Eudev not setting device node permissions/owner/group about eudev HOT 5 CLOSED

Does running udevadm trigger correct the permissions?

from eudev.

The problem is that you don't udevadm trigger late enough in the boot process. I couldn't pin point a good place to do it, but if you run udevadm trigger after boot is over, all the correct permissions and ownership are set. Note you don't need to mount /dev yourself and you should probably check mountpoint /dev before triggering. Sorry I can't help more, but I'm not familiar with the boot scripts.

from eudev.

ryao: it does

from eudev.

We do have to mount /dev ourselves, in the 'init' script in the initramfs, just before doing the switch_root:

mount -t devtmpfs devtmpfs /pup_new/dev

/pup_new will become / after the switch_root.

This explicit mount is required, we determined after some study of the subject.
A system that does not have an initramfs, that boots straight into the rootfs, does not have to explicitly mount /dev, the kernel does it.

By mounting it explicitly in the initramfs, the kernel will create it with all device nodes discovered so far.

After switch_root, the kernel will continue to populate /dev, as we replay uevents from /sys, by two means, after having started udevd (see /etc/rc.d/rc.sysinit):

```MODALIASES="ls /sys/bus/_/devices/_/modalias"
...
load udevd
...
#replay uevents from /sys...
for ONEMODALIAS in $MODALIASES
do
ONEPATH="`dirname $ONEMODALIAS`"
if [ -e ${ONEPATH}/uevent ];then
echo add > ${ONEPATH}/uevent #generates an 'add' uevent.
sleep 0.02
fi
done

2.
and by calls to 'udevadm trigger <parameters>'

...in other words, all (relevant) events are getting replayed from /sys

Ah, I might have solved it, or rather identified the basic cause.

I just looked in /dev, and /dev/ttyUSB0 does have the correct ownership. This node got created later on, when I plugged in my 3G USB stick after bootup.

Ok, it seems that the "focused" replay of events from /sys, that we do in /etc/rc.d/rc.sysinit, rather than a broad "udevadm trigger" to replay everything -- which we avoided for various reasons -- is the cause of this problem.

This is probably enough of a lead for me to fix it. Thanks for the feedback.

from eudev.

OK, I have implemented a fix in Puppy Linux.
We can consider this one closed.

from eudev.