Re: dcache: remove trylock loops (was Re: [BUG] lock_parent() breakage when used from shrink_dentry_list())

[Thomas Gleixner]

On Mon, 12 Mar 2018, Al Viro wrote:
> On Tue, Feb 27, 2018 at 06:16:28AM +0100, John Ogness wrote:
> What we get out of that is
> 	a) much saner semantics for d_move() et.al.
> 	b) saner behaviour of d_path() (see above)
> 	c) dentry can be IS_ROOT only if it has been
> such all along; that simplifies the hell out of analysis.

Nice.

> FWIW, there's another trylock loop on dentries - one in
> autofs get_next_positive_dentry().  Any plans re dealing
> with that one?

It's on our list and we wanted to wait for the final resolution of dcache
before tackling it. Do you have any immediate ide about that?

> I'd spent the last couple of weeks (when not being too sick
> for any work) going through dcache.c and related code; hopefully
> this time I will get the documentation into postable shape ;-/

Documentation is surely welcome. Paging that code back in is a mindboggling
exercise.

Thanks,

	tglx