Comment 19 for bug 818177

Revision history for this message
Steve Langasek (vorlon) wrote : Re: HP DL380G5 root disk mounted read-only on boot and boot fails

I'm not sure if there's a kernel bug here (there might be?), but I'm pretty sure there is a udev bug, as indicated by the linked branch. However, just adding an extra 'udevadm settle' to the initramfs before killing udev is not sufficient to fix the race conditions here; udevadm settle + udevadm control --exit is not atomic, which means any new events that arrive in between the two calls might still be lost by whatever's going on in udevadm control.

udevadm control --exit needs to be made robust, to ensure that any events already being processed by udev at the time it's called get fully processed, and not lost in the middle as seems to be happening now.