From: Boris Brezillon <boris.brezillon@free-electrons.com>
To: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Mark Rutland <mark.rutland@arm.com>,
Thomas Gleixner <tglx@linutronix.de>,
Jason Cooper <jason@lakedaemon.net>,
Nicolas Ferre <nicolas.ferre@atmel.com>,
Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>,
Alexandre Belloni <alexandre.belloni@free-electrons.com>,
Rob Herring <robh+dt@kernel.org>, Pawel Moll <Pawel.Moll@arm.com>,
Ian Campbell <ijc+devicetree@hellion.org.uk>,
Kumar Gala <galak@codeaurora.org>,
"devicetree@vger.kernel.org" <devicetree@vger.kernel.org>,
"Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
"linux-arm-kernel@lists.infradead.org"
<linux-arm-kernel@lists.infradead.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v4 3/5] irqchip: Add DT binding doc for the virtual irq demuxer chip
Date: Wed, 11 Feb 2015 16:03:17 +0100 [thread overview]
Message-ID: <20150211160317.6d6c64e5@bbrezillon> (raw)
In-Reply-To: <2437801.56xPyk3atd@vostro.rjw.lan>
On Wed, 11 Feb 2015 16:17:20 +0100
"Rafael J. Wysocki" <rjw@rjwysocki.net> wrote:
> On Wednesday, February 11, 2015 02:43:45 PM Mark Rutland wrote:
> > [...]
> >
> > > > > > > +static irqreturn_t __handle_irq_event_percpu(unsigned int irq, struct irqaction *action)
> > > > > > > +{
> > > > > > > + /*
> > > > > > > + * During suspend we must not call potentially unsafe irq handlers.
> > > > > > > + * See suspend_suspendable_actions.
> > > > > > > + */
> > > > > > > + if (unlikely(action->flags & IRQF_NO_ACTION))
> > > > > > > + return IRQ_NONE;
> > > > > >
> > > > > > Thomas was trying to avoid any new conditional code in the interrupt
> > > > > > handling path, that's why I added a suspended_action list in my
> > > > > > proposal.
> > > > > > Even if your 'unlikely' statement make things better I'm pretty sure it
> > > > > > adds some latency.
> > > > >
> > > > > I can see that we don't want to add more code here to keep things
> > > > > clean/pure, but I find it hard to believe that a single bit test and
> > > > > branch (for data that should be hot in the cache) are going to add
> > > > > measurable latency to a path that does pointer chasing to get to the
> > > > > irqaction in the first place. I could be wrong though, and I'm happy to
> > > > > benchmark.
> > > >
> > > > Again, I don't have enough experience to say this is (or isn't)
> > > > impacting irq handling latency, I'm just reporting what Thomas told me.
> > > >
> > > > >
> > > > > It would be possible to go for your list shuffling approach here while
> > > > > still keeping the flag internal and all the logic hidden away in
> > > > > kernel/irq/pm.c. I wasn't sure how actions could be manipulated during
> > > > > suspend, which made me wary of moving them to a separate list.
> > > >
> > > > Moving them to a temporary list on suspend and restoring them on
> > > > resume should not be a problem.
> > > > The only drawback I see is that actions might be reordered after the
> > > > first resume (anyway, relying on shared irq action order is dangerous
> > > > IMHO).
> > >
> > > We considered doing that too and saw some drawbacks (in addition to the
> > > reordering of actions you've mentioned). It added just too much complexity
> > > to the IRQ suspend-resume code.
> > >
> > > I, personally, would be fine with adding an IRQ flag to silence the
> > > warning mentioned by Alexandre. Something like IRQD_TIMER_SHARED that would
> > > be set automatically if someone requested IRQF_TIMER | IRQF_SHARED.
> > >
> > > Thoughts?
> >
> > Even if the timer driver does that, we still require the other handlers
> > sharing the line to do the right thing across suspend, no? So either
> > their actions need to be masked at suspend time, or the handlers need to
> > detect when they're called during suspend and return early.
>
> Well, the issue at hand is about things that share an IRQ with a timer AFAICS.
>
> That is odd enough already and I'd say everyone in that situation needs to
> be prepared to take the pain (including having to check if the device is not
> suspended in their interrupt handlers).
>
> And quite frankly they need to do that already, because we've never suspended
> timer IRQs.
>
> > So for the flag at request time approach to work, all the drivers using
> > the interrupt would have to flag they're safe in that context.
>
> Something like IRQF_"I can share the line with a timer" I guess? That wouldn't
> hurt and can be checked at request time even.
>
> > I'm not averse to having that (only a few drivers shuold be affected and
> > we can sanity check them now).
>
> Right.
Okay, if everyone agrees on this solution, then I'm fine with that too
(even if I'm a bit disappointed to have spent so much time on this
problem to eventually end-up with a simple IRQF_SHARED_TIMER flag :-().
--
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
next prev parent reply other threads:[~2015-02-11 15:03 UTC|newest]
Thread overview: 70+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-01-29 10:33 [PATCH v4 0/5] ARM: at91: fix irq_pm_install_action WARNING Boris Brezillon
2015-01-29 10:33 ` [PATCH v4 1/5] genirq: Authorize chained handlers to remain disabled when initialized Boris Brezillon
2015-01-29 10:33 ` [PATCH v4 2/5] irqchip: add virtual demultiplexer implementation Boris Brezillon
2015-02-10 15:00 ` Peter Zijlstra
2015-02-10 15:20 ` Boris Brezillon
2015-02-10 15:43 ` [PATCH] genirq: fix virtual irq demuxer related comments Boris Brezillon
2015-02-10 16:14 ` Peter Zijlstra
2015-02-20 16:12 ` Mark Rutland
2015-02-20 16:17 ` Peter Zijlstra
2015-02-10 15:48 ` [PATCH v4 2/5] irqchip: add virtual demultiplexer implementation Mark Rutland
2015-01-29 10:33 ` [PATCH v4 3/5] irqchip: Add DT binding doc for the virtual irq demuxer chip Boris Brezillon
2015-02-10 15:36 ` Mark Rutland
2015-02-10 15:52 ` Boris Brezillon
2015-02-10 16:06 ` Boris Brezillon
2015-02-10 16:16 ` Mark Rutland
2015-02-10 16:20 ` Boris Brezillon
2015-02-10 20:48 ` Mark Rutland
2015-02-11 8:53 ` Boris Brezillon
2015-02-11 11:11 ` Mark Rutland
2015-02-11 12:24 ` Boris Brezillon
2015-02-11 12:36 ` Mark Rutland
2015-02-11 13:38 ` Alexandre Belloni
2015-02-11 13:48 ` Mark Rutland
2015-02-11 14:55 ` Rafael J. Wysocki
2015-02-11 14:43 ` Mark Rutland
2015-02-11 15:17 ` Rafael J. Wysocki
2015-02-11 15:03 ` Boris Brezillon [this message]
2015-02-11 15:39 ` Rafael J. Wysocki
2015-02-11 15:23 ` Mark Rutland
2015-02-11 15:12 ` Mark Rutland
2015-02-11 15:51 ` Rafael J. Wysocki
2015-02-11 15:57 ` Mark Rutland
2015-02-11 16:15 ` Boris Brezillon
2015-02-11 16:32 ` Mark Rutland
2015-02-11 16:38 ` Boris Brezillon
2015-02-11 17:17 ` Mark Rutland
2015-02-20 14:22 ` Mark Rutland
2015-02-20 14:53 ` Boris Brezillon
2015-02-20 15:16 ` Mark Rutland
2015-02-23 17:00 ` Boris Brezillon
2015-02-23 18:14 ` Mark Rutland
2015-02-23 20:16 ` Boris Brezillon
2015-02-11 16:42 ` Rafael J. Wysocki
2015-02-11 16:28 ` Boris Brezillon
2015-02-11 17:13 ` Mark Rutland
2015-02-11 17:29 ` Boris Brezillon
2015-02-12 10:52 ` Mark Rutland
2015-02-12 11:09 ` Boris Brezillon
2015-02-12 11:23 ` Mark Rutland
2015-02-16 9:49 ` Peter Zijlstra
2015-02-16 9:28 ` Peter Zijlstra
2015-02-16 12:23 ` Mark Rutland
2015-02-19 1:16 ` Rafael J. Wysocki
2015-02-19 11:23 ` Mark Rutland
2015-02-19 22:35 ` Rafael J. Wysocki
2015-02-20 10:31 ` Mark Rutland
2015-02-24 1:02 ` Rafael J. Wysocki
2015-02-24 8:42 ` Boris Brezillon
2015-02-11 14:45 ` Boris Brezillon
2015-02-11 14:39 ` Rafael J. Wysocki
2015-02-11 9:11 ` Peter Zijlstra
2015-02-11 11:15 ` Mark Rutland
2015-02-11 14:31 ` Rafael J. Wysocki
2015-02-11 14:14 ` Mark Rutland
2015-02-11 15:07 ` Rafael J. Wysocki
2015-02-11 15:03 ` Mark Rutland
2015-02-11 14:34 ` Rafael J. Wysocki
2015-01-29 10:33 ` [PATCH v4 4/5] ARM: at91/dt: select VIRT_IRQ_DEMUX for all at91 SoCs Boris Brezillon
2015-01-29 10:33 ` [PATCH v4 5/5] ARM: at91/dt: define a virtual irq demultiplexer chip connected on irq1 Boris Brezillon
2015-02-09 15:47 ` [PATCH v4 0/5] ARM: at91: fix irq_pm_install_action WARNING Boris Brezillon
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=20150211160317.6d6c64e5@bbrezillon \
--to=boris.brezillon@free-electrons.com \
--cc=Pawel.Moll@arm.com \
--cc=alexandre.belloni@free-electrons.com \
--cc=devicetree@vger.kernel.org \
--cc=galak@codeaurora.org \
--cc=ijc+devicetree@hellion.org.uk \
--cc=jason@lakedaemon.net \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=nicolas.ferre@atmel.com \
--cc=plagnioj@jcrosoft.com \
--cc=rafael.j.wysocki@intel.com \
--cc=rjw@rjwysocki.net \
--cc=robh+dt@kernel.org \
--cc=tglx@linutronix.de \
/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).