On Mon, Aug 19, 2013 at 7:46 PM, Bryan Kadzban <br...@kadzban.is-a-geek.net> wrote: > Bruce Dubbs wrote: >> Ken Moffat wrote: >>> I'm about to do an almost [1] "by the book" build (-j1, keep static >>> libs until the end of chapter 6 which should at least make more of >>> the ld tests pass) so I rebuild my kernel to add >>> CONFIG_SCSI_DEBUG=y. Disaster : /dev/sda showed up with no >>> partitions, all my partitions previously on /dev/sda were now in >>> /dev/sdb so the root= line had become invalid, and /etc/fstab would >>> be invalid if I'd managed to boot. > > So, uh, here's where I plug /dev/disk/by-id/* symlinks. > > Combine with an initramfs, and it doesn't matter what the kernel decides > to randomly rename your device files on any single boot... :-P But > even without an initramfs, the only thing you have to worry about having > to manually change is the root= arg. > > (Or by-uuid for luks encrypted volumes, since luks provides a sane uuid > and you really want to make sure you're unlocking the right volume. The > uuid setup doesn't work for everything though.) > >> That's better than I got. I had consistent panics during boot. >> >>> Does using CONFIG_SCSI_DEBUG=m let the tests work ? (i.e. will the >>> module be loaded automatically when util-linux runs its testsuite >>> ?). >> >> That worked for me. I got it to pass all the tests. But note that >> it needs some packages beyond LFS. I don't recall which at the >> moment. I did not do the root tests in chroot, but after I rebooted. > > Personally, I am a fan of root=PARTUUID=blablabla. no initramfs/initrd required. Only works with GPT Partitions, not with MBR ones. The UUID is for the partition itself, not the filesystem's UUID.
-- Nathan Coulson (conathan) ------ Location: British Columbia, Canada Timezone: PST (-8) Webpage: http://www.nathancoulson.com -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
