Bug #20670: fork deadlocks in child process due to timer thread

• Status: Open
• Assignee: -
• Target version: -
• ruby -v: ruby 3.3.4 (2024-07-09 revision be1089c8ec) [x86_64-linux]
• Backport: 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN

[jhawthorn (John Hawthorn)]

We've been seeing an occasional failure in the Rails CI related to a test which forks and I managed to reduce it to the following reproduction.

Thread.new do
  loop { sleep 0.0001 }
end

1000.times do
  pid = fork{}
  Process.waitpid(pid)
rescue Exception
  Process.kill(:KILL, pid)
  raise
end

This hangs on Ruby 3.3 and HEAD (fairly reliably), but completes always on Ruby 3.2

In a debugger it seems like the timer thread acquires vm->ractor.sched.lock in the parent process just as the process is forking. The child process then ends up stuck inside of thread_sched_atfork when trying to acquire the same lock.

I've opened https://github.com/ruby/ruby/pull/11356 with a fix

[jhawthorn (John Hawthorn)]

• Related to Bug #19395: Process forking within non-main Ractor hits rb_bug() added

[ko1 (Koichi Sasada)]

Thank you!