Re: [proaudio] rtirq borked

[Dominique Michel <dominique.michel@xxxxxxxxxxxx>]

Le Thu, 14 Sep 2006 16:28:53 +0100,
karsten <k.gebbert@xxxxxxxxx> a écrit :

> On Thu, 2006-09-14 at 07:58 -0700, Mark Knecht wrote:
> > On 9/14/06, karsten <k.gebbert@xxxxxxxxx> wrote:
> > <SNIP>
> > > ps aux|grep 'IRQ 16'
> > >
> > >
> > > rambazamba ~ # ps aux|grep 'IRQ 16'
> > > root      1002  0.0  0.0      0     0 ?        S<   10:28   0:00 [IRQ
> > > 16]
> > > root      1003  0.0  0.0      0     0 ?        S<   10:28   0:00 [IRQ
> > > 16]
> > > root     18059  0.0  0.1   1624   608 pts/1    R+   16:08   0:00 grep
> > > --colour=auto IRQ 16
> > >
> > 
> > cat /proc/interrupts
> > 
> 
> that would give:
> 
> rambazamba ~ # cat /proc/interrupts 
>            CPU0       CPU1       
>   0:    5682767          0  IO-APIC        [........N/  0]-edge     pit
>   1:        355          0  IO-APIC        [........./  1]-edge
> i8042
>   7:          0          0  IO-APIC        [..P....M./  0]-edge
> parport0
>   8:          2          0  IO-APIC        [........./  0]-edge     rtc
>   9:      56844          0  IO-APIC        [........./  0]-level    acpi
>  12:        106          0  IO-APIC        [........./  0]-edge
> i8042
>  14:     137321          0  IO-APIC        [........./  0]-edge     ide0
>  15:     199137          0  IO-APIC        [........./  0]-edge     ide1
>  16:         42          0  IO-APIC        [........./  0]-level
> yenta, uhci_hcd:usb2, uhci_hcd:usb5, hdsp
>  17:          0          0  IO-APIC        [........./  0]-level
> Intel ICH5
>  18:          0          0  IO-APIC        [........./  0]-level
> uhci_hcd:usb4
>  19:        224          0  IO-APIC        [........./  0]-level
> ehci_hcd:usb1
>  20:          0          0  IO-APIC        [........./  0]-level
> uhci_hcd:usb3
>  21:     239176          0  IO-APIC        [........./  0]-level    eth0
>  22:          2          0  IO-APIC        [........./  0]-level
> ohci1394
>  23:          0          0  IO-APIC        [........./  0]-level
> wifi0
> NMI:          0          0 
> LOC:    5700756    5707083 
> ERR:          0
> MIS:          0
> 
> 
> 

I was having the same problem with my motherboard. To solve this, I removed
acpi from the bios and added "acpi=off" in the grub kernel line. (I think at it
is enough with "acpi=off" but was first trying with the bios.)

You will get a free IRQ 9, and it is the best IRQ for a sound card. After, it
was just to move the card in another slot and change the IRQ in the bios for
this slot to 9.

Another issue in your case is the yenta stuf. I don't know what it is, but it
will be good to have it on another IRQ as the usb.

A good way to have a stable rt system on an uniprocessor is to configure the
bios to use the PIC interface, and to configure the Bios and the cards (by
moving them from slot to slot) to not have any shared IRQ with the PIC
interface. After doing this, it is even not necessary to use the APIC
interface. I would even recommend to stay with the PIC interface because it
have less overrun.

Here is my /proc/interupts as example:
$ cat /proc/interrupts
           CPU0
  0:   12300905  XT-PIC         [........N/  0]  pit
  1:      21230  XT-PIC         [........./  4]  i8042
  2:          0  XT-PIC         [........N/  0]  cascade
  3:     159763  XT-PIC         [........./  0]  ohci_hcd:usb1
  4:      48438  XT-PIC         [........./  0]  eth0, ohci1394
  5:        132  XT-PIC         [........./  0]  ohci_hcd:usb2
  6:       7919  XT-PIC         [........./  0]  ohci_hcd:usb3, ehci_hcd:usb4
  7:          1  XT-PIC         [........./  0]  parport0
  8:          2  XT-PIC         [........./  0]  rtc
  9:    4829897  XT-PIC         [........./  0]  EMU10K1
 10:        132  XT-PIC         [........./  0]  bttv0
 11:     990768  XT-PIC         [........./  0]  nvidia
 14:      55792  XT-PIC         [........./  0]  ide0
 15:         74  XT-PIC         [........./  0]  ide1
NMI:          0
LOC:       1639
ERR:          0
MIS:          0

I am even able to use the proprietary 3D driver now. Before, nvidia was always
sharing the same interrupt as another hardware (generally the sound card or the
bttv), and the system was crashing often. It was necessary to use the free 2D
driver that doesn't use an IRQ.

This IRQ trimming is very important with a rt kernel if you want a stable
system.

Dominique