Re: [Qemu-devel] [PATCH v2 0/3] AioContext: ctx->dispatching is dead, all hail ctx->notify_me

["Richard W.M. Jones" <rjones@redhat.com>]

On Fri, Jul 17, 2015 at 11:30:38AM +0200, Paolo Bonzini wrote:
> error: kvm run failed Function not implemented
> PC=00000000bf671210  SP=00000000c00001f0
> X00=000000000a003e70 X01=0000000000000000 X02=00000000bf680548 X03=0000000000000030
> X04=00000000bbb5fc18 X05=00000000004b7770 X06=00000000bf721930 X07=000000000000009a
> X08=00000000bf716858 X09=0000000000000090 X10=0000000000000000 X11=0000000000000046
> X12=00000000a007e03a X13=0000000000000000 X14=0000000000000000 X15=0000000000000000
> X16=00000000bf716df0 X17=0000000000000000 X18=0000000000000000 X19=00000000bf6f5f18
> X20=0000000000000000 X21=0000000000000000 X22=0000000000000000 X23=0000000000000000
> X24=0000000000000000 X25=0000000000000000 X26=0000000000000000 X27=0000000000000000
> X28=0000000000000000 X29=0000000000000000 X30=0000000000000000 PSTATE=60000305 (flags -ZC-)

Vaguely reminiscent of this bug:

https://bugzilla.redhat.com/show_bug.cgi?id=1194366

(See comment 7 in particular)

> diff --git a/aio-posix.c b/aio-posix.c
> index 268d14d..d2011d0 100644
> --- a/aio-posix.c
> +++ b/aio-posix.c
> @@ -273,6 +273,13 @@ bool aio_poll(AioContext *ctx, bool blocking)
>          aio_context_acquire(ctx);
>      }
>  
> +    /* This should be optimized... */
> +    event_notifier_test_and_clear(&ctx->notifier);
> +
> +    if (blocking) {
> +        atomic_sub(&ctx->notify_me, 2);
> +    }
> +
>      /* if we have any readable fds, dispatch event */
>      if (ret > 0) {
>          for (i = 0; i < npfd; i++) {
> @@ -283,10 +290,6 @@ bool aio_poll(AioContext *ctx, bool blocking)
>      npfd = 0;
>      ctx->walking_handlers--;
>  
> -    if (blocking) {
> -        atomic_sub(&ctx->notify_me, 2);
> -    }
> -
>      /* Run dispatch even if there were no readable fds to run timers */
>      if (aio_dispatch(ctx)) {
>          progress = true;
> diff --git a/aio-win32.c b/aio-win32.c
> index 2bfd5f8..33809fd 100644
> --- a/aio-win32.c
> +++ b/aio-win32.c
> @@ -326,6 +326,10 @@ bool aio_poll(AioContext *ctx, bool blocking)
>          if (timeout) {
>              aio_context_acquire(ctx);
>          }
> +
> +        /* This should be optimized... */
> +        event_notifier_test_and_clear(&ctx->notifier);
> +
>          if (blocking) {
>              assert(first);
>              atomic_sub(&ctx->notify_me, 2);
> diff --git a/async.c b/async.c
> index 9204907..120e183 100644
> --- a/async.c
> +++ b/async.c
> @@ -202,6 +202,9 @@ aio_ctx_check(GSource *source)
>      AioContext *ctx = (AioContext *) source;
>      QEMUBH *bh;
>  
> +    /* This should be optimized... */
> +    event_notifier_test_and_clear(&ctx->notifier);
> +
>      atomic_and(&ctx->notify_me, ~1);
>      for (bh = ctx->first_bh; bh; bh = bh->next) {
>          if (!bh->deleted && bh->scheduled) {
> @@ -280,6 +280,10 @@ static void aio_rfifolock_cb(void *opaque)
>      aio_notify(opaque);
>  }
>  
> +static void event_notifier_dummy_cb(EventNotifier *e)
> +{
> +}
> +
>  AioContext *aio_context_new(Error **errp)
>  {
>      int ret;
> @@ -292,7 +296,7 @@ AioContext *aio_context_new(Error **errp)
>          return NULL;
>      }
>      g_source_set_can_recurse(&ctx->source, true);
> -    aio_set_event_notifier(ctx, &ctx->notifier, NULL);
> +    aio_set_event_notifier(ctx, &ctx->notifier, event_notifier_dummy_cb);
>      ctx->thread_pool = NULL;
>      qemu_mutex_init(&ctx->bh_lock);
>      rfifolock_init(&ctx->lock, aio_rfifolock_cb, ctx);

With this patch, I've got to 500 iterations without seeing the error.

Still testing it however ...

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://people.redhat.com/~rjones/virt-df/