Home | History | Annotate | Line # | Download | only in Support
      1 //===--- CrashRecoveryContext.cpp - Crash Recovery ------------------------===//
      2 //
      3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
      4 // See https://llvm.org/LICENSE.txt for license information.
      5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
      6 //
      7 //===----------------------------------------------------------------------===//
      8 
      9 #include "llvm/Support/CrashRecoveryContext.h"
     10 #include "llvm/Config/llvm-config.h"
     11 #include "llvm/Support/ErrorHandling.h"
     12 #include "llvm/Support/ExitCodes.h"
     13 #include "llvm/Support/ManagedStatic.h"
     14 #include "llvm/Support/Signals.h"
     15 #include "llvm/Support/ThreadLocal.h"
     16 #include <mutex>
     17 #include <setjmp.h>
     18 
     19 using namespace llvm;
     20 
     21 namespace {
     22 
     23 struct CrashRecoveryContextImpl;
     24 
     25 static ManagedStatic<
     26     sys::ThreadLocal<const CrashRecoveryContextImpl> > CurrentContext;
     27 
     28 struct CrashRecoveryContextImpl {
     29   // When threads are disabled, this links up all active
     30   // CrashRecoveryContextImpls.  When threads are enabled there's one thread
     31   // per CrashRecoveryContext and CurrentContext is a thread-local, so only one
     32   // CrashRecoveryContextImpl is active per thread and this is always null.
     33   const CrashRecoveryContextImpl *Next;
     34 
     35   CrashRecoveryContext *CRC;
     36   ::jmp_buf JumpBuffer;
     37   volatile unsigned Failed : 1;
     38   unsigned SwitchedThread : 1;
     39   unsigned ValidJumpBuffer : 1;
     40 
     41 public:
     42   CrashRecoveryContextImpl(CrashRecoveryContext *CRC) noexcept
     43       : CRC(CRC), Failed(false), SwitchedThread(false), ValidJumpBuffer(false) {
     44     Next = CurrentContext->get();
     45     CurrentContext->set(this);
     46   }
     47   ~CrashRecoveryContextImpl() {
     48     if (!SwitchedThread)
     49       CurrentContext->set(Next);
     50   }
     51 
     52   /// Called when the separate crash-recovery thread was finished, to
     53   /// indicate that we don't need to clear the thread-local CurrentContext.
     54   void setSwitchedThread() {
     55 #if defined(LLVM_ENABLE_THREADS) && LLVM_ENABLE_THREADS != 0
     56     SwitchedThread = true;
     57 #endif
     58   }
     59 
     60   // If the function ran by the CrashRecoveryContext crashes or fails, then
     61   // 'RetCode' represents the returned error code, as if it was returned by a
     62   // process. 'Context' represents the signal type on Unix; on Windows, it is
     63   // the ExceptionContext.
     64   void HandleCrash(int RetCode, uintptr_t Context) {
     65     // Eliminate the current context entry, to avoid re-entering in case the
     66     // cleanup code crashes.
     67     CurrentContext->set(Next);
     68 
     69     assert(!Failed && "Crash recovery context already failed!");
     70     Failed = true;
     71 
     72     if (CRC->DumpStackAndCleanupOnFailure)
     73       sys::CleanupOnSignal(Context);
     74 
     75     CRC->RetCode = RetCode;
     76 
     77     // Jump back to the RunSafely we were called under.
     78     if (ValidJumpBuffer)
     79       longjmp(JumpBuffer, 1);
     80 
     81     // Otherwise let the caller decide of the outcome of the crash. Currently
     82     // this occurs when using SEH on Windows with MSVC or clang-cl.
     83   }
     84 };
     85 } // namespace
     86 
     87 static ManagedStatic<std::mutex> gCrashRecoveryContextMutex;
     88 static bool gCrashRecoveryEnabled = false;
     89 
     90 static ManagedStatic<sys::ThreadLocal<const CrashRecoveryContext>>
     91        tlIsRecoveringFromCrash;
     92 
     93 static void installExceptionOrSignalHandlers();
     94 static void uninstallExceptionOrSignalHandlers();
     95 
     96 CrashRecoveryContextCleanup::~CrashRecoveryContextCleanup() {}
     97 
     98 CrashRecoveryContext::CrashRecoveryContext() {
     99   // On Windows, if abort() was previously triggered (and caught by a previous
    100   // CrashRecoveryContext) the Windows CRT removes our installed signal handler,
    101   // so we need to install it again.
    102   sys::DisableSystemDialogsOnCrash();
    103 }
    104 
    105 CrashRecoveryContext::~CrashRecoveryContext() {
    106   // Reclaim registered resources.
    107   CrashRecoveryContextCleanup *i = head;
    108   const CrashRecoveryContext *PC = tlIsRecoveringFromCrash->get();
    109   tlIsRecoveringFromCrash->set(this);
    110   while (i) {
    111     CrashRecoveryContextCleanup *tmp = i;
    112     i = tmp->next;
    113     tmp->cleanupFired = true;
    114     tmp->recoverResources();
    115     delete tmp;
    116   }
    117   tlIsRecoveringFromCrash->set(PC);
    118 
    119   CrashRecoveryContextImpl *CRCI = (CrashRecoveryContextImpl *) Impl;
    120   delete CRCI;
    121 }
    122 
    123 bool CrashRecoveryContext::isRecoveringFromCrash() {
    124   return tlIsRecoveringFromCrash->get() != nullptr;
    125 }
    126 
    127 CrashRecoveryContext *CrashRecoveryContext::GetCurrent() {
    128   if (!gCrashRecoveryEnabled)
    129     return nullptr;
    130 
    131   const CrashRecoveryContextImpl *CRCI = CurrentContext->get();
    132   if (!CRCI)
    133     return nullptr;
    134 
    135   return CRCI->CRC;
    136 }
    137 
    138 void CrashRecoveryContext::Enable() {
    139   std::lock_guard<std::mutex> L(*gCrashRecoveryContextMutex);
    140   // FIXME: Shouldn't this be a refcount or something?
    141   if (gCrashRecoveryEnabled)
    142     return;
    143   gCrashRecoveryEnabled = true;
    144   installExceptionOrSignalHandlers();
    145 }
    146 
    147 void CrashRecoveryContext::Disable() {
    148   std::lock_guard<std::mutex> L(*gCrashRecoveryContextMutex);
    149   if (!gCrashRecoveryEnabled)
    150     return;
    151   gCrashRecoveryEnabled = false;
    152   uninstallExceptionOrSignalHandlers();
    153 }
    154 
    155 void CrashRecoveryContext::registerCleanup(CrashRecoveryContextCleanup *cleanup)
    156 {
    157   if (!cleanup)
    158     return;
    159   if (head)
    160     head->prev = cleanup;
    161   cleanup->next = head;
    162   head = cleanup;
    163 }
    164 
    165 void
    166 CrashRecoveryContext::unregisterCleanup(CrashRecoveryContextCleanup *cleanup) {
    167   if (!cleanup)
    168     return;
    169   if (cleanup == head) {
    170     head = cleanup->next;
    171     if (head)
    172       head->prev = nullptr;
    173   }
    174   else {
    175     cleanup->prev->next = cleanup->next;
    176     if (cleanup->next)
    177       cleanup->next->prev = cleanup->prev;
    178   }
    179   delete cleanup;
    180 }
    181 
    182 #if defined(_MSC_VER)
    183 
    184 #include <windows.h> // for GetExceptionInformation
    185 
    186 // If _MSC_VER is defined, we must have SEH. Use it if it's available. It's way
    187 // better than VEH. Vectored exception handling catches all exceptions happening
    188 // on the thread with installed exception handlers, so it can interfere with
    189 // internal exception handling of other libraries on that thread. SEH works
    190 // exactly as you would expect normal exception handling to work: it only
    191 // catches exceptions if they would bubble out from the stack frame with __try /
    192 // __except.
    193 
    194 static void installExceptionOrSignalHandlers() {}
    195 static void uninstallExceptionOrSignalHandlers() {}
    196 
    197 // We need this function because the call to GetExceptionInformation() can only
    198 // occur inside the __except evaluation block
    199 static int ExceptionFilter(_EXCEPTION_POINTERS *Except) {
    200   // Lookup the current thread local recovery object.
    201   const CrashRecoveryContextImpl *CRCI = CurrentContext->get();
    202 
    203   if (!CRCI) {
    204     // Something has gone horribly wrong, so let's just tell everyone
    205     // to keep searching
    206     CrashRecoveryContext::Disable();
    207     return EXCEPTION_CONTINUE_SEARCH;
    208   }
    209 
    210   int RetCode = (int)Except->ExceptionRecord->ExceptionCode;
    211   if ((RetCode & 0xF0000000) == 0xE0000000)
    212     RetCode &= ~0xF0000000; // this crash was generated by sys::Process::Exit
    213 
    214   // Handle the crash
    215   const_cast<CrashRecoveryContextImpl *>(CRCI)->HandleCrash(
    216       RetCode, reinterpret_cast<uintptr_t>(Except));
    217 
    218   return EXCEPTION_EXECUTE_HANDLER;
    219 }
    220 
    221 #if defined(__clang__) && defined(_M_IX86)
    222 // Work around PR44697.
    223 __attribute__((optnone))
    224 #endif
    225 bool CrashRecoveryContext::RunSafely(function_ref<void()> Fn) {
    226   if (!gCrashRecoveryEnabled) {
    227     Fn();
    228     return true;
    229   }
    230   assert(!Impl && "Crash recovery context already initialized!");
    231   Impl = new CrashRecoveryContextImpl(this);
    232   __try {
    233     Fn();
    234   } __except (ExceptionFilter(GetExceptionInformation())) {
    235     return false;
    236   }
    237   return true;
    238 }
    239 
    240 #else // !_MSC_VER
    241 
    242 #if defined(_WIN32)
    243 // This is a non-MSVC compiler, probably mingw gcc or clang without
    244 // -fms-extensions. Use vectored exception handling (VEH).
    245 //
    246 // On Windows, we can make use of vectored exception handling to catch most
    247 // crashing situations.  Note that this does mean we will be alerted of
    248 // exceptions *before* structured exception handling has the opportunity to
    249 // catch it. Unfortunately, this causes problems in practice with other code
    250 // running on threads with LLVM crash recovery contexts, so we would like to
    251 // eventually move away from VEH.
    252 //
    253 // Vectored works on a per-thread basis, which is an advantage over
    254 // SetUnhandledExceptionFilter. SetUnhandledExceptionFilter also doesn't have
    255 // any native support for chaining exception handlers, but VEH allows more than
    256 // one.
    257 //
    258 // The vectored exception handler functionality was added in Windows
    259 // XP, so if support for older versions of Windows is required,
    260 // it will have to be added.
    261 
    262 #include "llvm/Support/Windows/WindowsSupport.h"
    263 
    264 static LONG CALLBACK ExceptionHandler(PEXCEPTION_POINTERS ExceptionInfo)
    265 {
    266   // DBG_PRINTEXCEPTION_WIDE_C is not properly defined on all supported
    267   // compilers and platforms, so we define it manually.
    268   constexpr ULONG DbgPrintExceptionWideC = 0x4001000AL;
    269   switch (ExceptionInfo->ExceptionRecord->ExceptionCode)
    270   {
    271   case DBG_PRINTEXCEPTION_C:
    272   case DbgPrintExceptionWideC:
    273   case 0x406D1388:  // set debugger thread name
    274     return EXCEPTION_CONTINUE_EXECUTION;
    275   }
    276 
    277   // Lookup the current thread local recovery object.
    278   const CrashRecoveryContextImpl *CRCI = CurrentContext->get();
    279 
    280   if (!CRCI) {
    281     // Something has gone horribly wrong, so let's just tell everyone
    282     // to keep searching
    283     CrashRecoveryContext::Disable();
    284     return EXCEPTION_CONTINUE_SEARCH;
    285   }
    286 
    287   // TODO: We can capture the stack backtrace here and store it on the
    288   // implementation if we so choose.
    289 
    290   int RetCode = (int)ExceptionInfo->ExceptionRecord->ExceptionCode;
    291   if ((RetCode & 0xF0000000) == 0xE0000000)
    292     RetCode &= ~0xF0000000; // this crash was generated by sys::Process::Exit
    293 
    294   // Handle the crash
    295   const_cast<CrashRecoveryContextImpl *>(CRCI)->HandleCrash(
    296       RetCode, reinterpret_cast<uintptr_t>(ExceptionInfo));
    297 
    298   // Note that we don't actually get here because HandleCrash calls
    299   // longjmp, which means the HandleCrash function never returns.
    300   llvm_unreachable("Handled the crash, should have longjmp'ed out of here");
    301 }
    302 
    303 // Because the Enable and Disable calls are static, it means that
    304 // there may not actually be an Impl available, or even a current
    305 // CrashRecoveryContext at all.  So we make use of a thread-local
    306 // exception table.  The handles contained in here will either be
    307 // non-NULL, valid VEH handles, or NULL.
    308 static sys::ThreadLocal<const void> sCurrentExceptionHandle;
    309 
    310 static void installExceptionOrSignalHandlers() {
    311   // We can set up vectored exception handling now.  We will install our
    312   // handler as the front of the list, though there's no assurances that
    313   // it will remain at the front (another call could install itself before
    314   // our handler).  This 1) isn't likely, and 2) shouldn't cause problems.
    315   PVOID handle = ::AddVectoredExceptionHandler(1, ExceptionHandler);
    316   sCurrentExceptionHandle.set(handle);
    317 }
    318 
    319 static void uninstallExceptionOrSignalHandlers() {
    320   PVOID currentHandle = const_cast<PVOID>(sCurrentExceptionHandle.get());
    321   if (currentHandle) {
    322     // Now we can remove the vectored exception handler from the chain
    323     ::RemoveVectoredExceptionHandler(currentHandle);
    324 
    325     // Reset the handle in our thread-local set.
    326     sCurrentExceptionHandle.set(NULL);
    327   }
    328 }
    329 
    330 #else // !_WIN32
    331 
    332 // Generic POSIX implementation.
    333 //
    334 // This implementation relies on synchronous signals being delivered to the
    335 // current thread. We use a thread local object to keep track of the active
    336 // crash recovery context, and install signal handlers to invoke HandleCrash on
    337 // the active object.
    338 //
    339 // This implementation does not attempt to chain signal handlers in any
    340 // reliable fashion -- if we get a signal outside of a crash recovery context we
    341 // simply disable crash recovery and raise the signal again.
    342 
    343 #include <signal.h>
    344 
    345 static const int Signals[] =
    346     { SIGABRT, SIGBUS, SIGFPE, SIGILL, SIGSEGV, SIGTRAP };
    347 static const unsigned NumSignals = array_lengthof(Signals);
    348 static struct sigaction PrevActions[NumSignals];
    349 
    350 static void CrashRecoverySignalHandler(int Signal) {
    351   // Lookup the current thread local recovery object.
    352   const CrashRecoveryContextImpl *CRCI = CurrentContext->get();
    353 
    354   if (!CRCI) {
    355     // We didn't find a crash recovery context -- this means either we got a
    356     // signal on a thread we didn't expect it on, the application got a signal
    357     // outside of a crash recovery context, or something else went horribly
    358     // wrong.
    359     //
    360     // Disable crash recovery and raise the signal again. The assumption here is
    361     // that the enclosing application will terminate soon, and we won't want to
    362     // attempt crash recovery again.
    363     //
    364     // This call of Disable isn't thread safe, but it doesn't actually matter.
    365     CrashRecoveryContext::Disable();
    366     raise(Signal);
    367 
    368     // The signal will be thrown once the signal mask is restored.
    369     return;
    370   }
    371 
    372   // Unblock the signal we received.
    373   sigset_t SigMask;
    374   sigemptyset(&SigMask);
    375   sigaddset(&SigMask, Signal);
    376   sigprocmask(SIG_UNBLOCK, &SigMask, nullptr);
    377 
    378   // Return the same error code as if the program crashed, as mentioned in the
    379   // section "Exit Status for Commands":
    380   // https://pubs.opengroup.org/onlinepubs/9699919799/xrat/V4_xcu_chap02.html
    381   int RetCode = 128 + Signal;
    382 
    383   // Don't consider a broken pipe as a crash (see clang/lib/Driver/Driver.cpp)
    384   if (Signal == SIGPIPE)
    385     RetCode = EX_IOERR;
    386 
    387   if (CRCI)
    388     const_cast<CrashRecoveryContextImpl *>(CRCI)->HandleCrash(RetCode, Signal);
    389 }
    390 
    391 static void installExceptionOrSignalHandlers() {
    392   // Setup the signal handler.
    393   struct sigaction Handler;
    394   Handler.sa_handler = CrashRecoverySignalHandler;
    395   Handler.sa_flags = 0;
    396   sigemptyset(&Handler.sa_mask);
    397 
    398   for (unsigned i = 0; i != NumSignals; ++i) {
    399     sigaction(Signals[i], &Handler, &PrevActions[i]);
    400   }
    401 }
    402 
    403 static void uninstallExceptionOrSignalHandlers() {
    404   // Restore the previous signal handlers.
    405   for (unsigned i = 0; i != NumSignals; ++i)
    406     sigaction(Signals[i], &PrevActions[i], nullptr);
    407 }
    408 
    409 #endif // !_WIN32
    410 
    411 bool CrashRecoveryContext::RunSafely(function_ref<void()> Fn) {
    412   // If crash recovery is disabled, do nothing.
    413   if (gCrashRecoveryEnabled) {
    414     assert(!Impl && "Crash recovery context already initialized!");
    415     CrashRecoveryContextImpl *CRCI = new CrashRecoveryContextImpl(this);
    416     Impl = CRCI;
    417 
    418     CRCI->ValidJumpBuffer = true;
    419     if (setjmp(CRCI->JumpBuffer) != 0) {
    420       return false;
    421     }
    422   }
    423 
    424   Fn();
    425   return true;
    426 }
    427 
    428 #endif // !_MSC_VER
    429 
    430 LLVM_ATTRIBUTE_NORETURN
    431 void CrashRecoveryContext::HandleExit(int RetCode) {
    432 #if defined(_WIN32)
    433   // SEH and VEH
    434   ::RaiseException(0xE0000000 | RetCode, 0, 0, NULL);
    435 #else
    436   // On Unix we don't need to raise an exception, we go directly to
    437   // HandleCrash(), then longjmp will unwind the stack for us.
    438   CrashRecoveryContextImpl *CRCI = (CrashRecoveryContextImpl *)Impl;
    439   assert(CRCI && "Crash recovery context never initialized!");
    440   CRCI->HandleCrash(RetCode, 0 /*no sig num*/);
    441 #endif
    442   llvm_unreachable("Most likely setjmp wasn't called!");
    443 }
    444 
    445 bool CrashRecoveryContext::throwIfCrash(int RetCode) {
    446 #if defined(_WIN32)
    447   // On Windows, the high bits are reserved for kernel return codes. Values
    448   // starting with 0x80000000 are reserved for "warnings"; values of 0xC0000000
    449   // and up are for "errors". In practice, both are interpreted as a
    450   // non-continuable signal.
    451   unsigned Code = ((unsigned)RetCode & 0xF0000000) >> 28;
    452   if (Code != 0xC && Code != 8)
    453     return false;
    454   ::RaiseException(RetCode, 0, 0, NULL);
    455 #else
    456   // On Unix, signals are represented by return codes of 128 or higher.
    457   // Exit code 128 is a reserved value and should not be raised as a signal.
    458   if (RetCode <= 128)
    459     return false;
    460   llvm::sys::unregisterHandlers();
    461   raise(RetCode - 128);
    462 #endif
    463   return true;
    464 }
    465 
    466 // FIXME: Portability.
    467 static void setThreadBackgroundPriority() {
    468 #ifdef __APPLE__
    469   setpriority(PRIO_DARWIN_THREAD, 0, PRIO_DARWIN_BG);
    470 #endif
    471 }
    472 
    473 static bool hasThreadBackgroundPriority() {
    474 #ifdef __APPLE__
    475   return getpriority(PRIO_DARWIN_THREAD, 0) == 1;
    476 #else
    477   return false;
    478 #endif
    479 }
    480 
    481 namespace {
    482 struct RunSafelyOnThreadInfo {
    483   function_ref<void()> Fn;
    484   CrashRecoveryContext *CRC;
    485   bool UseBackgroundPriority;
    486   bool Result;
    487 };
    488 } // namespace
    489 
    490 static void RunSafelyOnThread_Dispatch(void *UserData) {
    491   RunSafelyOnThreadInfo *Info =
    492     reinterpret_cast<RunSafelyOnThreadInfo*>(UserData);
    493 
    494   if (Info->UseBackgroundPriority)
    495     setThreadBackgroundPriority();
    496 
    497   Info->Result = Info->CRC->RunSafely(Info->Fn);
    498 }
    499 bool CrashRecoveryContext::RunSafelyOnThread(function_ref<void()> Fn,
    500                                              unsigned RequestedStackSize) {
    501   bool UseBackgroundPriority = hasThreadBackgroundPriority();
    502   RunSafelyOnThreadInfo Info = { Fn, this, UseBackgroundPriority, false };
    503   llvm_execute_on_thread(RunSafelyOnThread_Dispatch, &Info,
    504                          RequestedStackSize == 0
    505                              ? llvm::None
    506                              : llvm::Optional<unsigned>(RequestedStackSize));
    507   if (CrashRecoveryContextImpl *CRC = (CrashRecoveryContextImpl *)Impl)
    508     CRC->setSwitchedThread();
    509   return Info.Result;
    510 }
    511