On Tue, 28 Mar 2000, Michael H. Warfield wrote:
> > I guess the current configuration will be a major problem for upgrades.
>
> Yeah... Anyone using USB or a whole bunch of other "loose ends"
> are in for a rude awakening. "Loose ends" was someone else's term for
> it after I rolled devfs support into the Computone drivers and mentioned
> a lot of broken support even for things listed in devices.txt. Those
> updates were submitted but haven't even made it into the main kernel
> sources yet, so I'm pretty confident that devfs support is going to be
> really broken, when 2.4 is released, with a lot of drivers which just
> don't support it. There are a lot of things missing devfs support now
> and, if that isn't sufficient to get changes incorporated, it's going to
> be missing in the 2.4 rollout.
I quite disagree. I use numerous [different] USB devices and I don't have
any problems at all. On the contrary, without devfs I have the hassle of
figuring out which new mknod I need to update /dev. With devfs I haven't
had to worry about anything.
> Even after support for devfs is rolled into the drivers, the user
> space devsfd daemon is missing a lot of links from the traditional names
> to the devfs names. That means that a lot of existing apps and
> configurations are going to break, even if you are running devfsd. It
> creates a real uncomfortable critical dependency between enabling devfs
> and switching all of your configurations. Upgrades are going to be hell
> with lots of little devils in lots of little details.
I have the standard makecompat for devfsd. I have -one- application which
breaks. X. Compiling a kernel with devfs and putting devfsd onto a machine
which was sans devfs is a breeze. Compile, put devfsd and conf files on,
edit boot script. Reboot. All is well.
> The option of leaving /dev as is, then mounting devfs on /devfs and
> providing appropriate symlinks from one to the other answers a lot of those
> problems, but not all of them. It's a compromise that keeps the pain level
> of the transition to a minimum. It's also the way Sun handled this
> situation on Solaris, except there it's /dev and /devices where /devices
> has the magical entry points as defined by the device drivers.
I find mounting over /dev more enjoyable, easier, and overall more
simplistic.
-d
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 21:00:23 EST