Btrfs: don't allow device replace on RAID5/RAID6

This is not yet supported and causes crashes. One sad user reported
that it destroyed his filesystem.

One failure is in __btrfs_map_block+0xc1f calling kmalloc(0).

0x5f21f is in __btrfs_map_block (fs/btrfs/volumes.c:4923).
4918                            num_stripes = map->num_stripes;
4919                            max_errors = nr_parity_stripes(map);
4920
4921                            raid_map = kmalloc(sizeof(u64) * num_stripes,
4922                                               GFP_NOFS);
4923                            if (!raid_map) {
4924                                    ret = -ENOMEM;
4925                                    goto out;
4926                            }
4927

There might be more issues. Until this is really tested, don't allow
users to start the procedure on RAID5/RAID6 filesystems.

Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>

diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
index 7ba7b39..65241f3 100644 (file)
--- a/fs/btrfs/dev-replace.c
+++ b/fs/btrfs/dev-replace.c
@@ -313,6 +313,11 @@ int btrfs_dev_replace_start(struct btrfs_root *root,
        struct btrfs_device *tgt_device = NULL;
        struct btrfs_device *src_device = NULL;
 
+       if (btrfs_fs_incompat(fs_info, RAID56)) {
+               pr_warn("btrfs: dev_replace cannot yet handle RAID5/RAID6\n");
+               return -EINVAL;
+       }
+
        switch (args->start.cont_reading_from_srcdev_mode) {
        case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS:
        case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID: