Topic: Unknown VIA Southbridge

[ked]

Hello, I'm having a problem with slow PATA hard disk performance on SME 7.
 
I have 250 gb Seagate parallel ATA drives on an ASUS Mb - the ASUS K8V-VM - which has a VIA chipset (K8M890/VT8237A)
 
I have one HD as master on IDE channel 1, the other as master on IDE channel 2 and a LiteOn DVD burner as slave on IDE channel 2.

As part of the install SME set up software RAID 1 as desired.

However, the RAID resynchronization has been running for 18 hrs and is only at 38% - and the sy cpu in top sits at over 60%.

I have checked and can confirm that DMA is not enabled on either drive.

When I issue the "hdparm -d /dev/hda" command it appears to have no effect - " using_dma = 0 (0ff)" remains set.

I get the following messages at boot:

Mar  9 10:01:09 localhost kernel: VP_IDE: IDE controller at PCI slot 0000:00:0f.1
Mar  9 10:01:09 localhost kernel: VP_IDE: chipset revision 7
Mar  9 10:01:09 localhost kernel: VP_IDE: not 100% native mode: will probe irqs later
Mar  9 10:01:09 localhost kernel: VP_IDE: Unknown VIA SouthBridge, disabling DMA.

The VIA Soutbridge chip (VT8237A) is obviously the problem. On the install cd there is a Linux IDE_patch directory - but it only has RPMs for Fedora, Mandriva and SuSE.

I would value any advice as to whether I should:

1. upgrade to a later CentOS release.
2. ditch the mobo for another model/brand.
3. recompile the kernal (will the Fedora rpm work with RHEL/CentOS?) to support the 8237A.

Whilst I have found a fair number of users having issues with SATA RAID I can't find anyone else having a problem with PATA IDE performance on this chipset so am a little confused.

I am willing to try a kernal recompile but, given I'm putting this system together for a friends small business, would prefer to have as simple an install/reinstall process as possible.

Thanks in advance.

[Dirk]

Did you update to the latest kernel?

Dirk

[ked]

No, in the end I decided to stick with as standard an install of SME as possible - the server is for a small business belonging to a friend and I want to minimise my involvement from a support perspective.

So I wimped out and threw some cash at it. (only $39 AUD)

I got a PCI IDE card with a Silicon Image chipset that works with the current kernel - which has boosted throughput in the RAID resync from 1042k/sec to 50338k/sec.

Cheers.

Ken

[jameswilson]

i have this board too, and think it needs a kernel update

[ked]

Yes, from what I could find on the web it would seem that kernel version 2.6.20 should/will support the VIA southbridge, but my SME Server 7.1.2 is still on 2.6.9.

As I understand it, if I choose to upgrade to 2.6.20 from RedHat, future SME Server updates may either (a) fail or (b) revert the kernel back. Is this correct?

In the end I thought a cheap hardware solution was the easiest fix until SME/CentOS moves to the 2.6.20 kernel.

Cheers.

[jameswilson]

oh lol i was hoping a new sme cd would do it lol, obviously not

[TheBS]

As I understand it, if I choose to upgrade to 2.6.20 from RedHat,

That would be a kernel from Fedora Core, not Red Hat Enterprise Linux (RHEL).
Remember, SME Server is based on CentOS, which is a rebuild of RHEL.
RHEL anally backports all fixes and newer support to the original kernel release.
I.e., all RHEL 4 releases use kernel 2.6.9, even though newer features are backported to that version in newer kernel releases.

future SME Server updates may either (a) fail or (b) revert the kernel back. Is this correct?

In general, you never want to run a Fedora Core kernel on RHEL.

In the end I thought a cheap hardware solution was the easiest fix until SME/CentOS moves to the 2.6.20 kernel.

That will happen with RHEL 5.
Red Hat purposely avoids moving to a newer kernel -- even over the 5-7 year release duration -- of any RHEL release for maximum ABI/API compatibility.

[ked]

Thanks for that clarification, TheBS - I'm still finding my way through the RHEL-as-CentOS and RHEL vs Fedora thing!

RHEL annually backports all fixes and newer support to the original kernel release.

So if I understand you correctly, at some point in the next 12 months there will be a CentOS kernel upgrade (still at 2.6.9) based on the RHEL update release, which will hopefully contain support for the VIA Southbridge chip some of us are struggling with?

I assume that it will be delivered through the server-manager based software upgrade feature?

(edit: for clarity)