linux-kernel - Re: [PATCH v2] fs: make sure to fail try_to_unlazy() and try_to_unlazy() for LOOKUP

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20251217104854.GU1712166@ZenIV>
Date: Wed, 17 Dec 2025 10:48:54 +0000
From: Al Viro <viro@...iv.linux.org.uk>
To: Mateusz Guzik <mjguzik@...il.com>
Cc: brauner@...nel.org, jack@...e.cz, linux-kernel@...r.kernel.org,
	linux-fsdevel@...r.kernel.org, clm@...a.com
Subject: Re: [PATCH v2] fs: make sure to fail try_to_unlazy() and
 try_to_unlazy() for LOOKUP_CACHED

On Wed, Dec 17, 2025 at 11:13:22AM +0100, Mateusz Guzik wrote:

> I'm not arguing for drop_links() to change behavior, but for it to be
> renamed to something which indicates there is still potential
> symlink-related clean up to do.
> 
> As an outsider, a routine named drop_${whatever} normally suggests the
> ${whatever} is fully taken care of after the call, which is not the
> case here.

*shrug*

drop_link_bodies()?  Seriously, though - the real source of headache
had been a decision (mine, IIRC) to stash LOOKUP_CACHED logics into
legitimize_links() rather than having it done in both of its callers
(try_to_unlazy() and try_to_unlazy_next()).

It could be lifted into both callers instead, but we'd need to
	* remember that it should come *before* any references get
grabbed (and there would be a constant itch to move it down -
it would even work, and past legitimize_links() we wouldn't need
drop_links(), but it would make the thing potentially do seriously
blocking IO, contrary to LOOKUP_CACHED intent)
	* remember to keep drop_links()/resetting ->depth if done
before legitimize_links().
	* keep these two in sync.

If you have a cleaner solution for that, I'm all ears.  Really.
I'm not happy with details of that area (lazy to nonlazy transitions),
but so far I hadn't been able to make it more obviously correct.
And I'd looked at your original patches and missed that fun ;-/