| Message ID | 20221221145433.254805-1-lstoakes@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] virtio-blk: avoid kernel panic on VIRTIO_BLK_F_ZONED check | expand |
On Wed, Dec 21, 2022 at 02:54:33PM +0000, Lorenzo Stoakes wrote: > virtio zoned block device support is added by commit 0562d7bf1604 > ("virtio-blk: add support for zoned block devices") which adds > VIRTIO_BLK_F_ZONED to the features array in virtio_blk.c but makes it > conditional on CONFIG_BLK_DEV_ZONED. > > In it virtblk_probe() calls virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED) > unconditionally, which invokes virtio_check_driver_offered_feature(). > This function checks whether virtio_blk.feature_table (assigned to > the static features array) contains the specified feature enum, and if not > _causes a kernel panic_ via BUG(). > > This therefore means that failing to enable CONFIG_BLK_DEV_ZONED while > using virtio is a guaranteed kernel panic. Fix the issue by making the > feature test also conditional on CONFIG_BLK_DEV_ZONED. > > Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com> I think this was fixed by Message-ID: <20221220112340.518841-1-mst@redhat.com> > --- > drivers/block/virtio_blk.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c > index ff49052e26f7..94d210b10ebb 100644 > --- a/drivers/block/virtio_blk.c > +++ b/drivers/block/virtio_blk.c > @@ -1580,7 +1580,8 @@ static int virtblk_probe(struct virtio_device *vdev) > virtblk_update_capacity(vblk, false); > virtio_device_ready(vdev); > > - if (virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED)) { > + if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && > + virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED)) { > err = virtblk_probe_zoned_device(vdev, vblk, q); > if (err) > goto out_cleanup_disk; > -- > 2.39.0
On Wed, Dec 21, 2022 at 02:15:58PM -0500, Michael S. Tsirkin wrote: > I think this was fixed by > Message-ID: <20221220112340.518841-1-mst@redhat.com> You're right! Sorry I don't usually pay attention to virtio stuff, so missed that this was patched already, just noticed that my dev env was killed by the issue. I have tested your patch + confirmed working so will at least add a tested-by to it :)
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index ff49052e26f7..94d210b10ebb 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -1580,7 +1580,8 @@ static int virtblk_probe(struct virtio_device *vdev) virtblk_update_capacity(vblk, false); virtio_device_ready(vdev); - if (virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED)) { + if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && + virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED)) { err = virtblk_probe_zoned_device(vdev, vblk, q); if (err) goto out_cleanup_disk;
virtio zoned block device support is added by commit 0562d7bf1604 ("virtio-blk: add support for zoned block devices") which adds VIRTIO_BLK_F_ZONED to the features array in virtio_blk.c but makes it conditional on CONFIG_BLK_DEV_ZONED. In it virtblk_probe() calls virtio_has_feature(vdev, VIRTIO_BLK_F_ZONED) unconditionally, which invokes virtio_check_driver_offered_feature(). This function checks whether virtio_blk.feature_table (assigned to the static features array) contains the specified feature enum, and if not _causes a kernel panic_ via BUG(). This therefore means that failing to enable CONFIG_BLK_DEV_ZONED while using virtio is a guaranteed kernel panic. Fix the issue by making the feature test also conditional on CONFIG_BLK_DEV_ZONED. Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com> --- drivers/block/virtio_blk.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)