Guillaume Cottenceau wrote: >pascal <[EMAIL PROTECTED]> writes: > >>If you or a system configuration application put a 'rw' option in /etc/fstab >>on the ROOT Filesystem '/' by mistake then you do a mkinitrd, then lilo >>then you are in big troubles at next boot with the logic in /etc/rc.sysinit >>because at boot the ROOTFS is already mounted RW and Fsck cannot be done. It >>returns the error : >> >>" /dev/hda7 is mounted. e2fsck: Cannot continue, aborting. >> Failed to check filesystem. Do you want to repair the errors? (Y/N) >> (beware, you can loose data)" >> >>and then you are trapped in an endless reboot cycle because sysinit will >>always ask to repair an error that does not exist ... :) >> > >Yes, you're in trouble. The same as if you "dd if=/dev/zero >of=/dev/hda", etc. You can do a lot of bad things touching to >your system... > > >>The solutions are one or more of the following : >> >>* dont let mkinitrd create initrd images with ROOTFS option rw in /etc/fstab >>by default and give an option to do it >> > >I'm not gonna add a check for that situation, we'd need to add >another hundred of checks for errors that could happen with the >same frequency.. > > AFAIK root fs may need to be rw for umsdos and possibly other fses. But if initscrips do not support this theres no reason to support this in mkinitrd. To be consistent mkinitrd should do images that mount root ro IMHO. But if more flexibility is wanted an option to start with root rw could be included in initscripts.
Anyway both are minor issues and there are probably more important ones to solve. And you could always create a patch for the script which generates fstab to include a little comment on this for the confused newbies ;-)
