| 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
| ||
|
Message-ID: <20150212120917.44e58bf6@bbrezillon>
Date: Thu, 12 Feb 2015 12:09:17 +0100
From: Boris Brezillon <boris.brezillon@...e-electrons.com>
To: Mark Rutland <mark.rutland@....com>
Cc: "Rafael J. Wysocki" <rjw@...ysocki.net>,
Thomas Gleixner <tglx@...utronix.de>,
Jason Cooper <jason@...edaemon.net>,
Nicolas Ferre <nicolas.ferre@...el.com>,
Jean-Christophe Plagniol-Villard <plagnioj@...osoft.com>,
Alexandre Belloni <alexandre.belloni@...e-electrons.com>,
Rob Herring <robh+dt@...nel.org>,
Pawel Moll <Pawel.Moll@....com>,
Ian Campbell <ijc+devicetree@...lion.org.uk>,
Kumar Gala <galak@...eaurora.org>,
"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
"Rafael J. Wysocki" <rafael.j.wysocki@...el.com>,
"linux-arm-kernel@...ts.infradead.org"
<linux-arm-kernel@...ts.infradead.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v4 3/5] irqchip: Add DT binding doc for the virtual irq
demuxer chip
On Thu, 12 Feb 2015 10:52:15 +0000
Mark Rutland <mark.rutland@....com> wrote:
> [...]
>
> > > diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
> > > index d9b05b5..2b8ff50 100644
> > > --- a/include/linux/interrupt.h
> > > +++ b/include/linux/interrupt.h
> > > @@ -57,6 +57,9 @@
> > > * IRQF_NO_THREAD - Interrupt cannot be threaded
> > > * IRQF_EARLY_RESUME - Resume IRQ early during syscore instead of at device
> > > * resume time.
> > > + * IRQF_SHARED_TIMER_OK - Interrupt is safe to be shared with a timer. The
> > > + * handler may be called spuriously during suspend
> > > + * without issue.
> > > */
> > > #define IRQF_DISABLED 0x00000020
> > > #define IRQF_SHARED 0x00000080
> > > @@ -70,8 +73,10 @@
> > > #define IRQF_FORCE_RESUME 0x00008000
> > > #define IRQF_NO_THREAD 0x00010000
> > > #define IRQF_EARLY_RESUME 0x00020000
> > > +#define __IRQF_TIMER_SIBLING_OK 0x00040000
> > >
> > > #define IRQF_TIMER (__IRQF_TIMER | IRQF_NO_SUSPEND | IRQF_NO_THREAD)
> > > +#define IRQF_SHARED_TIMER_OK (IRQF_SHARED | __IRQF_TIMER_SIBLING_OK)
> > >
> > > /*
> > > * These values can be returned by request_any_context_irq() and
> > > diff --git a/kernel/irq/pm.c b/kernel/irq/pm.c
> > > index 3ca5325..e4ec91a 100644
> > > --- a/kernel/irq/pm.c
> > > +++ b/kernel/irq/pm.c
> > > @@ -28,6 +28,47 @@ bool irq_pm_check_wakeup(struct irq_desc *desc)
> > > }
> > >
> > > /*
> > > + * Check whether an interrupt is safe to occur during suspend.
> > > + *
> > > + * Physical IRQ lines may be shared between devices which may be expected to
> > > + * raise interrupts during suspend (e.g. timers) and those which may not (e.g.
> > > + * anything we cut the power to). Not all handlers will be safe to call during
> > > + * suspend, so we need to scream if there's the possibility an unsafe handler
> > > + * will be called.
> > > + *
> > > + * A small number of handlers are safe to be shared with timer interrupts, and
> > > + * we don't want to warn erroneously for these. Such handlers will not poke
> > > + * hardware that's not powered or call into kernel infrastructure not available
> > > + * during suspend. These are marked with __IRQF_TIMER_SIBLING_OK.
> > > + */
> > > +bool irq_safe_during_suspend(struct irq_desc * desc, struct irqaction *action)
> > > +{
> > > + const unsigned int safe_flags =
> > > + __IRQF_TIMER_SIBLING_OK | IRQF_NO_SUSPEND;
> > > +
> > > + /*
> > > + * If no-one wants to be called during suspend, or if everyone does,
> > > + * then there's no potential conflict.
> > > + */
> > > + if (!desc->no_suspend_depth)
> > > + return true;
> > > + if (desc->no_suspend_depth == desc->nr_actions)
> > > + return true;
Just another nit, can't we also return early when
desc->nr_actions == 1 (I mean, the handler cannot conflict with anything
since it is the only one registered) ?
> > > +
> > > + /*
> > > + * If any action hasn't asked to be called during suspend or is not
> > > + * happy to be called during suspend, we have a potential problem.
> > > + */
> > > + if (!(action->flags & safe_flags))
> > > + return false;
> > else if (!(action->flags & IRQF_NO_SUSPEND) ||
> > desc->no_suspend_depth > 1)
> > return true;
> >
> > Am I missing something or is the following loop only required if
> > we're adding an action with the IRQF_NO_SUSPEND flag set for the
> > first time ?
>
> With the check above we could return true incorrectly after the first
> time we return true. Consider adding the following in order to an empty
> desc:
>
> flags = IRQF_SHARED // safe, returns true
> flags = IRQF_NO_SUSPEND // unsafe, returns false
> flags = IRQF_NO_SUSPEND // unsafe, but returns true
Yep, you're right.
>
> Currently it shouldn't matter as the only caller is a WARN_ON_ONCE(),
> but it seems unfortunate to allow this.
Absolutely, forget about that, I guess we don't have to optimize that
test anyway.
>
> We'd also run the loop until we had at least two IRQF_NO_SUSPEND
> irqactions:
>
> flags = IRQF_SHARED_TIMER_OK // early return
> flags = IRQF_NO_SUSPEND // run loop
> flags = IRQF_SHARED_TIMER_OK // run loop
Hm, no, this one would return directly (it's an '||' operator not an
'&&' one), because we're not adding an IRQF_NO_SUSPEND handler here, and
adding IRQF_SHARED_TIMER_OK is always safe, isn't it ?
> flags = IRQF_SHARED_TIMER_OK // run loop
> flags = IRQF_SHARED_TIMER_OK // run loop
> flags = IRQF_NO_SUSPEND // don't run loop.
> flags = IRQF_SHARED_TIMER_OK // don't run loop
>
> I assume that we only have one IRQF_NO_SUSPEND action sharing the line
> anyway in your case?
Yep.
>
> Given that we'll only bother to run the test if there's a mismatch
> between desc->no_suspend_depth and desc->nr_actions, I don't think we
> win much. These cases should be rare in practice, the tests only
> performed when we request the irq, and there shouldn't be that many
> actions to loop over.
Sure, never mind, as I said, I'm not sure extra optimization is needed
here.
Regards,
Boris
--
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
--
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