[PATCH] btrfs: handle the BUG_ON in btrfs_num_devices()

[Anand Jain]

Its a logical bug if we hit fs_devices::num_devices == 1 and if the
replace is running because, as fs_devices::num_devices counts the in memory
devices, so it should include the replace target which is running as
indicated by the flag. If this happens return the -EINVAL back.

Suggested-by: Nikolay Borisov <nbori...@suse.com>
Signed-off-by: Anand Jain <anand.j...@oracle.com>
---
Hi,
 As it fixes the BUG_ON I have spun a new patch for this.
 Instead of -EINVAL should we use ASSERT?

 fs/btrfs/volumes.c | 27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 7359596ac8eb..ed2399caff80 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1855,9 +1855,11 @@ void btrfs_assign_next_active_device(struct btrfs_device 
*device,
 }
 
 /* Returns btrfs_fs_devices::num_devices minus replace device if any */
-static u64 btrfs_num_devices(struct btrfs_fs_info *fs_info)
+static int btrfs_num_devices(struct btrfs_fs_info *fs_info, u64 *num_devices)
 {
-       u64 num_devices = fs_info->fs_devices->num_devices;
+       int ret = 0;
+
+       *num_devices = fs_info->fs_devices->num_devices;
 
        /*
         * balance and replace co-exists in a scenario as below..
@@ -1867,12 +1869,13 @@ static u64 btrfs_num_devices(struct btrfs_fs_info 
*fs_info)
         */
        btrfs_dev_replace_read_lock(&fs_info->dev_replace);
        if (btrfs_dev_replace_is_ongoing(&fs_info->dev_replace)) {
-               BUG_ON(num_devices < 1);
-               num_devices--;
+               if (*num_devices < 1)
+                       ret = -EINVAL;
+               (*num_devices)--;
        }
        btrfs_dev_replace_read_unlock(&fs_info->dev_replace);
 
-       return num_devices;
+       return ret;
 }
 
 int btrfs_rm_device(struct btrfs_fs_info *fs_info, const char *device_path,
@@ -1886,7 +1889,12 @@ int btrfs_rm_device(struct btrfs_fs_info *fs_info, const 
char *device_path,
 
        mutex_lock(&uuid_mutex);
 
-       num_devices = btrfs_num_devices(fs_info);
+       ret = btrfs_num_devices(fs_info, &num_devices);
+       if (ret) {
+               btrfs_err(fs_info, "logical bug num_devices %llu < 0",
+                         num_devices);
+               return ret;
+       }
 
        ret = btrfs_check_raid_min_devices(fs_info, num_devices - 1);
        if (ret)
@@ -3755,7 +3763,12 @@ int btrfs_balance(struct btrfs_fs_info *fs_info,
                }
        }
 
-       num_devices = btrfs_num_devices(fs_info);
+       ret = btrfs_num_devices(fs_info, &num_devices);
+       if (ret) {
+               btrfs_err(fs_info, "hits a logical bug num_devices %llu < 0",
+                         num_devices);
+               return ret;
+       }
 
        allowed = BTRFS_AVAIL_ALLOC_BIT_SINGLE | BTRFS_BLOCK_GROUP_DUP;
        if (num_devices > 1)
-- 
2.7.0

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html