linux-kernel - Re: [PATCH] lockdep: Do no validate wait context for novalidate class

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

Message-ID: <20200820114036.GQ2674@hirez.programming.kicks-ass.net>
Date:   Thu, 20 Aug 2020 13:40:36 +0200
From:   peterz@...radead.org
To:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc:     linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...hat.com>,
        Will Deacon <will@...nel.org>, tglx@...utronix.de,
        kent.overstreet@...il.com
Subject: Re: [PATCH] lockdep: Do no validate wait context for novalidate class

On Mon, Jun 29, 2020 at 10:15:29PM +0200, Sebastian Andrzej Siewior wrote:
> The novalidate class is ignored in the lockchain validation but is
> considered in the wait context validation.
> If a mutex and a spinlock_t is ignored by using
> lockdep_set_novalidate_class() then both locks will share the same lock
> class. From the wait validation point of view the mutex will then appear
> like a spinlock_t and the validator will complain if another mutex will
> be acquired.
> 
> Ignore the nonvalidate locks from wait context checking.

Hurmph.. but how? There was only a single user... /me greps.

drivers/base/core.c:    lockdep_set_novalidate_class(&dev->mutex);
drivers/md/bcache/btree.c:      lockdep_set_novalidate_class(&b->lock);
drivers/md/bcache/btree.c:      lockdep_set_novalidate_class(&b->write_lock);

Urgh.. there's more now :-(

So write_lock, like dev->mutex is a mutex.

Kent, what's the story with b->lock? It appears to have lockdep
annotations, but then is also the novalidate class. Also neither of
these lockdep_set_novalidate_class() thingies have a comment.

Anyway, all 3 users should have the same wait context, so where is the
actual problem?