[PATCH] net/ftgmac100: Rewrite the driver

Cédric Le Goater clg at kaod.org
Wed Mar 29 03:41:41 AEDT 2017


>>>> It does that by, among other things, rewriting the receive and
>>>> transmit code, to both simplify the fast path as much as possible,
>>>> avoid flushing of the caches (the aspeed chips have really slow
>>>> little ARM cores), implementing support for fragmented sends,
>>>> fixing HW checksum generation (AST2500 only), etc...
>>>>
>>>> In addition, I've added netpoll support, tx timeout recovery,
>>>> better handling of link speed changes, multicast filter
>>>> and promisc support, various ethtool config additions etc...
>>>
>>> Nice. So I will need to update the QEMU model as it is quite 
>>> basic for the moment.
>>
>> You will need to support fragmented sends yes ;-)
> 
> among other things yes ... I will add that on the TODO list. Anyhow
> the model needed some cleanups. 

The fragmented tx was not taking into account the full size of 
the frame ... That is fixed now and I also added cksum offload 
support for the AST2500. So the new and old n/w drivers are now
supported in QEMU, in a very basic way. no nifty tuning.

I reworked a bit the way the rx queue is flushed but this is not 
enough to support iperf yet. it needs more work to sustain high 
throughput. U-Boot support would be nice also but I think we will 
need a more complete NCSI model. 

Cheers,

C. 
 



More information about the openbmc mailing list