On Thu, May 18, 2017 at 8:57 AM, Kevin Wolf <kw...@redhat.com> wrote: > Am 17.05.2017 um 22:16 hat Eric Blake geschrieben: >> On 05/17/2017 12:09 PM, Stefan Hajnoczi wrote: >> > Calling aio_poll() directly may have been fine previously, but this is >> > the future, man! >> >> lol >> >> > The difference between an aio_poll() loop and >> > BDRV_POLL_WHILE() is that BDRV_POLL_WHILE() releases the AioContext >> > around aio_poll(). >> > >> > This allows the IOThread to run fd handlers or BHs to complete the >> > request. Failure to release the AioContext causes deadlocks. >> > >> > Using BDRV_POLL_WHILE() partially fixes a 'savevm' hang with -object >> > iothread. >> >> I'm surprised at how many separate hangs we actually had! > > How hard would it be to write some test cases for this? Dataplane has > a serious lack of automated testing.
And this hang doesn't even require in-flight guest I/O, so it would be easy to reproduce in qemu-iotests if we add an IOThread mode. Stefan