On Thu, Oct 16, 2003 at 11:21:07PM +0200, Mich Lanners wrote: > On 16 Oct, this message from Chris Tillman echoed through cyberspace: > >> Chris, since you're rebuilding quik :-), maybe you can give this a > >> try as well? > > > > I can; I need to hook up a SCSI disk to really test it? Does the root > > partition/kernel have to be on the SCSI? > > Well, I asked myself that same question, and it does seem that the > second stage's ext2 code can only access SCSI disks??
Hmm. No, I have no problem booting /dev/hda6 with quik. I just meant, does the bug this is supposed to fix only appear on SCSI disks? Does it appear only when the root partition is on a SCSI disk? > In load_file() (second/file.c), we first make OF open the device where > we want to load the kernel from, and then ext2fs_open() with /dev/sdaX > as device path. Unless ext2fs_open doesn't care about the path itself? > > Ahh, that seems to be the case. We give the ext2 functions a pointer to > an array with our own IO functions, and those seem to just extract the > partition number (albeit again only a single digit...). See for instance > linux_open() in second/file.c I can't read C code as well as you obviously can. But I could test your patch; I just need to know the success/fail criteria. -- Debian GNU/Linux Operating System By the People, For the People Chris Tillman (a people instance) toff one at cox dot net