| Message ID | 20190731084655.7024-5-jasowang@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Fixes for metadata accelreation | expand |
On Wed, Jul 31, 2019 at 04:46:50AM -0400, Jason Wang wrote: > The vhost_set_vring_num_addr() could be called in the middle of > invalidate_range_start() and invalidate_range_end(). If we don't reset > invalidate_count after the un-registering of MMU notifier, the > invalidate_cont will run out of sync (e.g never reach zero). This will > in fact disable the fast accessor path. Fixing by reset the count to > zero. > > Reported-by: Michael S. Tsirkin <mst@redhat.com> Did Michael report this as well? > Fixes: 7f466032dc9e ("vhost: access vq metadata through kernel virtual address") > Signed-off-by: Jason Wang <jasowang@redhat.com> > drivers/vhost/vhost.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c > index 2a3154976277..2a7217c33668 100644 > +++ b/drivers/vhost/vhost.c > @@ -2073,6 +2073,10 @@ static long vhost_vring_set_num_addr(struct vhost_dev *d, > d->has_notifier = false; > } > > + /* reset invalidate_count in case we are in the middle of > + * invalidate_start() and invalidate_end(). > + */ > + vq->invalidate_count = 0; > vhost_uninit_vq_maps(vq); > #endif >
On 2019/7/31 下午8:41, Jason Gunthorpe wrote: > On Wed, Jul 31, 2019 at 04:46:50AM -0400, Jason Wang wrote: >> The vhost_set_vring_num_addr() could be called in the middle of >> invalidate_range_start() and invalidate_range_end(). If we don't reset >> invalidate_count after the un-registering of MMU notifier, the >> invalidate_cont will run out of sync (e.g never reach zero). This will >> in fact disable the fast accessor path. Fixing by reset the count to >> zero. >> >> Reported-by: Michael S. Tsirkin <mst@redhat.com> > Did Michael report this as well? Correct me if I was wrong. I think it's point 4 described in https://lkml.org/lkml/2019/7/21/25. Thanks > >> Fixes: 7f466032dc9e ("vhost: access vq metadata through kernel virtual address") >> Signed-off-by: Jason Wang <jasowang@redhat.com> >> drivers/vhost/vhost.c | 4 ++++ >> 1 file changed, 4 insertions(+) >> >> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c >> index 2a3154976277..2a7217c33668 100644 >> +++ b/drivers/vhost/vhost.c >> @@ -2073,6 +2073,10 @@ static long vhost_vring_set_num_addr(struct vhost_dev *d, >> d->has_notifier = false; >> } >> >> + /* reset invalidate_count in case we are in the middle of >> + * invalidate_start() and invalidate_end(). >> + */ >> + vq->invalidate_count = 0; >> vhost_uninit_vq_maps(vq); >> #endif >>
On Wed, Jul 31, 2019 at 09:29:28PM +0800, Jason Wang wrote: > > On 2019/7/31 下午8:41, Jason Gunthorpe wrote: > > On Wed, Jul 31, 2019 at 04:46:50AM -0400, Jason Wang wrote: > > > The vhost_set_vring_num_addr() could be called in the middle of > > > invalidate_range_start() and invalidate_range_end(). If we don't reset > > > invalidate_count after the un-registering of MMU notifier, the > > > invalidate_cont will run out of sync (e.g never reach zero). This will > > > in fact disable the fast accessor path. Fixing by reset the count to > > > zero. > > > > > > Reported-by: Michael S. Tsirkin <mst@redhat.com> > > Did Michael report this as well? > > > Correct me if I was wrong. I think it's point 4 described in > https://lkml.org/lkml/2019/7/21/25. I'm not sure what that is talking about But this fixes what I described: https://lkml.org/lkml/2019/7/22/554 Jason
On Wed, Jul 31, 2019 at 04:32:52PM -0300, Jason Gunthorpe wrote: > On Wed, Jul 31, 2019 at 09:29:28PM +0800, Jason Wang wrote: > > > > On 2019/7/31 下午8:41, Jason Gunthorpe wrote: > > > On Wed, Jul 31, 2019 at 04:46:50AM -0400, Jason Wang wrote: > > > > The vhost_set_vring_num_addr() could be called in the middle of > > > > invalidate_range_start() and invalidate_range_end(). If we don't reset > > > > invalidate_count after the un-registering of MMU notifier, the > > > > invalidate_cont will run out of sync (e.g never reach zero). This will > > > > in fact disable the fast accessor path. Fixing by reset the count to > > > > zero. > > > > > > > > Reported-by: Michael S. Tsirkin <mst@redhat.com> > > > Did Michael report this as well? > > > > > > Correct me if I was wrong. I think it's point 4 described in > > https://lkml.org/lkml/2019/7/21/25. > > I'm not sure what that is talking about > > But this fixes what I described: > > https://lkml.org/lkml/2019/7/22/554 > > Jason These are two reasons for a possible counter imbalance. Unsurprisingly they are both fixed if you reset the counter to 0.
On 2019/8/1 上午3:32, Jason Gunthorpe wrote: > On Wed, Jul 31, 2019 at 09:29:28PM +0800, Jason Wang wrote: >> On 2019/7/31 下午8:41, Jason Gunthorpe wrote: >>> On Wed, Jul 31, 2019 at 04:46:50AM -0400, Jason Wang wrote: >>>> The vhost_set_vring_num_addr() could be called in the middle of >>>> invalidate_range_start() and invalidate_range_end(). If we don't reset >>>> invalidate_count after the un-registering of MMU notifier, the >>>> invalidate_cont will run out of sync (e.g never reach zero). This will >>>> in fact disable the fast accessor path. Fixing by reset the count to >>>> zero. >>>> >>>> Reported-by: Michael S. Tsirkin <mst@redhat.com> >>> Did Michael report this as well? >> >> Correct me if I was wrong. I think it's point 4 described in >> https://lkml.org/lkml/2019/7/21/25. > I'm not sure what that is talking about > > But this fixes what I described: > > https://lkml.org/lkml/2019/7/22/554 > > Jason I'm sorry I miss this, will add your name as reported-by in the next version. Thanks
diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index 2a3154976277..2a7217c33668 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -2073,6 +2073,10 @@ static long vhost_vring_set_num_addr(struct vhost_dev *d, d->has_notifier = false; } + /* reset invalidate_count in case we are in the middle of + * invalidate_start() and invalidate_end(). + */ + vq->invalidate_count = 0; vhost_uninit_vq_maps(vq); #endif
The vhost_set_vring_num_addr() could be called in the middle of invalidate_range_start() and invalidate_range_end(). If we don't reset invalidate_count after the un-registering of MMU notifier, the invalidate_cont will run out of sync (e.g never reach zero). This will in fact disable the fast accessor path. Fixing by reset the count to zero. Reported-by: Michael S. Tsirkin <mst@redhat.com> Fixes: 7f466032dc9e ("vhost: access vq metadata through kernel virtual address") Signed-off-by: Jason Wang <jasowang@redhat.com> --- drivers/vhost/vhost.c | 4 ++++ 1 file changed, 4 insertions(+)