Path: utzoo!attcan!uunet!cbmvax!jesup
From: jesup@cbmvax.UUCP (Randell Jesup)
Newsgroups: comp.sys.amiga.tech
Subject: Re: FFS: The beating goes on...
Keywords: ffs fragile diskdoctor
Message-ID: <5738@cbmvax.UUCP>
Date: 19 Jan 89 21:53:55 GMT
References: <695@ibmpa.UUCP>
Reply-To: jesup@cbmvax.UUCP (Randell Jesup)
Organization: Commodore Technology, West Chester, PA
Lines: 104

In article <695@ibmpa.UUCP> saare@ibmpa.UUCP (John Saare) writes:
>    A500
>    Pac. Periph. SubSystem 500
>    Pac. Periph. Overdrive host adapter
>    Micron 2MB 
>    Adaptec 4000A SCSI to ST-506 xlator.
>    2x70MB disks (ST-506)(Only 1 is relevant).

>    - Using PP(Pac. Periph.) utilities of unknown vintage, but that are
>      at least 1.3 cognizant, I perform a low-level format of a drive
>      with an interleave of 2.  NOTE: There is no defect list available
>      for the drive.  The format completes with no errors reported.

	The Adaptec should find and map out any errors, though its scan is
less thorough than a manufacturers scan.  Why didn't the disks have bad
sector listing?  All st506 drives should have them!

>    - A mountlist entry is built for the drive.  I've heard that Interleave
>      should be set to 0 for a 1:1, 1 for a 2:1, etc....  I've tried
>      setting it to both n and n-1.  Results appear identical.  I've
>      tried deleting MaxTransfer and setting it to arbitrarily large
>      numbers e.g., 10000.  Because I'm using (was using) the Micron board, I've
>      set Mask to 7ffff to prevent DMA to fast RAM.  I've tried
>      removing the Micron board altogether (see above).

	Use 0 for interleave in the mountlist.  Maxtransfer should be set to
whatever PP suggests, since if it's too large your disk may be corrupted.
Maxtransfer is there largely because people wrote drivers that broke when you
gave them requests for more than 512 bytes at a time (note that maxtransfer is
in bytes, not sectors - the 1.3 docs are wrong).

>    - The file-system format completes with NO errors, i.e., "verifying"
>      never complains.  It is known that there are errors on this drive, but
>      not where.  It's annoying that NONE of these errors have been caught
>      up to this point.

	The adaptec should have caught them.  It did, as the verify produced no
errors.

>    - The system runs but disk xfer rates are very slow e.g. 27K/sec ala
>      diskperf.  Nothing I do improves disk performance.

	Try changing you interleave to 3.  Very low xfer rates are the sign 
of too small an interleave.  Small maxtransfers can affect it also, but don't
use values higher than the controller manufacturer recommends.

>    - One day, while compiling (Manx 3.6a for the detail freaks) I get
>      a R/W error on volume "SlowBirds".  I select cancel and then
>      quarantine the whole directory the error occurred in, using "protect".

	No guarantees it is actually allocated to that directory.

>    - I cross my fingers and continue to use the system over a period of
>      several days.  The system is rebooted before each day's use.  Everything
>      seems fine.

	No revalidation was done.

>    - Upon reboot, the disk-validator is run and WILL NOT GET PAST the 
>      allocated bad blocks, so I'm left with an invalid, read-only disk.

	No suprise.  It can't figure out what blocks are allocated and
what aren't.

>    - I wait until the Micron board returns (kudos to Micron for fast and
>      efficient service), re-install it and run DiskDoctor.  After ALL
>      this hoopla, the disk-validator is STILL unable to validate the
>      disk, running into the bad blocks and failing to procede.

	Have you tried Dave Haynie's latest DiskSalv?

>I've become disgusted and re"format"ted the whole mess.  I'm sure the problem
>will repeat itself.  So some questions (some are semi-rhetorical):

	If your drive is prone to errors, it probably will.

>    - It appears as though if a file-system is EVER formatted with ANY
>      bad blocks, you will be scr_w_d(I'll buy a vowel Vanna). Is this true?

	Amiga file handlers assume a perfect disk.  The _driver_ is supposed
to map out bad blocks, including those that occur during use.  Since the
Adaptec doesn't support the SCSI command REALLOCATE_BLOCK, the driver would
have to do the mapping itself.  It obviously doesn't.  This is a common
failing in disk devices on the amiga.

>    - The verification step of "format" appears to be mostly a waste
>      of cycles, doesn't it?

	It works fine.  You developed bad blocks later.

>      be of concern.  Effectively, permanently invalidating an entire
>      70MB filesystem because of a few bad sectors just isn't cricket!

	Personally, I use 10-15meg partitions.  Greatly reduces the chance of
any one problem hurting me, and makes backups easier (since several of the
partitions almost never change.)  If an FS doesn't validate, it's still
readable (you DO keep backups, don't you?)

	As for speed, try different interleaves and see what produces the best
speed.  Call PP, make sure your software is up to date, and get the right
value for maxtransfer.

-- 
Randell Jesup, Commodore Engineering {uunet|rutgers|allegra}!cbmvax!jesup