Description of problem: After 20040820 update I got a number of weird errors on startup culminating in X totally locking up machine with a blank screen. A closer investigation revealed that most of my /dev/ is gone. This is what I see: # rpm -V dev | grep missing | wc -l 19162 In whatever is left (not much) all block and character devices have 600 permissions and this includes things like /dev/null (!!!). X locks up because nothing exists in /dev/input/ and /dev/mouse also cannot be found. Luckily enough was left that I can still login. I happen to have udev installed but in /etc/rc.d/rc.sysinit there is the following code: if [ -f /etc/sysconfig/udev ];then . /etc/sysconfig/udev fi if [ "$USE_UDEV" = "yes" -a "$UDEV_RAMFS" = "yes" ]; then [ -x /sbin/start_udev ] && /sbin/start_udev fi and I do not have /etc/sysconfig/udev at this moment so this should not do anything. What is even more weird that x86_64 updated to the same level is not showing up any symptoms of that sort. I am somewhat baffled but something decided to mess up my /dev quite thorougly. A buggy %post script in some package and not initscripts at all? Version-Release number of selected component (if applicable): initscripts-7.64-1
A subject was supposed to say "... something ate most of /dev/". It seems that this gobbling is contagious.
The next strange thing. An attempt of recovery by reinstalling 'dev' package brought: Cannot install the dev package: mounted devfs detected. error: %pre(dev-3.8.3-1) scriptlet failed, exit status 1 error: install: %pre scriptlet failed (2), skipping dev-3.8.3-1 'devfs' was never in use on a machine in question but this gives a clue. This is a bogus message from %pre script in dev and responsible for that mess is a lack of synchronization between mkinitrd and initscripts which use different criteria when to employ 'udev' so I end up with very incomplete udev courtesy of initrd as there is no followup.
There is a fix for the udev start confusion in initscripts-7.65-1.
*** Bug 130516 has been marked as a duplicate of this bug. ***
This should be fixed in the final releases of udev-containing releases (such as FC3/FC4); moreover, there are notes about how to update to udev with yum. Ergo, closing.