On Dec 18,  4:47am, Andreas Wrede wrote:
} On Jul 21, 2007, at 12:40 , Andreas Wrede wrote:
} > On Jul 20, 2007, at 18:56 , Jonathan A. Kollasch wrote:
} >> On Fri, Jul 20, 2007 at 05:23:19PM -0400, Andreas Wrede wrote:
} >>>  I am trying to use gpt to create a GUID partition table on a 3TB  
} >>> Xserve
} >>>  RAID:
} >>>
} >>>  # gpt create /dev/sd0d
} >>>  gpt: unable to open device 'sd0d': Invalid argument
} >>>
} >>>  I assume this because in sys/dev/scsipi/sd.c does not set the  
} >>> properties.
} >>>  In -current, rev 1.260 of sd.c adds sd_set_properties().   
} >>> (4.0BETA's sd.c is
} >>>  at rev 1.258).
} >>>
} >>>  Is that all that's needed to get gpt to work on sd(4) type  
} >>> devices? If so,
} >>>  maybe that change can be pulled up to the 4.0 branch?
} >>
} >> Yes, a backport is trivial, I've actually done it.  I don't see  
} >> why it couldn't go into netbsd-4.

     I haven't been following this thread, but since I'm the author
of this change, I have requested a pullup.  It should appear in
netbsd-4 shortly.

} >> Also, you may want the appropriate gpt stuff for dk in the kernel:
} >>
} >> options DKWEDGE_AUTODISCOVER
} >> options DKWEDGE_METHOD_GPT
} >
} > Ok.  I guess, without it, I have to re-run the dkctl .. addwedge  
} > commands?
} >
} > BTW, pseudo-device   drvctl is only present in the ALL kernel and  
} > the DKWEDGE_... options are missing completely.  Should GENERIC not  
} > have drvctl? And the commented-out options?

     These aren't even in -current GENERICs.  I'll let someone else
decide if they should be added.

} > Also missing is the drvctl(4) man page, which is referenced from  
} > drvctl(8).

     drvctl(4) doesn't exist (yet).  I've requested a pullup of
drvctl(8) so that it will no longer referenece the non-existent
manpage.

}-- End of excerpt from Andreas Wrede