The following changes since commit c11d61271b9e6e7a1f0479ef1ca8fb55fa457a62:

  Update version for v2.11.0-rc3 release (2017-11-29 17:59:34 +0000)

are available in the git repository at:

  git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git tags/for_upstream

for you to fetch changes up to 75ba2ddb188fa07c3442446766782036e3085cba:

  pc: fix crash on attempted cpu unplug (2017-12-01 19:05:58 +0200)

----------------------------------------------------------------
pc, pci, virtio: fixes for rc3

A bunch of fixes all over the place.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

----------------------------------------------------------------
Chao Gao (1):
      i386/msi: Correct mask of destination ID in MSI address

Greg Kurz (1):
      vhost: fix error check in vhost_verify_ring_mappings()

Igor Mammedov (1):
      pc: fix crash on attempted cpu unplug

Marc-André Lureau (1):
      dump-guest-memory.py: fix No symbol "vmcoreinfo_find"

Maxime Coquelin (2):
      virtio: Add queue interface to restore avail index from vring used index
      vhost: restore avail index from vring used index on disconnection

Prasad J Pandit (1):
      virtio: check VirtQueue Vring object is set

 include/hw/i386/apic-msidef.h |  2 +-
 include/hw/virtio/virtio.h    |  1 +
 hw/i386/pc.c                  |  5 +++++
 hw/virtio/vhost.c             | 10 +++++++---
 hw/virtio/virtio.c            | 24 +++++++++++++++++++++---
 scripts/dump-guest-memory.py  | 12 +++++++-----
 6 files changed, 42 insertions(+), 12 deletions(-)

On 1 December 2017 at 17:08, Michael S. Tsirkin <mst@redhat.com> wrote:
> The following changes since commit c11d61271b9e6e7a1f0479ef1ca8fb55fa457a62:
>
>   Update version for v2.11.0-rc3 release (2017-11-29 17:59:34 +0000)
>
> are available in the git repository at:
>
>   git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git tags/for_upstream
>
> for you to fetch changes up to 75ba2ddb188fa07c3442446766782036e3085cba:
>
>   pc: fix crash on attempted cpu unplug (2017-12-01 19:05:58 +0200)
>
> ----------------------------------------------------------------
> pc, pci, virtio: fixes for rc3
>
> A bunch of fixes all over the place.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>
> ----------------------------------------------------------------

You can't send fixes for rc3 2 days after rc3 has been tagged :-)

> Chao Gao (1):
>       i386/msi: Correct mask of destination ID in MSI address
>
> Greg Kurz (1):
>       vhost: fix error check in vhost_verify_ring_mappings()
>
> Igor Mammedov (1):
>       pc: fix crash on attempted cpu unplug
>
> Marc-André Lureau (1):
>       dump-guest-memory.py: fix No symbol "vmcoreinfo_find"
>
> Maxime Coquelin (2):
>       virtio: Add queue interface to restore avail index from vring used index
>       vhost: restore avail index from vring used index on disconnection
>
> Prasad J Pandit (1):
>       virtio: check VirtQueue Vring object is set

Are any of these so important that we would absolutely refuse
to release without the fixes (ie they justify rolling an rc4
that we would otherwise not have needed) ?

thanks
-- PMM

On Fri, Dec 01, 2017 at 06:05:25PM +0000, Peter Maydell wrote:
> On 1 December 2017 at 17:08, Michael S. Tsirkin <mst@redhat.com> wrote:
> > The following changes since commit c11d61271b9e6e7a1f0479ef1ca8fb55fa457a62:
> >
> >   Update version for v2.11.0-rc3 release (2017-11-29 17:59:34 +0000)
> >
> > are available in the git repository at:
> >
> >   git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git tags/for_upstream
> >
> > for you to fetch changes up to 75ba2ddb188fa07c3442446766782036e3085cba:
> >
> >   pc: fix crash on attempted cpu unplug (2017-12-01 19:05:58 +0200)
> >
> > ----------------------------------------------------------------
> > pc, pci, virtio: fixes for rc3
> >
> > A bunch of fixes all over the place.
> >
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> >
> > ----------------------------------------------------------------
> 
> You can't send fixes for rc3 2 days after rc3 has been tagged :-)

Bad wording. I meant they are for bugs present in rc3.

> > Chao Gao (1):
> >       i386/msi: Correct mask of destination ID in MSI address
> >
> > Greg Kurz (1):
> >       vhost: fix error check in vhost_verify_ring_mappings()
> >
> > Igor Mammedov (1):
> >       pc: fix crash on attempted cpu unplug
> >
> > Marc-André Lureau (1):
> >       dump-guest-memory.py: fix No symbol "vmcoreinfo_find"
> >
> > Maxime Coquelin (2):
> >       virtio: Add queue interface to restore avail index from vring used index
> >       vhost: restore avail index from vring used index on disconnection
> >
> > Prasad J Pandit (1):
> >       virtio: check VirtQueue Vring object is set
> 
> Are any of these so important that we would absolutely refuse
> to release without the fixes (ie they justify rolling an rc4
> that we would otherwise not have needed) ?
> 
> thanks
> -- PMM

The msi one is less important it just happened to be queued a while ago
and I didn't want to rebase all testing. Others are crashers but they
don't affect everyone. So I wouldn't be sure, but there's also a
security fix in there, so yes, I suspect we are better off with rc4, and
if we do I think including others is justified (except maybe the msi
one, if you feel strongly I'll rebase and drop it).

-- 
MST

On 3 December 2017 at 04:56, Michael S. Tsirkin <mst@redhat.com> wrote:
> On Fri, Dec 01, 2017 at 06:05:25PM +0000, Peter Maydell wrote:
>> On 1 December 2017 at 17:08, Michael S. Tsirkin <mst@redhat.com> wrote:
>> > Chao Gao (1):
>> >       i386/msi: Correct mask of destination ID in MSI address
>> >
>> > Greg Kurz (1):
>> >       vhost: fix error check in vhost_verify_ring_mappings()
>> >
>> > Igor Mammedov (1):
>> >       pc: fix crash on attempted cpu unplug
>> >
>> > Marc-André Lureau (1):
>> >       dump-guest-memory.py: fix No symbol "vmcoreinfo_find"
>> >
>> > Maxime Coquelin (2):
>> >       virtio: Add queue interface to restore avail index from vring used index
>> >       vhost: restore avail index from vring used index on disconnection
>> >
>> > Prasad J Pandit (1):
>> >       virtio: check VirtQueue Vring object is set
>>
>> Are any of these so important that we would absolutely refuse
>> to release without the fixes (ie they justify rolling an rc4
>> that we would otherwise not have needed) ?

> The msi one is less important it just happened to be queued a while ago
> and I didn't want to rebase all testing. Others are crashers but they
> don't affect everyone. So I wouldn't be sure, but there's also a
> security fix in there, so yes, I suspect we are better off with rc4, and
> if we do I think including others is justified (except maybe the msi
> one, if you feel strongly I'll rebase and drop it).

The bar has to be set quite high here, because if it turns out
that there are problems with a bug fix then we are out of time
to rework or revert it. And the more fixes we throw in at the
last minute, even if they're individually simple, the more
likely that one of them turns out to have unexpected consequences.

So: were any of these bugs present in the 2.10 release? If so,
that strongly argues for not trying to fix them at this point.

With all of these patches plus David Gibson's, that would be
10 new patches in rc4. That is definitely more than would be
ideal.

thanks
-- PMM

On 4 December 2017 at 10:16, Peter Maydell <peter.maydell@linaro.org> wrote:
> On 3 December 2017 at 04:56, Michael S. Tsirkin <mst@redhat.com> wrote:
>> On Fri, Dec 01, 2017 at 06:05:25PM +0000, Peter Maydell wrote:
>>> Are any of these so important that we would absolutely refuse
>>> to release without the fixes (ie they justify rolling an rc4
>>> that we would otherwise not have needed) ?
>
>> The msi one is less important it just happened to be queued a while ago
>> and I didn't want to rebase all testing. Others are crashers but they
>> don't affect everyone. So I wouldn't be sure, but there's also a
>> security fix in there, so yes, I suspect we are better off with rc4, and
>> if we do I think including others is justified (except maybe the msi
>> one, if you feel strongly I'll rebase and drop it).
>
> The bar has to be set quite high here, because if it turns out
> that there are problems with a bug fix then we are out of time
> to rework or revert it. And the more fixes we throw in at the
> last minute, even if they're individually simple, the more
> likely that one of them turns out to have unexpected consequences.
>
> So: were any of these bugs present in the 2.10 release? If so,
> that strongly argues for not trying to fix them at this point.
>
> With all of these patches plus David Gibson's, that would be
> 10 new patches in rc4. That is definitely more than would be
> ideal.

After some discussion on IRC we decided that these should go in.

Thanks for the further information; pull applied to master.

-- PMM

On Mon, Dec 04, 2017 at 10:16:08AM +0000, Peter Maydell wrote:
> On 3 December 2017 at 04:56, Michael S. Tsirkin <mst@redhat.com> wrote:
> > On Fri, Dec 01, 2017 at 06:05:25PM +0000, Peter Maydell wrote:
> >> On 1 December 2017 at 17:08, Michael S. Tsirkin <mst@redhat.com> wrote:
> >> > Chao Gao (1):
> >> >       i386/msi: Correct mask of destination ID in MSI address
> >> >
> >> > Greg Kurz (1):
> >> >       vhost: fix error check in vhost_verify_ring_mappings()
> >> >
> >> > Igor Mammedov (1):
> >> >       pc: fix crash on attempted cpu unplug
> >> >
> >> > Marc-André Lureau (1):
> >> >       dump-guest-memory.py: fix No symbol "vmcoreinfo_find"
> >> >
> >> > Maxime Coquelin (2):
> >> >       virtio: Add queue interface to restore avail index from vring used index
> >> >       vhost: restore avail index from vring used index on disconnection
> >> >
> >> > Prasad J Pandit (1):
> >> >       virtio: check VirtQueue Vring object is set
> >>
> >> Are any of these so important that we would absolutely refuse
> >> to release without the fixes (ie they justify rolling an rc4
> >> that we would otherwise not have needed) ?
> 
> > The msi one is less important it just happened to be queued a while ago
> > and I didn't want to rebase all testing. Others are crashers but they
> > don't affect everyone. So I wouldn't be sure, but there's also a
> > security fix in there, so yes, I suspect we are better off with rc4, and
> > if we do I think including others is justified (except maybe the msi
> > one, if you feel strongly I'll rebase and drop it).
> 
> The bar has to be set quite high here, because if it turns out
> that there are problems with a bug fix then we are out of time
> to rework or revert it. And the more fixes we throw in at the
> last minute, even if they're individually simple, the more
> likely that one of them turns out to have unexpected consequences.
> 
> So: were any of these bugs present in the 2.10 release? If so,
> that strongly argues for not trying to fix them at this point.
> 
> With all of these patches plus David Gibson's, that would be
> 10 new patches in rc4. That is definitely more than would be
> ideal.
> 
> thanks
> -- PMM

None of these are regressions. Except for the msix one, they
do fix major bugs for features new in this release though.
I'll resend it without msix, pls go ahead and ignore if
you think a .1 stable release is a better place to put them.

-- 
MST