HEADSUP usb2 (usb4bsd) to become default in 2 weeks. [usb2 configuration]

[Bruce Cran]

On Sat, 27 Dec 2008 14:59:24 -0800
Sam Leffler <sam at freebsd.org> wrote:

> Hans Petter Selasky wrote:
> > On Saturday 27 December 2008, Dominic Fandrey wrote:
> >   
> >> Alfred Perlstein wrote:
> >>     
> >>> We're going to usher in the New Year with a new usb stack.
> >>>
> >>> Now is the time to test, test, test.
> >>>
> >>> It is also the time to point out anything missing from usb2 that
> >>> is in usb1. ...
> >>>
> >>> See attachment.
> >>> ...
> >>>
> >>>
> >>>  # USB Serial devices
> >>> -device		ucom		# Generic com ttys
> >>> -device		u3g		# USB-based 3G modems
> >>> (Option, Huawei, Sierra) ...
> >>>       
> >> It seems u3g doesn't show up in the usb2 stuff. For me
> >> this would be a real show stopper if I ran current.
> >>
> >>     
> >
> > The U3G driver has been ported to USB2. See "usb2_serial_3g".
> >
> >   
> 
> I think this sort of confusion is indicative of problems with the way 
> usb2 is configured.  I recently switched a machine to usb2 and I
> naively expected that doing this would just involve changing all
> items like
> 
> device   ums
> 
> to
> 
> device   ums2
> 
> when this didn't work I looked for manual pages but found none.  So
> then I looked in sys/conf/files and found that every driver has a new
> name; e.g. to get ums I now have to specify:
> 
> device   usb2_input_ms
> 
> but more than that I also have to include
> 
> device   usb2_input
> device   usb2_core
> 
> I suggest this is not a good idea.  It is possible through module 
> dependencies and config rules in files to handle all this 
> automatically.   Perhaps I've missed something here but I think that 
> after the switchover users should not need to alter their config
> files unless they are using new features.  This is all the more
> important in that all these names changes will require all
> documentation to be altered to match.
> 

Something that confused me for a while was when I forgot to add
usb2_controller on one machine but did have 
usb2_controller_[uhci,ehci,ohci].  The kernel built and ran, but USB
just didn't work until I kldload'd the controller modules.  Is there a
dependency mechanism to make the build fail when a mistake like this is
made?

The new devices should probably be added to NOTES too - there's a PR
open with a patch
(http://www.freebsd.org/cgi/query-pr.cgi?pr=conf/128590) but it needs
usb2s_core corrected to usb2_core before it can be committed.

-- 
Bruce Cran