This is a followup to the patch which skipped orphan inode processing
on a filesystem residing on a readonly bdev (think snapshot taken
with open, unlinked files). That avoided an oops.
Stephen had concerns about what happens if the bdev later goes readwrite,
and the filesystem is subsequently mounted remount,rw - if the orphan
list isn't processed then, things will be in bad shape.
The patch I sent earlier to process the orphan inode list on remount
was met with silence and/or slight suspicion (from akpm) so how
about this: just to plug the hole on what I think is a pretty rare
condition, simply disallow the ro->rw transition if we have unprocessed
orphan inodes.
Then we get this:
[root@magnesium ~]# mount -o remount,rw /mnt/test
mount: /mnt/test not mounted already, or bad option
[root@magnesium ~]# dmesg | tail -n 1
EXT3-fs: hda7: couldn't remount RDWR because of unprocessed orphan inode list. Please umount/remount instead.
Signed-off-by: Eric Sandeen <[email protected]>
Index: linux-2.6.19/fs/ext3/super.c
===================================================================
--- linux-2.6.19.orig/fs/ext3/super.c
+++ linux-2.6.19/fs/ext3/super.c
@@ -2350,6 +2350,22 @@ static int ext3_remount (struct super_bl
err = -EROFS;
goto restore_opts;
}
+
+ /*
+ * If we have an unprocessed orphan list hanging
+ * around from a previously readonly bdev mount,
+ * require a full umount/remount for now.
+ */
+ if (es->s_last_orphan) {
+ printk(KERN_WARNING "EXT3-fs: %s: couldn't "
+ "remount RDWR because of unprocessed "
+ "orphan inode list. Please "
+ "umount/remount instead.\n",
+ sb->s_id);
+ err = -EINVAL;
+ goto restore_opts;
+ }
+
/*
* Mounting a RDONLY partition read-write, so reread
* and store the current valid flag. (It may have