| Message ID | e45e00c3d31286c86b76693262266e702ed7f1a3.1598624685.git.josef@toxicpanda.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: hold the uuid_mutex for all close_fs_devices calls | expand |
On 28.08.20 г. 17:24 ч., Josef Bacik wrote: > My recent change to not take the device_list_mutex for closing devices > added a lockdep_assert_held(&uuid_mutex) to close_fs_devices. I then > went and verified all calls had that, except I overlooked > btrfs_close_devices() where we close seed devices. Fix this by holding > the uuid_mutex for this entire operation. > > 20cc6d129252 ("btrfs: do not hold device_list_mutex when closing devices") > Signed-off-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: Nikolay Borisov <nborisov@suse.com> > --- > fs/btrfs/volumes.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c > index 6f489245eec6..3f8bd1af29eb 100644 > --- a/fs/btrfs/volumes.c > +++ b/fs/btrfs/volumes.c > @@ -1183,13 +1183,13 @@ void btrfs_close_devices(struct btrfs_fs_devices *fs_devices) > close_fs_devices(fs_devices); > if (!fs_devices->opened) > list_splice_init(&fs_devices->seed_list, &list); > - mutex_unlock(&uuid_mutex); > > list_for_each_entry_safe(fs_devices, tmp, &list, seed_list) { > close_fs_devices(fs_devices); > list_del(&fs_devices->seed_list); > free_fs_devices(fs_devices); > } > + mutex_unlock(&uuid_mutex); > } > > static int open_fs_devices(struct btrfs_fs_devices *fs_devices, >
On Fri, Aug 28, 2020 at 10:24:57AM -0400, Josef Bacik wrote: > My recent change to not take the device_list_mutex for closing devices > added a lockdep_assert_held(&uuid_mutex) to close_fs_devices. I then > went and verified all calls had that, except I overlooked > btrfs_close_devices() where we close seed devices. Fix this by holding > the uuid_mutex for this entire operation. > > 20cc6d129252 ("btrfs: do not hold device_list_mutex when closing devices") This is still in misc-next so I won't apply it as a separate patch. You can eithere send a v2 of the patch or send it as an "improper" patch with reference to the patch that needs to be updated with the diff.
On 28/8/20 10:24 pm, Josef Bacik wrote: > My recent change to not take the device_list_mutex for closing devices > added a lockdep_assert_held(&uuid_mutex) to close_fs_devices. I then > went and verified all calls had that, except I overlooked > btrfs_close_devices() where we close seed devices. Fix this by holding > the uuid_mutex for this entire operation. > > 20cc6d129252 ("btrfs: do not hold device_list_mutex when closing devices") > Signed-off-by: Josef Bacik <josef@toxicpanda.com> > --- > fs/btrfs/volumes.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c > index 6f489245eec6..3f8bd1af29eb 100644 > --- a/fs/btrfs/volumes.c > +++ b/fs/btrfs/volumes.c > @@ -1183,13 +1183,13 @@ void btrfs_close_devices(struct btrfs_fs_devices *fs_devices) > close_fs_devices(fs_devices); > if (!fs_devices->opened) > list_splice_init(&fs_devices->seed_list, &list); > - mutex_unlock(&uuid_mutex); > > list_for_each_entry_safe(fs_devices, tmp, &list, seed_list) { > close_fs_devices(fs_devices); > list_del(&fs_devices->seed_list); > free_fs_devices(fs_devices); > } > + mutex_unlock(&uuid_mutex); > } > > static int open_fs_devices(struct btrfs_fs_devices *fs_devices, > Reviewed-by: Anand Jain <anand.jain@oracle.com> Thanks, Anand
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 6f489245eec6..3f8bd1af29eb 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1183,13 +1183,13 @@ void btrfs_close_devices(struct btrfs_fs_devices *fs_devices) close_fs_devices(fs_devices); if (!fs_devices->opened) list_splice_init(&fs_devices->seed_list, &list); - mutex_unlock(&uuid_mutex); list_for_each_entry_safe(fs_devices, tmp, &list, seed_list) { close_fs_devices(fs_devices); list_del(&fs_devices->seed_list); free_fs_devices(fs_devices); } + mutex_unlock(&uuid_mutex); } static int open_fs_devices(struct btrfs_fs_devices *fs_devices,
My recent change to not take the device_list_mutex for closing devices added a lockdep_assert_held(&uuid_mutex) to close_fs_devices. I then went and verified all calls had that, except I overlooked btrfs_close_devices() where we close seed devices. Fix this by holding the uuid_mutex for this entire operation. 20cc6d129252 ("btrfs: do not hold device_list_mutex when closing devices") Signed-off-by: Josef Bacik <josef@toxicpanda.com> --- fs/btrfs/volumes.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)