[Xenomai-help] irq conflict

[Xenomai-help] irq conflict

[Theo Veenker]

Hi all,

I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,
but the xenomai pactched kernel gives several of these messages:

   IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10

And some things (e.g. mouse, network) then don't work, not properly
anyway. The same kernel runs fine however on other systems.

Any clue what I can do about it?

Anyone running xenomai on a recent Compaq laptop?

Regards,

Theo

Re: [Xenomai-help] irq conflict

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 804 bytes --]

Theo Veenker wrote:
> Hi all,
> 
> I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
> and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,

Means the same kernel with I-pipe (and Xenomai) disabled or the same
kernel version and .config, but unpatched?

> but the xenomai pactched kernel gives several of these messages:
> 
>   IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10
> 
> And some things (e.g. mouse, network) then don't work, not properly
> anyway. The same kernel runs fine however on other systems.
> 
> Any clue what I can do about it?

Please provide bootlogs for both rt and non-rt kernel startup. And the
.config.

> 
> Anyone running xenomai on a recent Compaq laptop?
> 
> Regards,
> 
> Theo
> 
> 

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] irq conflict

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1429 bytes --]

Theo Veenker wrote:
> Jan Kiszka wrote:
>> Theo Veenker wrote:
>>> Hi all,
>>>
>>> I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
>>> and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,
>>
>> Means the same kernel with I-pipe (and Xenomai) disabled or the same
>> kernel version and .config, but unpatched?
> 
> The latter.
> 
>>
>>> but the xenomai pactched kernel gives several of these messages:
>>>
>>>   IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10
>>>
>>> And some things (e.g. mouse, network) then don't work, not properly
>>> anyway. The same kernel runs fine however on other systems.
>>>
>>> Any clue what I can do about it?
>>
>> Please provide bootlogs for both rt and non-rt kernel startup. And the
>> .config.
> 
> I 'solved' it by enabling ACPI (without processor module and others).
> Looks good. I don't see high latencies with the latency test. Under
> what conditions might ACPI affect real-time behaviour?

CONFIG_ACPI_PROCESSOR, because it allows deep sleep states with
significant wakeup latencies.

But back to your original issue: You mean your notebook booted fine
under vanilla Linux with ACPI disabled? That would mean I-pipe relies on
ACPI here without stating this anywhere. In that case /something/ needs
fixing: code, config deps, or doc.

Jan


PS: Please take care to not drop the list from CC.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] irq conflict

[Philippe Gerum]

On Thu, 2007-01-18 at 21:09 +0100, Jan Kiszka wrote:
> Theo Veenker wrote:
> > Jan Kiszka wrote:
> >> Theo Veenker wrote:
> >>> Hi all,
> >>>
> >>> I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
> >>> and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,
> >>
> >> Means the same kernel with I-pipe (and Xenomai) disabled or the same
> >> kernel version and .config, but unpatched?
> > 
> > The latter.
> > 
> >>
> >>> but the xenomai pactched kernel gives several of these messages:
> >>>
> >>>   IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10
> >>>
> >>> And some things (e.g. mouse, network) then don't work, not properly
> >>> anyway. The same kernel runs fine however on other systems.
> >>>
> >>> Any clue what I can do about it?
> >>
> >> Please provide bootlogs for both rt and non-rt kernel startup. And the
> >> .config.
> > 
> > I 'solved' it by enabling ACPI (without processor module and others).
> > Looks good. I don't see high latencies with the latency test. Under
> > what conditions might ACPI affect real-time behaviour?
> 
> CONFIG_ACPI_PROCESSOR, because it allows deep sleep states with
> significant wakeup latencies.
> 
> But back to your original issue: You mean your notebook booted fine
> under vanilla Linux with ACPI disabled? That would mean I-pipe relies on
> ACPI here without stating this anywhere. In that case /something/ needs
> fixing: code, config deps, or doc.
> 

It doesn't, at least, not willingly. Before drawing any conclusion,
please try 2.6.19 with the same configuration first: the generic irq
layer may have had a positive influence on the I-pipe in this area.

> Jan
> 
> 
> PS: Please take care to not drop the list from CC.
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
-- 
Philippe.

Re: [Xenomai-help] irq conflict

[Theo Veenker]

Philippe Gerum wrote:
> On Thu, 2007-01-18 at 21:09 +0100, Jan Kiszka wrote:
> 
>>Theo Veenker wrote:
>>
>>>Jan Kiszka wrote:
>>>
>>>>Theo Veenker wrote:
>>>>
>>>>>Hi all,
>>>>>
>>>>>I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
>>>>>and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,
>>>>
>>>>Means the same kernel with I-pipe (and Xenomai) disabled or the same
>>>>kernel version and .config, but unpatched?
>>>
>>>The latter.
>>>
>>>
>>>>>but the xenomai pactched kernel gives several of these messages:
>>>>>
>>>>>  IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10
>>>>>
>>>>>And some things (e.g. mouse, network) then don't work, not properly
>>>>>anyway. The same kernel runs fine however on other systems.
>>>>>
>>>>>Any clue what I can do about it?
>>>>
>>>>Please provide bootlogs for both rt and non-rt kernel startup. And the
>>>>.config.
>>>
>>>I 'solved' it by enabling ACPI (without processor module and others).
>>>Looks good. I don't see high latencies with the latency test. Under
>>>what conditions might ACPI affect real-time behaviour?
>>
>>CONFIG_ACPI_PROCESSOR, because it allows deep sleep states with
>>significant wakeup latencies.
>>
>>But back to your original issue: You mean your notebook booted fine
>>under vanilla Linux with ACPI disabled? That would mean I-pipe relies on
>>ACPI here without stating this anywhere. In that case /something/ needs
>>fixing: code, config deps, or doc.
>>
> 
> 
> It doesn't, at least, not willingly. Before drawing any conclusion,
> please try 2.6.19 with the same configuration first: the generic irq
> layer may have had a positive influence on the I-pipe in this area.

I'd love to try it, but there is no 2.6.19 patch for 2.2.5. When 2.3
was released I tried using that, but failed (unable to load the posix
module or something I don't remember exactly). Since I had only a few
days to convert all our systems (for language and speech research)
from RTLinux to Xenomai I decided to stay at 2.2.5 until quieter
times. Hope to figure it out soon.

Theo

Re: [Xenomai-help] irq conflict

[Theo Veenker]

Jan Kiszka wrote:
> Theo Veenker wrote:
> 
>>Jan Kiszka wrote:
>>
>>>Theo Veenker wrote:
>>>
>>>>Hi all,
>>>>
>>>>I need to run xenomai on a HP Compaq nc6320. I use kernel 2.6.17.14
>>>>and xenomai 2.2.5. The non-realtime kernel runs fine on this laptop,
>>>
>>>Means the same kernel with I-pipe (and Xenomai) disabled or the same
>>>kernel version and .config, but unpatched?
>>
>>The latter.
>>
>>
>>>>but the xenomai pactched kernel gives several of these messages:
>>>>
>>>>  IRQ routing conflict for xxxx.xx.xx.x, have irq 11, want irq 10
>>>>
>>>>And some things (e.g. mouse, network) then don't work, not properly
>>>>anyway. The same kernel runs fine however on other systems.
>>>>
>>>>Any clue what I can do about it?
>>>
>>>Please provide bootlogs for both rt and non-rt kernel startup. And
> 
> the
> 
>>>.config.
>>
>>I 'solved' it by enabling ACPI (without processor module and others).
>>Looks good. I don't see high latencies with the latency test. Under
>>what conditions might ACPI affect real-time behaviour?
> 
> 
> CONFIG_ACPI_PROCESSOR, because it allows deep sleep states with
> significant wakeup latencies.
> 
> But back to your original issue: You mean your notebook booted fine
> under vanilla Linux with ACPI disabled? That would mean I-pipe relies on
> ACPI here without stating this anywhere. In that case /something/ needs
> fixing: code, config deps, or doc.

Just tried that. Things don't work (properly) with acpi=off (jumpy mouse,
no network, sound loops). I tried the ubuntu 6.06 2.6.15-23-386 kernel,
a vanilla 2.6.17.14 kernel and a xenomai patched 2.6.17.14 kernel. All
with same result. So that's reassuring I guess.

I recently converted my stuff from RTLinux to Xenomai and just assumed it
would be saver to turn off ACPI for real-time operation. Didn't know some
systems actually rely on ACPI.

> 
> Jan
> 
> 
> PS: Please take care to not drop the list from CC.

I'm sorry. The mistake occurred to me the moment I clicked send.

Theo