ghost
commented
8 months ago Tagging subscribers to this area: @dotnet/gc See info in area-owners.md if you want to be subscribed.
Issue Details
DotNet Runtime: 6.0.21 GC mode: Concurrent workstation vertarget: ```Windows 10 Version 19044 MP (24 procs) Free x64 Product: WinNt, suite: SingleUserTS Edition build lab: 19041.1.amd64fre.vb_release.191206-1406 Debug session time: Fri Jan 5 09:31:06.000 2024 (UTC + 0:00) System Uptime: 5 days 9:58:21.986 Process Uptime: 0 days 4:13:58.000 Kernel time: 0 days 0:00:03.000 User time: 0 days 0:01:44.000 ``` We've started running into sporadic infinite hangs during the initialisation phase of our application. We've taken several dumps spaced apart in time and it seems that each of them are stuck executing GC logic (the stacks I've attached below are the same in each of them). The events I've been able to trace are (please correct me if I've missed something) 1. The allocating thread triggers a GC. 1. The call out to the GC suspends all other managed threads (the managed threads have calls to GCHeap::WaitUntilGCComplete in their stack). 1. The allocating thread progresses to the point where it makes a check to ensure the BGC (background gc) thread has been started (it hasn’t). 1. It then attempts to start the BGC thread waiting for it to initialise with an infinite timeout [CreateSuspendableThread](https://github.com/dotnet/runtime/blob/v6.0.21/src/coreclr/vm/gcenv.ee.cpp#L1425) 1. The BGC thread is started but is stuck inside some calls to the loader that are made as a part of thread initialization (ThreadStartedEvent is therefore never set). 1. All managed threads are then stuck until the BGC thread initialises. 1. Infinite hang Any ideas on what could cause this/ things to investigate? We're looking at upgrading to 6.0.26 but not at a major version upgrade at this time. I'm happy to provide more information from the dump if needed but unfortunately I can't share the file in full. ``` 0:084> kp # Child-SP RetAddr Call Site 00 00000049`099f6938 00007ff9`b30930ce ntdll!NtWaitForSingleObject+0x14 01 00000049`099f6940 00007ff9`52215f2f KERNELBASE!WaitForSingleObjectEx+0x8e 02 00000049`099f69e0 00007ff9`422752e0 mso!OfficeSpace::Root::FCanAskControlUsers+0x7f 03 (Inline Function) --------`-------- coreclr!CLREventWaitHelper2(void)+0x6 [D:\a\_work\1\s\src\coreclr\vm\synch.cpp @ 372] 04 00000049`099f6a30 00007ff9`4238e2fb coreclr!CLREventWaitHelper(void * handle = 0x00000000`00000000, unsigned long dwMilliseconds = 1, int alertable = 0n161442832)+0x20 [D:\a\_work\1\s\src\coreclr\vm\synch.cpp @ 397] 05 (Inline Function) --------`-------- coreclr!CLREventBase::WaitEx(void)+0x10 [D:\a\_work\1\s\src\coreclr\vm\synch.cpp @ 466] 06 (Inline Function) --------`-------- coreclr!CLREventBase::Wait(void)+0x10 [D:\a\_work\1\s\src\coreclr\vm\synch.cpp @ 412] 07 00000049`099f6a90 00007ff9`422daa1f coreclr!`anonymous namespace'::CreateSuspendableThread(
| Author: | Jordan-Osborn |
|---|---|
| Assignees: | - |
| Labels: | `area-GC-coreclr`, `untriaged` |
| Milestone: | - |
jkotas
Jordan-Osborn
dotnet-policy-service[bot]
DotNet Runtime: 6.0.21 GC mode: Concurrent workstation vertarget:
We've started running into sporadic infinite hangs during the initialisation phase of our application. We've taken several dumps spaced apart in time and it seems that each of them are stuck executing GC logic (the stacks I've attached below are the same in each of them).
The events I've been able to trace are (please correct me if I've missed something)
Any ideas on what could cause this/ things to investigate? We're looking at upgrading to 6.0.26 but not at a major version upgrade at this time. I'm happy to provide more information from the dump if needed but unfortunately I can't share the file in full.
.NET BGC thread