Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Report error and abort in CollectiveMutex during stack unwinding #14356

Merged
merged 2 commits into from
Nov 1, 2022
Merged

Report error and abort in CollectiveMutex during stack unwinding #14356

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8695e98
Remove communication during stack unwinding.
gassmoeller Oct 17, 2022
57c6048
Abort when exception is uncaught.
gassmoeller Oct 20, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
48 changes: 48 additions & 0 deletions source/base/mpi.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1078,6 +1078,46 @@ namespace Utilities



namespace internal
{
namespace CollectiveMutexImplementation
{
/**
* Abort, should there be an exception being processed (see the error
* message).
*/
void
check_exception()
{
#ifdef DEAL_II_WITH_MPI
# if __cpp_lib_uncaught_exceptions >= 201411
// std::uncaught_exception() is deprecated in c++17
if (std::uncaught_exceptions() != 0)
# else
if (std::uncaught_exception() == true)
# endif
{
std::cerr
<< "---------------------------------------------------------\n"
<< "An exception was thrown inside a section of the program\n"
<< "guarded by a CollectiveMutex.\n"
<< "Because a CollectiveMutex guards critical communication\n"
<< "handling the exception would likely\n"
<< "deadlock because only the current process is aware of the\n"
<< "exception. To prevent this deadlock, the program will be\n"
<< "aborted.\n"
<< "---------------------------------------------------------"
<< std::endl;

MPI_Abort(MPI_COMM_WORLD, 1);
}
#endif
}
} // namespace CollectiveMutexImplementation
} // namespace internal



CollectiveMutex::CollectiveMutex()
: locked(false)
, request(MPI_REQUEST_NULL)
Expand All @@ -1089,6 +1129,10 @@ namespace Utilities

CollectiveMutex::~CollectiveMutex()
{
// First check if this destructor is called during exception handling
// if so, abort.
internal::CollectiveMutexImplementation::check_exception();

Assert(
!locked,
ExcMessage(
Expand Down Expand Up @@ -1138,6 +1182,10 @@ namespace Utilities
{
(void)comm;

// First check if this function is called during exception handling
// if so, abort. This can happen if a ScopedLock is destroyed.
internal::CollectiveMutexImplementation::check_exception();

Assert(
locked,
ExcMessage(
Expand Down