linux-kernel - Re: [RFC PATCH 12/30] rcu: Prepare rcu_read_[un]lock

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: <20181017005520.GA146369@joelaf.mtv.corp.google.com>
Date:   Tue, 16 Oct 2018 17:55:20 -0700
From:   Joel Fernandes <joel@...lfernandes.org>
To:     Frederic Weisbecker <frederic@...nel.org>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Peter Zijlstra <peterz@...radead.org>,
        "David S . Miller" <davem@...emloft.net>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        "Paul E . McKenney" <paulmck@...ux.vnet.ibm.com>,
        Ingo Molnar <mingo@...nel.org>,
        Frederic Weisbecker <fweisbec@...il.com>,
        Mauro Carvalho Chehab <mchehab@...pensource.com>,
        rostedt@...dmis.org
Subject: Re: [RFC PATCH 12/30] rcu: Prepare rcu_read_[un]lock_bh() for
 handling softirq mask

On Wed, Oct 17, 2018 at 02:44:19AM +0200, Frederic Weisbecker wrote:
> On Mon, Oct 15, 2018 at 10:28:44PM -0700, Joel Fernandes wrote:
> > > diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
> > > index f8ec3d4..490358c 100644
> > > --- a/crypto/pcrypt.c
> > > +++ b/crypto/pcrypt.c
> > > @@ -73,12 +73,13 @@ struct pcrypt_aead_ctx {
> > >  static int pcrypt_do_parallel(struct padata_priv *padata, unsigned int *cb_cpu,
> > >  			      struct padata_pcrypt *pcrypt)
> > >  {
> > > +	unsigned int bh;
> > >  	unsigned int cpu_index, cpu, i;
> > >  	struct pcrypt_cpumask *cpumask;
> > >  
> > >  	cpu = *cb_cpu;
> > >  
> > > -	rcu_read_lock_bh();
> > > +	bh = rcu_read_lock_bh();
> > >  	cpumask = rcu_dereference_bh(pcrypt->cb_cpumask);
> > >  	if (cpumask_test_cpu(cpu, cpumask->mask))
> > >  			goto out;
> > > @@ -95,7 +96,7 @@ static int pcrypt_do_parallel(struct padata_priv *padata, unsigned int *cb_cpu,
> > >  	*cb_cpu = cpu;
> > >  
> > >  out:
> > > -	rcu_read_unlock_bh();
> > > +	rcu_read_unlock_bh(bh);
> > >  	return padata_do_parallel(pcrypt->pinst, padata, cpu);
> > >  }
> > 
> > This complicates the RCU API for -bh and doesn't look pretty at all. Is there
> > anything better we can do so we don't have to touch existing readers at all?
> 
> Indeed, so I'm going to give up with the idea of converting all the callers
> in once, this is unmaintainable anyway. I'll keep the RCU API as is for now,
> ie: disable all softirqs, and we'll see later if we need per vector granularity.
> Surely that would be too fun to handle, with per vector quiescent states and grace
> periods ;-)

Cool, sounds good.

> > 
> > Also, I thought softirqs were kind of a thing of the past, and threaded
> > interrupts are the more preferred interrupt bottom halves these days,
> > especially for -rt. Maybe that was just wishful thinking on my part :-)
> 
> We all wish that. I think it was the plan but threaded IRQs involve context
> switches and IIUC it's the border that's hard to cross on some performance
> measurements.

Ok, thanks.

- Joel