> On Sat, 2 Dec 2000, Lars Munch Christensen wrote:
> > On Thu, Nov 30, 2000 at 08:01:37AM -0800, Ahmon Dancy wrote:
> > > All I have to do to hang the system (on 2.4.0-test11 and 2.2.17) is:
(by
> the
> > > way, I also made a uniprocessor kernel and this still would hang)
> > >
> > > dd if=/dev/hdg of=/dev/null
> > >
> > > Just before it does, the kernel prints either:
> > > hdg: timeout waiting for DMA
> > > or
> > > ida_dmaproc: chipset_supported ide_dma_timeout func only: 14
> >
> > I have also had this problem. After trying almost everything I removed
> > my HD's from the HPT366 and I have been running stable ever since. I not
> > a good solution but it works.
>
> Okay. This seems to be the consensus. I've removed my drives from the
HPT366
> controller and things seem to be working fine now. It's unfortunate,
though.
> I was getting at nice 27MB/sec transfer rate on the HPT.. now I'm down to
20.
I've tried the dd-ing for several hours with various combinations of IDE
options, set by hdparm(8).
No joy! I cannot reproduce the problem :) :) :) , though I used to
suffer from lockups when I first built the machine, and I too like Tuomas,
cut back on my plans when I hit them.
It's worth using the HPT if you can keep a stable system, although past list
members have disabled it, and used Promise IDE cards instead, which gave
less trouble. There's been a whole long trail of "solutions", all of which
probably just mask something borderline about the BP6. They include, fans
and thermal paste on the BX chip, meaty power supplies, Bios updates
(including fooling with microcode), suspicions about Matrox cards, and
obviously the Rev 1.1 board Capacitance fiasco did not enhance Abit's
reputation one bit. Somebody attempted to move to the MSI board, which
turned out to be worse, so finally the 2nd CPU had to be dropped to get
stability.
I'm running :
Debian (potato), with 2.2.17 + IDE patch by Andre Hedrik
qq_01beta BIOS
2x450Mhz @100MHz FSB, 300A Celerons
128MB PC100 RAM, one DIMM
1 I|BM 20GB ATA-66 HD on the HPT366 as hde, with UDMA4 mode enabled.
1 IBM 20GB ATA-66 HD as hda boot disk UDMA2 mode.
Nvidia TNT2 AGP card
Sparkle 300W p/s with a 1.0 revision board.
This list has used vast amounts of bandwidth discussing BP6 lockups. I'm
surprised I don't have it at the moment, was planning to use a second disk
on the HPT366, and have ordered new RAM, which have both been fingered as
possible causes in the past.
Theres some discussion of it in the FAQ, There's a FAQ for the BP6
http://www.snurgle.org/~griffon/bp6-linux.html and
a mini-HOWTO for the HPT366 ATA-66 controller
http://www.csie.ntu.edu.tw/~b6506063/hpt366/ .
Basically, I'm going to try and get the lock ups, shifting disks over, then
the RAM, and see at what point (if?) my BP6 becomes unstable. It's been
disappointingly solid, I'm at the point I'd like an excuse to go and buy
myself an upgrade, though at work I'm looking at kernel oops, with an Asus
AV7 + Tbird combo, some things never change ....
Think, if I can't run 2 disks off the HPT stably, I'll just buy a Promise
card, this post was reassuring :
> > I have an Abit BP6 (two 433 Celerons, not overclocked!) with four IBM
> > 37.5 GB EIDE hard drives connected to the HPT366 UDMA66 controllers (hde
> > thru hdh). I am running RH61 and 2.2.14 kernel with the most recent IDE
> > patch. The controller and all four hard drives are detected and I can
> > cfdisk them (fdisk only sees 4 GBs...go figure). I can mkfs and mount
> > them without a problem, but under heavy usage, I occasionally get "hdX:
> > Timeout waiting for DMA." This will either hard lock the machine (with
> > HD activity light on) or the kernel will kick the controller back to PIO
> > 4. I have a similar setup using a Promise Ultra66 card that works
> > without a problem in UDMA66.
> Are you on the Abit Linux mailing list?
> The guy who wrote the driver for the HPT366 made a couple of
> interesting posts on it today.
>
> Basically he said there were some possibilities for deadlock at present.
> I assume the IDE patch was for HPT366 support? Perhaps you'ld have
> more luck with running in UDMA33 mode, the HPT366 Mini-HOWTO
> explains this.
Thanks for the info. Yeah, the IDE patch is Hedrick's which I pulled
from one of the kernel mirrors. It handles the HPT and others. I will
take a look at what he had to say. Honestly, I just gave up on the
HPT366. I installed a Promise Ultra66 and everything "just worked."
The $40 for the Promise was not worth the trouble or possible
instability. That RAID system has been in service since Monday...110 GB
RAID5 server for under $2k...it was a big hit.
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
