On 26.12.15 19:55, Leif Lindholm wrote:
On Fri, Dec 25, 2015 at 10:25:22AM +0100, Andreas Färber wrote:
Am 25.12.2015 um 10:02 schrieb Alexander Graf: [snip]
The reason I implemented "bootefi" was really because it's the natural fit into how U-Boot handles all other formats today. I don't think this is going to be the last patch set around EFI support.
I think what Matwey was suggesting is integrating your "bootefi" into the standard "distro" boot sequence environment, so that it probes each device for an EFI binary and if it finds one runs load and bootefi, without the need for any boot.scr.
That would be a follow-up patch.
It however conflicts with your idea of having some potentially board-specific code mess with "fdt addr" command before running "bootefi".
This could however be resolved by moving to a model where the device-tree was considered a component of the firmware (which is how we treat it in UEFI). If U-Boot had an awareness (if it does not already) of an FDT being something it should have and know about, then this could trivially be passed onto bootefi. This could means a "default_fdt" environment variable which is loaded automatically before bootcmd is executed.
That part is already the case in my patches :).
The difficult bit is what to do when it doesn't work, because the kernel guys consider a dtb to be a "Linux configuration" framework rather than the "hardware description" framework it's supposed to be.
Alex