Re: [PATCH v3 06/33] util/async: aio_co_schedule(): support reschedule in same ctx

[Paolo Bonzini]

On 09/06/21 12:24, Vladimir Sementsov-Ogievskiy wrote:
> Thanks, I'll base v4 of nbd patches on it.
>
> I now run make check. test-aio-multithread crashes on assertion:

With the patch I've sent it doesn't, so hopefully you can go ahead.

Paolo

> (gdb) bt
> #0  0x00007f4af8d839d5 in raise () from /lib64/libc.so.6
> #1  0x00007f4af8d6c8a4 in abort () from /lib64/libc.so.6
> #2  0x00007f4af8d6c789 in __assert_fail_base.cold () from /lib64/libc.so.6
> #3  0x00007f4af8d7c026 in __assert_fail () from /lib64/libc.so.6
> #4  0x000055daebfdab95 in aio_poll (ctx=0x7f4ae0000b60, blocking=true) 
> at ../util/aio-posix.c:567
> #5  0x000055daebea096c in iothread_run (opaque=0x55daed81bc90) at 
> ../tests/unit/iothread.c:91
> #6  0x000055daebfc6c4a in qemu_thread_start (args=0x55daed7d9940) at 
> ../util/qemu-thread-posix.c:521
> #7  0x00007f4af8f1a3f9 in start_thread () from /lib64/libpthread.so.0
> #8  0x00007f4af8e47b53 in clone () from /lib64/libc.so.6
> (gdb) fr 4
> #4  0x000055daebfdab95 in aio_poll (ctx=0x7f4ae0000b60, blocking=true) 
> at ../util/aio-posix.c:567
> 567         assert(in_aio_context_home_thread(ctx == 
> iohandler_get_aio_context() ?
> (gdb) list
> 562          *
> 563          * aio_poll() may only be called in the AioContext's thread. 
> iohandler_ctx
> 564          * is special in that it runs in the main thread, but that 
> thread's context
> 565          * is qemu_aio_context.
> 566          */
> 567         assert(in_aio_context_home_thread(ctx == 
> iohandler_get_aio_context() ?
> 568                                           qemu_get_aio_context() : 
> ctx));
> 569
> 570         qemu_lockcnt_inc(&ctx->list_lock);