[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Uwe Wojak]

Hallo,

we use a mini PCIe card (9390) with a miniPCe to PCIe adapter on a X86 
system and a system with a Xilinix Microblaze CPU.
As the performance on the Microblaze is bad (ping times up to 10 sec), 
we examined the interrupt service routine.

On Microblaze *and* on X86 we see about 10 "spurious" interrupts per 
second.
In this cases in ar9003_hw_get_isr (ar9003_mac.c) the sync_cause bit 
"AR_INTR_SYNC_HOST1_PERR" is set.
The driver ignores this.

Does AR_INTR_SYNC_HOST1_PER indicate a PICe parity error ?

We use Kernel 3.1.4.

Regards
Uwe Wojak

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Mohammed Shafi]

On Wed, Dec 7, 2011 at 4:36 PM, Uwe Wojak <uwe.wojak@gmx.de> wrote:
> Hallo,
>
> we use a mini PCIe card (9390) with a miniPCe to PCIe adapter on a X86
> system and a system with a Xilinix Microblaze CPU.
> As the performance on the Microblaze is bad (ping times up to 10 sec),
> we examined the interrupt service routine.
>
> On Microblaze *and* on X86 we see about 10 "spurious" interrupts per
> second.
> In this cases in ar9003_hw_get_isr (ar9003_mac.c) the sync_cause bit
> "AR_INTR_SYNC_HOST1_PERR" is set.
> The driver ignores this.

its a fatal interrupt (referred in ar9002_mac.c) and we have to reset
the chip. not sure why we are getting this in ar9390

>
> Does AR_INTR_SYNC_HOST1_PER indicate a PICe parity error ?
>
> We use Kernel 3.1.4.
>
> Regards
> Uwe Wojak
>
>
>
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel



-- 
shafi

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Mohammed Shafi]

On Wed, Dec 7, 2011 at 5:14 PM, Mohammed Shafi <shafi.wireless@gmail.com> wrote:
> On Wed, Dec 7, 2011 at 4:36 PM, Uwe Wojak <uwe.wojak@gmx.de> wrote:
>> Hallo,
>>
>> we use a mini PCIe card (9390) with a miniPCe to PCIe adapter on a X86
>> system and a system with a Xilinix Microblaze CPU.
>> As the performance on the Microblaze is bad (ping times up to 10 sec),
>> we examined the interrupt service routine.
>>
>> On Microblaze *and* on X86 we see about 10 "spurious" interrupts per
>> second.
>> In this cases in ar9003_hw_get_isr (ar9003_mac.c) the sync_cause bit
>> "AR_INTR_SYNC_HOST1_PERR" is set.
>> The driver ignores this.
>
> its a fatal interrupt (referred in ar9002_mac.c) and we have to reset
> the chip. not sure why we are getting this in ar9390

i will send a patch to reset the chip once we receive in ar9390 also..
please see if helps

>
>>
>> Does AR_INTR_SYNC_HOST1_PER indicate a PICe parity error ?
>>
>> We use Kernel 3.1.4.
>>
>> Regards
>> Uwe Wojak
>>
>>
>>
>> _______________________________________________
>> ath9k-devel mailing list
>> ath9k-devel at lists.ath9k.org
>> https://lists.ath9k.org/mailman/listinfo/ath9k-devel
>
>
>
> --
> shafi



-- 
shafi

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Mohammed Shafi]

On Wed, Dec 7, 2011 at 5:29 PM, Mohammed Shafi <shafi.wireless@gmail.com> wrote:
> On Wed, Dec 7, 2011 at 5:14 PM, Mohammed Shafi <shafi.wireless@gmail.com> wrote:
>> On Wed, Dec 7, 2011 at 4:36 PM, Uwe Wojak <uwe.wojak@gmx.de> wrote:
>>> Hallo,
>>>
>>> we use a mini PCIe card (9390) with a miniPCe to PCIe adapter on a X86
>>> system and a system with a Xilinix Microblaze CPU.
>>> As the performance on the Microblaze is bad (ping times up to 10 sec),
>>> we examined the interrupt service routine.
>>>
>>> On Microblaze *and* on X86 we see about 10 "spurious" interrupts per
>>> second.
>>> In this cases in ar9003_hw_get_isr (ar9003_mac.c) the sync_cause bit
>>> "AR_INTR_SYNC_HOST1_PERR" is set.
>>> The driver ignores this.

please try with the attached patch in latest compat wireless and see if it helps



-- 
shafi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ar9003-fatal-intr.patch
Type: text/x-diff
Size: 1549 bytes
Desc: not available
Url : http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20111207/75e62815/attachment-0001.patch 

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Uwe Wojak]

Am 07.12.2011 13:44, schrieb Mohammed Shafi:
> On Wed, Dec 7, 2011 at 5:29 PM, Mohammed Shafi<shafi.wireless@gmail.com>  wrote:
>> On Wed, Dec 7, 2011 at 5:14 PM, Mohammed Shafi<shafi.wireless@gmail.com>  wrote:
>>> On Wed, Dec 7, 2011 at 4:36 PM, Uwe Wojak<uwe.wojak@gmx.de>  wrote:
>>>> Hallo,
>>>>
>>>> we use a mini PCIe card (9390) with a miniPCe to PCIe adapter on a X86
>>>> system and a system with a Xilinix Microblaze CPU.
>>>> As the performance on the Microblaze is bad (ping times up to 10 sec),
>>>> we examined the interrupt service routine.
>>>>
>>>> On Microblaze *and* on X86 we see about 10 "spurious" interrupts per
>>>> second.
>>>> In this cases in ar9003_hw_get_isr (ar9003_mac.c) the sync_cause bit
>>>> "AR_INTR_SYNC_HOST1_PERR" is set.
>>>> The driver ignores this.
> please try with the attached patch in latest compat wireless and see if it helps
>
>
Hallo Mohammed,

thank you for your patch. After applying it the chip is reset when the error occurs (i added debug output),
but the error still appears about 10 times a second.

Do you know what "AR_INTR_SYNC_HOST1_PERR" means ?
I feel that there is something wrong with our miniPCIe Adapters or with the network card
(but i see the effect on two different ones from different vendors).

May there be some chip setup missing for 9390 ?

Thank you for your help
Uwe

[ath9k-devel] Atheros 9390 and AR_INTR_SYNC_HOST1_PERR

[Mohammed Shafi]

Hallo Mohammed,

Hi Uwe,

>
> thank you for your patch. After applying it the chip is reset when the error occurs (i added debug output),
> but the error still appears about 10 times a second.

bad luck, looks like some issue i thought doing a chip reset would
have brought back the chip to sane state

>
> Do you know what "AR_INTR_SYNC_HOST1_PERR" means ?
> I feel that there is something wrong with our miniPCIe Adapters or with the network card
> (but i see the effect on two different ones from different vendors).

I really don't know the exact expansion.. i just saw in the code base
it looks like "seriour pkt errors" not sure what it means. also it
seems chip reset is not required for this interrupt in AR9003
chip-sets. some one reported some time back AR9390 is working fine in
x86 but no so well in Powerpc.

>
> May there be some chip setup missing for 9390 ?

don't know let me check it out.

>
> Thank you for your help
> Uwe
>
>
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel



-- 
shafi