //
 108          // See also https://github.com/google/sanitizers/issues/488
 109
 110          // Bug 1614605 - permanent
 111          "deadlock:SanctionsTestServer\n"
 112          "deadlock:OCSPStaplingServer\n"
 113          // Bug 1643087 - permanent
 114          "deadlock:BadCertAndPinningServer\n"
 115          // Bug 1606804 - permanent
 116          "deadlock:cert_storage::SecurityState::open_db\n"
 117          "deadlock:cert_storage::SecurityState::add_certs\n"
 118          // Bug 1651770 - permanent
 119          "deadlock:mozilla::camera::LockAndDispatch\n"
 120          // Bug 1606804 - permanent
 121          "deadlock:third_party/rust/rkv/src/env.rs\n"
 122          // Bug 1680655 - permanent
 123          "deadlock:EncryptedClientHelloServer\n"
 124          // Bug 1682861 - permanent
 125          "deadlock:nsDOMWindowUtils::CompareCanvases\n"
 126
 127
 128
 129
 130
 131          // Benign races in third-party code (should be fixed upstream)
 132
 133          // No Bug - permanent
 134          // No Upstream Bug Filed!
 135          //
 136          // SIMD Initialization in libjpeg, potentially runs
 137          // initialization twice, but otherwise benign. Init
 138          // routine itself is in native assembler.
 139          "race:init_simd\n"
 140          "race:simd_support\n"
 141          "race:jsimd_can_ycc_rgb\n"
 142          // Bug 1615228 - permanent
 143          // No Upstream Bug Filed!
 144          //
 145          // Likely benign race in ipc/chromium/ where we set
 146          // `message_loop_` to `NULL` on two threads when stopping
 147          // a thread at the same time it is already finishing.
 148          "race:base::Thread::Stop\n"
 149          // Bug 1615569 - permanent
 150          // No Upstream Bug Filed!
 151          //
 152          // NSS is using freebl from two different threads but freebl isn't
 153          // that threadsafe.
 154          "race:mp_exptmod.max_window_bits\n"
 155          // Bug 1652499 - permanent
 156          // No Upstream Bug Filed!
 157          //
 158          // Likely benign race in webrtc.org code - race while updating the
 159          // minimum log severity.
 160          "race:Loggable\n"
 161          "race:UpdateMinLogSeverity\n"
 162          // Bug 1652174 - permanent
 163          // Upstream Bug: https://github.com/libevent/libevent/issues/777
 164          //
 165          // Likely benign write-write race in libevent to set a sticky boolean
 166          // flag to true.
 167          "race:event_debug_mode_too_late\n"
 168
 169          // Bug 1653618 - permanent
 170          // Upstream Bug: https://github.com/sctplab/usrsctp/issues/507
 171          //
 172          // Might lead to scheduled timers in libusrsctp getting dropped?
 173          "race:sctp_handle_tick\n"
 174          "race:sctp_handle_sack\n"
 175          // Bug 1648604 - permanent
 176          // Upstream Bug: https://github.com/sctplab/usrsctp/issues/482
 177          //
 178          // Likely benign race in libusrsctp allocator during a free.
 179          "race:system_base_info\n"
 180          // Benign lock-order-inversion in libusrsctp
 181          // No upstream bug filed!
 182          "deadlock:sctp_add_to_readq\n"
 183
 184          // Bug 1153409 - permanent
 185          // No Upstream Bug Filed!
 186          //
 187          // Probably benign - sqlite has a few optimizations where it does
 188          // racy reads and then does properly synchronized integrity checks
 189          // afterwards. Some concern of compiler optimizations messing this
 190          // up due to "volatile" being too weak for this.
 191          "race:third_party/sqlite3/*\n"
 192          "deadlock:third_party/sqlite3/*\n"
 193          // Bug 1674770 - permanent
 194          // Upstream Bug: https://github.com/Amanieu/parking_lot/issues/257
 195          //
 196          // parking_lot using incorrect atomic orderings in RwLock, upstream
 197          // fix already up for review.
 198          "race:StrongRuleNode::ensure_child\n"
 199          // No Bug - permanent
 200          // Upstream Bugs:
 201          //
 202          //  * https://github.com/rayon-rs/rayon/issues/812
 203          //  * https://github.com/crossbeam-rs/crossbeam/issues/589
 204          //
 205          // Probably a false-positive from crossbeam's deque not being
 206          // understood by tsan.
 207          "race:crossbeam_deque*::resize\n"
 208          "race:crossbeam_deque*::push\n"
 209          "race:crossbeam_deque*::write\n"
 210          "race:crossbeam_deque*::read\n"
 211          "race:crossbeam_deque*::steal\n"
 212          // Bug 1805819 - permanent
 213          // No Upstream Bug Filed!
 214          //
 215          // False positive in libc's tzset_internal
 216          // See https://crbug.com/379738 also
 217          "race:tzset_internal\n"
 218
 219
 220
 221
 222
 223          // The rest of these suppressions are miscellaneous issues in gecko
 224          // that should be investigated and ideally fixed.
 225
 226          // Bug 1671574 - Permanent
 227          // The StartupCache thread intentionally races with the main thread to
 228          // trigger OS-level paging. It is never joined with the main thread.
 229          "thread:StartupCache\n"
 230
 231          // Bug 1734262 - Permanent
 232          // When spawning async processes, we create a helper thread to wait for
 233          // the process to terminate in order to asynchronously report the exit
 234          // code to Gecko. This thread waits on a syscall for the process to end,
 235          // which means there's no easy way to cancel and join it during Gecko
 236          // shutdown. Suppress thread leak reports for this thread.
 237          "thread:CreateMonitorThread\n"
 238
 239          // Bug 1601600
 240          "race:SkARGB32_Blitter\n"
 241          "race:SkARGB32_Shader_Blitter\n"
 242          "race:SkARGB32_Opaque_Blitter\n"
 243          "race:SkRasterPipelineBlitter\n"
 244          "race:Clamp_S32_D32_nofilter_trans_shaderproc\n"
 245          "race:SkSpriteBlitter_Memcpy\n"
 246
 247          // Bug 1606800
 248          "race:CallInitFunc\n"
 249
 250          // Bug 1606803
 251          "race:ipv6_is_present\n"
 252
 253          // Bug 1615123
 254          "race:_dl_deallocate_tls\n"
 255          "race:__libc_memalign\n"
 256
 257          // Bug 1664803
 258          "race:Sampler::sSigHandlerCoordinator\n"
 259
 260          // Bug 1656068
 261          "race:WebRtcAec_Create\n"
 262
 263          // No Bug - Logging bug in Mochitests
 264          "race:mochitest/ssltunnel/ssltunnel.cpp\n"
 265
 266          // This thread does not seem to be stopped/joined.
 267          // ImageBridgeChild should be turned back into a background
 268          // task queue in bug 1647628, in which case these suppressions
 269          // can be removed.
 270          "race:mozilla::layers::ImageBridgeChild::ShutDown\n"
 271
 272          // Bug 1652530
 273          "mutex:XErrorTrap\n"
 274
 275          // Bug 1671601
 276          "race:CamerasParent::ActorDestroy\n"
 277          "race:CamerasParent::DispatchToVideoCaptureThread\n"
 278
 279          // Bug 1623541
 280          "race:VRShMem::PullSystemState\n"
 281          "race:VRShMem::PushSystemState\n"
 282          "race:VRShMem::PullBrowserState\n"
 283          "race:VRShMem::PushBrowserState\n"
 284
 285          // Bug 1682951
 286          "race:storage::Connection::Release\n"
 287
 288          // Bug 1683357
 289          "race:image::ImageSurfaceCache::SuggestedSizeInternal\n"
 290          "race:image::RasterImage::SetMetadata\n"
 291          "race:image::RasterImage::GetWidth\n"
 292
 293          // Bug 1722721 - This is a benign race creating worker/SW compositor threads.
 294          "race:webrender::profiler::register_thread\n"
 295
 296          // Bug 1722721 - This is a false positive during SW-WR rendering.
 297          "race:scale_blit\n"
 298
 299          "race:mozilla::gl::MesaMemoryLeakWorkaround\n"
 300
 301          // Bug 1733908
 302          "race:js::wasm::Code::bestTier\n"
 303          "race:js::wasm::Code::commitTier2\n"
 304          "race:js::wasm::Code::setTier2\n"
 305          "race:js::wasm::Code::setAndBorrowTier2\n"
 306
 307          // Bug 1755449
 308          // The Glean init thread is used to perform I/O and other blocking operations.
 309          // It is never joined with the main thread, but this is being re-evaluated.
 310          "thread:glean::initialize\n"
 311
 312          // Bug 1822605 - permanent
 313          // A race exists in libvulkan_lvp.so.  This was previously addressed in bug
 314          // 1816713. However, libvulkan_lvp.so is unloaded so a called_from_lib
 315          // suppression cannot be used.
 316          "race:libvulkan_lvp.so\n"
 317
 318       // End of suppressions.
 319       ;  // Please keep this semicolon.
 320 }
 321 // clang-format on
 322 #endif  // _MSC_VER