btrfs: do not initialize dev replace for bad dev root
authorJosef Bacik <josef@toxicpanda.com>
Thu, 11 Mar 2021 16:23:16 +0000 (11:23 -0500)
committerDavid Sterba <dsterba@suse.com>
Wed, 17 Mar 2021 18:42:14 +0000 (19:42 +0100)
While helping Neal fix his broken file system I added a debug patch to
catch if we were calling btrfs_search_slot with a NULL root, and this
stack trace popped:

  we tried to search with a NULL root
  CPU: 0 PID: 1760 Comm: mount Not tainted 5.11.0-155.nealbtrfstest.1.fc34.x86_64 #1
  Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/22/2020
  Call Trace:
   dump_stack+0x6b/0x83
   btrfs_search_slot.cold+0x11/0x1b
   ? btrfs_init_dev_replace+0x36/0x450
   btrfs_init_dev_replace+0x71/0x450
   open_ctree+0x1054/0x1610
   btrfs_mount_root.cold+0x13/0xfa
   legacy_get_tree+0x27/0x40
   vfs_get_tree+0x25/0xb0
   vfs_kern_mount.part.0+0x71/0xb0
   btrfs_mount+0x131/0x3d0
   ? legacy_get_tree+0x27/0x40
   ? btrfs_show_options+0x640/0x640
   legacy_get_tree+0x27/0x40
   vfs_get_tree+0x25/0xb0
   path_mount+0x441/0xa80
   __x64_sys_mount+0xf4/0x130
   do_syscall_64+0x33/0x40
   entry_SYSCALL_64_after_hwframe+0x44/0xa9
  RIP: 0033:0x7f644730352e

Fix this by not starting the device replace stuff if we do not have a
NULL dev root.

Reported-by: Neal Gompa <ngompa13@gmail.com>
CC: stable@vger.kernel.org # 5.11+
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/dev-replace.c

index 3a9c1e0..d05f735 100644 (file)
@@ -81,6 +81,9 @@ int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
        struct btrfs_dev_replace_item *ptr;
        u64 src_devid;
 
+       if (!dev_root)
+               return 0;
+
        path = btrfs_alloc_path();
        if (!path) {
                ret = -ENOMEM;