Skip to contentSkip to conversation

Via KT600 support?

DISCUSSION Cory Bell 3 participants 4 messages Nov 12, 2003 – Nov 13, 2003

lore
Share & tools
lore

Cory Bell

author
#0 lore raw
Quoted 2 lines · Joe Harrington

I am having show-stopper difficulties getting Linux to run on my Asus A7V600 motherboard, which sports the Via KT600 chipset.

I also have an Asus A7V600. I'm unsure of the board revision, but I'm running BIOS 1005.

Quoted 5 lines

During install of Fedora Core 1, Fedora Core test3, Red Hat 9, and Debian 3.0r1, the install fails at a random point, generally during the non-interactive package loading phase. The most recent kernel with the problem is kernel-2.4.22-1.2115.nptl in the Fedora Core 1 release. The problem is 100% reproducible.

I installed FC1 from CD with no problems.

Quoted 6 lines

VC messages indicate problems in different programs each time. Generally the failure happens during package installation, but sometimes it happens earlier. After the first indication of a problem, one can generally still switch VCs, but eventually that stops working, too. Frequently, there are several programs indicating problems in the VC screens.

What are you seeing? Segfaults? Panics? Have you tried the memtest86 boot option (from the FC1 CDs)?

Quoted 7 lines

The hardware:

Asus A7V600 mobo (VIA KT600 chipset)
AMD 2800+XP CPU
1536 MB DDR333 (3 sticks)
IDE disks (250GB WD)
IDE CD/DVD reader

I have:
Asus A7V600
Athlon XP 2600+ (Barton)
512MB DDR333 (PC2700)
Lite-On IDE DVD-ROM
Lite-On IDE CD-RW
Maxtor PATA 160GB
Chaintech Geforce 4 Ti4800 (AGP 8x)
No SATA

Quoted 4 lines

Things I have tried:
noapic nolapic acpi=off pci=noacpi allowcddma nodma in various combinations and individually. None worked. I also changed "APIC" to "PIC" in the BIOS, to no avail.

Mine works fine in APIC mode and with ACPI.

Quoted 6 lines

I have tested and replaced the hardware extensively over the past several months of chasing this problem down. The problem is not damaged hardware. I suspect it is a problem with chipset support in the kernel. However, I have not been able to find a reliable source of information about the support status of various chipsets and motherboard features in the kernel.

Everything is supported fine except:
1) The OSS driver doesn't like the onboard AC-97 Audio - I installed ALSA 0.9.8 with "options snd-via82xx dxs_support=2".
2) The sk98lin driver included in 2.4.22 is very old and doesn't support the onboard "3Com 3C940" (actually a Syskonnect chip). 2.4.23-rc1 contains the latest driver, or you can patch 2.4.22.

ALSA: http://www.alsa-project.org/
Syskonnect Driver: http://www.syskonnect.com/syskonnect/support/driver/d0102_driver.html

One caveat: It appears that the kernel included with FC1 was not compiled with the same version of gcc that is included in FC1. This means you will need to compile your own kernel (or rebuild the FC1 kernel SRPM). I got an instant panic when trying to insert a locally-compiled sk98lin module into the supplied kernel, so I just downloaded the latest 2.4.23 test release.

-Cory

#1 lore raw
Quoted 2 lines

I also have an Asus A7V600. I'm unsure of the board revision, but I'm running BIOS 1005.

I am glad to see a positive report with this botherboard (actually, that was a typo but it's perfectly true, so I'll leave it). The problem has turned out to be faulty memory, so I apologize to the list for wasting your time. So nobody complains, I used 4 different sticks of SimpleTech memory in this machine, all with the same results. They passed 24 hours of memtest86 with no errors. But, replacing them with some Corsair memory fixed the problem. A Simpletech rep says the batch that my 4 sticks come from was not a particularly bad batch, but they are sending boards from another batch that they use in more applications in hope that it fixes the problem. They also say this chipset is particularly finicky about memory. Needless to say, I won't be sending them any more of my hard-earned cash: If it says PC2700 on it, it should perform that way. This was not "value RAM" to me.

In case anyone else has similar trouble, I'll answer some of your questions so it's easier to identify:

Quoted 9 lines
Quoted 6 lines

VC messages indicate problems in different programs each time. Generally the failure happens during package installation, but sometimes it happens earlier. After the first indication of a problem, one can generally still switch VCs, but eventually that stops working, too. Frequently, there are several programs indicating problems in the VC screens.

What are you seeing? Segfaults? Panics? Have you tried the memtest86 boot option (from the FC1 CDs)?

I ran memtest86 for a day with no errors. Is there something that can be done to make its tests more definitive? Clearly this tool didn't find the problem, but a few minutes of transferring files did.

VC4 gave messages including:
EIP is at (some process or kernel routine name, different each time) then a stack and call trace (again, different each time) then sometimes

Kernel panic: 
also sometimes
Kernel BUG in buffer.c:

sometimes there are several sets of data including the EIP message, stack trace, etc. for different programs/kernel routines at once.

Sometimes the installer just quits, saying it exited abnormally, and it does a graceful shutdown without any VC4 errors.

The clues in your report will help a lot now that I have this thing working. I wish there were a website for such info, like there is for laptops.

Thanks again for your report.

--jh--

#2 lore raw
In reply to Cory Bell
Quoted 7 lines

One caveat: It appears that the kernel included with FC1 was not compiled with the same version of gcc that is included in FC1. This means you will need to compile your own kernel (or rebuild the FC1 kernel SRPM). I got an instant panic when trying to insert a locally-compiled sk98lin module into the supplied kernel, so I just downloaded the latest 2.4.23 test release.

Install the gcc32 rpm and compile the kernel with CC=gcc32 instead.

// Stefan

#3 lore raw
Quoted 9 lines · Stefan Smietanowski
Quoted 7 lines

One caveat: It appears that the kernel included with FC1 was not compiled with the same version of gcc that is included in FC1. This means you will need to compile your own kernel (or rebuild the FC1 kernel SRPM). I got an instant panic when trying to insert a locally-compiled sk98lin module into the supplied kernel, so I just downloaded the latest 2.4.23 test release.

Quoted 3 lines

Install the gcc32 rpm and compile the kernel with CC=gcc32 instead.

// Stefan

That is, compile the driver you want compiled that way.

// Stefan