9 Oct
2006
9 Oct
'06
4:13 p.m.
Hi list.
I have a question regarding IRQ-assignment for an on-board card identified as an Intel
82801DB-ICH4.
It is a crappy card, I know, and you probably ask yourselves why on earth I would want to
put any effort
in configuring such sub-standard equipment. The question, fortunately, is a bit broader
than that. I'm writing some swedish docs
on how to set up a Linux-system as a DAW, and see the need to do it thoroughly. :)
As of now the card shares IRQ with the eth0-card, which isn't super if one is to set
RT-prio 99 on it.
Is there a simple way in setting a specific IRQ without the need for a super-BIOS that
does it for you? Since I have
a laptop the option of shuffling the PCI-cards isn't really viable. :)
I think I remember from my early days of kernel-hacking that it is possible to set IRQ
with modprobe options, but it
was a while ago. I searched through the kernel tree in the /sound/pci/intel_8x0.c file to
no avail.
Any feedback is welcome, with the sole exception of "Why?" ;)
Regards,
Mathias
9 Oct
9 Oct
4:47 p.m.
On Mon, 2006-10-09 at 20:13 +0200, Mathias Friman wrote:
Hi list.
I have a question regarding IRQ-assignment for an on-board card identified as an Intel
82801DB-ICH4.
It is a crappy card, I know, and you probably ask yourselves why on earth I would want to
put any effort
in configuring such sub-standard equipment. The question, fortunately, is a bit broader
than that. I'm writing some swedish docs
on how to set up a Linux-system as a DAW, and see the need to do it thoroughly. :)
As of now the card shares IRQ with the eth0-card, which isn't super if one is to set
RT-prio 99 on it.
Is there a simple way in setting a specific IRQ without the need for a super-BIOS that
does it for you? Since I have
a laptop the option of shuffling the PCI-cards isn't really viable. :)
I think I remember from my early days of kernel-hacking that it is possible to set IRQ
with modprobe options, but it
was a while ago. I searched through the kernel tree in the /sound/pci/intel_8x0.c file to
no avail.
its basically not possible. PC architecture has changed quite a lot in
the last few years, and AFAIK you can no longer get the kernel to change
IRQ assignments. it seems stupid, i know, but i dug into this rather
deeply a year or so back, and basically the kernel picks up the IO-APIC
routings and does not modify them, even with the many flags that appear
to be related to possibly modifying them :( there is a really cool perl
script that actually analyses IO-APIC routings via the /proc filesystem,
and it confirmed for me what i was worried about.
if your BIOS doesn't allow it, you're out of luck *AFAIK*. if i am
wrong, i will be happy to hear more, since i face similar issues on my
laptop.
--p
5:07 p.m.
On Mon, 2006-10-09 at 14:46 -0400, Paul Davis wrote:
its basically not possible. PC architecture has
changed quite a lot in
the last few years, and AFAIK you can no longer get the kernel to change
IRQ assignments. it seems stupid, i know, but i dug into this rather
deeply a year or so back, and basically the kernel picks up the IO-APIC
routings and does not modify them, even with the many flags that appear
to be related to possibly modifying them :( there is a really cool perl
script that actually analyses IO-APIC routings via the /proc filesystem,
and it confirmed for me what i was worried about.
if your BIOS doesn't allow it, you're out of luck *AFAIK*. if i am
wrong, i will be happy to hear more, since i face similar issues on my
laptop.
Has it ever been possible to change the IRQs of onboard devices? I
thought it was hard wired.
Lee
5:28 p.m.
On Mon, 2006-10-09 at 15:07 -0400, Lee Revell wrote:
On Mon, 2006-10-09 at 14:46 -0400, Paul Davis wrote:
the wiring into the IO-APIC is board-level, but the IO-APIC is
programmable and thus you can change the actual IRQ that the device
generates, in theory.
however, the kernel at this time doesn't contain code that tries to
reprogram the IO-APIC in this area of its function, so you get what it
gives you.
i believe.
its basically not possible. PC architecture has
changed quite a lot in
the last few years, and AFAIK you can no longer get the kernel to change
IRQ assignments. it seems stupid, i know, but i dug into this rather
deeply a year or so back, and basically the kernel picks up the IO-APIC
routings and does not modify them, even with the many flags that appear
to be related to possibly modifying them :( there is a really cool perl
script that actually analyses IO-APIC routings via the /proc filesystem,
and it confirmed for me what i was worried about.
if your BIOS doesn't allow it, you're out of luck *AFAIK*. if i am
wrong, i will be happy to hear more, since i face similar issues on my
laptop.
Has it ever been possible to change the IRQs of onboard devices? I
thought it was hard wired. 
10 Oct
10 Oct
4:32 a.m.
Paul Davis wrote:
On Mon, 2006-10-09 at 15:07 -0400, Lee Revell wrote:
But at this point, the interrupt lines from onboard and external PCI
devices are already merged, so changing the IRQ would just move
both devices.
When an onboard device and a PCI card conflict, you have to move the
card to another slot (or, better, to fix the driver(s)).
Regards,
Clemens
Has it ever been possible to change the IRQs of
onboard devices? I
thought it was hard wired.
the wiring into the IO-APIC is board-level, but the IO-APIC is
programmable and thus you can change the actual IRQ that the device
generates, in theory.
12:18 p.m.
On Tue, 2006-10-10 at 08:32 +0200, Clemens Ladisch wrote:
Paul Davis wrote:
just as a followup, ryan on #ardour pointed out the kernel boot argument
"acpi_irq_balance" which results in *much* better distribution of IRQs
among devices on my laptop. i still have the builtin soundcrap, plus the
yenta and HDSP driver on the same IRQ, but i used to have the ethernet
and two other devices there as well. others who tried it reported
improvements as well.
--p
On Mon, 2006-10-09 at 15:07 -0400, Lee Revell
wrote:
But at this point, the interrupt lines from onboard and external PCI
devices are already merged, so changing the IRQ would just move
both devices.
When an onboard device and a PCI card conflict, you have to move the
card to another slot (or, better, to fix the driver(s)). Has it ever been possible to change the IRQs of
onboard devices? I
thought it was hard wired.
the wiring into the IO-APIC is board-level, but the IO-APIC is
programmable and thus you can change the actual IRQ that the device
generates, in theory.
4:36 p.m.
On Tue, 2006-10-10 at 10:17 -0400, Paul Davis wrote:
I tried with acpi_irq_balance and acpi_irq_pci=3,4,5,6,9,10,11,12,13,14
but that didn't change a thing. I still have my soundcard on the same
IRQ as my eth0, IRQ 11. It is also noteworthy that in the kernel
Documentation-directory there is a file "kernel-parameters.txt" which
lists all, surprise, kernel parameters that can be passed to the kernel
at boot-time. According to that list, it is possible to set IRQ on most,
if not all, OSS-drivers. It is also possible to set IO-addr and DMA, but
seemingly not in ALSA. Is it just a driver design issue or are the
systems in themselves that different?
Regards,
Mathias
But at this
point, the interrupt lines from onboard and external PCI
devices are already merged, so changing the IRQ would just move
both devices.
When an onboard device and a PCI card conflict, you have to move the
card to another slot (or, better, to fix the driver(s)).
just as a followup, ryan on #ardour pointed out the kernel boot argument
"acpi_irq_balance" which results in *much* better distribution of IRQs
among devices on my laptop. i still have the builtin soundcrap, plus the
yenta and HDSP driver on the same IRQ, but i used to have the ethernet
and two other devices there as well. others who tried it reported
improvements as well.
4:50 p.m.
On Tue, 2006-10-10 at 20:41 +0200, Mathias Friman wrote:
On Tue, 2006-10-10 at 10:17 -0400, Paul Davis wrote:
I tried with acpi_irq_balance and acpi_irq_pci=3,4,5,6,9,10,11,12,13,14
but that didn't change a thing. I still have my soundcard on the same
IRQ as my eth0, IRQ 11. It is also noteworthy that in the kernel
Documentation-directory there is a file "kernel-parameters.txt" which
lists all, surprise, kernel parameters that can be passed to the kernel
at boot-time. According to that list, it is possible to set IRQ on most,
if not all, OSS-drivers. It is also possible to set IO-addr and DMA, but
seemingly not in ALSA. Is it just a driver design issue or are the
systems in themselves that different?
This is not a software but a hardware issue. Those OSS driver
parameters don't actually set the hardware IRQ, IO address and DMA
channel - they are for old non-PnP hardware where the driver cannot
probe the hardware configuration.
The IRQ sharing between your ethernet and sound card is almost certainly
hard wired.
Lee
But at
this point, the interrupt lines from onboard and external PCI
devices are already merged, so changing the IRQ would just move
both devices.
When an onboard device and a PCI card conflict, you have to move the
card to another slot (or, better, to fix the driver(s)).
just as a followup, ryan on #ardour pointed out the kernel boot argument
"acpi_irq_balance" which results in *much* better distribution of IRQs
among devices on my laptop. i still have the builtin soundcrap, plus the
yenta and HDSP driver on the same IRQ, but i used to have the ethernet
and two other devices there as well. others who tried it reported
improvements as well.
6727
days inactive
6728
days old
linux-audio-user@lists.linuxaudio.org
7 comments
4 participants
participants (4)
-
Clemens Ladisch -
Lee Revell -
Mathias Friman -
Paul Davis