Weird cold boot problems with Abit KT7 motherboard

Weird cold boot problems with Abit KT7 motherboard

[Pavel Roskin]

Hello!

I have noticed several anomalies with Abit KT7 motherboard.  They all
happen after power on.  First reboot from Linux (using the reboot command
or reset button) usually fixes all the problems.  Sometimes two or three
resets are needed before the motherboard starts working properly.  In two
cases (of about 20) the motherboard started working properly right after
powering up.

I tried removing some cards and disabling the on-board USB.  Since cold
boot is needed to reproduce the problems, there are many things I could
have tried but didn't because I don't want to ruin other hardware by
constant power cycling.

I just hope the description below will rings a bell with somebody, and
I'll gladly test the suggestions.  If not, the motherboard will go to the
dumpster.

Anomalous state will refer to the state after the cold boot.  Normal state
is what I have after sufficient number of resets.  Following effects are
observed:

1) Variable PCI ID of ATI Rage graphics card.

In the anomalous state, the PCI ID of the card is 1002:5247 (ATI
Technologies Inc Rage 128 RG).  In the normal state, it's 1002:5246 (ATI
Technologies Inc Rage 128 RF/SG AGP).  It's also 1002:5246 on another
machine, both after cold boot and reset.  No other bytes in the PCI
configuration space of the card are different.

Another AGP card, ELSA Gloria, doesn't exhibit this anomaly on the same
motherboard.  Its PCI configuration space is unchanged.

2) "Bad slots" for 3com 3c900.

In the anomalous state, the 3com 3c900 doesn't work in certain slots, but
does work in others.  The card can send the data, but doesn't receive
anything.  The interrupt count stops at 94 and doesn't increase afterwards
even if I ping in both directions to and from another machine.

3) Hang on reading PCI configuration space of TI PCI1410.

In the anomalous state, the system hangs if TI PCI1410 (CardBus bridge) is
present.  The hang happens on this line in drivers/pci/probe.c:

pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);

PCI_CB_SUBSYSTEM_VENDOR_ID is 0x40.  Reading the configuration space below
0x40 works fine.  If I comment out this and the next lines, I can boot the
system to the command line.  XFree86 hangs.  lspci also hangs.  I can read
individual files from /proc/bus/pci, but not /proc/bus/pci/00/0f.0, which
corresponds to the CardBus bridge.  Reading that file beyond byte 64
causes the system to hang.

4) Host bridge detects parity error.

In the anomalous state, the PERR (parity error) flag is not set for the
Host bridge 00:00.0.  It is set in the normal state.


I put some data here: http://www.red-bean.com/~proski/kt7/

Most interesting should be lspci-vvvxxx.diff, which is the difference
between "lspci -vvvxxx" outputs for anomalous and normal modes.
lspci-nvvvxxx.diff is the same for "lspci -nvvvxxx".  Kernel log in the
normal mode is in dmesg.  dmesg.bad shows hang while probing TI PCI1410 on
cold boot.  interrupts, iomem and ioports are from /proc/, taken in the
anomalous mode after TI PCI1410 was removed.

I'm using Linux 2.6.6, but I've seen variable PCI ID of ATI Rage for years
with many kernels.  I just never had a chance to look deeper.

I know that the KT7 motherboard is not designed to run Athlon XP 2000+
(which is underclocked to 1.25 GHz), but the same problems existed with
Athlon 1.2 GHz on the same motherboard.

The BIOS is the latest, 07/11/2002-8363-686A-6A6LMA19C-A9.

-- 
Regards,
Pavel Roskin

Re: Weird cold boot problems with Abit KT7 motherboard

[Chuck Wolber]

On Wed, 12 May 2004, Pavel Roskin wrote:

> I have noticed several anomalies with Abit KT7 motherboard.  They all
> happen after power on.  First reboot from Linux (using the reboot
> command or reset button) usually fixes all the problems.  Sometimes two
> or three resets are needed before the motherboard starts working
> properly.  In two cases (of about 20) the motherboard started working
> properly right after powering up.

%< SNIP

> I just hope the description below will rings a bell with somebody, and
> I'll gladly test the suggestions.  If not, the motherboard will go to the
> dumpster.


We just recently saw a similar report with the same motherboard. The
system apparently boots fine under FC2. This is only second-hand
information though, I have not personally tested this.

-Chuck


-- 
http://www.quantumlinux.com 
 Quantum Linux Laboratories, LLC.
 ACCELERATING Business with Open Technology

 "The measure of the restoration lies in the extent to which we apply 
  social values more noble than mere monetary profit." - FDR

Re: Weird cold boot problems with Abit KT7 motherboard

[Grzegorz Kulewski]

On Wed, 12 May 2004, Pavel Roskin wrote:

> Hello!
> 
> I have noticed several anomalies with Abit KT7 motherboard.  They all
> happen after power on.  First reboot from Linux (using the reboot command
> or reset button) usually fixes all the problems.  Sometimes two or three
> resets are needed before the motherboard starts working properly.  In two
> cases (of about 20) the motherboard started working properly right after
> powering up.
> 
> [...]

Can you test your machine in both bad and good state with memtest86 for 
few hours? (It is included for example in latest Gentoo live cd). What 
about other cpu and memory intensive benchmarks? (Can you do 3 or 5 
[depending on memory amount] simulateus full botstraps of gcc 3 without 
problems or compile full Gentoo unstable on it? - the best benchmarks I 
know of.)


Grzegorz Kulewski

Re: Weird cold boot problems with Abit KT7 motherboard

[Andy Whitcroft]

--On 12 May 2004 19:13 -0400 Pavel Roskin <proski@gnu.org> wrote:

> Hello!
> 
> I have noticed several anomalies with Abit KT7 motherboard.  They all
> happen after power on.  First reboot from Linux (using the reboot command
> or reset button) usually fixes all the problems.  Sometimes two or three
> resets are needed before the motherboard starts working properly.  In two
> cases (of about 20) the motherboard started working properly right after
> powering up.

Hmmm, I have one of these pieces of shit ... I have a boot issue with mine
on Windows too.  Graphics card doesn't initialise correctly on most boots.
Basically the rumours out there are that this mobo is marginal in an
undefined part of the agp spec to do with power levels, the reset line and
probabally the phase of the moon.  I have noticed that my machine will boot
more reliably when the ethernet card is not connected to the hub.  In my
case a long manual reset (pressy button) just after power on seem to fix
this from then on.  There is a site somewhere which shows you how to solder
some wire on which 'helps the +3v line' or something.

I hate this mobo.  But I'm too lazy to replace it.

-apw

Re: Weird cold boot problems with Abit KT7 motherboard

[Flavio Stanchina]

Pavel Roskin wrote:
> I have noticed several anomalies with Abit KT7 motherboard. [...]

Your board is broken in some way. I have one right next to me at this 
very moment and it's been running Linux and occasionally Windows XP 
(only for games, of course) for years without a glitch. I don't remember 
if I ever put my only 3c900 in it, but a couple of my 3c905s certainly 
did and they never show problems.

I also have a KT7A which has been my workstation for 3 years and it 
never had any problems like the ones you're reporting, but that's 
different enough to not count as an example.

-- 
Ciao, Flavio

Re: Weird cold boot problems with Abit KT7 motherboard

[Pavel Roskin]

On Wed, 12 May 2004, Chuck Wolber wrote:

> On Wed, 12 May 2004, Pavel Roskin wrote:
>
> > I have noticed several anomalies with Abit KT7 motherboard.  They all
> > happen after power on.  First reboot from Linux (using the reboot
> > command or reset button) usually fixes all the problems.  Sometimes two
> > or three resets are needed before the motherboard starts working
> > properly.  In two cases (of about 20) the motherboard started working
> > properly right after powering up.
>
> %< SNIP
>
> > I just hope the description below will rings a bell with somebody, and
> > I'll gladly test the suggestions.  If not, the motherboard will go to the
> > dumpster.
>
>
> We just recently saw a similar report with the same motherboard. The
> system apparently boots fine under FC2. This is only second-hand
> information though, I have not personally tested this.

The kernel from Fedora testing (kernel-2.6.5-1.358.i686) indeed passed the
PCI probe on the cold booted machine with the TI PCI1410 bridge inserted
(it would have booted if not missing reiserfs support).  However, I was
unable to reproduce the cold boot problems I described with any kernel
since then.  Everything is working properly today.

I just didn't want to miss a chance to fix support for this motherboard.
But from my today's experience, from the replies I got and from further
search on the web, it appears unlikely that there is anything we can do in
the kernel to work around the cold boot problems with Abit KT7.

Thanks to everybody who replied!

-- 
Regards,
Pavel Roskin