Re: [PATCH v10 2/5] ext4: introduce ext4_put_ea_inode() for safe deferred iput

[Zhou, Yun]

On 6/27/26 00:53, Jan Kara wrote:
> On Thu 25-06-26 23:29:38, Yun Zhou wrote:
> > +
> > +/*
> > + * Drain all pending deferred EA inode iputs.  Must be called before
> > + * freeing resources that eviction depends on (quota, block allocator).
> > + * Loops because worker iput may trigger eviction that re-queues.
> > + */
>
> Can you please explain how iput of EA inode can trigger iput of another EA
> inode? So far I don't think that's possible but perhaps I'm missing
> something.
>
> > +static inline void ext4_drain_ea_inode_work(struct ext4_sb_info *sbi)
> > +{
> > +     while (flush_delayed_work(&sbi->s_ea_inode_work) ||
> > +            !llist_empty(&sbi->s_ea_inode_to_free))
> > +             ;
> > +}

On a corrupted fs, an evicted ea_inode might still have xattrs and point 
to another ea inode, which creates a potential "re-arm" scenario. Of 
course, this situation is extremely rare. Therefore, I plan to remove 
this redundant loop logic for flush_delayed_work. Instead, I will add a 
simple EXT4_EA_INODE_FL flag check right before calling 
ext4_xattr_delete_inode(), as shown below:

+       if (!(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL)) {
+               err = ext4_xattr_delete_inode(handle, inode, extra_credits);
+               if (err) {

This will be submitted as a separate, optional enhancement patch. What 
do you think?

BR,
Yun