RX-fixup question.

[Markus Gothe]

For briding you want two differently adresses.

Especially if you want to do extra magic with tools like ebtables.

A private source adress is to prefer. A shared adress would cause me headache when not briding. ;-)

I will look at the WiMaX driver. Thx for the feedback.

//M

Skickat ifrån min mobil.
Den 5 nov 2014 10:36 skrev =?ISO-8859-1?Q?Bj=F8rn_Mork?= <bjorn at mork.no>:
>
> Markus Gothe <markus at dovado.com> writes: 
>
> > Thanks for the information... 
> > Then I should start working on a decent mac source address fix for it. Ive applied an u8-string to usbnet.h if using QMI or MBIM. 
> > 
> > Any opnions on this approach? 
>
> I am not sure there is much gain in making this a usbnet feature. Is it 
> relevant to any other usbnet drivers than those two? 
>
> I did some tests using the local device address as both source and dest, 
> but I guess that doesn't work for bridging?  It will probably want 
> different valid addresses? 
>
> Hmm, qmi_wwan already has the "default_modem_addr", which actually is 
> used as source address by some devices.  Reusing for our source address 
> seems like an obvious simplification.  Any reason not to do this? 
>
> Grepping a bit around the kernel I found this driver which does similar 
> header tricks: drivers/net/wimax/i2400m/.  But that driver stores a per 
> device random src_mac_addr in the device private data. I believe this is 
> overkill for our purpose.  A shared static address will do just 
> fine. Won't it? 
>
>
>
> Bjørn