udev problems

Kevin P. Fleming kpfleming at linuxfromscratch.org
Tue May 25 20:23:24 PDT 2004

Bryan Kadzban wrote:

> Whichever script calls udevstart seems to start it up.  I'm getting the
> same results on my machine (I'm using the 2.6+udev hint, with the
> lfs-bootscripts-2.1.1-pre3 udev script instead of the one in the hint,
> plus a patch to the udevstart binary to not set the UDEV_NO_SLEEP env.
> variable, so that udevstart waits until udev is finished populating
> /dev before exiting).

I don't see that here, doing LFS unstable builds with lfs-bootscripts-2.1.1.

> When I was first trying to get udev going, running udevstart
> manually to populate my test device tree (which was not stored in /dev)
> seemed to spawn off the udevd process.

I have never seen that, that's strange. I wonder if udev itself can 
start udevd if the config file tells it to... I've never looked.

> I don't see anything in the source of udevstart, though, that does it
> directly -- from comments in the udevd manpage, it looks like hotplug
> calls udevsend (which makes sense: udev's Makefile installs a symlink to
> udevsend in /etc/hotplug.d/default/), which talks to udevd.  According
> to the udevstart source, it starts up udevd in at least one place
> (presumably if it isn't running, though I haven't verified that).

I think you meant s/udevstart/udevsend/ in that last sentence; I agree 
that the udevstart source doesn't have any direct way of starting udevd, 
it doesn't even know about udevd.

> What I'm not sure of is why running udevstart generated hotplug events.
> Maybe it was just coincidence...

Could be, generating device nodes really shouldn't cause any more 
hotplug events.

More information about the lfs-dev mailing list