ext4: brelse all indirect buffer in ext4_ind_remove_space()
authorzhangyi (F) <yi.zhang@huawei.com>
Sat, 23 Mar 2019 15:43:05 +0000 (11:43 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Sat, 23 Mar 2019 15:43:05 +0000 (11:43 -0400)
All indirect buffers get by ext4_find_shared() should be released no
mater the branch should be freed or not. But now, we forget to release
the lower depth indirect buffers when removing space from the same
higher depth indirect block. It will lead to buffer leak and futher
more, it may lead to quota information corruption when using old quota,
consider the following case.

 - Create and mount an empty ext4 filesystem without extent and quota
   features,
 - quotacheck and enable the user & group quota,
 - Create some files and write some data to them, and then punch hole
   to some files of them, it may trigger the buffer leak problem
   mentioned above.
 - Disable quota and run quotacheck again, it will create two new
   aquota files and write the checked quota information to them, which
   probably may reuse the freed indirect block(the buffer and page
   cache was not freed) as data block.
 - Enable quota again, it will invoke
   vfs_load_quota_inode()->invalidate_bdev() to try to clean unused
   buffers and pagecache. Unfortunately, because of the buffer of quota
   data block is still referenced, quota code cannot read the up to date
   quota info from the device and lead to quota information corruption.

This problem can be reproduced by xfstests generic/231 on ext3 file
system or ext4 file system without extent and quota features.

This patch fix this problem by releasing the missing indirect buffers,
in ext4_ind_remove_space().

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: stable@kernel.org
fs/ext4/indirect.c

index c2225f0d31b511bbb0c62b176324e4f483437638..02c39524bda2a76fdf83500752e615c478d59cd2 100644 (file)
@@ -1390,10 +1390,14 @@ end_range:
                                           partial->p + 1,
                                           partial2->p,
                                           (chain+n-1) - partial);
-                       BUFFER_TRACE(partial->bh, "call brelse");
-                       brelse(partial->bh);
-                       BUFFER_TRACE(partial2->bh, "call brelse");
-                       brelse(partial2->bh);
+                       while (partial > chain) {
+                               BUFFER_TRACE(partial->bh, "call brelse");
+                               brelse(partial->bh);
+                       }
+                       while (partial2 > chain2) {
+                               BUFFER_TRACE(partial2->bh, "call brelse");
+                               brelse(partial2->bh);
+                       }
                        return 0;
                }