From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sebastian Andrzej Siewior Subject: Re: [patchlet] locking/rt: fix rt_read_lock() lockdep annotation. Date: Fri, 9 May 2014 15:27:13 +0200 Message-ID: <20140509132713.GH29014@linutronix.de> References: <20140503170110.GA10652@linutronix.de> <1399449247.5142.14.camel@marge.simpson.net> <20140507094807.229d88d8@gandalf.local.home> <1399471264.5248.5.camel@marge.simpson.net> <1399472415.5248.15.camel@marge.simpson.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Cc: Steven Rostedt , linux-rt-users , Thomas Gleixner , John Kacur To: Mike Galbraith Return-path: Received: from www.linutronix.de ([62.245.132.108]:35956 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750744AbaEIN1Q (ORCPT ); Fri, 9 May 2014 09:27:16 -0400 Content-Disposition: inline In-Reply-To: <1399472415.5248.15.camel@marge.simpson.net> Sender: linux-rt-users-owner@vger.kernel.org List-ID: * Mike Galbraith | 2014-05-07 16:20:15 [+0200]: >> > This looks like it reverse applies? >> Nope, I thought Thomas's whacked the one he did on purpose. >I should learn to speak lockdep. How about this instead? > >locking/rt: fix rt_read_trylock() lockdep annotation. > >rt-rw-lockdep-annotations.patch dropped a rwlock_acquire_read > >Signed-off-by: Mike Galbraith >--- > kernel/locking/rt.c | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) > >--- a/kernel/locking/rt.c >+++ b/kernel/locking/rt.c >@@ -217,7 +217,7 @@ int __lockfunc rt_read_trylock(rwlock_t > migrate_disable(); > ret = rt_mutex_trylock(lock); > if (ret) >- rwlock_acquire(&rwlock->dep_map, 0, 1, _RET_IP_); >+ rwlock_acquire_read(&rwlock->dep_map, 0, 1, _RET_IP_); Think we should drop rwlock_acquire_read() and use rwlock_acquire() instead (like the previous patch) because on -RT there is no difference between a read and a write lock. Or is there more to it? > else > migrate_enable(); > Sebastian