I believe the problem is that atapibus doesn't seem to want to connect to
usb mass devices anymore.  I have a compact flash card reader that
attached as sd0 on atapibus0 on umass0 on usb1:

  | +-ohci0     
  | | +-usb1    
  | |   +-uhub1 
  | |     +-ums0
  | |     | +-wsmouse0
  | |     +-umass0
  | |       +-atapibus0
  | |         +-sd0

But since the new scsipi stuff was checked in it now connects as uk0 on
scsibus0 on umass0.

  | +-ohci0
  | | +-usb1
  | |   +-uhub1
  | |     +-ums0
  | |     | +-wsmouse0
  | |     +-umass0
  | |       +-scsibus0
  | |         +-uk0

The umass.c code didn't change, and my config file has:

sd*     at scsibus? target ? lun ?      # SCSI disk drives
sd*     at atapibus? drive ? flags 0x0000       # ATAPI disk drives

So, it should at least connect as sd0 instead of uk0.

The problem is that the compactflash reader is a UFI device (mostly) and
since UFI is pretty much just a subset of ATAPI it works really well when
attached as ATAPI but fairly poorly as SCSI without a lot of
translations.  The umass.c code was recently modified to make UFI devices
connect as ATAPI instead of SCSI for this very reason.  So, something in
the new scsipi code has broken the atapibus attachment.

++Brett;


On Tue, 1 May 2001, Manuel Bouyer wrote:

>On Tue, May 01, 2001 at 06:15:29PM +0900, Hirohito Higuchi wrote:
>> Hello,people!
>> My name is H.Higuchi.
>> I am using (probabily) NetBSD-i386-current 1.5U.
>> and I got the following problem.
>> ----
>> scsibus0 at umass0 channel0 :2 targets,1 luns per target
>> scsibus0 :waiting 2seconds for device 2 settle....
>> uk0 at scsibus0 target 1 lun 0:<IBM USB CDROM,20B4 >
>> scsi0 5/cdrom removable 
>> uk0 :unknown device
>
>Could you double check that you have 'cd* at scsibus? target ? lun ?'
>in your kernel config file ?
>This should attach as cd0 not uk0
>
>--
>Manuel Bouyer <bouyer@antioche.eu.org>
>--
>
>