Skip to content

[rocprofiler-sdk] Fix signal-handler re-entrancy deadlock and enable app-abort testsadd a thread re-entrancy guard #6875

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

Closed
minseobshin11 wants to merge 1 commit into from
Closed

[rocprofiler-sdk] Fix signal-handler re-entrancy deadlock and enable app-abort testsadd a thread re-entrancy guard #6875

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions projects/rocprofiler-sdk/source/lib/rocprofiler-sdk-tool/tool.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3674,6 +3674,32 @@ rocprofv3_error_signal_handler(int signo, siginfo_t* info, void* ucontext)
this_func,
signo);

// Guard against recursive re-entry on the same thread. This occurs when a
// chained signal handler (e.g. one installed by the HSA runtime *after*
// rocprofv3 installed its own handler) invokes the previously-registered
// handler, which is this very function. Without this guard the std::call_once
// below would be re-entered by the same thread while it is still executing
// the call_once callable, which deadlocks (recursive use of std::once_flag is
// undefined behavior). When re-entered, the finalization has already been
// performed (or is in progress) by this thread, so we simply break the chain
// loop and proceed to termination.
static thread_local bool _in_signal_handler = false;
if(_in_signal_handler)
{
ROCP_WARNING << fmt::format(
"[PPID={}][PID={}][TID={}][{}] rocprofv3 re-entered signal handler for {} (chained "
"handler looped back); terminating",
this_ppid,
this_pid,
this_tid,
this_func,
signo);
if(signal_handler_exit) ::quick_exit(signo);
::raise(signo);
return;
}
_in_signal_handler = true;

static auto _once = std::once_flag{};
std::call_once(_once, [&]() {
auto get_children = [&this_pid]() {
Expand Down
4 changes: 1 addition & 3 deletions projects/rocprofiler-sdk/tests/rocprofv3/aborted-app/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,7 @@ string(REPLACE "LD_PRELOAD=" "ROCPROF_PRELOAD=" PRELOAD_ENV
if(ROCPROFILER_MEMCHECK STREQUAL "ThreadSanitizer")
set(IS_DISABLED ON)
else()
# set(IS_DISABLED OFF)
set(IS_DISABLED ON) # this test is currently unstable so we are disabling it
# unconditionally for now
set(IS_DISABLED OFF)
endif()

# app-abort
Expand Down
Loading