Am Do, 27.10.2011, 11:12 schrieb Prafulla Wadaskar:
If I understood it correctly, In current implementation, if only egiga1 device is enabled
on the board,
then it will assign MAC address associated with environment
variable
"ethaddr".
yes but the current mvgbe driver will check eth1addr and set it to a random value if not set.
Yes, but this may be avoided by defining CONFIG_SKIP_LOCAL_MAC_RANDOMIZATION as done for edminiv2 board.
Ok, but that is not my problem ;) And even if this macro is set, the wrong environment variable will still be set (in this case not a randomized mac address, but :00:00:<devnum> suffix).
This patch will make environment variable "egiga1" available
in such case.
Right?
mh? This patch will use the same enumeration as the net/eth.c code in eth_initialize(). At least if there is no other ethernet driver than mvgbe. So ethaddr is set to a random value instead of eth1addr, which eth_initialize() then use the set the mac address.
If so, then this is not the case with only mvgbe, it is
applicable for all
network drivers.
yeah other drivers also set eth(N)addr sometimes, which suffers from the same problem. maybe a driver could provide some callback to initialize a macaddress or eth_register returns the device index which in turn could be used to get the proper ethNaddr environment variable.
I think right place to provide solution for this problem is net/eth.c. The suggested change is logical but doing this will affect all u-boot users. May be some more opinion on this will be helpful.
ok, so the second option still applies, that is eth_register() returns a device index. the current eth_register always returns zero, in fact all drivers in the uboot source don't check the return code. so here we could return the device index.
@wolfgang: any opinion on that? or any other idea how to pass the ethernet device number to a network driver?