Can mdev replace udev in an ordinary linux distro?

[Tom Gundersen]

Hi guys,

I have been following this thread, and I wonder if someone could
explain to me what is the real reason for wanting a replacement for
udev?

I saw some passing references to udev being slow, could someone
elaborate? I have spent some time looking at boot speed (and hence
udev speed), and found that if you remove any distro-specific
hacks/rules from udev things speed up a lot. Furthermore, any slowness
is usually in apps called by third-party rules. I'd be interested in
seeing some more details about the problems people are running into.
For what it's worth I'm using a Pentium 4 for testing, so hardly top
of the line, but maybe problems are easier to see on ever lower end
systems.

Some misconceptions about udev I spotted in the tread so far:

/usr: this is not a problem with udev, but rather with third party
programs. Udev works perfectly fine with a separately /usr, but plenty
of third party programs (that install udev rules files) will fail. For
more info see: <http://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken>.
Udev is not going to force anyone to do anything with /usr, if
anything, you might at most be required to maintain a simple patch.
The problems with third-party programs will not go away on its own
though...

redhat: udev was created by a novell employee and the current
maintainer only very recently moved from novell to redhat (I think it
is counterproductive to bring who employs whom into the discussion
though).

devtmpfs: the reason to use devtmpfs that I have heard is that it
allows the information about devices that the kernel already has, to
be used directly rather than figuring it out again in userspace (so
you do half the work). It also avoids some race-conditions with node
creation if nodes are created from userspace. I'm not an expert on
this though, so I might be misstaken. That said, it might be worth
looking into in more detail before deciding that devtmpfs is no good.

Cheers,

Tom