On 2018-08-13 16:43, Paolo Bonzini wrote: > On 13/08/2018 04:19, Max Reitz wrote: >> Signed-off-by: Max Reitz <mre...@redhat.com> > > Locking AioContext should not be needed anywhere here. mirror_run is > called via aio_co_enter or aio_co_wake, so the lock is actually already > taken every time you call aio_context_acquire. > > It was needed only because AIO callbacks do *not* take the lock, but > it's not needed anymore since the conversion to coroutines.
Uh, nice. Thanks for letting me know! :-) I'll drop the locks in v2. Max
signature.asc
Description: OpenPGP digital signature