Re: 2.4.{1,2pre1} oops in via82cxxx_audio (?)

[rui.sousa@conexant.com]

[-- Attachment #1: Type: text/plain, Size: 1345 bytes --]





On 05/02/01 20:55 Jeff Garzik wrote:

> rui.sousa@conexant.com wrote:
> >
> > On Mon, 5 Feb 2001, Matthew Harrell wrote:
> >
> > > : Ouch.  After applying the attached patch, do any of the assertions
> > > : trigger?  (You should get a message 'Assertion failed! ...' right
before
> > > : the oops)
> > >
> > > : +   assert (chan->sgtable != NULL);
> > >
> > > Yep, I get this one "chan->sgtable != NULL".  I have no idea what
this means
> > > but I got it one out of the two times I tried.
> > >
> >
> > Is there any other device on the same irq?
>
> Yes, it looks like:
>
>   Matthew Harrell wrote:
>   >   PCI: Found IRQ 10 for device 00:14.5
>   >   PCI: The same IRQ used for device 00:04.0
>
> The attached patch, sent to Matthew privately, apparently has fixed his
> problem.  Right now it looks like an out-of-band interrupt...   The
> interrupt is enabled via request_irq, and its shared so the interrupt
> handler will be called.  However the channel isn't active so the SG
> table hasn't been allocated yet.

But your interrupt status register should indicate that it wasn't the
sound device that generated the interrupt...


Matthew, can you try the attached patched and report the output?
You should apply it on a clean 2.4.1 (without the patches Jeff sent you).
Try only sound playback.

Rui Sousa


(See attached file: patch)

[-- Attachment #2: patch --]
[-- Type: application/octet-stream, Size: 706 bytes --]

--- via82cxxx_audio.c.orig	Mon Feb  5 21:17:19 2001
+++ via82cxxx_audio.c	Mon Feb  5 21:23:13 2001
@@ -1598,6 +1598,7 @@
 {
 	struct via_info *card = dev_id;
 	u32 status32;
+	u8 status;
 
 	/* to minimize interrupt sharing costs, we use the SGD status
 	 * shadow register to check the status of all inputs and
@@ -1609,6 +1610,12 @@
 		return;
 
 	DPRINTK ("intr, status32 == 0x%08X\n", status32);
+
+	if (card->ch_out.sgtable == NULL) {
+	        status = inb (card->baseaddr + VIA_BASE0_PCM_OUT_CHAN) & (VIA_SGD_FLAG | VIA_SGD_EOL | VIA_SGD_STOPPED);
+		printk("%x %x\n", status32, status);
+		return;
+	}
 
 	/* synchronize interrupt handling under SMP.  this spinlock
 	 * goes away completely on UP