probs on 6.2-prerelease

[Kris Kennaway]

On Tue, Sep 26, 2006 at 06:14:51PM -0700, Jeremy Chadwick wrote:
> On Wed, Sep 27, 2006 at 10:46:28AM +1000, Michael Vince wrote:
> > Yeah for some reason I couldn't do that, I can't even remember all the 
> > reasons now but  the main reason is because I have a USB keyboard, for 
> > some reason I can use the USB keyboard in the BIOS but as soon as the 
> > boot loader starts with the 10sec menu I loose all control of the keyboard.
> > Booting into fixit mode from the CDROM loads the USB drivers (as well as 
> > not hanging up the OS anywhere) so I could mount the filesystem and fix it.
> 
> I can confirm this problem, and have been fighting with it for quite
> some time.  BIOS is configured for "USB Legacy" support (which should
> enable the keyboard in such environments).  boot0, boot2, and loader
> all accept the keyboard just fine.  Once the kernel loads, it loads
> the usb layer, detects the keyboard (via ukbd), and of course pops
> up the "Pick a shell" single-user prompt.  But no keyboard input.
> 
> This behaviour happened only on my Intel machine; my current AMD
> machine does not do this.
> 
> The problem as I saw it (which may be an incorrect prognosis, since
> my AMD box says the same thing as the Intel via dmesg, but behaves
> fine) was that the FreeBSD kernel was considering the default keyboard
> device as atkbd* rather than ukbd*.  It wasn't a problem outside
> of single-user, since usbd would load and do some magic to make it
> all work... but usbd doesn't get loaded in single-user.  :-)
> 
> The workaround I found was to use some loader.conf tweaks to disable
> specific AT/PS2 keyboard devices.  I believe I used something like
> this, but I can't remember (and don't have loader.conf from my Intel
> box :( -- And yes, I know psm is for the mouse):
> 
> hint.atkbdc.0.disabled="1"
> hint.atkbd.0.disabled="1"
> hint.psm.0.disabled="1"
> 
> This oversight really needs to be addressed.  I'll gladly buy anyone
> who wishes to solve it a new USB keyboard to test/debug with.

Wasn't the kbdmux framework introduced some time ago to solve this?

Kris