diff options
| author | John Hawthorn <[email protected]> | 2024-08-08 02:10:15 -0700 |
|---|---|---|
| committer | John Hawthorn <[email protected]> | 2024-08-13 11:52:24 -0700 |
| commit | 87a85550edd786665e081b355c6af62c4854b1d7 (patch) | |
| tree | 81b94db34a4716b8b1d4a6aed6b54b6d3f7b0ff5 /thread_pthread.c | |
| parent | 908305a2977856264a932cad02437c30f08c365b (diff) | |
Re-initialize vm->ractor.sched.lock after fork
Previously under certain conditions it was possible to encounter a
deadlock in the forked child process if ractor.sched.lock was held.
Co-authored-by: Nathan Froyd <[email protected]>
Notes
Notes:
Merged: https://github.com/ruby/ruby/pull/11356
Diffstat (limited to 'thread_pthread.c')
| -rw-r--r-- | thread_pthread.c | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/thread_pthread.c b/thread_pthread.c index a4363aec48..3ebb67aaa3 100644 --- a/thread_pthread.c +++ b/thread_pthread.c @@ -1556,6 +1556,7 @@ thread_sched_atfork(struct rb_thread_sched *sched) } vm->ractor.sched.running_cnt = 0; + rb_native_mutex_initialize(&vm->ractor.sched.lock); // rb_native_cond_destroy(&vm->ractor.sched.cond); rb_native_cond_initialize(&vm->ractor.sched.cond); rb_native_cond_initialize(&vm->ractor.sched.barrier_complete_cond); |