[U-Boot] [PATCH] net: asix: Fix ASIX 88772B with driver model

[Marek Vasut]

On 08/11/2016 12:28 PM, Alban Bedel wrote:
> On Thu, 11 Aug 2016 11:26:23 +0200
> Marek Vasut <marex at denx.de> wrote:
> 
>> On 08/11/2016 10:52 AM, Alban Bedel wrote:
>>> On Tue, 9 Aug 2016 14:32:14 +0200
>>> Marek Vasut <marex at denx.de> wrote:
>>>
>>>> On 08/09/2016 02:14 PM, Marcel Ziswiler wrote:
>>>>> On Thu, 2016-08-04 at 11:12 +0200, Marek Vasut wrote:
>>>>>> On 08/04/2016 11:07 AM, Alban Bedel wrote:
>>>>>>>
>>>>>>> On Wed, 3 Aug 2016 15:23:30 +0000
>>>>>>> Marcel Ziswiler <marcel.ziswiler at toradex.com> wrote:
>>>>>>>
>>>>>>>>
>>>>>>>> On Wed, 2016-08-03 at 15:51 +0200, Marek Vasut wrote:
>>>>>>>>>
>>>>>>>>> On 08/03/2016 11:46 AM, Alban Bedel wrote:
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On Wed, 3 Aug 2016 09:00:42 +0200
>>>>>>>>>> Marek Vasut <marex at denx.de> wrote:
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 08/03/2016 07:32 AM, Alban Bedel wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Commit 147271209a9d ("net: asix: fix operation without
>>>>>>>>>>>> eeprom")
>>>>>>>>>>>> added a special handling for ASIX 88772B that enable
>>>>>>>>>>>> another
>>>>>>>>>>>> type of header. This break the driver in DM mode as the
>>>>>>>>>>>> extra
>>>>>>>>>>>> handling
>>>>>>>>>>>> needed in the receive path is missing.
>>>>>>>>>>> So add the extra handling ?
>>>>>>>>>> I can do that too, but I though u-boot preferred to avoid
>>>>>>>>>> useless
>>>>>>>>>> code.
>>>>>>>>> Yes, if it is useless.
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> However this new header mode is not required and only
>>>>>>>>>>>> seems to
>>>>>>>>>>>> increase the code complexity, so this patch revert this
>>>>>>>>>>>> part of
>>>>>>>>>>>> commit 147271209a9d.
>>>>>>>>>>> Why is it not required ?
>>>>>>>>>> It works fine without, since 2012. In fact this change is not
>>>>>>>>>> even
>>>>>>>>>> mentioned in the log of commit 147271209a9d, so I really
>>>>>>>>>> don't know
>>>>>>>>>> why
>>>>>>>>>> it was added in the first place. As can be seen in the revert
>>>>>>>>>> all
>>>>>>>>>> it
>>>>>>>>>> does is adding 2 bytes to the USB packets that are then just
>>>>>>>>>> skipped.
>>>>>>>>>> Seems pretty useless to me.
>>>>>>>>> I would like to get some feedback on this from Marcel, since he
>>>>>>>>> added
>>>>>>>>> this stuff.
>>>>>>>> Yes, sorry. I just came back from vacation and started looking
>>>>>>>> into it
>>>>>>>> now. As far as I remember on our hardware without this Ethernet
>>>>>>>> did not
>>>>>>>> quite work reliably. This also means that with driver model so
>>>>>>>> far it
>>>>>>>> does not work for us which I fed back to Simon once but so far
>>>>>>>> this has
>>>>>>>> not been resolved. That fix came from some early U-Boot work done
>>>>>>>> by
>>>>>>>> Antmicro way back and I am missing some of the history.
>>>>>>> Then I'll do a new patch that just fix the driver model receive
>>>>>>> path.
>>>>>> Hold on. Marcel, can you maybe test if removing this code has any
>>>>>> impact
>>>>>> on the behavior now ?
>>>>>
>>>>> Sorry for the delay. I tested Alban's patch now both on Toradex Colibri
>>>>> T20 as well as T30 and its on-module ASIX USB-to-Ethernet chip actually
>>>>> works perfectly aside from the occasional EHCI timed out on TD -
>>>>> token=0x88008d80 Rx: failed to receive: -5 message which last I checked
>>>>> with Simon is still unresolved but was already there long before any of
>>>>> the driver model work started.
>>>>>
>>>>> Tested-by: Marcel Ziswiler <marcel.ziswiler at toradex.com>
>>>>> Tested-on: Colibri T20/T30 on Colibri Evaluation Board
>>>>>
>>>
>>> Will this be applied for the upcoming release?
>>
>> Yeah. Why the hurry though ?
> 
> I was just wondering because all the other patches I submitted have
> been applied but this one still seems to be on hold.

Well because this one was broken, so I had to throw it away. Please do
make sure next time that the stuff builds using buildman.


-- 
Best regards,
Marek Vasut