Ben Nizette <[EMAIL PROTECTED]> wrote: > > On Tue, 2008-09-02 at 14:09 +0200, Haavard Skinnemoen wrote: > > Right. I was thinking maybe we could extend the boot protocol to pass > > partitioning information from u-boot to the kernel -- that should make > > it possible to run the same kernel on boards with different flash > > layout as long as the in-flash u-boot is consistent with the flash > > layout. > > If we do end up with libfdt on avr32 then we solve our problems here > too, we can encapsulate this partitioning in the fdt image.
Yes, we can probably do that. I guess it all depends on what we want to do first. > But, as I > understand it, that involves not only the actual libfdt support but > making all avr32 driver nice and OF compatible. At least, if we want > the fdt support to be _useful_ then that needs to happen (CMIIW). Or we could just add a glue layer which creates platform_devices based on the fdt information. Perhaps pass along a pointer to the device tree node for drivers that want to peek at it. PowerPC has a of_node in struct dev_archdata for this purpose I believe. > > One problem this doesn't solve is that if you load an older kernel > > which doesn't understand the extended boot protocol, it will keep using > > the old flash layout. > > Doesn't the MTD system have facility to override partitioning > information from the boot arguments? If so then any big u-boots can > just ship with such a clause in their default bootargs, job done. Have > to look in to that... Yes. But it won't fix boards which already have a valid environment and just want to upgrade u-boot. We could perhaps fix up the bootargs and add partitioning information when preparing to boot the kernel... Haavard _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
