On Mon, 4 Dec 1995 14:36:46 -0800 
 gdonl@gv.ssi1.com (Don Lewis) wrote:

 > It appears to have been a bum disk, but why didn't the kernel report
 > any SCSI errors?  The only manifestations were error returns in userland
 > and silent data corruption on the disk that eventually resulted in
 > panics.  If the disk was just corrupting data on it's own while reporting
 > that it was happy, then open() shouldn't fail with EIO.  When I ran
 > dump, it reported some unreadable blocks (the block numbers looked OK),
 > but the kernel didn't complain, and I could run dd on on the raw disk
 > without error.

It is fairly well known that the hp300 SCSI code could stand some 
improvement.  Also, there are times where a device may not report and 
error and/or the controller cannot detect it.

I'm slowing getting the hp300 port in a state where we can switch to new 
config.  Once that's done, I or some very kind volunteer (hint, hint) 
will redo the hp300 scsi controller driver so that it can work with the 
MI scsi code.  The MI scsi code has undergone active development and 
improvement for some time, and is in a much better state in terms of the 
configuration model and error handling.

"We're getting there."

--------------------------------------------------------------------------
Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                               Home: 408.866.1912
NAS: M/S 258-6                                          Work: 415.604.0935
Moffett Field, CA 94035                                Pager: 415.428.6939