[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150507133022.7da17fbb@thinkpad-w530>
Date: Thu, 7 May 2015 13:30:22 +0200
From: David Hildenbrand <dahi@...ux.vnet.ibm.com>
To: Ingo Molnar <mingo@...nel.org>
Cc: Peter Zijlstra <peterz@...radead.org>,
linux-kernel@...r.kernel.org, mingo@...hat.com,
yang.shi@...driver.com, bigeasy@...utronix.de,
benh@...nel.crashing.org, paulus@...ba.org,
akpm@...ux-foundation.org, heiko.carstens@...ibm.com,
schwidefsky@...ibm.com, borntraeger@...ibm.com, mst@...hat.com,
tglx@...utronix.de, David.Laight@...LAB.COM, hughd@...gle.com,
hocko@...e.cz, ralf@...ux-mips.org, herbert@...dor.apana.org.au,
linux@....linux.org.uk, airlied@...ux.ie, daniel.vetter@...el.com,
linux-mm@...ck.org, linux-arch@...r.kernel.org
Subject: Re: [PATCH RFC 01/15] uaccess: count pagefault_disable() levels in
pagefault_disabled
>
> * David Hildenbrand <dahi@...ux.vnet.ibm.com> wrote:
>
> > > On Thu, May 07, 2015 at 12:50:53PM +0200, David Hildenbrand wrote:
> > > > Just to make sure we have a common understanding (as written in my cover
> > > > letter):
> > > >
> > > > Your suggestion won't work with !CONFIG_PREEMPT (!CONFIG_PREEMPT_COUNT). If
> > > > there is no preempt counter, in_atomic() won't work.
> > >
> > > But there is, we _always_ have a preempt_count, and irq_enter() et al.
> > > _always_ increment the relevant bits.
> > >
> > > The thread_info::preempt_count field it never under PREEMPT_COUNT
> > > include/asm-generic/preempt.h provides stuff regardless of
> > > PREEMPT_COUNT.
> > >
> > > See how __irq_enter() -> preempt_count_add(HARDIRQ_OFFSET) ->
> > > __preempt_count_add() _always_ just works.
> > >
> > > Its only things like preempt_disable() / preempt_enable() that get
> > > munged depending on PREEMPT_COUNT/PREEMPT.
> > >
> >
> > Sorry for the confusion. Sure, there is always the count.
> >
> > My point is that preempt_disable() won't result in an in_atomic() == true
> > with !PREEMPT_COUNT, so I don't see any point in adding in to the pagefault
> > handlers. It is not reliable.
>
> That's why we have the preempt_count_inc()/dec() methods that are
> always available.
>
> So where's the problem?
My point:
Getting rid of PREEMPT_COUNT (and therefore always doing
preempt_count_inc()/dec()) will make preempt_disable() __never__ be a NOP.
So with !CONFIG_PREEMPT we will do preemption stuff that is simply not needed.
Two concepts that share one mechanism. I think this is broken.
David
--
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/
Powered by blists - more mailing lists