Description of problem: If the /proc directory is missing, the boot sequence doesn't try to create it before attempting to mount the proc filesystem onto it. Version-Release number of selected component (if applicable): Any version (FC4, FC5 and CentOS 4 all exhibit this problem) How reproducible: Always Steps to Reproduce: 1. Install a Red Hat-based distro (e.g. FC5, FC6T1, RHEL4, CentOS etc) onto a spare partition on your hard drive. 2. Mount that partition in your "original" OS and delete the /proc directory in that freshly installed distro. 3. Reboot and select the freshly installed OS to be booted via GRUB. Actual results: A mount error message is displayed when trying to mount /proc (because of the missing directory) and then the OS hangs a little further down in the boot sequence (in the bit where it initialises storage). Expected results: The boot sequence should detect the missing /proc directory and try to recreate it (and maybe display a warning/info message that it's done so) before trying to mount /proc onto it. Additional info: Yes, it's unlikely anyone would delete the proc directory, particularly on the currently running system (are you even allowed to...I've never dared trying!), but it's clear that early creation of the missing /proc dir is needed. A quick look at the initscripts code suggests that /etc/rc.d/rc.sysinit could do with some love - it currently reads: # Mount /proc and /sys (done here so volume labels can work with fsck) mount -n -t proc /proc /proc Can I suggest that at least a mkdir is placed before the mount command - e.g.: # Mount /proc and /sys (done here so volume labels can work with fsck) mkdir -p -m 555 /proc mount -n -t proc /proc /proc Yes, that's probably the one-liner no-brainer fix, but you can get fancier with tests for the directory and warning messages if you want...
The root filesystem is read-only, so that won't work at that point.