Re: kern/38643: [dM] st tape drive loses data

[Havard Eidnes <he%NetBSD.org@localhost>]

The following reply was made to PR kern/38643; it has been noted by GNATS.

From: Havard Eidnes <he%NetBSD.org@localhost>
To: gnats-bugs%NetBSD.org@localhost, mouse%Rodents.Montreal.QC.CA@localhost
Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
 netbsd-bugs%netbsd.org@localhost, yamt%NetBSD.org@localhost
Subject: Re: kern/38643: [dM] st tape drive loses data
Date: Mon, 01 Sep 2008 20:06:58 +0200 (CEST)

 Hi,
 
 prompted by discussion elsewhere, I had a brief glance at this
 bug.
 
 I don't have any suggestions for a fix, but I may have an
 observation (which I'm sure you've thought of yourself as well,
 but it's good to have it written down anyway).
 
 On nearly all NetBSD ports, the MAXPHYS constant is 64KB, and any
 physical I/O operation can at most transfer this much data in one
 operation.
 
 1048577 is an odd size to read, if you were expecting an integral
 number of 10KB tape blocks -- it's 16 * 64KB + 1 (why the +1?).
 
 I suspect your first big read operation is being broken up into
 separate MAXPHYS-sized requests, instead of just a single block
 being read.  I'm not sufficiently familiar with the code to
 suggest where that split is being done, though, although I think
 I would take a long hard look at the changes to kern_physio.c
 between the two releases, and perhaps have a chat with yamt@
 (CC'ed here), as he has been in there making some significant
 changes between netbsd-3-0-RELEASE and netbsd-4-0-RELEASE.
 
 (...which means that even though I've taken a cursory look, I
 wasn't able to pinpoint where the problem is...)
 
 Regards,
 
 - H=E5vard