Skip to content

DPL: fix race condition with the delivery of region events #11961

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

Merged
davidrohr merged 2 commits into from
Sep 26, 2023
Merged

DPL: fix race condition with the delivery of region events #11961

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bede0bf
DPL: fix race condition with the delivery of region events
ktf Sep 26, 2023
dddc517
Update Framework/Core/src/DataProcessingDevice.cxx
ktf Sep 26, 2023
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
15 changes: 12 additions & 3 deletions Framework/Core/src/DataProcessingDevice.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -957,9 +957,12 @@ void DataProcessingDevice::InitTask()
LOG(detail) << "ptr: " << info.ptr;
LOG(detail) << "size: " << info.size;
LOG(detail) << "flags: " << info.flags;
context.expectedRegionCallbacks -= 1;
// Now we check for pending events with the mutex,
// so the lines below are atomic.
pendingRegionInfos.push_back(info);
// We always want to handle these on the main loop
context.expectedRegionCallbacks -= 1;
// We always want to handle these on the main loop,
// so we awake it.
ServiceRegistryRef ref{registry};
uv_async_send(ref.get<DeviceState>().awakeMainThread);
});
Expand Down Expand Up @@ -991,11 +994,17 @@ void DataProcessingDevice::InitTask()
// We will get there.
this->fillContext(mServiceRegistry.get<DataProcessorContext>(ServiceRegistry::globalDeviceSalt()), deviceContext);

auto hasPendingEvents = [&mutex = mRegionInfoMutex, &pendingRegionInfos = mPendingRegionInfos](DeviceContext& deviceContext) {
std::lock_guard<std::mutex> lock(mutex);
return (pendingRegionInfos.empty() == false) || deviceContext.expectedRegionCallbacks > 0;
};
/// We now run an event loop also in InitTask. This is needed to:
/// * Make sure region registration callbacks are invoked
/// on the main thread.
/// * Wait for enough callbacks to be delivered before moving to START
while (deviceContext.expectedRegionCallbacks > 0 && uv_run(state.loop, UV_RUN_ONCE)) {
while (hasPendingEvents(deviceContext)) {
// Wait for the callback to signal its done, so that we do not busy wait.
uv_run(state.loop, UV_RUN_ONCE);
// Handle callbacks if any
{
std::lock_guard<std::mutex> lock(mRegionInfoMutex);
Expand Down