You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
It looks like the issue with the previous handling of stack overflow was that the alternate stack was too small. The guard page was included in the size of the stack when it should have been added to the size of the stack. This allocates an additional page for the alternate stack and seems to fix the issue for me.
I wasn't able to reproduce the issues in local builds, but I downloaded the exes from the failing CI tests and they segfaulted every 10 runs or so. After editing the binary to allocate more for the signal stack, I didn't see any more segfaults.
It looked like the issue was that we would hit a segfault and enter the SIGSEGV handler right as a GC pause was sending SIG34. Control would go to the GC signal handler first and would mess with the stack (I couldn't figure out exactly what it did though) and when control returned to the SIGSEGV handler, the first pushq in the method prelude caused another segfault and the program crashed.
I think that we cannot just run the hardware exception handler on the alternate stack like we do now. In coreclr, we switch off the alternate stack as soon as we figure out it is not a stack overflow. Here we don't, so if the catch handler ends up calling a deep call chain, we would get stack overflow.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
ghost
closed this
It looks like the issue with the previous handling of stack overflow was that the alternate stack was too small. The guard page was included in the size of the stack when it should have been added to the size of the stack. This allocates an additional page for the alternate stack and seems to fix the issue for me.
I wasn't able to reproduce the issues in local builds, but I downloaded the exes from the failing CI tests and they segfaulted every 10 runs or so. After editing the binary to allocate more for the signal stack, I didn't see any more segfaults.
It looked like the issue was that we would hit a segfault and enter the SIGSEGV handler right as a GC pause was sending SIG34. Control would go to the GC signal handler first and would mess with the stack (I couldn't figure out exactly what it did though) and when control returned to the SIGSEGV handler, the first
pushqin the method prelude caused another segfault and the program crashed.