VIA 686b Bug - once again :(

VIA 686b Bug - once again :(

[Volker Dierks]

Hello folks,

my first mail to this famous list .. ;)

OK, 2 month ago, I bought a Abit KG7 
MOBO (VIA 686b southbridge) and
yesterday I ended up with the second
data corruption within these two month.

This one isn't so bad as the first one ..
I can't find any 400GB files with the
rescue system and also no partition has
gone. But on normal system boot, there
are errors with rc.S and when kdm
normally should start, the system stops
to do anything. CTRL+ALT+DEL is
configured to reboot the system - when
pressed it is going into runlevel 6,
but hangs around after that.

I set _no_ special settings with hdparm.
hdparm -c -d /dev/hda gives
DMA = 1
32 bit = 1

Some words about my system:
Debian GNU/Linux (sid)
Kernel 2.4.10
Abit KG7 (lite)
Athlon 1.4GHz (step c)
Soundblaster Live (value)
1 IBM DTLA 20 GB
Dawicontrol DC-2974 PCI - AM53C974
(CDROM and CDR are SCSI)

I searched 1 hour on google but
the only solution seems to be a
fix with loadlin and pciset from
Thilo <Free.Spirit@gmx.net>

So my questions is:
I'm going to buy a 3ware 6410(B)
IDE raid controller .. can I suspect
a failure safe system (in aspect to
the 686b problems) when all discs
are connected to the 3ware
controller?

Greetings from Germany
volker
-- 
GNU/Linux
Let loose the Daemon inside...

Re: VIA 686b Bug - once again :(

[Kelledin Tane]

> So my questions is:
> I'm going to buy a 3ware 6410(B)
> IDE raid controller .. can I suspect
> a failure safe system (in aspect to
> the 686b problems) when all discs
> are connected to the 3ware
> controller?

I would certainly expect so.  I have a Gigabyte GA-7DX with the infamous VIA
686B southbridge...and an SBLive! Value...and I have no IDE devices at all
(complete SCSI).  I have never encountered data corruption.

Just out of curiosity...is this IBM drive a 75GXP model?  What does IBM's
Drive Fitness Test tell you about it?  I ask because the 75GXP is IMHO a
flawed product line.

Kelledin
-----------------
[ Kelledin@Valhalla ~ ] # kill -9 1
init: Just what do you think you're doing, Dave?

Re: VIA 686b Bug - once again :(

[Vojtech Pavlik]

On Sun, Oct 21, 2001 at 09:34:35AM -0500, Kelledin Tane wrote:

> > So my questions is:
> > I'm going to buy a 3ware 6410(B)
> > IDE raid controller .. can I suspect
> > a failure safe system (in aspect to
> > the 686b problems) when all discs
> > are connected to the 3ware
> > controller?
> 
> I would certainly expect so.  I have a Gigabyte GA-7DX with the infamous VIA
> 686B southbridge...and an SBLive! Value...and I have no IDE devices at all
> (complete SCSI).  I have never encountered data corruption.

Corruption *has* been reported on some scsi-only 686b systems, though. 

> Just out of curiosity...is this IBM drive a 75GXP model?  What does IBM's
> Drive Fitness Test tell you about it?  I ask because the 75GXP is IMHO a
> flawed product line.

Well, I have a 30GB drive from the 75GXP family. It works just fine so far.

-- 
Vojtech Pavlik
SuSE Labs

Re: VIA 686b Bug - once again :(

[Krzysztof Halasa]

Vojtech Pavlik <vojtech@suse.cz> writes:

> > I would certainly expect so.  I have a Gigabyte GA-7DX with the infamous
> > VIA
> > 686B southbridge...and an SBLive! Value...and I have no IDE devices at all
> > (complete SCSI).  I have never encountered data corruption.
> 
> Corruption *has* been reported on some scsi-only 686b systems, though. 

Right, there were few reports, including mine. As it turned out, disabling
MPS 1.4 and probably setting something related to SDRAM (don't remember
exactly) fixed the problems.
-- 
Krzysztof Halasa
Network Administrator

Re: VIA 686b Bug - once again :(

[Rupa Schomaker]

Kelledin Tane <feralgod@home.com> writes:

>> So my questions is:
>> I'm going to buy a 3ware 6410(B)
>> IDE raid controller .. can I suspect
>> a failure safe system (in aspect to
>> the 686b problems) when all discs
>> are connected to the 3ware
>> controller?
> 
> I would certainly expect so.  I have a Gigabyte GA-7DX with the infamous VIA
> 686B southbridge...and an SBLive! Value...and I have no IDE devices at all
> (complete SCSI).  I have never encountered data corruption.


I experienced constant and reproducable data corruption on two via
motherboards.  One using 686 and one 686b.  However, not with the IDe
drives.  This was with a DDS4 scsi tape drive attached to either a
Adaptec or Advansys scsi controller.  A 2.5G file dump to tape would
never restore the same.  One the Adaptec card, I would get 1 or 2
blocks of 64bytes that would differ.  On the Advansys it would be 1 or
2 blocks of 63 bytes.

Switch to a PIII-500 on a BX motherboard and have had no problems.
The VIA motherboards are now in windows machines where the flakiness
of the OS is even worse than the hardware.

-- 
-rupa

Re: VIA 686b Bug - once again :(

[Lorenzo Marcantonio]

On Mon, 22 Oct 2001, Rupa Schomaker wrote:

> Adaptec or Advansys scsi controller.  A 2.5G file dump to tape would
> never restore the same.  One the Adaptec card, I would get 1 or 2
> blocks of 64bytes that would differ.  On the Advansys it would be 1 or
> 2 blocks of 63 bytes.

Some month ago (about at 2.4.6 kernel) I've got the same problem with my
DDS3 (the TAPE CORRUPTION thread). 64 bytes more or less aligned at page
boundary. Tought it was the tape driver, it was an Adaptec driver issue
(back to when you needed Berkeley DB to compile the firmware)!
Now it works perfectly (and I've got an infamous Asus A7V...). BTW no HDD
on secondary IDE, only a CDROM :)

-- Lorenzo Marcantonio

Re: VIA 686b Bug - once again :(

[Stefan Smietanowski]

Hi.

> So my questions is:
> I'm going to buy a 3ware 6410(B)
> IDE raid controller .. can I suspect

You should be aware of that 3ware is dropping all their IDE RAID 
controllers so if you want it, buy it fast. They're going out of 
production. That is... if you want a product that won't be supported RSN.

// Stefan

Re: VIA 686b Bug - once again :(

[Federico Sevilla III]

On Sun, 21 Oct 2001 at 17:05, Stefan Smietanowski wrote:
> You should be aware of that 3ware is dropping all their IDE RAID
> controllers so if you want it, buy it fast. They're going out of
> production. That is... if you want a product that won't be supported
> RSN.

Eh? This is interesting. I haven't heard of anything like this before, and
can't find anything in their website. Maybe you can point me to where you
got this bit of information? Rather alarming because I've got a 3ware
controller myself.

Thanks.

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: <http://jijo.leathercollection.ph/jijo.gpg>

Re: VIA 686b Bug - once again :(

[Alan Cox]

> Eh? This is interesting. I haven't heard of anything like this before, and
> can't find anything in their website. Maybe you can point me to where you
> got this bit of information? Rather alarming because I've got a 3ware
> controller myself.

I heard this straight from a 3ware representative

Re: VIA 686b Bug - once again :(

[Stefan Smietanowski]

Hi.

>>You should be aware of that 3ware is dropping all their IDE RAID
>>controllers so if you want it, buy it fast. They're going out of
>>production. That is... if you want a product that won't be supported
>>RSN.
>>
> 
> Eh? This is interesting. I haven't heard of anything like this before, and
> can't find anything in their website. Maybe you can point me to where you
> got this bit of information? Rather alarming because I've got a 3ware
> controller myself.

The Linux XFS development list had a discussion about it a few weeks 
ago. I even think someone from 3ware joined in or if someone had a quote 
from someone at 3ware. But just go there and you'll see the site has 
changed. The only graphic on the front page is now their IP storage 
thing. The Palisade.

Go to Products on their page. Notice the lack of information on the 
Escalades.

http://www.3ware.com, but I think you got that already.

Only info on Escalades is:

" Inquires regarding Escalade Storage Switch products should be directed 
to 650.327.8600 or info@3ware.com."

and a picture...

// Stefan

Re: VIA 686b Bug - once again :(

[Federico Sevilla III]

On Sun, 21 Oct 2001 at 13:26, Volker Dierks wrote:
> OK, 2 month ago, I bought a Abit KG7 MOBO (VIA 686b southbridge) and
> yesterday I ended up with the second data corruption within these two
> month.

Just so you know I've got an Asus motherboard also with the VIA 686b
Southbridge. I'm using 2.4.12-xfs and it's been doing pretty well. I don't
use the IDE controller on the drive that often (althoug the kernel is
built with support in particlar for the VIA chipset) as it's only for the
CD-ROM drive. I have a 3ware 6400 and four IBM hard drives.

First of all the IBM-DTLA-307030 drives are bad. Really bad. I've had one
hard drive (of four in hardware RAID5) go down a number of times already.
Fortunately the IBM DFT (Drive Fitness Tool) has been able to "repair" the
drive and so I'm still using it now.

I also ran into XFS corruption before, but this was because of a firmware
bug that got triggered when a drive went down during filesystem activity.
I flashed the firmware on the 3ware controller and the problem went away.
It had nothing to do with XFS or the kernel (although every 3ware firmware
upgrade comes with a driver upgrade and hence a kernel update).

> I'm going to buy a 3ware 6410(B) IDE raid controller .. can I suspect
> a failure safe system (in aspect to the 686b problems) when all discs
> are connected to the 3ware controller?

I recommend you get the 7xxx series instead. There are a number of
improvements there and you might overall be much happier with that.
Especially if you decide to go RAID5, as the 6xxx controller and buffer
are too small to do decent RAID5 (so the performance of my system sucks).

RAID is also not "failure safe". You still need regular backups. But it
helps because it allows you to recover from small hardware failure that
would otherwise have been significant. Like one hard drive going down.
Normally I have enough time to pull out the drive (I have a hot swap bay,
made in Taiwan or China, and the 3ware controller handles IDE hot swap
nicely), plug it in a workstation, reboot with the IBM DFT disk, and fix
it. Then I plug the drive back in and rebuild the array on the fly. This
whole process takes a few hours, but is hopefully invisible to most users
(although performance drops significantly because of the drive activity).

It's hard to get replacements from IBM in my country, though. So if the
DFT can't fix a drive some day, I'll just have to replace it. And with
RAID5, if any two drives go down at the same time (or one after the other
as long as you have more than one drive down), you're still dead. :(

Good luck.

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: <http://jijo.leathercollection.ph/jijo.gpg>

Re: VIA 686b Bug - once again :(

[clemens]

On Sun, Oct 21, 2001 at 01:26:11PM -0000, Volker Dierks wrote:

> I searched 1 hour on google but
> the only solution seems to be a
> fix with loadlin and pciset from
> Thilo <Free.Spirit@gmx.net>

have a look at:

http://www.viahardware.com/686bfaq.shtm
http://www.au-ja.org/review-kt133a-1.phtml (german)

basically you have to find a way to set some specific PCI registers of the
northbridge. 
check out the write handler of this is /proc/bus/pci/00/00.0
it's at linux/drivers/pci/proc.c:proc_bus_pci_write

basically it looks like you have to seek to the "register" via lseek and
then write the control value.
shouldn't be that hard.

in worst case you could write a 20-line linux kernel module to set the
few required pci registers.

clemens

Re: VIA 686b Bug - once again :(

[Federico Sevilla III]

On Sun, 21 Oct 2001 at 19:11, Alan Cox wrote:
> > Eh? This is interesting. I haven't heard of anything like this before, and
> > can't find anything in their website. Maybe you can point me to where you
> > got this bit of information? Rather alarming because I've got a 3ware
> > controller myself.
>
> I heard this straight from a 3ware representative

Ack! :(

Thanks for validating Stefan's news.

I presume (hope) support via the open-source driver already with the Linux
community will continue as usual, even in the worst case scenario where
3ware decides to drop support completely?

:)

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: <http://jijo.leathercollection.ph/jijo.gpg>