rpjday <[EMAIL PROTECTED]> writes: > On Sun, 26 May 2002, Joachim Breuer wrote: > >> For those running custom kernels with ext3 as a module: Also remember >> to rebuild the initrd (after editing fstab), otherwise the root root >> fs cannot be mounted... Not insurmountable using the install cd/rescue >> mode, but unneccesary hassle. > > technically speaking, this is not true. even without the ext3 module > in the initrd.img, the boot will work just fine. given my curiosity > with the different run levels, i've been experimenting with just what > i can get away with in terms of booting/troubleshooting, and i tested > just that. > > the boot works fine, and if you check /var/log/dmesg afterwards, you > see the line: > > VFS: Mounted root (ext2 filesystem) readonly. > > so even though the root filesystem is an ext3 with the associated > .journal file, there's nothing wrong with the kernel mounting it > as an ext2 to start with, at which point, since the /lib/modules > directories are on that filesystem, the ext3.o module can be loaded > out of there. and afterwards, "lsmod" confirms that the ext3 module > has been loaded. > > i tested this by just deleting the "initrd" out of grub.conf -- > no problem. i suspect, of course, that if you were missing other > initrd modules like scsi or raid, you'd have trouble. > > i'm hoping i tested this the right way -- anyone else want to > confirm?
The problem occurs if your initrd is assuming an ext3 filesystem and there no longer is a journal inode on the drive. mount fails. If it was the root fs it's at least annoying to fiddle the initrd back to ext2 by booting from cd. - So my "unintentional test" was exactly the opposite case as yours. Sorry if I was unclear about that: You'll run into problems when your initrd assumes an ext3 root filesystem and the device contains an ext2 filesystem. At least on my rh 7.2 boxen the initrds that get generated with ext2 in /etc/fstab vs. ext3 in /etc/fstab are *not* identical. Mounting an ext3 filesystem (on disk) with the ext2 kernel driver works, the other way around did not work for me. (ext3: Cannot initialize journal - or something like that). I do not know whether the root filesystem type is always "built into" the initrd, or only (as in my case) when the root fs needs to be loaded as a module. Or maybe depending on something completely different. I also do not know whether there is maybe a simple kernel command line flag to override the built-in setting. So long, Joe -- "I use emacs, which might be thought of as a thermonuclear word processor." -- Neal Stephenson, "In the beginning... was the command line" _______________________________________________ Redhat-list mailing list [EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-list
