Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

[Dieter Nützel]

On Friday, 24. March 2002 22:18:21 David wrote:
> Hi,
>
> First off, I'd like to thank Andre Hedrick for the link to:
> http://www.tecchannel.de/hardware/817/index.html
>
> It was very helpful in making me understand why the performance of my 
> newly bought Adaptec 2400A IDE-Raid card sucked so badly in combination 
> with my VIA KT133 based board.
>
> I have done some testing (on the Win platform since that's where the 
> patches to remedy this situation are available) that shows quite nicely 
> how the PCI bus gets totally overrun by transfer rates in excess of 
> approximately 74MB/s and instead slows down.
>
> Short summary
> =============
> Expected rate           Experienced rate
> 84MB/s 
>                 50MB/s
> 80MB/s 
>                 64MB/s
> 74MB/s 
>                 74MB/s
> (If someone is very interested, mail me and I can publish some transfer 
> rate graphs online)

Can you please redo with latest VIA fix applied?

PrePatch:   http://www.tecchannel.de/hardware/817/5.html
Via Patch:  http://www.viaarena.com/?PageID=66#raid
Post Patch: http://www.tecchannel.de/hardware/817/11.html

rpp1.02.zip

Thanks,
	Dieter

-- 
Dieter Nützel
Graduate Student, Computer Science

University of Hamburg
Department of Computer Science
@home: Dieter.Nuetzel@hamburg.de

Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

[David Härdeman]

Hi all,

I got some mails regarding my first mail and I'll try to respond to the 
feedback...

Mark Hahn wrote:
 >>newly bought Adaptec 2400A IDE-Raid card sucked so badly
 >>in combination with my VIA KT133 based board.
 >
 >
 > that's a bizarre combination: quite old/cheap motherboard with a
 > fairly expensive HW raid card.
 >

Yes, but that was the box that was available for testing currently. 
Thats life :-)

 > of course, you could do everyone a big favor by finding a tool that
 > would dump the chipset's PCI config space under windows, to see what
 > the patch changes.  under Linux, of course, lspci does this.

Yes, I could do that (using WPCREDIT), but I'd like to know that it 
would be helpful to some kernel hackers first, so mail me and I'll do it.

If someone *is* interested, I'll dump bus0,dev0,func0 (Host Bridge) and 
bus0,dev1,func0 (PCI-PCI bridge) with and without patches installed, is 
there any other device that should be included (like IDE-controller)?



Dieter Nützel wrote:
 > Can you please redo with latest VIA fix applied?

I did try with the latest VIA patch applied (it gave approx 4Mb/s in 
increase). However, some more research has uncovered yet more hardware 
issues that may (in combination with the VIA issues) be the "real" 
perpetrator.

Apparently (by looking at the links at the end of this mail), SeaGate 
Barracuda IV's and RAID is a big mistake. I will RMA the drives and buy 
some other brand before I continue benchmarking the 2400A card.

Thanks to Mark, Dieter and Ville.

Regards,
David

Barracuda IV problem links
==========================
http://forums.storagereview.com/viewtopic.php?t=2154
http://forums.storagereview.net/viewtopic.php?t=430
http://forums.storagereview.net/viewtopic.php?t=85
http://www.viahardware.com/faq/kt7/faqhpt370.htm#barracuda
http://www.baluma.com/hardware/almacenamiento/barracudaIV/home.asp
http://www.msi.com.tw/forum/index.php?target=article&tid=6899&page=2
http://www.geocities.com/teomorell/st380021a/winbench.htm
http://forums.storagereview.net/viewtopic.php?t=1333

Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

[David Härdeman]

Mark Cooke wrote:

> If I run a simultaneous one on hdc and hde, the combined rate tops 
> out at 50MB again.  Hence, the limitation isn't the raid card.  Or at 
> least I'd be exceedingly surprised.
> 

The bugs that exist in VIA chipsets and Barracuda drives have already 
exceedingly surprised me many, many times :-)

I have done some dumping on the via chipset and this what I've come up 
with is available (in a cooked format) at:
http://www.student.nada.kth.se/~i99_hnd/via/

I hope that this is useful in some way for you?.

Regards,
David

Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

[Mark Cooke]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 2442 bytes --]

On Mon, 1 Apr 2002, David Härdeman wrote:

> Date: Mon, 01 Apr 2002 18:33:04 +0200
> From: David Härdeman <david@2gen.com>
> To: Mark Cooke <mpc@star.sr.bham.ac.uk>
> Cc: LinuxKernel <linux-kernel@vger.kernel.org>
> Subject: Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load
>     togeter
> 
> Mark Cooke wrote:
> 
> > If I run a simultaneous one on hdc and hde, the combined rate tops 
> > out at 50MB again.  Hence, the limitation isn't the raid card.  Or at 
> > least I'd be exceedingly surprised.
> 
> The bugs that exist in VIA chipsets and Barracuda drives have already 
> exceedingly surprised me many, many times :-)

Agreed.

> I have done some dumping on the via chipset and this what I've come up 
> with is available (in a cooked format) at:
> http://www.student.nada.kth.se/~i99_hnd/via/
> 
> I hope that this is useful in some way for you?.

Yes.  It's interesting that comparing the differences in north bridge
registers between windows unpatched, and the latest 1.05 raid patch is
just register 0xb7 changing.  Perhaps I'll be 'brave/foolish' and just
twiddle it with setpci after backing up.  Do you have documentation 
for these registers ?  I've been kicking through via's site looking 
for chipset datasheets to no avail.

I toasted one of my raid disks trying to get Win98se installed over
the weekend to do similar testing here.  Didn't like the size of the
disk I tried to install it to, and I ended up with a zapped partition
table and who knows what else.  Thankfully the software raid 5 coped
perfectly (thanks guys!) and the system was back after the rebuild
time.

I've a spare (small) disk to try this with my system today or
tomorrow.  Which ABit bios are you using on that test system currently
?  I put the newest ABios bios (7P vs 65)  on my system (KT7A-RAID
v1.3)  and it appears to have gained me about 3-4MB/sec over the PCI
bus. (54MB/sec vs 50MB/sec)

Attached files are my raw pci data for 00:0 and 00:1, for your 
reference.  It's different enough to your data that it's not easy
to compare though.

Cheers,

Mark

-- 
+-------------------------------------------------------------------------+
Mark Cooke                  The views expressed above are mine and are not
Systems Programmer          necessarily representative of university policy
University Of Birmingham    URL: http://www.sr.bham.ac.uk/~mpc/
+-------------------------------------------------------------------------+

[-- Attachment #2: Type: TEXT/plain, Size: 950 bytes --]

# lspci -xxx -s 00:0
00:00.0 Host bridge: VIA Technologies, Inc. VT8363/8365 [KT133/KM133] (rev 03)
00: 06 11 05 03 06 00 10 a2 03 00 00 06 00 08 00 00
10: 08 00 00 c0 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 7b 14 01 a4
30: 00 00 00 00 a0 00 00 00 00 00 00 00 00 00 00 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 17 a3 6b b4 47 09 20 20 88 00 08 10 18 20 20 20
60: 0f aa 02 20 51 51 51 01 45 0c 43 0f 08 5f 00 00
70: da 88 cc 0c 0e 80 d2 00 01 b4 19 02 00 00 00 00
80: 0f 44 00 00 00 00 00 00 03 00 0c 12 00 00 00 00
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 32 00 00
a0: 02 c0 20 00 03 02 00 1f 02 01 00 00 6f 12 00 00
b0: c9 ef 98 74 30 ee 80 07 67 00 00 00 00 00 00 00
c0: 01 00 02 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 03 04 11 22 00 00 00 00 80 91 06

[-- Attachment #3: Type: TEXT/plain, Size: 944 bytes --]

# lspci -xxx -s 00:1
00:01.0 PCI bridge: VIA Technologies, Inc. VT8363/8365 [KT133/KM133 AGP]
00: 06 11 05 83 07 00 30 a2 00 00 04 06 00 00 01 00
10: 00 00 00 00 00 00 00 00 00 01 01 00 f0 00 00 00
20: 00 d8 f0 d9 00 d0 f0 d7 00 00 00 00 00 00 00 00
30: 00 00 00 00 80 00 00 00 00 00 00 00 00 00 0c 00
40: cb cd 08 44 25 72 05 83 00 00 00 00 00 00 00 00
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
80: 01 00 02 02 00 00 00 00 00 00 00 00 00 00 00 00
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load togeter

[Henrique de Moraes Holschuh]

On Mon, 01 Apr 2002, Mark Cooke wrote:
> On Mon, 1 Apr 2002, David Härdeman wrote:
> > Date: Mon, 01 Apr 2002 18:33:04 +0200
> > From: David Härdeman <david@2gen.com>
> > To: Mark Cooke <mpc@star.sr.bham.ac.uk>
> > Cc: LinuxKernel <linux-kernel@vger.kernel.org>
> > Subject: Re: Kernel 2.4.17 with VT8367 [KT266] crashes on heavy ide load
> >     togeter
> > 
> > Mark Cooke wrote:
> > > If I run a simultaneous one on hdc and hde, the combined rate tops 
> > > out at 50MB again.  Hence, the limitation isn't the raid card.  Or at 
> > > least I'd be exceedingly surprised.
> > 
> > The bugs that exist in VIA chipsets and Barracuda drives have already 
> > exceedingly surprised me many, many times :-)

Yes. For the VIA side, I have this in my rc.S stuff:

#!/bin/sh
PATH=/sbin:/bin:/usr/sbin:/usr/bin
export PATH

echo -n "Optimizing hardware configuration: "

if command -v setpci >/dev/null 2>&1; then
    echo -n "PCI"
    #
    # Optimize PCI latency for IDE controllers
    #
    setpci -d 1106:0571 latency_timer=60  >/dev/null 2>&1
    setpci -d 105a:*	latency_timer=60  >/dev/null 2>&1
    echo "."
else
    echo '(lspci/setpci not available!)'
fi

Normal latency set by the BIOS is 32, which is too damn small for IDE.

You may have to tweak the PCI ids a bit. You want all storage controllers
(both chipset, Promise and any extra cards). The IDs up there are for my VIA
kt133, and Promise PDC20265.

Oh, some of the PCI 'optimizations' in BIOS must be enabled for that to
actually help a bit. Stuff like the PCI caches.

VIA PCI is utter crap. I am not buying anything of theirs ever again :(

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh