On Wed, Jul 15, 2015 at 07:13:23PM +0200, Paolo Bonzini wrote:
> This patch rewrites the ctx->dispatching optimization, which was the cause
> of some mysterious hangs that could be reproduced on aarch64 KVM only.
> The hangs were indirectly caused by aio_poll() and in particular by
> flash memory u
On 16/07/2015 11:14, Kevin Wolf wrote:
>>> With this information, I understand that what has changed is that the
>>> > > return value of g_main_context_iteration() changes from true before this
>>> > > patch (had the aio_notify() from aio_set_fd_handler() pending) to false
>>> > > after the patch
On Thu, Jul 16, 2015 at 01:29:20PM +0800, Fam Zheng wrote:
> On Wed, 07/15 19:13, Paolo Bonzini wrote:
> > This patch rewrites the ctx->dispatching optimization
>
> I lost track, but what was the justification of this optimization?
To avoid signalling the event notifier when it's not necessary.
Am 16.07.2015 um 00:59 hat Paolo Bonzini geschrieben:
> On 15/07/2015 22:14, Kevin Wolf wrote:
> > > index 233d8f5..ae7c6cf 100644
> > > --- a/aio-win32.c
> > > +++ b/aio-win32.c
> > > @@ -318,11 +313,11 @@ bool aio_poll(AioContext *ctx, bool blocking)
> > > first = true;
> > >
> > > /*
On 16/07/2015 07:29, Fam Zheng wrote:
>> This patch rewrites the ctx->dispatching optimization
>
> I lost track, but what was the justification of this optimization?
qemu_bh_schedule's call to aio_notify were showing in the profile as 20%
or so.
Paolo
> Anyway, awesome debugging and explanat
On Wed, 07/15 19:13, Paolo Bonzini wrote:
> This patch rewrites the ctx->dispatching optimization
I lost track, but what was the justification of this optimization?
Anyway, awesome debugging and explanations!
Fam
On 15/07/2015 22:14, Kevin Wolf wrote:
> > index 233d8f5..ae7c6cf 100644
> > --- a/aio-win32.c
> > +++ b/aio-win32.c
> > @@ -318,11 +313,11 @@ bool aio_poll(AioContext *ctx, bool blocking)
> > first = true;
> >
> > /* wait until next event */
> > -while (count > 0) {
> > +do {
Am 15.07.2015 um 19:13 hat Paolo Bonzini geschrieben:
> This patch rewrites the ctx->dispatching optimization, which was the cause
> of some mysterious hangs that could be reproduced on aarch64 KVM only.
> The hangs were indirectly caused by aio_poll() and in particular by
> flash memory updates's
comments below
On 07/15/15 19:13, Paolo Bonzini wrote:
> This patch rewrites the ctx->dispatching optimization, which was the cause
> of some mysterious hangs that could be reproduced on aarch64 KVM only.
> The hangs were indirectly caused by aio_poll() and in particular by
> flash memory updates'
This patch rewrites the ctx->dispatching optimization, which was the cause
of some mysterious hangs that could be reproduced on aarch64 KVM only.
The hangs were indirectly caused by aio_poll() and in particular by
flash memory updates's call to blk_write(), which invokes aio_poll().
Fun stuff: they
10 matches
Mail list logo
