Message ID | 20130610143842.GB4838@stefanha-thinkpad.redhat.com |
---|---|
State | New |
Headers | show |
Il 10/06/2013 10:38, Stefan Hajnoczi ha scritto: > On Mon, Jun 10, 2013 at 02:25:57PM +0200, Stefan Hajnoczi wrote: >> @@ -1427,26 +1456,18 @@ void bdrv_close_all(void) >> void bdrv_drain_all(void) >> { >> BlockDriverState *bs; >> - bool busy; >> - >> - do { >> - busy = qemu_aio_wait(); >> >> + while (bdrv_requests_pending_all()) { >> /* FIXME: We do not have timer support here, so this is effectively >> * a busy wait. >> */ >> QTAILQ_FOREACH(bs, &bdrv_states, list) { >> if (!qemu_co_queue_empty(&bs->throttled_reqs)) { >> qemu_co_queue_restart_all(&bs->throttled_reqs); >> - busy = true; >> } >> } >> - } while (busy); >> >> - /* If requests are still pending there is a bug somewhere */ >> - QTAILQ_FOREACH(bs, &bdrv_states, list) { >> - assert(QLIST_EMPTY(&bs->tracked_requests)); >> - assert(qemu_co_queue_empty(&bs->throttled_reqs)); >> + qemu_aio_wait(); >> } >> } > > tests/ide-test found an issue here: block.c invokes callbacks from a BH > so we may not yet have completed the request when this loop terminates. > > Kevin: can you fold in this patch? > > diff --git a/block.c b/block.c > index 31f7231..e176215 100644 > --- a/block.c > +++ b/block.c > @@ -1469,6 +1469,9 @@ void bdrv_drain_all(void) > > qemu_aio_wait(); > } > + > + /* Process pending completion BHs */ > + aio_poll(qemu_get_aio_context(), false); > } aio_poll could require multiple iterations, this is why the old code was starting with "busy = qemu_aio_wait()". You can add a while loop around the new call, or do something more similar to the old code, along the lines of "do { QTAILQ_FOREACH(...) ...; busy = bdrv_request_pending_all(); busy |= aio_poll(qemu_get_aio_context(), busy); } while(busy);".
On Thu, Jun 13, 2013 at 10:33:58AM -0400, Paolo Bonzini wrote: > Il 10/06/2013 10:38, Stefan Hajnoczi ha scritto: > > On Mon, Jun 10, 2013 at 02:25:57PM +0200, Stefan Hajnoczi wrote: > >> @@ -1427,26 +1456,18 @@ void bdrv_close_all(void) > >> void bdrv_drain_all(void) > >> { > >> BlockDriverState *bs; > >> - bool busy; > >> - > >> - do { > >> - busy = qemu_aio_wait(); > >> > >> + while (bdrv_requests_pending_all()) { > >> /* FIXME: We do not have timer support here, so this is effectively > >> * a busy wait. > >> */ > >> QTAILQ_FOREACH(bs, &bdrv_states, list) { > >> if (!qemu_co_queue_empty(&bs->throttled_reqs)) { > >> qemu_co_queue_restart_all(&bs->throttled_reqs); > >> - busy = true; > >> } > >> } > >> - } while (busy); > >> > >> - /* If requests are still pending there is a bug somewhere */ > >> - QTAILQ_FOREACH(bs, &bdrv_states, list) { > >> - assert(QLIST_EMPTY(&bs->tracked_requests)); > >> - assert(qemu_co_queue_empty(&bs->throttled_reqs)); > >> + qemu_aio_wait(); > >> } > >> } > > > > tests/ide-test found an issue here: block.c invokes callbacks from a BH > > so we may not yet have completed the request when this loop terminates. > > > > Kevin: can you fold in this patch? > > > > diff --git a/block.c b/block.c > > index 31f7231..e176215 100644 > > --- a/block.c > > +++ b/block.c > > @@ -1469,6 +1469,9 @@ void bdrv_drain_all(void) > > > > qemu_aio_wait(); > > } > > + > > + /* Process pending completion BHs */ > > + aio_poll(qemu_get_aio_context(), false); > > } > > aio_poll could require multiple iterations, this is why the old code was > starting with "busy = qemu_aio_wait()". You can add a while loop around > the new call, or do something more similar to the old code, along the > lines of "do { QTAILQ_FOREACH(...) ...; busy = > bdrv_request_pending_all(); busy |= aio_poll(qemu_get_aio_context(), > busy); } while(busy);". Good idea, the trick is to use busy as the blocking flag. I will resend with the fix to avoid making this email thread too messy. Stefan
diff --git a/block.c b/block.c index 31f7231..e176215 100644 --- a/block.c +++ b/block.c @@ -1469,6 +1469,9 @@ void bdrv_drain_all(void) qemu_aio_wait(); } + + /* Process pending completion BHs */ + aio_poll(qemu_get_aio_context(), false); } /* make a BlockDriverState anonymous by removing from bdrv_state list.