Re: [PATCH 3/6] tracefs: dentry lookup crapectomy

[Linus Torvalds]

On Tue, 30 Jan 2024 at 16:23, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Jan 30, 2024 at 11:03:52AM -0800, Linus Torvalds wrote:
> >  {
> >       struct eventfs_inode *ei;
> >
> > -     mutex_lock(&eventfs_mutex);
> >       do {
> >               // The parent is stable because we do not do renames
> >               dentry = dentry->d_parent;
> > @@ -247,7 +246,6 @@ static struct eventfs_inode *eventfs_find_events(struct dentry *dentry)
> >               }
> >               // Walk upwards until you find the events inode
> >       } while (!ei->is_events);
> > -     mutex_unlock(&eventfs_mutex);
>
> Unless I'm missing something, you've just lost exclusion with
> removals (not that the original hadn't been suspicious in that
> respect - what's to protect ei past that mutex_unlock?

No, the mutex is actually taken up the call chain, and removing it
here is fixing a deadlock.

Ie we have the mutex taken in eventfs_root_lookup(), and then that
goes either through lookup_dir_entry() or lookup_file_dentry(), before
it gets to update_inode_attr().

That series is not very clean, I'm afraid.

          Linus