mdev race conditions

Bernard Blackham bernard at largestprime.net
Wed Jun 4 09:04:16 UTC 2008


On Wed, Jun 04, 2008 at 04:27:31AM -0400, Mike Frysinger wrote:
> [udev] does a *lot* of stuff that people keep asking to add to mdev.
> historically i declined anything but the bare min stuff as imo,
> we're wasting time reinventing a perfectly good wheel.

I'm not asking for features. I just want reliability. If people use
mdev through /proc/sys/kernel/hotplug as it is currently documented,
they will hit race conditions 1% of the time which may slip through
their testing. (Luckily, we're hitting it about 20% of the time
here).

> > I'd like someone to answer my original proposal of whether or not a
> > simple daemon running in a loop would suffice. I imagine the resulting
> > binary would be a mere few kilobytes in size, as *all* it needs to do is
> > serialize requests and pass them to mdev. Nothing more.
> 
> writing a standalone binary does not fit the busybox design.

Sorry, what I meant to say is that the resulting addition of mdevd
to the busybox binary would only add a few kilobytes to its size.

> i doubt mdevd would be terribly difficult to do with real netlink
> sockets/daemon.

And that's exactly what I'm proposing. I think I'll just do it. I'll
let you know how I go :)

Cheers,
Bernard.




More information about the busybox mailing list