ext4: prevent online resize with backup superblock
authorTheodore Ts'o <tytso@mit.edu>
Sat, 27 Dec 2014 04:58:21 +0000 (23:58 -0500)
committerTheodore Ts'o <tytso@mit.edu>
Sat, 27 Dec 2014 04:58:21 +0000 (23:58 -0500)
Prevent BUG or corrupted file systems after the following:

mkfs.ext4 /dev/vdc 100M
mount -t ext4 -o sb=40961 /dev/vdc /vdc
resize2fs /dev/vdc

We previously prevented online resizing using the old resize ioctl.
Move the code to ext4_resize_begin(), so the check applies for all of
the resize ioctl's.

Reported-by: Maxim Malkov <malkov@ispras.ru>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
fs/ext4/resize.c

index bf76f405a5f91df5f276e2a2d1dcf2935178d4a7..8a8ec6293b195f16623e716342463979427b3156 100644 (file)
@@ -23,6 +23,18 @@ int ext4_resize_begin(struct super_block *sb)
        if (!capable(CAP_SYS_RESOURCE))
                return -EPERM;
 
+       /*
+        * If we are not using the primary superblock/GDT copy don't resize,
+         * because the user tools have no way of handling this.  Probably a
+         * bad time to do it anyways.
+         */
+       if (EXT4_SB(sb)->s_sbh->b_blocknr !=
+           le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) {
+               ext4_warning(sb, "won't resize using backup superblock at %llu",
+                       (unsigned long long)EXT4_SB(sb)->s_sbh->b_blocknr);
+               return -EPERM;
+       }
+
        /*
         * We are not allowed to do online-resizing on a filesystem mounted
         * with error, because it can destroy the filesystem easily.
@@ -758,18 +770,6 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
                       "EXT4-fs: ext4_add_new_gdb: adding group block %lu\n",
                       gdb_num);
 
-       /*
-        * If we are not using the primary superblock/GDT copy don't resize,
-         * because the user tools have no way of handling this.  Probably a
-         * bad time to do it anyways.
-         */
-       if (EXT4_SB(sb)->s_sbh->b_blocknr !=
-           le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) {
-               ext4_warning(sb, "won't resize using backup superblock at %llu",
-                       (unsigned long long)EXT4_SB(sb)->s_sbh->b_blocknr);
-               return -EPERM;
-       }
-
        gdb_bh = sb_bread(sb, gdblock);
        if (!gdb_bh)
                return -EIO;