Ethernet reinit on phyCORE hangs the device

Ethernet reinit on phyCORE hangs the device

[Andrej Picej]

Hi all,

when testing barebox ethernet on i.MX6 (phyCORE) we encountered a bug:
the ethernet re-init on barebox doesn't work! If we do:
$ ifup eth0
$ ifdown eth0
and then:
$ ifup eth0
$ ping HOST_IP
the device hangs and can only be saved by resetting the board manually.

We narrowed the problem down to the "fec_tx_task_enable". It seems that 
enabling tx task with writing to TDAR reg hangs the device shortly after.

Did you guys maybe encounter something similar on i.MX6 boards or on any 
other boards?

I guess this is a really specific problem as people don't normally do 
ifup/ifdown with ethernet in bootloader :D.

Do you have any idea what could be the reason or how could this be solved?
Any useful information would be greatly appreciated.

Using barebox v2022.02.

Best Regards,
Andrej


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: Ethernet reinit on phyCORE hangs the device

[Sascha Hauer]

On Tue, Jun 14, 2022 at 09:12:23AM +0200, Andrej Picej wrote:
> Hi all,
> 
> when testing barebox ethernet on i.MX6 (phyCORE) we encountered a bug:
> the ethernet re-init on barebox doesn't work! If we do:
> $ ifup eth0
> $ ifdown eth0
> and then:
> $ ifup eth0
> $ ping HOST_IP
> the device hangs and can only be saved by resetting the board manually.

I can confirm the behaviour is the same here.

> 
> We narrowed the problem down to the "fec_tx_task_enable". It seems that
> enabling tx task with writing to TDAR reg hangs the device shortly after.
> 
> Did you guys maybe encounter something similar on i.MX6 boards or on any
> other boards?
> 
> I guess this is a really specific problem as people don't normally do
> ifup/ifdown with ethernet in bootloader :D.

Nope :)

> 
> Do you have any idea what could be the reason or how could this be solved?
> Any useful information would be greatly appreciated.

Part of the problem is that fec_halt() tears down the device completely.
The counterpart of that is fec_init(), but that is called from during
probe only. I think fec_init() should be called from fec_open() as well.

A quick test revealed this is really only part of the problem. The next
one is that fec_update_linkspeed() is not called on second open, so the
fec speed registers will not get their correct value.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: Ethernet reinit on phyCORE hangs the device

[Andrej Picej]

Hi Sascha,

On 14. 06. 22 10:27, Sascha Hauer wrote:
> On Tue, Jun 14, 2022 at 09:12:23AM +0200, Andrej Picej wrote:
>> Hi all,
>>
>> when testing barebox ethernet on i.MX6 (phyCORE) we encountered a bug:
>> the ethernet re-init on barebox doesn't work! If we do:
>> $ ifup eth0
>> $ ifdown eth0
>> and then:
>> $ ifup eth0
>> $ ping HOST_IP
>> the device hangs and can only be saved by resetting the board manually.
> 
> I can confirm the behaviour is the same here.

Great.

> 
>>
>> We narrowed the problem down to the "fec_tx_task_enable". It seems that
>> enabling tx task with writing to TDAR reg hangs the device shortly after.
>>
>> Did you guys maybe encounter something similar on i.MX6 boards or on any
>> other boards?
>>
>> I guess this is a really specific problem as people don't normally do
>> ifup/ifdown with ethernet in bootloader :D.
> 
> Nope :)
> 
>>
>> Do you have any idea what could be the reason or how could this be solved?
>> Any useful information would be greatly appreciated.
> 
> Part of the problem is that fec_halt() tears down the device completely.
> The counterpart of that is fec_init(), but that is called from during
> probe only. I think fec_init() should be called from fec_open() as well.

Hmm I thought I already tried that, guess not...

> 
> A quick test revealed this is really only part of the problem. The next
> one is that fec_update_linkspeed() is not called on second open, so the
> fec speed registers will not get their correct value.

Ok, I'll check that out.
Thanks for your help so far.

Andrej