Skip to content

Conversation

@zhaomaosu
Copy link
Contributor

By default, address sanitizer will inline call for setting private shadow memory with small size. However, if work group size is too large, the private shadow memory may allocate failed. We need to check if shadow base is null before trying to poison it.

By default, address sanitizer will inline call for setting private
shadow memory with small size. However, if work group size is too large,
the private shadow memory may allocate failed. We need to check if
shadow base is null before trying to poison it.
@zhaomaosu zhaomaosu requested a review from a team as a code owner July 29, 2024 07:55
Copy link
Contributor

@yingcong-wu yingcong-wu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm, but let's wait for another approval.

…ize.cpp

Co-authored-by: Yang Zhao <allanzyne@outlook.com>
@zhaomaosu
Copy link
Contributor Author

Hi @intel/llvm-gatekeepers, this PR is ready to be merged. And pre-ci failures are unrelated to this change.

@zhaomaosu
Copy link
Contributor Author

Hi @intel/llvm-gatekeepers, could you please help merge this PR? Thanks.

@sommerlukas sommerlukas merged commit f203826 into intel:sycl Jul 31, 2024
@zhaomaosu zhaomaosu deleted the devsan-force-outline branch July 31, 2024 07:46
@sommerlukas
Copy link
Contributor

@zhaomaosu This PR failed in post-commit, can you take a look if the failure is related to your changes:
https://github.com/intel/llvm/actions/runs/10175954748/job/28144955957

Specifically:

Timed Out Tests (1):
  SYCL :: Plugin/level_zero_usm_residency.cpp

@zhaomaosu
Copy link
Contributor Author

Hi @sommerlukas, I believe this failure is unrelated to my change. My change only affect device sanitizer. And the failed test doesn't use device sanitizer.

@sommerlukas
Copy link
Contributor

Reported the unrelated failure in #14863

AlexeySachkov pushed a commit to AlexeySachkov/llvm that referenced this pull request Nov 26, 2024
…ry (intel#14818)

By default, address sanitizer will inline call for setting private
shadow memory with small size. However, if work group size is too large,
the private shadow memory may allocate failed. We need to check if
shadow base is null before trying to poison it.

---------

Co-authored-by: Yang Zhao <allanzyne@outlook.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants