Xref: utzoo comp.sys.amiga:75754 comp.sys.amiga.tech:17429
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!cbmvax!jesup
From: jesup@cbmvax.commodore.com (Randell Jesup)
Newsgroups: comp.sys.amiga,comp.sys.amiga.tech
Subject: Re: HD Errors
Message-ID: <17129@cbmvax.commodore.com>
Date: 5 Jan 91 03:35:50 GMT
References: <139834@pyramid.pyramid.com> <37568@cup.portal.com>
Reply-To: jesup@cbmvax.commodore.com (Randell Jesup)
Organization: Commodore, West Chester, PA
Lines: 44

In article <37568@cup.portal.com> thad@cup.portal.com (Thad P Floryan) writes:
>Consider what happens if a block presently associated with a file gets "mapped
>out".  Bingo, inconsistent file.  The SCSI OS can be instructed (using tools
>such as the ones I mentioned can be found at "adaptex") to automatically map
>out bad blocks upon writing, and the OS (i.e. filesystem) would never even know
>about it, since the SCSI OS still presents "sequential"(ly numbered blocks) to
>the device driver.

	Well, some SCSI drives can.  Last I checked, it was a vendor-specific
command (which quantum has), though it may have found it's way into SCSI-2.

>  In my experience, bad blocks are "found" when doing a read
>after write operation; the SCSI devices can detect this "automagically" if
>instructed to do so and thus the user won't see any problem.  Of course, if
>later a block is found "bad" during a read, then you better have a recent
>backup!  :-)

	Once again, I think this is a vendor-specific command.  SCSI can also
tell you on a read that the data was bad but the ECC recovered it, and quantums
will remap the block (without losing data).  If you use HDToolBox to Verify
a drive ("Verify data on drive", SCSI Verify command), it will notice
recovered read errors (and full errors, of course), and let you map them out.

	The Amiga standard is that bad-block mapping be handled at the device
level or lower (which happens to work quite well with how SCSI is set up), and
that the FS is presented with an array of fully usable blocks, which of course
is why bad-block mapping isn't in the FS (it would be silly to have it there
AND in the SCSI device itself).

	The scsidirect standard makes it easy to write programs to send
arbitrary SCSI commands, regardless of the controller you own (if they follow
the standard, of course).  Therefore, it's easy to write utilities that set up
things like the quantum auto-remapping feature (or do just about anything
else).  I know at least two people have programs to do all sorts of scsidirect
tricks, but they don't seem to have released them (or I missed them, quite
easy given how busy I am).  It _really_ is quite simple to use scsidirect if
you know exec I/O and have some sort of drive/scsi manual.

-- 
Randell Jesup, Keeper of AmigaDos, Commodore Engineering.
{uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.commodore.com  BIX: rjesup  
The compiler runs
Like a swift-flowing river
I wait in silence.  (From "The Zen of Programming")  ;-)