On Wed, Sep 26, 2018 at 10:23:25 +0200, Paolo Bonzini wrote:
> On 24/09/2018 20:46, Emilio G. Cota wrote:
> > Applying this on my local tree is deadlocking icount, since
> > cpu_update_icount is called from cpu_get_icount_raw_locked:
> >
> > #6 cpu_update_icount (cpu=<optimized out>) at /data/src/qemu/cpus.c:257
> > #7 0x000055a6fbc7ae5c in cpu_get_icount_raw_locked () at
> > /data/src/qemu/cpus.c:271
> > #8 0x000055a6fbc7ae99 in cpu_get_icount_locked () at
> > /data/src/qemu/cpus.c:279
> > #9 0x000055a6fbc7b3ac in cpu_get_icount () at /data/src/qemu/cpus.c:302
> > #10 0x000055a6fc0f3a05 in qemu_clock_get_ns
> > (type=type@entry=QEMU_CLOCK_VIRTUAL) at /data/src/qemu/util/qemu-timer.c:601
> >
> > I am however not sure what Paolo's queued tree looks like, so I
> > might be missing something.
>
> No, you're not missing anything.
>
> Looking at other callers of cpu_update_icount, this should be the fix:
>
> diff --git a/cpus.c b/cpus.c
(snip)
This does indeed fix the deadlock. Feel free to add my
Tested-by: Emilio G. Cota <c...@braap.org>
in the eventual patch.
Thanks,
Emilio
