Re: [PATCH 2/2] Btrfs: fix deadlock on sb->s_umount when doing umount

On tue, 6 Dec 2011 05:49:06 +0000, Al Viro wrote:

quoted

+void btrfs_writeback_inodes_sb_nr(struct btrfs_root *root,
+				  unsigned long nr_pages)
+{
+	struct super_block *sb = root->fs_info->sb;
+
+	if (writeback_in_progress(sb->s_bdi))
+		return;
+
+	/*
+	 * If we can not get s_umount, it means the fs is on remounting or
+	 * umounting. At this time, we just sync all the delalloc file.
+	 */
+	if (down_read_trylock(&sb->s_umount)) {
+		writeback_inodes_sb_nr(sb, nr_pages);
+		up_read(&sb->s_umount);
+	} else {
+		btrfs_start_delalloc_inodes(root, 0);
+		btrfs_wait_ordered_extents(root, 0, 0);
+	}
+}

If that can race with umount, what prevents sb, its ->s_bdi et.al. being freed
under you?

In fact, it happened. See the following mail.

   http://marc.info/?l=linux-btrfs&m=131495252725296&w=2

The above function is called when some one want to modify the meta-data.
Btrfs will wait until all the meta-data operations end, and then free ->s_bdi
and the other objects. So we needn't worry about those objects.
(Maybe I misunderstood what you said. If yes, I'm sorry)

Thanks
Miao

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