Re: UDMA is NFG on Ultra33 (CMD646 in disguise)

jason.harp@mail.ray.ca
Wed, 25 Nov 98 14:46:38 -0800

Mark,


When you say you get corruption what exactly do you mean. Is it that
the filesystem become unusable quickly or rather is it subtle problems
with the filesystem (and data) as time goes on?

I've been using the Ultra33 cards with ultra2 for some time and have
seen no problems. How can I check for the corruption you speak of?
Fsck does not show any problem after several months of operation. The
occasional ultra dma ECC correction does take place under heavy load
but this is likely due to my extended length of IDE cable. (as long
as 24" in length. 6" longer then they should be) My system is as
follows:

Supermicro P6DLH (9 pci / 3 ISA slots)
2.0.31 with patches for Ultra33
3 DEC FDDI cards
2 Promise Ultra33 cards 1.25 BIOS
1 AIC78xx SCSI
8 Quantum 12GB Bigfoot TX drives
1 Seagate 3.2 Boot disk

Linux is configured with NO modules and I use RAID5 across the two
promise controllers.

As for the speed problems I simple copy the bytes from the primary
promise PCI area to the secondary area ( This suggestion was
originally from Gadi some 7 months ago)


I had no end of problems with Promise Ultra33 cards with BIOS 1.2[0-4]

Upgrades for the BIOS are FREE if you report a problem to Promise that
is BIOS related (i.e. two promise cards before 1.25 could not share
the same IRQ)


Jason Harp





______________________________ Reply Separator _________________________________
Subject: UDMA is NFG on Ultra33 (CMD646 in disguise)
Author: Mark Lord <mlord@pobox.com> at IAMS-MIME
Date: 24/11/98 2:14 AM


Hi Guys.

I dunno what to make of it, other than
"UDMA is broken on the Promise Ultra33" card.

I have had one of these cards for over a year now,
but only ever used it for secondary drives.

I recently moved my primary filesystem over to
a drive on the Ultra33 card, and Boom!

Data corruption.

I have experimented extensively with it now,
verifying correct initialization of the chipset,
and playing with the PCI timing bits in the "System Control"
register -- no lucko.

The corruption seems to happen only when there is contention
between the Ultra33 and another PCI bus-master.

The card still seems to work flawlessly in PIO (PCI target) mode,
so that's exactly how I'll be using it from now own.
Much slower though!

I suppose it is no wonder that the card is buggy,
given that the PDC20246 chip is actually a re-labelled CMD646 chip
(pin 2 at power-on selects between CMD646 and PDC20246 modes).

Everybody remember the CMD640..? Same designers worked on this one. 
-- 
mlord@pobox.com
     
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in 
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
     






-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/