| Message ID | 20200601071956.18006-1-chen.zhang@intel.com |
|---|---|
| State | New |
| Headers | show |
| Series | block/quorum.c: Decrease child index when del_child | expand |
On Mon 01 Jun 2020 09:19:56 AM CEST, Zhang Chen wrote: > From: Zhang Chen <chen.zhang@intel.com> > > Fix this bug: > colo: Can not recover colo after svm failover twice > https://bugs.launchpad.net/bugs/1881231 > > The child index still be hold when it be deleted, the max num is 32. > > Reported-by: Ye.Zou <ye.zou@zstack.io> > Signed-off-by: Zhang Chen <chen.zhang@intel.com> > --- > block/quorum.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/block/quorum.c b/block/quorum.c > index 7cf7ab1546..f71bd4e19d 100644 > --- a/block/quorum.c > +++ b/block/quorum.c > @@ -1099,6 +1099,7 @@ static void quorum_del_child(BlockDriverState *bs, BdrvChild *child, > (s->num_children - i - 1) * sizeof(BdrvChild *)); > s->children = g_renew(BdrvChild *, s->children, --s->num_children); > bdrv_unref_child(bs, child); > + s->next_child_index--; > > bdrv_drained_end(bs); > } As I explained a few weeks ago this patch is not correct. quorum_del_child() allows you to remove any child from the Quorum device, so nothing guarantees that next_child_index-1 is free. https://lists.gnu.org/archive/html/qemu-block/2020-05/msg00634.html Berto
On Mon, 01 Jun 2020 12:38:32 +0200 Alberto Garcia <berto@igalia.com> wrote: > As I explained a few weeks ago this patch is not correct. > quorum_del_child() allows you to remove any child from the Quorum > device, so nothing guarantees that next_child_index-1 is free. > > https://lists.gnu.org/archive/html/qemu-block/2020-05/msg00634.html > > Berto > Hi, Did you have a look at my series? There it's fixed properly: https://lore.kernel.org/qemu-devel/9df6b3723ec30cb749ceaa555d82a29a6d79496d.1589199922.git.lukasstraub2@web.de/ Regards, Lukas Straub
On Mon 01 Jun 2020 08:12:00 PM CEST, Lukas Straub wrote: >> As I explained a few weeks ago this patch is not correct. >> quorum_del_child() allows you to remove any child from the Quorum >> device, so nothing guarantees that next_child_index-1 is free. >> >> https://lists.gnu.org/archive/html/qemu-block/2020-05/msg00634.html >> > Hi, > Did you have a look at my series? There it's fixed properly: > https://lore.kernel.org/qemu-devel/9df6b3723ec30cb749ceaa555d82a29a6d79496d.1589199922.git.lukasstraub2@web.de/ Oh, I somehow overlooked that, thanks for the reminder! Berto
Oh, I missed the patch detail. I just reviewed overall view on your series. Looks your patch is good for me. Thanks Zhang Chen > -----Original Message----- > From: Lukas Straub <lukasstraub2@web.de> > Sent: Tuesday, June 2, 2020 2:12 AM > To: Alberto Garcia <berto@igalia.com> > Cc: Zhang, Chen <chen.zhang@intel.com>; Kevin Wolf <kwolf@redhat.com>; > Max Reitz <mreitz@redhat.com>; qemu-dev <qemu-devel@nongnu.org>; > Jason Wang <jasowang@redhat.com>; Zhanghailiang > <zhang.zhanghailiang@huawei.com>; Dr . David Alan Gilbert > <dgilbert@redhat.com>; Zhang Chen <zhangckid@gmail.com> > Subject: Re: [PATCH] block/quorum.c: Decrease child index when del_child > > On Mon, 01 Jun 2020 12:38:32 +0200 > Alberto Garcia <berto@igalia.com> wrote: > > As I explained a few weeks ago this patch is not correct. > > quorum_del_child() allows you to remove any child from the Quorum > > device, so nothing guarantees that next_child_index-1 is free. > > > > https://lists.gnu.org/archive/html/qemu-block/2020-05/msg00634.html > > > > Berto > > > > Hi, > Did you have a look at my series? There it's fixed properly: > https://lore.kernel.org/qemu- > devel/9df6b3723ec30cb749ceaa555d82a29a6d79496d.1589199922.git.lukasstr > aub2@web.de/ > > Regards, > Lukas Straub
diff --git a/block/quorum.c b/block/quorum.c index 7cf7ab1546..f71bd4e19d 100644 --- a/block/quorum.c +++ b/block/quorum.c @@ -1099,6 +1099,7 @@ static void quorum_del_child(BlockDriverState *bs, BdrvChild *child, (s->num_children - i - 1) * sizeof(BdrvChild *)); s->children = g_renew(BdrvChild *, s->children, --s->num_children); bdrv_unref_child(bs, child); + s->next_child_index--; bdrv_drained_end(bs); }