Re: [PATCH v3] vfs: don't evict uninitialized inode

[Marc Dionne <marc.c.dionne@gmail.com>]

On Tue, Jul 24, 2018 at 10:01 AM, Miklos Szeredi <mszeredi@redhat.com> wrote:
> iput() ends up calling ->evict() on new inode, which is not yet initialized
> by owning fs.  So use destroy_inode() instead.
>
> Add to sb->s_inodes list only if inode is not in I_CREATING state (meaning
> that it wasn't allocated with new_inode(), which already does the
> insertion).
>
> Reported-by: Al Viro <viro@zeniv.linux.org.uk>
> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
> Fixes: 80ea09a002bf ("vfs: factor out inode_insert5()")
> ---
>  fs/inode.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/fs/inode.c b/fs/inode.c
> index 04dd7e0d5142..0aa5b29b6f87 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -1050,6 +1050,7 @@ struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
>  {
>         struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval);
>         struct inode *old;
> +       bool creating = inode->i_state & I_CREATING;
>
>  again:
>         spin_lock(&inode_hash_lock);
> @@ -1083,6 +1084,8 @@ struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
>         inode->i_state |= I_NEW;
>         hlist_add_head(&inode->i_hash, head);
>         spin_unlock(&inode->i_lock);
> +       if (!creating)
> +               inode_sb_list_add(inode);
>  unlock:
>         spin_unlock(&inode_hash_lock);
>
> @@ -1117,12 +1120,13 @@ struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
>         struct inode *inode = ilookup5(sb, hashval, test, data);
>
>         if (!inode) {
> -               struct inode *new = new_inode(sb);
> +               struct inode *new = alloc_inode(sb);
>
>                 if (new) {
> +                       new->i_state = 0;
>                         inode = inode_insert5(new, hashval, test, set, data);
>                         if (unlikely(inode != new))
> -                               iput(new);
> +                               destroy_inode(new);
>                 }
>         }
>         return inode;
> --
> 2.14.3
>

FYI with this patch (now merged) I'm seeing warnings whenever an
object is created in an overlayfs mount:

[  842.152673] list_add double add: new=ffff88017efe03d8,
prev=ffff88015c07ad88, next=ffff88017efe03d8.
[  842.152687] WARNING: CPU: 4 PID: 7592 at lib/list_debug.c:31
__list_add_valid+0x6e/0x80

The call stack looks like this, for the file creation case:

[  842.152746]  inode_sb_list_add+0x4e/0x90
[  842.152753]  ? ovl_inode_test+0x20/0x20 [overlay]
[  842.152757]  ? ovl_get_redirect_xattr+0x140/0x140 [overlay]
[  842.152759]  inode_insert5+0x13e/0x1f0
[  842.152764]  ovl_get_inode.cold.16+0x38/0x44 [overlay]
[  842.152768]  ovl_instantiate+0x75/0x130 [overlay]
[  842.152773]  ovl_create_or_link+0x1c9/0x7d0 [overlay]
[  842.152776]  ? ovl_alloc_inode+0x1b/0x80 [overlay]
[  842.152778]  ? inode_sb_list_add+0x4e/0x90
[  842.152782]  ? ovl_fill_inode+0xd8/0x150 [overlay]
[  842.152787]  ovl_create_object+0xa1/0xd0 [overlay]
[  842.152791]  ovl_create+0x23/0x30 [overlay]

.. where the inode passed to inode_insert5 originally comes from
new_inode (ovl_create_object ->  ovl_new_inode -> new_inode), was
already added to the sb list, and doesn't have I_CREATING set.

Originally seen from docker's use of overlayfs, but easily
reproducible by creating a simple overlay of 2 ext4 directories and
creating a new file or directory in the overlay.

Marc