ext4: check and assert if marking an no_delete evicting inode dirty
authorZhang Yi <yi.zhang@huawei.com>
Wed, 29 Jun 2022 11:26:47 +0000 (19:26 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 12 Jan 2023 10:59:01 +0000 (11:59 +0100)
commit 318cdc822c63b6e2befcfdc2088378ae6fa18def upstream.

In ext4_evict_inode(), if we evicting an inode in the 'no_delete' path,
it cannot be raced by another mark_inode_dirty(). If it happens,
someone else may accidentally dirty it without holding inode refcount
and probably cause use-after-free issues in the writeback procedure.
It's indiscoverable and hard to debug, so add an WARN_ON_ONCE() to
check and detect this issue in advance.

Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220629112647.4141034-2-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/ext4/inode.c

index f449c12..dc6311a 100644 (file)
@@ -338,6 +338,12 @@ stop_handle:
        ext4_xattr_inode_array_free(ea_inode_array);
        return;
 no_delete:
+       /*
+        * Check out some where else accidentally dirty the evicting inode,
+        * which may probably cause inode use-after-free issues later.
+        */
+       WARN_ON_ONCE(!list_empty_careful(&inode->i_io_list));
+
        if (!list_empty(&EXT4_I(inode)->i_fc_list))
                ext4_fc_mark_ineligible(inode->i_sb, EXT4_FC_REASON_NOMEM, NULL);
        ext4_clear_inode(inode);        /* We must guarantee clearing of inode... */