I agree with Brad's points here, especially the user
point of view. We're going to have lots of user email
about /proc/bus/usb not mounted unless we find a
solution to this.
I don't see the proc fs as something that we can make
_required_, however, so what are the alternatives?
I just modified usb/inode.c to mount usbdevfs
just after it creates /proc/bus/usb [and to umount
it just before it kills /proc/bus/usb, but that
part doesn't work since rmmod sees use_count == 1].
[Oh, it doesn't show up in mount(8) output either...
but I can cd /proc/bus/usb and cat drivers and cat devices.]
Maybe we could do something like this conditionally,
although it goes against lots of Unix/Linux principals
of not auto-mounting (altho that's what devfs does also).
~Randy
___________________________________________________
|Randy Dunlap Intel Corp., DAL Sr. SW Engr.|
|randy.dunlap.at.intel.com 503-696-2055|
|NOTE: Any views presented here are mine alone |
|and may not represent the views of my employer. |
|_________________________________________________|
> -----Original Message-----
> From: Brad Hards [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, May 04, 2000 1:50 AM
> To: Jim Gettys
> Cc: Thomas Sailer; [EMAIL PROTECTED]; Matthew Dharm; [EMAIL PROTECTED];
> [EMAIL PROTECTED]
> Subject: Re: [linux-usb] X support for USB devices...
>
>
> Jim Gettys wrote:
> > My claim is that finding out if devices come or go
> > should be a reliable part of the system: having to explicitly mount
> > a file system is fragile, and I don't like fragile systems that
> > break in mysterious ways. I think that /proc/bus/usb
> should always be
> > there whenever there is USB in your system...
>
> On usbdevfs, see my comments below. I think that using
> /proc/bus/usb is
> OK for development, but a more general approach is required for later
> kernel additions, like perhaps an extension to Vojtech's input system.
> Remember that X won't care about certain devices, and X is
> not the only game in town
>
> The user point of view: once you get the right combination of magic
> module.conf and /etc/fstab, you never think about it again. Just like
> proc, which is _not_ guarenteed to be mounted, although
> almost everyone
> does mount it on /proc. However until you get it sorted, it is a
> nightmare.
>
> The newbie helper / documenter point of view: a lot of people
> stuff this
> up, or don't understand why they need to mount it, or don't
> have a clue
> what mount and filesystem are. We need to help them more.
>
> The developer point of view: the system needs to support arbitrary
> loading and unloading of modules, including the usbdevfs function for
> developers working on extending usbdevfs.
>
> This is a hard set of requirements to cover completely, but I
> think the
> user part is the most important, and we need to find a
> stable, safe and easy (for the user) way to do this.
>
> Brad
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
