Re: Linux 3.15 .. and continuation of merge window

["J. R. Okajima" <hooanon05g@gmail.com>]

Al Viro:
> So I suspect that the right fix is a bit trickier - in addition to check
> on the fast path (i.e. when trylock gets us the lock on parent), we need
> to
> 	* get rcu_read_lock() before dropping ->d_lock.
> 	* check if dentry is already doomed right after taking rcu_read_lock();
> if not, any value we might see in ->d_parent afterwards will point to object
> not freed until we drop rcu_read_lock.
>
> IOW, something like the delta below.  Comments?

I will try testing later.
For now, as a comment before testing, the patch looks weird for me. It
checks d_lockref.count twice during d_lockref.lock held. It must be the
same result, isn't it? Or does it mean that denty can be handled by
lockref_mark_dead() even if d_lockref.lock is held?


J. R. Okajima