Hi all,
Well seeing Nick's patch spurred me to play with devfs this morning (thanks
Nick for the devfs=nomount kernel line option, makes it a lot easier to
test things out that way.)
As it is today (2.3.99-pre4-1), the usb-serial driver works just fine with
devfs without any changes. But I do have some questions about how usb and
devfs should work together (I suppose I should ask this of the devfs
mailing list, but I thought I would start out here.)
The usb-serial driver, when loaded, creates 16 nodes (ttyUSB0 - ttyUSB15)
in the main devfs directory. Namespace question: Is this the proper place
to put them? There isn't a usb subdirectory yet, and I don't think we
really need one (the input patches of Vojtech put the input devices in the
root devfs directory.)
And there's the issue that the usb-serial driver creates these nodes _when
the driver is loaded_ not when the device is plugged in. I had been looking
at changing this for a while now (only registering the node with the tty
layer when a device is actually found), but had not seen a good reason to.
Now it looks like this is necessary. Does anyone else think that this
should be changed? (if so, I suppose the acm driver should also be
changed.) That way, the ttyUSB?? nodes would only show up when a device was
found, and only those nodes that were actually connected would show up
(which I thought was the reason for devfs in the first place.)
And since the tty layer already does devfs_register_chrdev, why should the
individual serial driver also have to do a tty_register_devfs? (only a very
few serial drivers actually call tty_register_devfs: console, pty, and
serial)
thanks,
greg k-h
[EMAIL PROTECTED]
p.s. Apologies if 2 copies get sent to the list, I'm having mail
routing problems from work.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]