Re: [PATCH v7 3/4] ext4: introduce ext4_put_ea_inode() for safe deferred iput

[Jan Kara]

On Mon 22-06-26 18:06:23, Zhou, Yun wrote:
> Hi Honza,
> 
> On 6/18/26 02:42, Jan Kara wrote:
> > 
> > Allocating ext4_ea_iput_entry for dropping each inode is somewhat wasteful.
> > I want to suggest another scheme (somewhat more involved but more efficient
> > scheme):
> > 
> > 1) Create a VFS helper bool iput_if_not_last(struct inode *inode) which
> > drops inode reference if it is not the last one (and returns true in that
> > case). Basically:
> > 
> > bool iput_if_not_last(struct inode *inode)
> > {
> >          return atomic_add_unless(&inode->i_count, -1, 1);
> > }
> > 
> > This needs to be a separate patch as it should get vetting from VFS
> > maintainers.
> After taking a closer look, it seems that this function doesn't need to be
> added to the VFS layer — at least not for now. For example, we could
> directly inline atomic_add_unless() into ext4_put_ea_inode():
> 
> void ext4_put_ea_inode(struct super_block *sb, struct inode *inode)
> {
>     if (!inode)
>         return;
>     if (atomic_add_unless(&inode->i_count, -1, 1))
>         return;
>     llist_add(&EXT4_I(inode)->i_ea_iput_node,
>         &EXT4_SB(sb)->s_ea_inode_to_free);
>     schedule_delayed_work(&EXT4_SB(sb)->s_ea_inode_work, 1);
> }
> 
> This way, we avoid submitting an isolated patch to the VFS layer that
> currently has only one user (ext4). If there is indeed a real need for it in
> the future, we can always submit a follow-up patch to refactor it.
> 
> What do you think?

No, please provide a proper helper in VFS for this. We don't really want
filesystems to play games with inode refcount without proper abstraction in
VFS. That causes longterm maintenance issues.

								Honza
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR