Use main Ractor EC on threads without EC #4108
Closed
Conversation
This file contains hidden or 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
jhawthorn
force-pushed
the
stackprof_null_ec_crash
branch
from
69b4671 to
5cff4b9
Compare
Previously in threads which weren't created to running Ruby (which includes the timer thread), calling rb_postponed_job_register_one as well as probably similar hooks would cause a segfault due to the now-thread-local EC being NULL. This was particularly a problem for profiling tools (like stackprof), which rely on signal handlers, which may be sent to any thread in the process including those not managed by Ruby. [Fixes #17573] Co-authored-by: John Crepezzi <john.crepezzi@gmail.com>
jhawthorn
force-pushed
the
stackprof_null_ec_crash
branch
from
5cff4b9 to
ffd0431
Compare
XrXr
added a commit
that referenced
this pull request
Nov 23, 2021
After 5680c38, postponed job APIs now expect to be called on native threads not managed by Ruby and handles getting a NULL execution context. However, in debug builds the change runs into an assertion failure with GET_EC() which asserts that EC is non-NULL. Avoid the assertion failure by passing `false` for `expect_ec` instead as the intention is to handle when there is no EC. Add a test from John Crepezzi and John Hawthorn to exercise this situation. See GH-4108 See GH-5094 [Bug #17573] Co-authored-by: John Hawthorn <john@hawthorn.email> Co-authored-by: John Crepezzi <john.crepezzi@gmail.com>
matzbot
pushed a commit
that referenced
this pull request
Nov 23, 2021
…Backport #17573] Use valid `ec` for postponed job. Postponed job can be registered from non-Ruby thread, which means `ec` in TLS can be NULL. In this case, use main thread's `ec` instead. See #4108 and #4336 --- vm_trace.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) Avoid assert failure when NULL EC is expected After 5680c38, postponed job APIs now expect to be called on native threads not managed by Ruby and handles getting a NULL execution context. However, in debug builds the change runs into an assertion failure with GET_EC() which asserts that EC is non-NULL. Avoid the assertion failure by passing `false` for `expect_ec` instead as the intention is to handle when there is no EC. Add a test from John Crepezzi and John Hawthorn to exercise this situation. See GH-4108 See GH-5094 [Bug #17573] Co-authored-by: John Hawthorn <john@hawthorn.email> Co-authored-by: John Crepezzi <john.crepezzi@gmail.com> --- ext/-test-/postponed_job/postponed_job.c | 31 ++++++++++++++++++++++++++ test/-ext-/postponed_job/test_postponed_job.rb | 7 ++++++ vm_trace.c | 2 +- 3 files changed, 39 insertions(+), 1 deletion(-)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Previously in threads which weren't created to running Ruby (which includes the timer thread), calling rb_postponed_job_register_one as well as probably similar hooks would cause a segfault due to the now-thread-local EC being NULL.
This was particularly a problem for profiling tools (like stackprof), which rely on signal handlers, which may be sent to any thread in the process including those not managed by Ruby.