accidental startup deadlock caught in gdb #59538
Description
on 7de5585
(gdb) thread apply all bt
Thread 4 (Thread 0x7f01bc2cd640 (LWP 791445) "julia"):
#0 0x00007f01c304c646 in julia_lock_2766 () at locks-mt.jl:81
#1 0x00007f01c3363723 in lock () at condition.jl:75
#2 julia_init_perthread_2775 () at lock.jl:883
#3 0x00007f01c391215e in julia_getindex_2768 () at lock.jl:864
#4 0x00007f01c36a4a40 in workqueue_for () at task.jl:923
#5 julia_enq_work_2734 () at task.jl:947
#6 0x00007f01c2b1d47c in julia_#schedule#597_2726 () at task.jl:1027
#7 0x00007f01c39dffe0 in schedule () at task.jl:1014
#8 julia_notify_2722 () at condition.jl:165
#9 0x00007f01c396ed89 in julia_#notify#402_2805 () at condition.jl:159
#10 0x00007f01c33db396 in julia_notify_2802 () at condition.jl:159
#11 0x00007f01c3363de6 in notify () at condition.jl:159
#12 julia_init_perthread_2775 () at lock.jl:927
#13 0x00007f01c391215e in julia_getindex_2768 () at lock.jl:864
#14 0x00007f01c35929d3 in workqueue_for () at task.jl:923
#15 julia_wait_2815 () at task.jl:1191
#16 0x00007f01c33c2525 in julia_task_done_hook_66526 () at task.jl:847
#17 0x00007f01c369a8bf in jfptr_task_done_hook_66527 () from /home/vtjnash/julia-globals-invokelatest/usr/lib/julia/sys.so
#18 0x00007f01d3ddc78c in jl_apply (args=0x7f01bc2ccce0, nargs=2) at /workspace/src/julia.h:2370
#19 jl_finish_task (ct=ct@entry=0x7f01c1e08010) at /workspace/src/task.c:349
#20 0x00007f01d3e134d9 in jl_threadfun (arg=0x823680) at /workspace/src/scheduler.c:122
#21 0x00007f01d4868ac3 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#22 0x00007f01d48fa850 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81
Thread 3 (Thread 0x7f01c2a0e640 (LWP 791419) "julia"):
#0 0x00007f01d481721a in __GI___sigtimedwait (set=set@entry=0x7f01c2a0d970, info=info@entry=0x7f01c2a0d9f0, timeout=timeout@entry=0x0) at ../sysdeps/unix/sysv/linux/sigtimedwait.c:61
#1 0x00007f01d481729b in __GI___sigwaitinfo (set=set@entry=0x7f01c2a0d970, info=info@entry=0x7f01c2a0d9f0) at ../sysdeps/unix/sysv/linux/sigwaitinfo.c:25
#2 0x00007f01d3e1f9db in signal_listener (arg=<optimized out>) at /workspace/src/signals-unix.c:976
#3 0x00007f01d4868ac3 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#4 0x00007f01d48fa850 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81
Thread 2 (Thread 0x7f01d47afd00 (LWP 791418) "iou-sqp-791416"):
#0 0x0000000000000000 in ?? ()
Backtrace stopped: Cannot access memory at address 0x0
Thread 1 (Thread 0x7f01d47afd00 (LWP 791416) "julia"):
#0 0x00007f01c304c646 in julia_lock_2766 () at locks-mt.jl:81
#1 0x00007f01c3363723 in lock () at condition.jl:75
#2 julia_init_perthread_2775 () at lock.jl:883
#3 0x00007f01c391215e in julia_getindex_2768 () at lock.jl:864
#4 0x00007f01c35929d3 in workqueue_for () at task.jl:923
#5 julia_wait_2815 () at task.jl:1191
#6 0x00007f01c2bca791 in julia_#wait#401_2808 () at condition.jl:141
#7 0x00007f01c3363b5b in wait () at condition.jl:136
#8 julia_init_perthread_2775 () at lock.jl:904
#9 0x00007f01c391215e in julia_getindex_2768 () at lock.jl:864
#10 0x00007f01c36a4968 in workqueue_for () at task.jl:923
#11 julia_enq_work_2734 () at task.jl:956
#12 0x00007f01c2b197f6 in schedule () at task.jl:970
#13 macro expansion () at threadingconstructs.jl:522
#14 julia_start_profile_listener_41336 () at Base.jl:354
#15 0x00007f01c363147a in julia___init___41279 () at Base.jl:391
#16 0x00007f01c3a1e8f9 in jfptr___init___41280 () from /home/vtjnash/julia-globals-invokelatest/usr/lib/julia/sys.so
#17 0x00007f01d3df3b3e in jl_apply (args=0x7fff64b83c68, nargs=1) at /workspace/src/julia.h:2370
#18 jl_module_run_initializer (m=0x7f01c8e6c3a0 <jl_system_image_data+85615520>) at /workspace/src/toplevel.c:68
#19 0x00007f01d3dda998 in _finish_jl_init_ (sysimage=..., ct=<optimized out>, ptls=<optimized out>) at /workspace/src/init.c:632
#20 0x00007f01d3ddbbab in ijl_init_ (sysimage=...) at /workspace/src/init.c:784
#21 0x00007f01d3e1de12 in jl_repl_entrypoint (argc=<optimized out>, argv=0x7fff64b84058) at /workspace/src/jlapi.c:1131
#22 0x0000000000401079 in main (argc=<optimized out>, argv=<optimized out>) at /workspace/cli/loader_exe.c:58
Base.Workqueue = Base.IntrusiveLinkedListSynchronized{Task}(queue=Base.IntrusiveLinkedList{Task}(head=nothing, tail=nothing), lock=Base.Threads.SpinLock(owned=0))
Base.Workqueues = Base.OncePerThread{Base.IntrusiveLinkedListSynchronized{Task}, Type{Base.IntrusiveLinkedListSynchronized{Task}}}(xs=GenericMemory{:atomic, Base.IntrusiveLinkedListSynchronized{Task}, Core.AddrSpace{Core}(0x00)}(2, 0x7f01c1de4020)[
#<null>,
Base.IntrusiveLinkedListSynchronized{Task}(queue=Base.IntrusiveLinkedList{Task}(head=nothing, tail=nothing), lock=Base.Threads.SpinLock(owned=0))], ss=GenericMemory{:atomic, UInt8, Core.AddrSpace{Core}(0x00)}(2, 0x7f01c1de0020)[0x00, 0x01], initializer=Base.IntrusiveLinkedListSynchronized{Task})
Base.PerThreadLock = Base.GenericCondition{Base.Threads.SpinLock}(waitq=Base.IntrusiveLinkedList{Task}(head=nothing, tail=nothing), lock=Base.Threads.SpinLock(owned=1))
Note the deadlock loop here workqueue_for => notify => workqueue_for. Seems like we need a wait/notify for ThreadSynchronizer that keeps the spin lock behavior, since the normal behavior of accessing the scheduler may de-synchronize the threads instead