PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

[Daniel Laird]

Hi all.

I have been happily using Linux 2.6.22.1 for ages on PNX8550 (STB810).
 I have recently decided to step up and move onto Linux 2.6.24 series.
However I am not getting very far. :-(
It seems that all the clock stuff has changed again (since 2.6.22.1)
and this is causing issues.
The board crashes as soon as local_irq_enable is called in main.c

I was wondering if anyone out there might also be running on an
STB810/JBS PNX8550 based system and have any ideas as to why I am
crashing.
I know that PNX8550 does not enable the R4K Clock source stuff as the
chip is a bit 'special' and requires the two timers to be used instead
of one.

As far as I can see I think my interrupts are not being setup.

Any help, ideas etc are appreciated.

Daniel Laird

RE: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

[Nico Coesel]

Daniel,
I've had similar problems on the AU1100 platform. It turned out the
interrupt enumeration was faulty. You might want to check this.

Nico Coesel
 

> -----Oorspronkelijk bericht-----
> Van: linux-mips-bounce@linux-mips.org 
> [mailto:linux-mips-bounce@linux-mips.org] Namens Daniel Laird
> Verzonden: donderdag 28 februari 2008 13:58
> Aan: linux-mips@linux-mips.org
> Onderwerp: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?
> 
> Hi all.
> 
> I have been happily using Linux 2.6.22.1 for ages on PNX8550 (STB810).
>  I have recently decided to step up and move onto Linux 2.6.24 series.
> However I am not getting very far. :-(
> It seems that all the clock stuff has changed again (since 
> 2.6.22.1) and this is causing issues.
> The board crashes as soon as local_irq_enable is called in main.c
> 
> I was wondering if anyone out there might also be running on 
> an STB810/JBS PNX8550 based system and have any ideas as to 
> why I am crashing.
> I know that PNX8550 does not enable the R4K Clock source 
> stuff as the chip is a bit 'special' and requires the two 
> timers to be used instead of one.
> 
> As far as I can see I think my interrupts are not being setup.
> 
> Any help, ideas etc are appreciated.
> 
> Daniel Laird
> 
> 

RE: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

[Nico Coesel]

Daniel,
I've had similar problems on the AU1100 platform. It turned out the
interrupt enumeration was faulty. You might want to check this.

Nico Coesel
 

> -----Oorspronkelijk bericht-----
> Van: linux-mips-bounce@linux-mips.org 
> [mailto:linux-mips-bounce@linux-mips.org] Namens Daniel Laird
> Verzonden: donderdag 28 februari 2008 13:58
> Aan: linux-mips@linux-mips.org
> Onderwerp: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?
> 
> Hi all.
> 
> I have been happily using Linux 2.6.22.1 for ages on PNX8550 (STB810).
>  I have recently decided to step up and move onto Linux 2.6.24 series.
> However I am not getting very far. :-(
> It seems that all the clock stuff has changed again (since 
> 2.6.22.1) and this is causing issues.
> The board crashes as soon as local_irq_enable is called in main.c
> 
> I was wondering if anyone out there might also be running on 
> an STB810/JBS PNX8550 based system and have any ideas as to 
> why I am crashing.
> I know that PNX8550 does not enable the R4K Clock source 
> stuff as the chip is a bit 'special' and requires the two 
> timers to be used instead of one.
> 
> As far as I can see I think my interrupts are not being setup.
> 
> Any help, ideas etc are appreciated.
> 
> Daniel Laird
> 
> 

Re: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

[Ralf Baechle]

On Thu, Feb 28, 2008 at 12:57:47PM +0000, Daniel Laird wrote:

> I have been happily using Linux 2.6.22.1 for ages on PNX8550 (STB810).
>  I have recently decided to step up and move onto Linux 2.6.24 series.
> However I am not getting very far. :-(
> It seems that all the clock stuff has changed again (since 2.6.22.1)
> and this is causing issues.

Yes, it's a bit special and I don't have any possibility to test the
PNX code so I left it mostly untouched when rewriting the timer code.
Sorry about that but eventually I had to go ahead.  Vitaly Wool
eventually submitted a few patches to fix things but as your testing
seems to show that wasn't quite enough.  That far about the history
of the changes.

> The board crashes as soon as local_irq_enable is called in main.c
> 
> I was wondering if anyone out there might also be running on an
> STB810/JBS PNX8550 based system and have any ideas as to why I am
> crashing.
> I know that PNX8550 does not enable the R4K Clock source stuff as the
> chip is a bit 'special' and requires the two timers to be used instead
> of one.

csrc-r4k.c and cevt-r4k.c assume a standard compliant R4000-style
c0_count and c0_compare register.  The PNX chips violate the expected
behaviour badly so can't use these functions.

But that's not very much of a problem.  The timer code is module and
it is easy to write something like csrc-pnx.c and cevt-pnx.c to drive
a PNX style count/compare timer.

Will look over the code to see if I can spot what crashes the PNXes.

  Ralf

Re: PNX8550 Broken on Linux 2.6.24 - Interrupt issues?

[Sergei Shtylyov]

Hello.

Ralf Baechle wrote:

>>I have been happily using Linux 2.6.22.1 for ages on PNX8550 (STB810).
>> I have recently decided to step up and move onto Linux 2.6.24 series.
>>However I am not getting very far. :-(

    Heh, Daniel, you're not alone -- the Alchemy code got much more breakage 
before 2.6.24... :-)

>>The board crashes as soon as local_irq_enable is called in main.c

>>I was wondering if anyone out there might also be running on an
>>STB810/JBS PNX8550 based system and have any ideas as to why I am
>>crashing.
>>I know that PNX8550 does not enable the R4K Clock source stuff as the
>>chip is a bit 'special' and requires the two timers to be used instead
>>of one.

> csrc-r4k.c and cevt-r4k.c assume a standard compliant R4000-style
> c0_count and c0_compare register.  The PNX chips violate the expected
> behaviour badly so can't use these functions.

> But that's not very much of a problem.  The timer code is module and
> it is easy to write something like csrc-pnx.c and cevt-pnx.c to drive
> a PNX style count/compare timer.

    The PNX counter 0 can still be used as the standards clocksource with some 
ad-hockery (setting comparator to all ones and hooking the interrupt to clear 
it) but Vitaly went another way using counter 1 for clocksource, and counter 0 
as clockevent...

> Will look over the code to see if I can spot what crashes the PNXes.

>   Ralf

WBR, Sergei