From: Mike Frysinger Date: Mon, 11 May 2009 12:56:59 -0400 Subject: [U-Boot] [PATCH 1/4] net: extend the netdev to have a common way to set the hw mac address In-Reply-To: <20090511163728.GB32661@game.jcrosoft.org> References: <1242023969-13542-1-git-send-email-plagnioj@jcrosoft.com> <200905111224.32766.vapier@gentoo.org> <20090511163728.GB32661@game.jcrosoft.org> Message-ID: <200905111257.00648.vapier@gentoo.org> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Monday 11 May 2009 12:37:28 Jean-Christophe PLAGNIOL-VILLARD wrote: > On 12:24 Mon 11 May , Mike Frysinger wrote: > > On Monday 11 May 2009 10:24:54 Jean-Christophe PLAGNIOL-VILLARD wrote: > > > On 15:26 Mon 11 May , Wolfgang Denk wrote: > > > > In message Jean-Christophe wrote: > > > > > > there is no need for that on purpose. the network layer will > > > > > > recheck the environment every time it gets initialized, and it > > > > > > gets initialized with every network step. this is already > > > > > > documented in doc/README.enetaddr. > > > > > > > > > > Unfortunatly we need it to set the enetaddr in the register of the > > > > > soc evenif there is no ethernet transfer as for the macb on at91 or > > > > > the dm9000. This how the kernel pass the mac address for the > > > > > bootloader. > > > > > > > > You mean, this is how U-Boot is supposed to pass the MAC address to > > > > the Kernel, right? > > > > > > > > Well, strictly speaking this is (1) in direct violation of U-Boot's > > > > design principles which say NOT to initialize things it does not use > > > > itself, and (2) a bug in the Linux kernel that should be fixed there. > > > > > > > > In this case it may be just writing to a register without further > > > > initialization of the hardware, but if we allow for that you can bet > > > > that tomorrow someone will try and add code that requires more > > > > extensive initialization of the hardware for the same purpose, and he > > > > will argue that his code must go in because there is preexisting code > > > > that does the same. > > > > > > this exactly what is done only write mac hw register > > > without init the eth stack > > > > > > before we do need to init the mac and detect the phy if we want to > > > store now we have to call the set_hw_enetaddr > > > > then we agree you're doing it wrong and this patch is not needed/wanted > > > > there is this FAQ, but i think we can add more info to it: > > http://www.denx.de/wiki/view/DULG/EthernetDoesNotWorkInLinux > > It does not work ether your solution blatantly ignores the issues raised in the FAQ. you cant push a change without clearing addressing those. > how do you deal with multiple nic? the ATAG/FDT tree handles most arches. for the rest you have two major issues to address: - how do i pass it to the next OS ? * have the OS read the u-boot environment * have the OS read the command line - how does the OS set it ? * add a board hook to set it * set it via user space and normal methods (`ifconfig hw`) > how do set a mac for NFS Rootfs? use initramfs > how do you boot a *mainline* kernel? i dont know why you're emphasizing mainline. if it's your board and if you implement the change correctly, submit it to mainline. -mike -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 836 bytes Desc: This is a digitally signed message part. Url : http://lists.denx.de/pipermail/u-boot/attachments/20090511/2e4100ca/attachment.pgp