Re: [Qemu-devel] coroutines: block: Co-routine re-entered recursively when migrating disk with iothreads

[Fam Zheng]

On Tue, 05/24 11:05, Jason J. Herne wrote:
> Thread 13 (Thread 0x3ff989ff910 (LWP 29452)):
> #0  0x000003ff99abe2c0 in raise () from /lib64/libc.so.6
> #1  0x000003ff99abfc26 in abort () from /lib64/libc.so.6
> #2  0x0000000080427d80 in qemu_coroutine_enter (co=0x9c5a4120, opaque=0x0)
> at /root/kvmdev/qemu/util/qemu-coroutine.c:112
> #3  0x000000008032246e in nbd_restart_write (opaque=0x9c5897b0) at
> /root/kvmdev/qemu/block/nbd-client.c:114
> #4  0x00000000802b3a1c in aio_dispatch (ctx=0x9c530770) at
> /root/kvmdev/qemu/aio-posix.c:341
> #5  0x00000000802b4332 in aio_poll (ctx=0x9c530770, blocking=true) at
> /root/kvmdev/qemu/aio-posix.c:479
> #6  0x0000000080155aba in iothread_run (opaque=0x9c530200) at
> /root/kvmdev/qemu/iothread.c:46
> #7  0x000003ff99c87c2c in start_thread () from /lib64/libpthread.so.0
> #8  0x000003ff99b8ec9a in thread_start () from /lib64/libc.so.6

This is the continuation of write request to the NBD target....

> Thread 1 (Thread 0x3ff9a6f2a90 (LWP 29433)):
> #0  0x000003ff99c8d68a in pthread_cond_wait@@GLIBC_2.3.2 () from
> /lib64/libpthread.so.0
> #1  0x000000008040932e in qemu_cond_wait (cond=0x9c530800, mutex=0x9c5307d0)
> at /root/kvmdev/qemu/util/qemu-thread-posix.c:123
> #2  0x0000000080426a38 in rfifolock_lock (r=0x9c5307d0) at
> /root/kvmdev/qemu/util/rfifolock.c:59
> #3  0x00000000802a1f72 in aio_context_acquire (ctx=0x9c530770) at
> /root/kvmdev/qemu/async.c:373
> #4  0x00000000802b3f54 in aio_poll (ctx=0x9c530770, blocking=true) at
> /root/kvmdev/qemu/aio-posix.c:415
> #5  0x000000008031e7ac in bdrv_flush (bs=0x9c59b5c0) at
> /root/kvmdev/qemu/block/io.c:2470
> #6  0x00000000802a8e6e in bdrv_close (bs=0x9c59b5c0) at
> /root/kvmdev/qemu/block.c:2134
> #7  0x00000000802a9966 in bdrv_delete (bs=0x9c59b5c0) at
> /root/kvmdev/qemu/block.c:2341
> #8  0x00000000802ac7c6 in bdrv_unref (bs=0x9c59b5c0) at
> /root/kvmdev/qemu/block.c:3376
> #9  0x0000000080315340 in mirror_exit (job=0x9c956ed0, opaque=0x9c9570d0) at
> /root/kvmdev/qemu/block/mirror.c:494
> #10 0x00000000802afb52 in block_job_defer_to_main_loop_bh
> (opaque=0x9c90dc10) at /root/kvmdev/qemu/blockjob.c:476

... while this is the completion of mirror. They are not supposed to happen
together. Either the job is completed too early, or the nbd_restart_write
function is invoked incorrectly.

I'll see if I can reproduce it here.

Fam