linux-kernel - Re: [PATCH] spinlock: __raw_spin_is_locked() should return true for UP

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]

Date:	Wed, 19 Aug 2009 11:38:21 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Olivier Galibert <galibert@...ox.com>
Cc:	Linus Torvalds <torvalds@...ux-foundation.org>,
	Kumar Gala <galak@...nel.crashing.org>,
	Steven Rostedt <rostedt@...dmis.org>,
	linux-kernel@...r.kernel.org, mingo@...e.hu, tglx@...utronix.de,
	linuxppc-dev@...abs.org
Subject: Re: [PATCH] spinlock: __raw_spin_is_locked() should return true
 for UP

On Wed, 2009-08-19 at 11:31 +0200, Olivier Galibert wrote:
> On Tue, Aug 18, 2009 at 07:40:16PM -0700, Linus Torvalds wrote:
> > 
> > 
> > On Tue, 18 Aug 2009, Kumar Gala wrote:
> > > 
> > > I agree its a little too easy to abuse spin_is_locked.  However we should be
> > > consistent between spin_is_locked on UP between with and without
> > > CONFIG_DEBUG_SPINLOCK enabled.
> > 
> > No we shouldn't.
> > 
> > With CONFIG_DEBUG_SPINLOCK, you have an actual lock variable for debugging 
> > purposes, so spin_is_locked() can clearly return a _valid_ answer, and 
> > should do so.
> > 
> > Without DEBUG_SPINLOCK, there isn't any answer to return.
> > 
> > So there's no way we can or should be consistent. In one case an answer 
> > exists, in another one the answer is meaningless and doesn't exist.
> 
> I always thought behaviour should be consistent between code with
> debugging on and code without.  Otherwise you may end up with cases of
> "it starts working when I turn on debugging" which are a pain to fix.
> Has something changed?
> 
> Or in other words, do you think lockdep should try solving deadlocks
> instead of just reporting them for instance?

The point is spin_is_locked() is a broken interface in that respect. Its
plain impossible to give the right answer.


Suppose there's code doing:

  /*
   * Ensure we don't have foo lock taken, because that would cause
   * lock inversion under bar lock.
   */
  BUG_ON(spin_is_locked(&foo));
  spin_lock(&bar);

and other code doing:

  /*
   * Ensure we've got foo locked because it protects bar
   */
  BUG_ON(!spin_is_locked(&foo));
  bar = fancy;

What value should you return when locks don't exist (which is the case
for UP)?


There simply is no right answer other than: don't use spin_is_locked().
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/