btrfs: fix qgroup data rsv leak caused by falloc failure

[BUG]
When running fsstress with only falloc workload, and a very low qgroup
limit set, we can get qgroup data rsv leak at unmount time.

 BTRFS warning (device dm-0): qgroup 0/5 has unreleased space, type 0 rsv 20480
 BTRFS error (device dm-0): qgroup reserved space leaked

The minimal reproducer looks like:

  #!/bin/bash
  dev=/dev/test/test
  mnt="/mnt/btrfs"
  fsstress=~/xfstests-dev/ltp/fsstress
  runtime=8

  workload()
  {
          umount $dev &> /dev/null
          umount $mnt &> /dev/null
          mkfs.btrfs -f $dev > /dev/null
          mount $dev $mnt

          btrfs quota en $mnt
          btrfs quota rescan -w $mnt
          btrfs qgroup limit 16m 0/5 $mnt

          $fsstress -w -z -f creat=10 -f fallocate=10 -p 2 -n 100 \
  		-d $mnt -v > /tmp/fsstress

          umount $mnt
          if dmesg | grep leak ; then
		echo "!!! FAILED !!!"
  		exit 1
          fi
  }

  for (( i=0; i < $runtime; i++)); do
          echo "=== $i/$runtime==="
          workload
  done

Normally it would fail before round 4.

[CAUSE]
In function insert_prealloc_file_extent(), we first call
btrfs_qgroup_release_data() to know how many bytes are reserved for
qgroup data rsv.

Then use that @qgroup_released number to continue our work.

But after we call btrfs_qgroup_release_data(), we should either queue
@qgroup_released to delayed ref or free them manually in error path.

Unfortunately, we lack the error handling to free the released bytes,
leaking qgroup data rsv.

All the error handling function outside won't help at all, as we have
released the range, meaning in inode io tree, the EXTENT_QGROUP_RESERVED
bit is already cleared, thus all btrfs_qgroup_free_data() call won't
free any data rsv.

[FIX]
Add free_qgroup tag to manually free the released qgroup data rsv.

Reported-by: Nikolay Borisov <nborisov@suse.com>
Reported-by: David Sterba <dsterba@suse.cz>
Fixes: 9729f10a608f ("btrfs: inode: move qgroup reserved space release to the callers of insert_reserved_file_extent()")
CC: stable@vger.kernel.org # 5.10+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 77182be403c56bd706f677325adcf54bb2dc1ec2..ea5ede619220cbf99fffc3ce4efa2d6d5e7f5db9 100644 (file)
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -9895,7 +9895,7 @@ static struct btrfs_trans_handle *insert_prealloc_file_extent(
                                                  file_offset, &stack_fi,
                                                  true, qgroup_released);
                if (ret)
-                       return ERR_PTR(ret);
+                       goto free_qgroup;
                return trans;
        }
 
@@ -9910,17 +9910,31 @@ static struct btrfs_trans_handle *insert_prealloc_file_extent(
        extent_info.insertions = 0;
 
        path = btrfs_alloc_path();
-       if (!path)
-               return ERR_PTR(-ENOMEM);
+       if (!path) {
+               ret = -ENOMEM;
+               goto free_qgroup;
+       }
 
        ret = btrfs_replace_file_extents(&inode->vfs_inode, path, file_offset,
                                     file_offset + len - 1, &extent_info,
                                     &trans);
        btrfs_free_path(path);
        if (ret)
-               return ERR_PTR(ret);
-
+               goto free_qgroup;
        return trans;
+
+free_qgroup:
+       /*
+        * We have released qgroup data range at the beginning of the function,
+        * and normally qgroup_released bytes will be freed when committing
+        * transaction.
+        * But if we error out early, we have to free what we have released
+        * or we leak qgroup data reservation.
+        */
+       btrfs_qgroup_free_refroot(inode->root->fs_info,
+                       inode->root->root_key.objectid, qgroup_released,
+                       BTRFS_QGROUP_RSV_DATA);
+       return ERR_PTR(ret);
 }
 
 static int __btrfs_prealloc_file_range(struct inode *inode, int mode,