linux-kernel - Re: [v3.4-rc1] ACPI regression bisected

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.LFD.2.02.1204031001510.2542@ionos>
Date:	Tue, 3 Apr 2012 10:11:36 +0200 (CEST)
From:	Thomas Gleixner <tglx@...utronix.de>
To:	Andi Kleen <ak@...ux.intel.com>
cc:	Jörg Otte <jrg.otte@...glemail.com>,
	linux-kernel@...r.kernel.org, Len Brown <lenb@...nel.org>
Subject: Re: [v3.4-rc1] ACPI regression bisected

On Mon, 2 Apr 2012, Andi Kleen wrote:

> >  ACPI: Added _OSI(Processor Aggregator Device)
> >  ACPI: EC: EC description table is found, configuring boot EC
> > 
> > Comparing this with the displays of kernel v3.3 the next expected
> > outputs would be:
> > 
> >  ACPI: Interpreter enabled
> >  ACPI: (supports S0 S5)
> > 
> > 
> > I bisected the problem to commit:
> > 6fe0d0628245fdcd6fad8b837c81e8f7ebc3364d  "ACPI: Make ACPI interrupt threaded"
> 
> Hmm, that's odd. It should not really change any semantics. I assume you double
> checked you hit the right commit?

Well it does change the semantics depending on the type of the
interrupt. The above commit does:

-       if (request_irq(irq, acpi_irq, IRQF_SHARED, "acpi", acpi_irq)) {
+       if (request_threaded_irq(irq, NULL, acpi_irq, IRQF_SHARED, "acpi",
+                                acpi_irq)) {

which creates a threaded handler and the primary handler is the
default core one which merily returns IRQ_WAKE_THREAD. Though that
leaves the interrupt line unmasked and for level type interrupts this
is not working. So you either need a primary handler which masks the
irq at the device level or you can add IRQF_ONESHOT to let the core
code deal with the mask/unmask business. See patch below.

Now the problem with this approach is that if this is really a shared
interrupt the core code will either reject this request or the one of
the other device which shares that interrupt line, depending which one
request first. The reason is that we require all devices to agree on
that flag - otherwise a long running/waiting/sleeping threaded handler
might block out the other device for a long time.

If it's never shared, nothing to worry about. Otherwise you need to
provide a primary handler which silences the interrupt at the device
level and unmask it when your threaded handler is done.

Thanks,

	tglx

diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
index ba14fb9..1b81bc9 100644
--- a/drivers/acpi/osl.c
+++ b/drivers/acpi/osl.c
@@ -607,7 +607,8 @@ acpi_os_install_interrupt_handler(u32 gsi, acpi_osd_handler handler,

 	acpi_irq_handler = handler;
 	acpi_irq_context = context;
-	if (request_threaded_irq(irq, NULL, acpi_irq, IRQF_SHARED, "acpi",
+	if (request_threaded_irq(irq, NULL, acpi_irq,
+				 IRQF_SHARED | IRQF_ONESHOT, "acpi",
 				 acpi_irq)) {
 		printk(KERN_ERR PREFIX "SCI (IRQ%d) allocation failed\n", irq);
 		acpi_irq_handler = NULL;
--
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/