Comment 113 for bug 19171

Created an attachment (id=3944)
syslog from failed boot, with full debugging from udevd start

Added syslog contributed from Charles Henderson; this contains the boot
sequence of a machine that didn't plug /dev/input/mice. Full udev debugging
was enabled and syslog started early enough to capture it all (other than
udevstart, which we've already ruled out).

Just a quick read shows some fun; we get 10 udevsend events all at once all of
which try to start a udevd. Noteworth is that /class/input/mice is amongst
them, and reaches the udevd -- but that the udevd throws it away because
netlink is already active by this point.

Next up, compare this with my logs to see why the event _isn't_ thrown away for
me.