Re: [PATCH net-next] net: stmmac: stm32: Set TSO/TBS Tx queues default settings

[Maxime Chevallier <maxime.chevallier@bootlin.com>]

Hi,

On 8/21/26 15:41, Maxime Chevallier wrote:
> Hi Kurt,

(Sorry Kurt this has nothing to do with your patch, I'm following-up with
some stuff I said on that thread)

>> However, we need to keep track of what SoC implementation can enable TBS
>> on which queue. For gmac4 we can probably reuse tbs_en and for xgmac
>> there's is capability field tbs_ch_num. Let's see.
> 
> Thanks for tackling this, let me know if you need testing. In the past weeks
> I've been looking deep in my drawers as well as on various online shops, and
> I'm starting to have a farily comprehensive set of stmmac-enabled devices
> (although I couldn't get my hands on XGMAC, AgileX5 devkits appear to be way
> outside my budget).
Made an intersting finding yesterday, maybe it's already known stuff, but it turns
out it's easier to find a DRIVER for XGMAC than an actual device...

Looking at register layouts and at some then-removed copyright notices from the history,
it seems we have not one, not two, but three different drivers upstream for the same
DWXGMAC IP :

 - stmmac's dwxgmac
 - AMD's XGBE drives a designware XGMAC
 - Samsung's SXGBE also drives a designware XGMAC

No wonder the xgbe selftests were doing the same as the stmmac selftests :)

I think reuniting the codebases isn't gonna happen in the near future, however it
could be a good idea to hack stmmac to work on amd HW could be a good way to validate
the stmmac dwxgmac code, with another reference driver...

Maxime