Re: [rcu] [ INFO: suspicious RCU usage. ]

From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>,
	Fengguang Wu <fengguang.wu@intel.com>, LKP <lkp@01.org>,
	linux-kernel@vger.kernel.org,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>,
	linux-arm-kernel@lists.infradead.org,
	Arnd Bergmann <arnd@arndb.de>, MarkRutland <mark.rutland@arm.com>
Subject: Re: [rcu] [ INFO: suspicious RCU usage. ]
Date: Wed, 4 Feb 2015 08:28:14 -0800	[thread overview]
Message-ID: <20150204162814.GG5370@linux.vnet.ibm.com> (raw)
In-Reply-To: <1423066256.24415.13.camel@AMDC1943>

On Wed, Feb 04, 2015 at 05:10:56PM +0100, Krzysztof Kozlowski wrote:
> On śro, 2015-02-04 at 07:56 -0800, Paul E. McKenney wrote:
> > On Wed, Feb 04, 2015 at 04:22:28PM +0100, Krzysztof Kozlowski wrote:
> > > 
> > > Actually the timeout versions but I think that doesn't matter.
> > > The wait_on_bit will busy-loop with testing for the bit. Inside the loop
> > > it calls the 'action' which in my case will be bit_wait_io_timeout().
> > > This calls schedule_timeout().
> > 
> > Ah, good point.
> > 
> > > See proof of concept in attachment. One observed issue: hot unplug from
> > > commandline takes a lot more time. About 7 seconds instead of ~0.5.
> > > Probably I did something wrong.
> > 
> > Well, you do set the timeout to five seconds, and so if the condition
> > does not get set before the surviving CPU finds its way to the
> > out_of_line_wait_on_bit_timeout(), you are guaranteed to wait for at
> > least five seconds.
> >
> > One alternative approach would be to have a loop around a series of
> > shorter waits.  Other thoughts?
> 
> Right! That was the issue. It seems it works. I'll think also on
> self-adapting interval as you said below. I'll test it more and send a
> patch.

Sounds good!

Are you doing ARM, ARM64, or both?  I of course vote for both.  ;-)

							Thanx, Paul

> Best regards,
> Krzysztof
> 
> > 
> > > > You know, this situation is giving me a bad case of nostalgia for the
> > > > old Sequent Symmetry and NUMA-Q hardware.  On those platforms, the
> > > > outgoing CPU could turn itself off, and thus didn't need to tell some
> > > > other CPU when it was ready to be turned off.  Seems to me that this
> > > > self-turn-off capability would be a great feature for future systems!
> > > 
> > > There are a lot more issues with hotplug on ARM...
> > 
> > Just trying to clean up this particular corner at the moment.  ;-)
> > 
> > > Patch/RFC attached.
> > 
> > Again, I believe that you will need to loop over a shorter timeout
> > in order to get reasonable latencies.  If waiting a millisecond at
> > a time is an energy-efficiency concern (don't know why it would be
> > in this rare case, but...), then one approach would be to start
> > with very short waits, then increase the wait time, for example,
> > doubling the wait time on each pass through the loop would result
> > in a smallish number of wakeups, but would mean that you waited
> > no more than twice as long as necessary.
> > 
> > Thoughts?
> 
> 