From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Julian Anastasov <ja@ssi.bg>
Cc: Peter Zijlstra <peterz@infradead.org>,
Simon Horman <horms@verge.net.au>,
Eric Dumazet <eric.dumazet@gmail.com>,
Ingo Molnar <mingo@redhat.com>,
lvs-devel@vger.kernel.org, netdev@vger.kernel.org,
netfilter-devel@vger.kernel.org, linux-kernel@vger.kernel.org,
Pablo Neira Ayuso <pablo@netfilter.org>,
Dipankar Sarma <dipankar@in.ibm.com>
Subject: Re: [PATCH v2 1/2] sched: Add cond_resched_rcu_lock() helper
Date: Wed, 1 May 2013 12:04:12 -0700 [thread overview]
Message-ID: <20130501190412.GU3780@linux.vnet.ibm.com> (raw)
In-Reply-To: <alpine.LFD.2.00.1305012004360.3397@ja.ssi.bg>
On Wed, May 01, 2013 at 09:22:08PM +0300, Julian Anastasov wrote:
>
> Hello,
>
> On Wed, 1 May 2013, Peter Zijlstra wrote:
>
> > On Wed, May 01, 2013 at 05:22:05PM +0300, Julian Anastasov wrote:
> >
> > > 2. Same without need_resched because cond_resched already
> > > performs the same checks:
> > >
> > > static void inline cond_resched_rcu_lock(void)
> > > {
> > > #ifndef CONFIG_PREEMPT_RCU
> > > rcu_read_unlock();
> > > cond_resched();
> > > rcu_read_lock();
> > > #endif
> > > }
> >
> > Ah so the 'problem' with this last version is that it does an unconditional /
> > unnessecary rcu_read_unlock().
>
> It is just a barrier() for the non-preempt case.
>
> > The below would be in line with all the other cond_resched*() implementations.
>
> ...
>
> > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > index 802a751..fd2c77f 100644
> > --- a/include/linux/sched.h
> > +++ b/include/linux/sched.h
> > @@ -2449,6 +2449,13 @@ extern int __cond_resched_softirq(void);
> > __cond_resched_softirq(); \
> > })
> >
> > +extern int __cond_resched_rcu(void);
> > +
> > +#define cond_resched_rcu() ({ \
> > + __might_sleep(__FILE__, __LINE__, 0); \
>
> I see your goal. But digging into __might_sleep()
> I see that rcu_sleep_check() will scream for the non-preempt
> case because we are under rcu_read_lock.
>
> What about such inline version:
>
> static void inline cond_resched_rcu(void)
> {
> #ifndef CONFIG_PREEMPT_RCU
> rcu_read_unlock();
> __might_sleep(__FILE__, __LINE__, 0);
> cond_resched();
> rcu_read_lock();
> #else
> __might_sleep(__FILE__, __LINE__, 0);
> rcu_lockdep_assert(rcu_preempt_depth() == 1,
> "Illegal cond_resched_rcu() context");
The above requires that include/linux/sched.h be included. This might
be OK, but please check the current intended uses.
Thanx, Paul
> #endif
> }
>
> It will restrict to single RCU lock level for all
> RCU implementations. But we don't have _cond_resched_rcu
> helper for two reasons:
>
> - __might_sleep uses __FILE__, __LINE__
> - only cond_resched generates code, so need_resched() is
> avoided
>
> > + __cond_resched_rcu(); \
> > +})
> > +
> > /*
> > * Does a critical section need to be broken due to another
> > * task waiting?: (technically does not depend on CONFIG_PREEMPT,
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index 7d7901a..2b3b4e6 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -4358,6 +4358,20 @@ int __sched __cond_resched_softirq(void)
> > }
> > EXPORT_SYMBOL(__cond_resched_softirq);
> >
> > +int __sched __cond_resched_rcu(void)
> > +{
> > +#ifndef CONFIG_PREEMPT_RCU
> > + if (should_resched()) {
> > + rcu_read_unlock();
> > + __cond_resched();
> > + rcu_read_lock();
> > + return 1;
> > + }
> > +#endif
> > + return 0;
> > +}
> > +EXPORT_SYMBOL(__cond_resched_rcu);
> > +
>
> Regards
>
> --
> Julian Anastasov <ja@ssi.bg>
>
next prev parent reply other threads:[~2013-05-01 19:04 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-04-30 2:52 [PATCH v2 0/2] sched: Add cond_resched_rcu_lock() helper Simon Horman
2013-04-30 2:52 ` [PATCH v2 1/2] " Simon Horman
2013-04-30 7:12 ` Julian Anastasov
2013-04-30 7:29 ` Simon Horman
2013-04-30 7:52 ` Julian Anastasov
2013-05-01 9:10 ` Peter Zijlstra
2013-05-01 12:46 ` Paul E. McKenney
2013-05-01 14:32 ` Paul E. McKenney
2013-05-02 7:27 ` Peter Zijlstra
2013-05-01 15:17 ` Peter Zijlstra
2013-05-01 15:29 ` Eric Dumazet
2013-05-01 15:59 ` Peter Zijlstra
2013-05-01 16:02 ` Paul E. McKenney
2013-05-01 16:57 ` Peter Zijlstra
2013-05-01 17:30 ` Paul E. McKenney
2013-05-01 14:22 ` Julian Anastasov
2013-05-01 15:55 ` Peter Zijlstra
2013-05-01 18:22 ` Julian Anastasov
2013-05-01 19:04 ` Paul E. McKenney [this message]
2013-05-02 7:26 ` Peter Zijlstra
2013-05-02 10:06 ` Julian Anastasov
2013-05-02 15:54 ` Julian Anastasov
2013-05-02 17:32 ` Paul E. McKenney
2013-05-02 18:55 ` Julian Anastasov
2013-05-02 19:24 ` Julian Anastasov
2013-05-02 19:34 ` Paul E. McKenney
2013-05-02 20:19 ` Julian Anastasov
2013-05-02 22:31 ` Paul E. McKenney
2013-05-03 7:52 ` Julian Anastasov
2013-05-03 16:30 ` Paul E. McKenney
2013-05-03 17:04 ` Peter Zijlstra
2013-05-03 17:34 ` Paul E. McKenney
2013-05-03 18:09 ` Peter Zijlstra
2013-05-03 17:47 ` Julian Anastasov
2013-05-04 7:23 ` Julian Anastasov
2013-05-04 18:03 ` Paul E. McKenney
2013-04-30 2:52 ` [PATCH v2 2/2] ipvs: Use cond_resched_rcu_lock() helper when dumping connections Simon Horman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130501190412.GU3780@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=dipankar@in.ibm.com \
--cc=eric.dumazet@gmail.com \
--cc=horms@verge.net.au \
--cc=ja@ssi.bg \
--cc=linux-kernel@vger.kernel.org \
--cc=lvs-devel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=netdev@vger.kernel.org \
--cc=netfilter-devel@vger.kernel.org \
--cc=pablo@netfilter.org \
--cc=peterz@infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).