initrd and 2.6.0-test8
Quoted 28 lines · Walt H
Quoted 12 lines
Quoted 12 lines
Hi,
Seems that something changed between test7 and test8 regarding initrd or romfs support. I'm using highly modularized 2.6.0 kernel which has all filesystems beside romfs compiled as modules (romfs is compiled inside of kernel).
Modules for my rootfs are loaded from initrd (which is image with romfs as filesystem) but starting from test8 kernel is not able to mount initrd filesystem - stops with typical message about not being able to mount rootfs.
Quoted 5 lines
Quoted 2 lines
cset test7 from 20031012_0407 is known to be ok so something was changed later
I noticed this happened in 2.6.0-test6-mm4. Backing out this patch fixes it in the short-term.
Even better would be to report the bug ;-/
I can't reproduce it here. 2.6.0-test8 vanilla, so far (last 15 minutes) tried with
* compressed initrd image
* plain ext2
and I'll try romfs as soon as I hunt down mkfs for that animal. All appears to be working...
Quoted 4 lines
What did it say before the "typical message"? Specifically, were there any lines starting with RAMDISK:?
.config would be also useful.
But do you have HDD-controller parts as modules or built-in? According to the changelog message, if initrd is detected as 'usual initrd' (e.g. not initramfs), then it's copied to /dev/initrd on rootfs. But as I understand, there is no such thing as rootfs as long as it isn't mounted (ide/scsi modules are not loaded
--
Witold Kręcicki (adasi) adasi [at] culm.net
GPG key: 7AE20871
http://www.culm.net
Quoted 5 lines · Witold Krecicki
But do you have HDD-controller parts as modules or built-in? According to the changelog message, if initrd is detected as 'usual initrd' (e.g. not initramfs), then it's copied to
/dev/initrdon rootfs. But as I understand, there is no such thing as rootfs as long as it isn't mounted (ide/scsi modules are not loaded
What happens in the case of a devfs-based system where the real rootfs can't be mounted till after the initrd is loaded, because / is on an LVM volume? I know my laptop crashed-and-burned on this, even though the IDE and LVM modules were built into the kernel....
Quoted 11 lines · On Sat, Oct 18, 2003 at 03:41:43PM -0400, [email protected] wrote:
Quoted 5 lines · Witold Krecicki
But do you have HDD-controller parts as modules or built-in? According to the changelog message, if initrd is detected as 'usual initrd' (e.g. not initramfs), then it's copied to
/dev/initrdon rootfs. But as I understand, there is no such thing as rootfs as long as it isn't mounted (ide/scsi modules are not loadedWhat happens in the case of a devfs-based system where the real rootfs can't be mounted till after the initrd is loaded, because / is on an LVM volume?
Guys, rootfs is there from the very beginning. It's just a ramfs - doesn't require any drivers, is writable and is there. We mount ramdisks, final root, devfs, etc. on it. Presense of drivers doesn't matter - we don't need any drivers to access that beast, since it just sits in pagecache. No backing device at all.
Quoted 2 lines
I know my laptop crashed-and-burned on this, even though the IDE
and LVM modules were built into the kernel....
Details? How long ago had it started, .config, boot log, etc...