IRQ 15 not assigned, but still unable to use it - in MSDOS it works

IRQ 15 not assigned, but still unable to use it - in MSDOS it works

[Erik Rull]

Hi all,

I've worked a lot on bringing the IRQ 15 to work for my embedded system but 
it still fails.

My kernel is the 2.6.29.4 - I've already updated from 2.6.23.3 because 
there were some bugs in some ata drivers.
My system is a embedded system with Intel ATOM CPU and Intel 965 chipset.
I need to access some very old ISA-Bus boards. Before I used a Pentium-M 
Board that worked perfect.
The new board seems to have problems disabling the IRQ 15 in linux. BIOS 
has IRQ 15 reserved for legacy hardware and the IDE port behind it is 
disabled - so IRQ15 should be really free.
The 2.6.23.3 found an IDE device and said afterwards that it is disabled 
but kept the resources assigned without releasing them - so I was not able 
to use the IRQ15 with my own software.
With the 2.6.29.4 I don't get this any longer but the IRQ15 doesn't work 
either. Other interrupts are fine (5 or 10) but they are already used by 
other hardware components.
Now the interesting part: in MSDOS I'm able to see the IRQ15 from the ISA 
board as a source....

So I think that linux somewhere takes the IRQ15 and does not allow me to 
use it - and IRQ sharing seems not to be possible in this case.

I need some ideas where to continue my search - I think that it is 
somewhere in the libata that blocks this interrupt. /proc/interrupts shows 
up no interrupt neither that #15 is used.

Any ideas?

If you need more information please let me know.

Thanks a lot!

Best regards,

Erik

Re: IRQ 15 not assigned, but still unable to use it - in MSDOS it works

[Robert Hancock]

On 06/16/2010 05:30 AM, Erik Rull wrote:
> Hi all,
>
> I've worked a lot on bringing the IRQ 15 to work for my embedded system
> but it still fails.
>
> My kernel is the 2.6.29.4 - I've already updated from 2.6.23.3 because
> there were some bugs in some ata drivers.
> My system is a embedded system with Intel ATOM CPU and Intel 965 chipset.
> I need to access some very old ISA-Bus boards. Before I used a Pentium-M
> Board that worked perfect.
> The new board seems to have problems disabling the IRQ 15 in linux. BIOS
> has IRQ 15 reserved for legacy hardware and the IDE port behind it is
> disabled - so IRQ15 should be really free.
> The 2.6.23.3 found an IDE device and said afterwards that it is disabled
> but kept the resources assigned without releasing them - so I was not
> able to use the IRQ15 with my own software.
> With the 2.6.29.4 I don't get this any longer but the IRQ15 doesn't work
> either. Other interrupts are fine (5 or 10) but they are already used by
> other hardware components.
> Now the interesting part: in MSDOS I'm able to see the IRQ15 from the
> ISA board as a source....
>
> So I think that linux somewhere takes the IRQ15 and does not allow me to
> use it - and IRQ sharing seems not to be possible in this case.
>
> I need some ideas where to continue my search - I think that it is
> somewhere in the libata that blocks this interrupt. /proc/interrupts
> shows up no interrupt neither that #15 is used.

If no driver attaches to the device and nothing is reported in 
/proc/interrupts then nothing should be blocking the interrupt 
driver-wise. I suspect the problem might be elsewhere, like the IRQ 
routing or something..

Re: IRQ 15 not assigned, but still unable to use it - in MSDOS it works

[Erik Rull]

Robert Hancock wrote:
> On 06/16/2010 05:30 AM, Erik Rull wrote:
>>
>> I need some ideas where to continue my search - I think that it is
>> somewhere in the libata that blocks this interrupt. /proc/interrupts
>> shows up no interrupt neither that #15 is used.
>
> If no driver attaches to the device and nothing is reported in
> /proc/interrupts then nothing should be blocking the interrupt
> driver-wise. I suspect the problem might be elsewhere, like the IRQ
> routing or something..

Hm interesting. Is there a way to debug/see the IRQ routing somehow or 
somewhere? I've found another possiblity of watching incoming interrupts 
via /proc/stats but there is also nothing happening (The number that 
belongs to IRQ15 remains on 0).

Best regards,

Erik