Libertas driver performance on Samsung platform

[Ravi Mahadevappa]

Hi Jacky,

Thanks for the response. In our case, these are in place already.
Maybe I am not using the right tools to measure the throughput. I will
try to check with iperf.

Best Regards,
Ravi

On Sun, Nov 22, 2009 at 5:40 PM, Jacky Lam <jackylam.gsl at gmail.com> wrote:
> Dear Ravi,
>
>        The problem we have met is our platform driver use polling when dealing
> SDIO interrupt. The solution is provided by tommy. I will describe the
> solution as follows:
>
>        1) Your mmc host capability need to enable MMC_CAP_SDIO_IRQ.
>        2) In mmc host operations (struct mmc_host_ops), you have to implement
> the callback enable_sdio_irq.
>
>        Also, the firmware version we are using is v9 as Dan said.
>
>        Hope this help.
>
> Br,
> Jacky
>
> Ravi Mahadevappa 提到:
>> Hi Dan, Jacky,
>>
>> We are facing the same problem. Could you please elaborate a bit on this
>> issue.
>> What is the firmware version and the driver version I should use?
>> Where do I look to verify that the interrupt support is enabled?
>>
>> Thanks,
>> Ravi
>>
>> On 11/15/2009 10:43 PM, Jacky Lam wrote:
>>> Dear Dan,
>>>
>>>      I just talk to tommy and find that the problem really happened in the
>>> interrupt support as you said. Thanks for all advice from you.
>>>
>>> BR,
>>> Jacky
>>>
>>> Dan Williams 提到:
>>>
>>>> On Mon, 2009-11-16 at 09:43 +0800, Jacky Lam wrote:
>>>>
>>>>> Dan Williams 提到:
>>>>>
>>>>>> On Fri, 2009-11-13 at 13:12 +0800, Jacky Lam wrote:
>>>>>>
>>>>>>> Dear all,
>>>>>>>
>>>>>>>  I have read the previous mail about the low performance of Samsung
>>>>>>> BSP's SDIO platform driver when working with Libertas. I want to know if
>>>>>>> anyone have tried the kernel mainstream's Samsung driver and have any
>>>>>>> improvement on that?
>>>>>>>
>>>>>> Does the Samsung SDHC have quirks or limitations with respect to DMA,
>>>>>> block size, transfer width, etc?
>>>>>>
>>>>>> Dan
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>> Dear Dan,
>>>>>
>>>>>    I look through the source code and can't find any "special" limitation
>>>>> for the DMA, block size, transfer width.....Maybe I am not familiar
>>>>> enough with Linux MMC stack. Could you tell me if there is any
>>>>> requirement for the MMC stack on the host's DMA, block size, transfer
>>>>> width..etc? So, I can pinpoint my investigation?
>>>>>
>>>> There's no requirement in the stack, but the libertas driver and
>>>> firmware expect multi-block transfer ability and interrupt support.  The
>>>> libertas firmware will fail to load if if the host controller doesn't
>>>> support multi-block, which we just found to be a limitation of one of
>>>> the i.MX platform for example.
>>>>
>>>> Dan
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>> _______________________________________________
>>> libertas-dev mailing list
>>> libertas-dev at lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/libertas-dev
>>>
>