[U-Boot] [REGRESSION] Ethernet broken on Colibri T20/T30

[Simon Glass]

Hi,

On 18 January 2016 at 10:42, Stephen Warren <swarren at wwwdotorg.org> wrote:
> On 01/16/2016 05:51 PM, Marcel Ziswiler wrote:
>>
>> Hi Simon
>>
>> As mentioned before I noticed Ethernet (on-module USB ASIX chip) to be
>> broken on master while it still worked fine in v2016.01. I kind of
>> remember having once noticed something along those lines when testing
>> some of your early dm stuff but could not find our discussion about it
>> anymore plus I don't think it did actually lead anywhere.
>>
>> FWIW it currently crashes as follows (showing Colibri T20 but the same
>> happens on Colibri T30):
>>
>> U-Boot SPL 2016.01-00205-gb8c5b47 (Jan 17 2016 - 01:31:48)
>
> ...
>
>> Colibri T20 # usb start
>> starting USB...
>> USB0:   USB EHCI 1.00
>> USB1:   USB EHCI 1.00
>> USB2:   USB EHCI 1.00
>> scanning bus 1 for devices... 1 USB Device(s) found
>> scanning bus 2 for devices...
>> Warning: asix_eth using MAC address from ROM
>> 2 USB Device(s) found
>> scanning bus 0 for devices... 2 USB Device(s) found
>> Colibri T20 # dhcp
>> BOOTP broadcast 1
>> EHCI timed out on TD - token=0x88008d80
>> Rx: failed to receive: -5
>> BOOTP broadcast 2
>> EHCI timed out on TD - token=0x8008d80
>> Rx: failed to receive: -5
>> BOOTP broadcast 3
>> EHCI timed out on TD - token=0x88008d80
>> Rx: failed to receive: -5
>> BOOTP broadcast 4
>> EHCI timed out on TD - token=0x8008d80
>> Rx: failed to receive: -5
>> BOOTP broadcast 5
>> EHCI timed out on TD - token=0x88008d80
>> Rx: failed to receive: -5
>> BOOTP broadcast 6
>> EHCI timed out on TD - token=0x8008d80
>> Rx: failed to receive: -5
>>
>> Retry time exceeded; starting again
>> data abort
>> pc : [<1fb4d194>]          lr : [<1fbb9784>]
>> reloc pc : [<0012d194>]    lr : [<00199784>]
>> sp : 1d61c5e0  ip : 6aa665ea     fp : 1d62bca0
>> r10: 00000000  r9 : 1d61fee0     r8 : 1d62bca0
>> r7 : 00000000  r6 : 1d6312e0     r5 : 1f6da8e0  r4 : 00000000
>> r3 : 1c585760  r2 : 03155180     r1 : 00000000  r0 : caa661ca
>> Flags: NzCv  IRQs off  FIQs off  Mode SVC_32
>> Resetting CPU ...
>>
>> resetting ...
>>
>> Do you have any clue what could be going on?
>
>
> FWIW, I tried that same commit on Beaver with an asix USB Ethernet device
> and didn't see that problem, even after aligning the config file for Beaver
> more closely with colibri_t30 (e.g. disabling PCIe support, copying all the
> config options that colibri_t30 has but beaver doesn't into beaver's config
> file).
>
> I wonder if it's anything to do with having 3 USB controllers active rather
> than 2 (although that seems unlikely; I can see going from 1->2 causing
> issues, but once 2 work, adding more seems like it should). Perhaps there's
> a DRAM bandwidth issue, but then I can't see why the patches between
> v2016.01 and the commit you mentioned would affect that.

I have seen the EHCI timeout on and off. I thought it might be due to
cache misalignment such that we wail to talk to the hardware. I have
never dug into it though, or at least not successfully. If you have a
repeatable case it might be worth taking a look.

Regards,
Simon