Re: More info on hard lock-ups with MIDI on envy24, x86_64 machine

[Lee Revell <rlrevell@joe-job.com>]

On Thu, 2005-08-11 at 16:43 +0100, Daniel James wrote:
> Hi Lee,
> 
> > I guess you have tried to trigger the crash from a console as well
> > as X
> 
> I've done that now. If I boot the system, switch to a console with 
> ctrl-alt-f1 and run nothing but:
> 
> strace seq24
> 
> it starts to go wrong after: 
> 
> open ("/dev/snd/seq", O_RDWR)
> 
> There are a bunch of lines beginning with ioctl, then the killer blow 
> is dealt by:
> 
> ioctl (3, CDROM_DEBUG or SNDRV_SEQ_IOCTL_SUBSCRIBE_PORT
> 
> (This is retyped from hand-written notes of course). The machine is 
> then locked up solid - it's very reproducible too, same output every 
> time. 
> 
> > and there's no Oops output? 
> 
> Nothing at all appears in the console or in /var/log/messages. Should 
> I try increasing the log level or something like that?
> 

Try this (not even compile tested) patch.  Right now it looks like the
interrupt handler could loop forever if something weird happened.  Now
it will at least printk, and try to clear the interrupt pending
register.  This is shamelessly copied from the emu10k1 IRQ handler.

Index: alsa-kernel/pci/ice1712/ice1712.c
===================================================================
RCS file: /cvsroot/alsa/alsa-kernel/pci/ice1712/ice1712.c,v
retrieving revision 1.68
diff -u -r1.68 ice1712.c
--- alsa-kernel/pci/ice1712/ice1712.c	3 Aug 2005 12:40:11 -0000	1.68
+++ alsa-kernel/pci/ice1712/ice1712.c	11 Aug 2005 16:16:42 -0000
@@ -410,13 +410,14 @@
 static irqreturn_t snd_ice1712_interrupt(int irq, void *dev_id, struct pt_regs *regs)
 {
 	ice1712_t *ice = dev_id;
-	unsigned char status;
+	unsigned char status, orig_status;
 	int handled = 0;
 
 	while (1) {
 		status = inb(ICEREG(ice, IRQSTAT));
 		if (status == 0)
 			break;
+		orig_status = status;
 		handled = 1;
 		if (status & ICE1712_IRQ_MPU1) {
 			if (ice->rmidi[0])
@@ -472,6 +473,20 @@
 				snd_pcm_period_elapsed(ice->playback_con_substream);
 			outb(ICE1712_IRQ_CONPBK, ICEREG(ice, IRQSTAT));
 		}
+		if (status) {
+			unsigned char bits;
+			snd_printk(KERN_ERR "ice1712: unhandled interrupt: 0x%08x\n", status);
+			bits = ICE1712_IRQ_MPU1 |
+			       ICE1712_IRQ_TIMER |
+			       ICE1712_IRQ_MPU2 |
+			       ICE1712_IRQ_PROPCM |
+			       ICE1712_IRQ_FM |
+			       ICE1712_IRQ_PBKDS |
+			       ICE1712_IRQ_CONCAP |
+			       ICE1712_IRQ_CONPBK;
+			outb(bits, ICEREG(ice, IRQSTAT));
+		}
+		outb(orig_status, ICEREG(ice, IRQSTAT));
 	}
 	return IRQ_RETVAL(handled);
 }




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf