netdev - Re: [PATCH] b44: fix resume, request

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: <201010120808.05713.james@albanarts.com>
Date:	Tue, 12 Oct 2010 08:08:05 +0100
From:	James Hogan <james@...anarts.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Gary Zambrano <zambrano@...adcom.com>,
	Jiri Pirko <jpirko@...hat.com>,
	FUJITA Tomonori <fujita.tomonori@....ntt.co.jp>,
	Hauke Mehrtens <hauke@...ke-m.de>,
	Larry Finger <Larry.Finger@...inger.net>,
	netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
	"David S. Miller" <davem@...emloft.net>
Subject: Re: [PATCH] b44: fix resume, request_irq after hw reset

On Tuesday 12 October 2010 00:34:40 Andrew Morton wrote:
> On Tue, 12 Oct 2010 00:22:12 +0100
> 
> James Hogan <james@...anarts.com> wrote:
> > This driver was hanging on resume because it was requesting a shared irq
> > that it wasn't ready to immediately handle, which was tested in
> > request_irq because of the CONFIG_DEBUG_SHIRQ config option. The
> > interrupt handler tried to read the interrupt status register but for
> > some reason it hung the system.
> > 
> > The request_irq is now moved a bit later after resetting the hardware
> > which seems to fix it.
> > 
> > Signed-off-by: James Hogan <james@...anarts.com>
> > ---
> > 
> >  drivers/net/b44.c |   12 ++++++------
> >  1 files changed, 6 insertions(+), 6 deletions(-)
> > 
> > diff --git a/drivers/net/b44.c b/drivers/net/b44.c
> > index 1e620e2..dbba981 100644
> > --- a/drivers/net/b44.c
> > +++ b/drivers/net/b44.c
> > @@ -2296,12 +2296,6 @@ static int b44_resume(struct ssb_device *sdev)
> > 
> >  	if (!netif_running(dev))
> >  	
> >  		return 0;
> > 
> > -	rc = request_irq(dev->irq, b44_interrupt, IRQF_SHARED, dev->name, 
dev);
> > -	if (rc) {
> > -		netdev_err(dev, "request_irq failed\n");
> > -		return rc;
> > -	}
> > -
> > 
> >  	spin_lock_irq(&bp->lock);
> >  	
> >  	b44_init_rings(bp);
> > 
> > @@ -2309,6 +2303,12 @@ static int b44_resume(struct ssb_device *sdev)
> > 
> >  	netif_device_attach(bp->dev);
> >  	spin_unlock_irq(&bp->lock);
> > 
> > +	rc = request_irq(dev->irq, b44_interrupt, IRQF_SHARED, dev->name, 
dev);
> > +	if (rc) {
> > +		netdev_err(dev, "request_irq failed\n");
> > +		return rc;
> > +	}
> > +
> > 
> >  	b44_enable_ints(bp);
> >  	netif_wake_queue(dev);
> 
> OK, running the interrupt handler before b44_init_hw() is presumably
> the problem here.
> 
> The hardware surely won't be generating interrupts until we've run
> b44_init_hw() and b44_enable_ints(), so this patch really is only to
> keep CONFIG_DEBUG_SHIRQ happy.

For me it's mainly to keep CONFIG_DEBUG_SHIRQ happy (Fedora has this switched 
on), but since it's a shared IRQ, there is still a chance it could be 
called before enabling it's own interrupts by a different device on the same 
IRQ.

It makes sense to me why it's disabling the IRQ now, in case another device 
triggers it when it cannot handle it safely. I also tried calling the 
interrupt directly before the free_irq in the suspend function to check that 
it wasn't being done too late, and it didn't fail, so possibly it is the core 
suspension that makes it start failing until it is brought back up properly.

Cheers
James
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html