On Thu, Sep 20, 2001 at 11:20:26AM -0400, Dan Malek wrote: > > Stefan Roese wrote: > > > #ifndef __CPCI405_H > > #define __CPCI405_H > > > > #include <linux/config.h> > > > > #include <asm/ppcboot.h> > > > > #endif /* __CPCI405_H */ > > This is fine, Tom said it was an empty file.........
Well, that's empty to me. :) ppcboot.h should include linux/config.h and that's all it does... > > Dan, you are right that we (PPCBoot) people will have some homework to do, > > but right now our only way to Linux is via the bd_info struct! > > No, the way to Linux is to _always_ use a zImage with an attached bootloader. > PPCboot can continue to pass whatever it wants, or the bootloader can find > the information in other ways as it does today with many boards. Booting > a vmlinux image and trying to keep track of that private interface between > the bootloader and the kernel is difficult. We already have more changes > planned in addition to the bi_recs interface. Much of the code found today > in the head.S files is going to migrate to the bootloaders. In the x86 world it's always true that we have a bit of a loader tacked on someplace. In some parts of the PPC world, we can't/don't for various reasons (technical and/or ingrained). For example, I don't see a problem with something which takes a vmlinux.gz and boots it. As long as we get the information we need, we shouldn't care. The in-kernel loaders are the easiest to get changed. But we should define what we're going to do before we do it so that yaboot/ppcboot and apus can figure out how much re-working they'll need and whatnot.. -- Tom Rini (TR1265) http://gate.crashing.org/~trini/ ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
