On 2025/11/04 21:18, Paolo Bonzini wrote:
On 11/4/25 09:38, Thomas Huth wrote:
Thread 4 (Thread 0x7f31bd1ff6c0 (LWP 89223) "qemu-system-alp"):
#0 0x00007f31c47150cd in syscall () at /lib64/libc.so.6
#1 0x00005593dd2b578d in qemu_futex_wait (f=0x5593ddad9e50
<rcu_call_ready_event>, val=4294967295) at /home/thuth/devel/qemu/
include/qemu/futex.h:47
#2 0x00005593dd2b59a1 in qemu_event_wait (ev=0x5593ddad9e50
<rcu_call_ready_event>) at ../../home/thuth/devel/qemu/util/event.c:162
#3 0x00005593dd2c12e3 in call_rcu_thread (opaque=0x0) at ../../home/
thuth/devel/qemu/util/rcu.c:304
The RCU thread is simply waiting.
Thread 3 (Thread 0x7f31bc8fd6c0 (LWP 89224) "qemu-system-alp"):
#0 0x00007f31c469c462 in __syscall_cancel_arch () at /lib64/libc.so.6
#1 0x00007f31c469075c in __internal_syscall_cancel () at /lib64/
libc.so.6
#2 0x00007f31c46907a4 in __syscall_cancel () at /lib64/libc.so.6
#3 0x00007f31c470a7c6 in ppoll () at /lib64/libc.so.6
#4 0x00007f31c6916890 in g_main_context_iterate_unlocked.isra () at /
lib64/libglib-2.0.so.0
#5 0x00007f31c6916a4f in g_main_loop_run () at /lib64/libglib-2.0.so.0
#6 0x00005593dd0d1ab0 in iothread_run (opaque=0x559405a567a0)
at ../../ home/thuth/devel/qemu/iothread.c:70
#7 0x00005593dd2b3311 in qemu_thread_start (args=0x559405a571a0)
at ../../home/thuth/devel/qemu/util/qemu-thread-posix.c:393
#8 0x00007f31c4693f54 in start_thread () at /lib64/libc.so.6
#9 0x00007f31c471732c in __clone3 () at /lib64/libc.so.6
This iothread is doing nothing.
Thread 2 (Thread 0x7f3137fff6c0 (LWP 89225) "qemu-system-alp"):
#0 0x00007f31c469c462 in __syscall_cancel_arch () at /lib64/libc.so.6
#1 0x00007f31c469075c in __internal_syscall_cancel () at /lib64/
libc.so.6
#2 0x00007f31c46907a4 in __syscall_cancel () at /lib64/libc.so.6
#3 0x00007f31c470b2be in write () at /lib64/libc.so.6
#4 0x00005593dd2af441 in event_notifier_set (e=0x559405a56a54)
at ../../home/thuth/devel/qemu/util/event_notifier-posix.c:117
#5 0x00005593dd2cdcde in aio_notify (ctx=0x559405a56980) at ../../
home/ thuth/devel/qemu/util/async.c:506
In this backtrace the CPU is waking up the main loop (thread 1), but the
main loop is running so I don't think it's really a deadlock. It's more
likely that the replay is not matching the record, or there's a similar
reason why the replay is not proceeding.
I agree. It is more likely that debugging the replay code instead of the
RCU change will lead to the real cause.
Regards,
Akihiko Odaki
