On Thu, 13 Nov 2008 09:30:51 -0800 Ben Warren biggerbadderben@gmail.com wrote:
Heiko Schocher wrote:
Hello Ben
Ben Warren wrote:
Heiko Schocher wrote:
Check the presence of the PIGGY on the keymile boards mgcoge, mgsuvd and kmeter1. If the PIGGY is not present, dont register this Ethernet device.
Signed-off-by: Heiko Schocher hs@denx.de
This looks like useful stuff to have, but I'd prefer that you put the check logic in board_eth_init() rather than adding to the individual device drivers. I know the 8260 SCC driver is the older style, which precludes the use of board_eth_init, but I'll convert it if you're able to test.
Unfortunately, this approach won't work. First of all, the 82xx SCC driver is now initialized in cpu_eth_init(), which knows nothing about board-specific peculiarities like the PIGGY. Secondly, the HDLC driver for Keymile has to be initialized in board_eth_init(), and it has nothing to do with the PIGGY. Putting the check in board_eth_init() would break it completely. I looked at Heiko's latest patch and couldn't figure out a way to cleanly differentiate between initializing the HDLC driver and checking whether the PIGGY was present fo the other ENET drivers.
Yes, I could test such a change for you, but hmm... I am not sure, if board_eth_init () is the right place for my purpose. I need for every Ethernet device a selection, if this device is present or not. Correct me if I am wrong, but it looks like board_eth_init () is not made for this purpose. (Ok, I can do a specific device init in board_eth_init (), but then we must do something, that prevents that the device is again initialized in eth_initialize () ...
board_eth_init() was introduced for exactly this sort of thing. Have a look at the net repo (I've sent a pull request to Wolfgang so the current changes will make it into the 12.2008 release). There aren't any device initializations left in eth_initialize(), so there's no issue of a device being initialized twice. The goal is for all devices to be started by cpu_eth_int() or board_eth_init().
Hmm... while writing this it comes a idea in my mind: we could move all the *_initialize functions in eth_initialize () in a seperate function, say eth_hardware_init() and maybe making this function "weak", so a board writer can write his own eth_hardware_init() ... in such a function, I could check which device is present, and only initialize the present devices ... what do you think?
That's what board_eth_init() and cpu_eth_init() are for. In addition, I forgot to mention that a couple of days ago Gary Jennejohn submitted a patch for changing the 82xx SCC driver over to CONFIG_NET_MULTI style.
Correct, and it must explicitly call ethernet_present() - see above.
--- Gary Jennejohn ********************************************************************* DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office@denx.de *********************************************************************