linux-kernel - Re: [PATCH 0/5] Fix regression introduced by set_irq

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]

Date:	Tue, 20 Oct 2015 16:23:33 +0200
From:	Thomas Petazzoni <thomas.petazzoni@...e-electrons.com>
To:	Russell King - ARM Linux <linux@....linux.org.uk>
Cc:	Jason Cooper <jason@...edaemon.net>,
	Lior Amsalem <alior@...vell.com>, Andrew Lunn <andrew@...n.ch>,
	Tawfik Bayouk <tawfik@...vell.com>,
	Marc Zyngier <marc.zyngier@....com>,
	linux-kernel@...r.kernel.org, Nadav Haklai <nadavh@...vell.com>,
	Gregory Clement <gregory.clement@...e-electrons.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	linux-arm-kernel@...ts.infradead.org,
	Sebastian Hesselbarth <sebastian.hesselbarth@...il.com>
Subject: Re: [PATCH 0/5] Fix regression introduced by set_irq_flags()
 removal

Russell,

On Tue, 20 Oct 2015 15:17:36 +0100, Russell King - ARM Linux wrote:

> However, this is rather worrying.  NOAUTOEN is supposed to avoid enabling
> the interrupt when the interrupt is claimed.
> 
> If, as a result of Rob's patch, we now have a load of IRQs which are
> marked with NOAUTOEN which weren't, that's quite a large regression -
> possibly one which hasn't been properly found (not everyone tests -rc
> kernels) and we may be better to revert Rob's patch to avoid lots of
> breakge being reported when 4.3 is released.

I believe the problem is only for per-CPU interrupts. We have
IRQ_NOAUTOEN set for per-CPU interrupts because:

static inline void irq_set_percpu_devid_flags(unsigned int irq)
{
        irq_set_status_flags(irq,
                             IRQ_NOAUTOEN | IRQ_PER_CPU | IRQ_NOTHREAD |
                             IRQ_NOPROBE | IRQ_PER_CPU_DEVID);
}

Calling set_irq_flags() used to have the effect of *clearing* the
IRQ_NOAUTOEN flag:

-void set_irq_flags(unsigned int irq, unsigned int iflags)
-{
-       unsigned long clr = 0, set = IRQ_NOREQUEST | IRQ_NOPROBE | IRQ_NOAUTOEN;
-
-       if (irq >= nr_irqs) {
-               pr_err("Trying to set irq flags for IRQ%d\n", irq);
-               return;
-       }
-
-       if (iflags & IRQF_VALID)
-               clr |= IRQ_NOREQUEST;
-       if (iflags & IRQF_PROBE)
-               clr |= IRQ_NOPROBE;
-       if (!(iflags & IRQF_NOAUTOEN))
-               clr |= IRQ_NOAUTOEN;
-       /* Order is clear bits in "clr" then set bits in "set" */
-       irq_modify_status(irq, clr, set & ~clr);
-}

I.e, unless you were passing IRQF_NOAUTOEN in your set_irq_flags()
invocation, set_irq_flags() was automatically clearing the IRQ_NOAUTOEN
bit.

But this is really only a per-CPU interrupt problem, which probably
limits the potential regressions caused by Rob's change.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android 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/