Bug #21860: Process.fork: the child may deadlock on `th->interrupt_lock` in `threadptr_interrupt_exec_cleanup`

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 3.4.4 (2025-05-14 revision a38531fd3f) +PRISM [aarch64-linux]
• Backport: 3.2: WONTFIX, 3.3: DONE, 3.4: DONE, 4.0: DONTNEED

[byroot (Jean Boussier)]

We recently observed some deadlocked processes. These got deadlock during child initialization right after Process.fork.

Based on the gdb session, the child is deadlocked in threadptr_interrupt_exec_cleanup, which suggest th->interrupt_lock was locked by another thread in the parent, and wasn't reinitialized in the children before calling threadptr_interrupt_exec_cleanup.

(gdb) bt 20 
#0  0x0000ffff855a09c0 in __lll_lock_wait () from /lib64/libc.so.6
#1  0x0000ffff855a6d60 in pthread_mutex_lock@@GLIBC_2.17 () from /lib64/libc.so.6
#2  0x0000aaaaaeec04c8 [PAC] in rb_native_mutex_lock (lock=<optimized out>) at /ruby-3.4.4/thread_pthread.c:116
#3  threadptr_interrupt_exec_cleanup (th=<optimized out>) at thread.c:6052
#4  thread_cleanup_func_before_exec (th_ptr=0xffff35803000) at thread.c:514
#5  thread_cleanup_func (atfork=1, th_ptr=0xffff35803000) at thread.c:524
#6  terminate_atfork_i (current_th=0xffff84e1e000, th=0xffff35803000) at thread.c:4769
#7  rb_thread_atfork_internal (atfork=<optimized out>, th=0xffff84e1e000) at thread.c:4736
#8  rb_thread_atfork () at thread.c:4779
#9  0x0000aaaaaee110fc [PAC] in after_fork_ruby (pid=0) at process.c:1693
#10 rb_fork_ruby (status=status@entry=0x0) at process.c:4253
#11 0x0000aaaaaee11154 [PAC] in proc_fork_pid () at process.c:4266
#12 rb_proc__fork (_obj=<optimized out>) at process.c:4313
#13 0x0000aaaaaeefeb7c [PAC] in vm_call_cfunc_with_frame_ (stack_bottom=0xffff84e7a1c8, argv=0xffff84e7a1d0, argc=0, calling=<optimized out>, reg_cfp=0xffff84f79150, ec=0xffff84e31050) at /ruby-3.4.4/vm_insnhelper.c:3794
#14 vm_call_cfunc_with_frame (ec=0xffff84e31050, reg_cfp=0xffff84f79150, calling=<optimized out>) at /ruby-3.4.4/vm_insnhelper.c:3840
#15 0x0000aaaaaef1aa44 [PAC] in vm_sendish (method_explorer=<optimized out>, block_handler=<optimized out>, cd=<optimized out>, reg_cfp=<optimized out>, ec=<optimized out>) at /ruby-3.4.4/vm_callinfo.h:415
#16 vm_exec_core (ec=0xffff84e31050) at /ruby-3.4.4/insns.def:1063
#17 0x0000aaaaaef0aa58 [PAC] in rb_vm_exec (ec=ec@entry=0xffff84e31050) at vm.c:2595
#18 0x0000aaaaaef0fe48 [PAC] in vm_call0_body (ec=ec@entry=0xffff84e31050, calling=calling@entry=0xffffea5a1e78, argv=argv@entry=0x0) at /ruby-3.4.4/vm_eval.c:225
#19 0x0000aaaaaef136e8 in vm_call0_cc (kw_splat=0, cc=0xfffeee654508, argv=<optimized out>, argc=0, id=27393, recv=281472906794680, ec=0xffff84e31050) at /ruby-3.4.4/vm_eval.c:101

[byroot (Jean Boussier)]

Oh, it has been fixed by https://github.com/ruby/ruby/pull/12981

[byroot (Jean Boussier)]

• Backport changed from 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN, 4.0: UNKNOWN to 3.2: WONTFIX, 3.3: REQUIRED, 3.4: REQUIRED, 4.0: DONTNEED

[byroot (Jean Boussier)]

• Status changed from Open to Closed

3.4 backport PR: https://github.com/ruby/ruby/pull/16060
3.3 backport PR: https://github.com/ruby/ruby/pull/16061

[luke-gru (Luke Gruber)]

• Backport changed from 3.2: WONTFIX, 3.3: REQUIRED, 3.4: REQUIRED, 4.0: DONTNEED to 3.2: WONTFIX, 3.3: DONE, 3.4: REQUIRED, 4.0: DONTNEED

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 3.2: WONTFIX, 3.3: DONE, 3.4: REQUIRED, 4.0: DONTNEED to 3.2: WONTFIX, 3.3: DONE, 3.4: DONE, 4.0: DONTNEED

ruby_3_4: merged at 43771bb0efcd139acd9112a770e8b8d719118dce.