Hi Mike,
Mike Frysinger wrote:
<snip> >> not specific to either of these drivers, so if we did choose to make this >> behavior optional via some define, it would make more sense to do it in >> the common eth code rather than copying & pasting it everywhere. >> > Agreed. Do you think you have time and resources to craft such a patch? >
net/eth.c:eth_initialize() already has the logic to handle this, but that only applies to NET_MULTI drivers. and most do not take advantage of it. i think the documentation should be updated like so: in the driver function that calls eth_register(), the driver should initialize dev->enetaddr to the MAC found in the hardware (if possible) and then applicable drivers should be fixed. if we agree on this route, i can do a quick scan of the net drivers and post relevant patches.
That'd be great. You've done good work bringing consistency to the code.
in the case of a mismatch, we would see from the common eth code: Warning: Blackfin EMAC MAC addresses don't match: Address in SROM is 0a:0a:0a:0a:0a:0a Address in environment is 00:e0:22:fe:44:ec
the smsc drivers however are not in the NET_MULTI category -- they dont use any of the common ethernet stack. so once they are converted to NET_MULTI, they'll get this functionality for free (when exactly were we adding #warning about non-NET_MULTI usage ?). so rather than expend effort on restoring duplicate code, how about interested parties convert the driver ;). -mike
The #warning patch is done but not submitted. Pretty lame, I know, but it's been a crazy summer. I have the SMSC driver mostly ported but don't have any hardware to test on. I guess there are 3 days left in this merge window to get the ball rolling... If anybody else has already done this, please submit.
regards, Ben