On Mon, Jan 04, 2016 at 10:20:11AM -0800, Dan Williams wrote: > Force a filesystem shutdown when the backing device is known to be dead. > I.e. blk_queue_enter() permanently returns -ENODEV. > > Cc: xfs@xxxxxxxxxxx > Suggested-by: Dave Chinner <david@xxxxxxxxxxxxx> > Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx> > --- > fs/block_dev.c | 3 ++- > fs/xfs/xfs_super.c | 9 +++++++++ > include/linux/fs.h | 2 ++ > 3 files changed, 13 insertions(+), 1 deletion(-) > > diff --git a/fs/block_dev.c b/fs/block_dev.c > index 739e43a37e64..7d6c66148948 100644 > --- a/fs/block_dev.c > +++ b/fs/block_dev.c > @@ -1806,7 +1806,7 @@ int __invalidate_device(struct block_device *bdev, bool kill_dirty) > } > EXPORT_SYMBOL(__invalidate_device); > > -static void generic_bdi_gone(struct super_block *sb) > +void generic_bdi_gone(struct super_block *sb) > { > struct inode *inode, *_inode = NULL; > > @@ -1832,6 +1832,7 @@ static void generic_bdi_gone(struct super_block *sb) > spin_unlock(&sb->s_inode_list_lock); > iput(_inode); > } > +EXPORT_SYMBOL(generic_bdi_gone); That should be in the previous patch. > void shutdown_partition(struct gendisk *disk, int partno) > { > diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c > index 36bd8825bfb0..63c36508e9db 100644 > --- a/fs/xfs/xfs_super.c > +++ b/fs/xfs/xfs_super.c > @@ -1618,6 +1618,14 @@ xfs_fs_free_cached_objects( > return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan); > } > > +static void > +xfs_fs_bdi_gone( xfs_fs_shutdown > + struct super_block *sb) > +{ > + xfs_force_shutdown(XFS_M(sb), SHUTDOWN_DEVICE_REQ); > + generic_bdi_gone(sb); > +} > + This is wrong. we have to unmap the DAX inodes during *every* shutdown that XFS executes. Hence it needs to be done inside xfs_do_force_shutdown(), not in addition to the shutdown when the block device is pulled. i.e. something like this in xfs_do_force_shutdown(): */ if (xfs_log_force_umount(mp, logerror)) return; + + /* + * If DAX is in use, we have to unmap all direct access + * virtual mappings to ensure nothing more gets written + * directly from userspace. This will force them to refault + * and that will result in them detecting the shutdown + * condition and hence will fail appropriately. + */ + unmap_dax_inodes(mp->m_super); if (flags & SHUTDOWN_CORRUPT_INCORE) { xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT > static const struct super_operations xfs_super_operations = { > .alloc_inode = xfs_fs_alloc_inode, > .destroy_inode = xfs_fs_destroy_inode, > @@ -1632,6 +1640,7 @@ static const struct super_operations xfs_super_operations = { > .show_options = xfs_fs_show_options, > .nr_cached_objects = xfs_fs_nr_cached_objects, > .free_cached_objects = xfs_fs_free_cached_objects, > + .bdi_gone = xfs_fs_bdi_gone, > }; > > static struct file_system_type xfs_fs_type = { > diff --git a/include/linux/fs.h b/include/linux/fs.h > index 0e201ed38045..b1e8e049e4b8 100644 > --- a/include/linux/fs.h > +++ b/include/linux/fs.h > @@ -2265,6 +2265,7 @@ extern struct super_block *freeze_bdev(struct block_device *); > extern void emergency_thaw_all(void); > extern int thaw_bdev(struct block_device *bdev, struct super_block *sb); > extern int fsync_bdev(struct block_device *); > +extern void generic_bdi_gone(struct super_block *sb); previous patch. -Dave. -- Dave Chinner david@xxxxxxxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-block" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html