ServerWorks CSB6 DMA problems

ServerWorks CSB6 DMA problems

[Gian-Carlo Pascutto]

Hi all,

is anyone aware of the status of ServerWorks CSB6's chipset DMA support?

On a machine with a RHEL3 kernel (2.4.21-27.0.1.EL), the kernel seems to 
enable UDMA33 on the first connected disk (Seagate 7200.7) but uses PIO 
mode on the second (Maxtor 6Y080P0).

I found out this is for good reasons because after enabling UDMA100 on 
both disks, via hdparm, filesystem corruption quickly resulted.

Googling turns up a lot of talk about buggy CSB4 chipsets, but nothing 
about the CSB6, actually several posts claiming it should work ok. The 
above experience and the kernel's selection suggest that's not the whole 
story though.

Basically, I'm curious if there's any *safe* way to get some more 
performance out of this configuration on Linux, or if I'm SOL. I mean, 
the Maxtor in PIO mode works at about 2Mb/s... :-/

-- 
GCP

Re: ServerWorks CSB6 DMA problems

[Alan Cox]

On Gwe, 2005-01-14 at 21:39, Gian-Carlo Pascutto wrote:
> On a machine with a RHEL3 kernel (2.4.21-27.0.1.EL), the kernel seems to 
> enable UDMA33 on the first connected disk (Seagate 7200.7) but uses PIO 
> mode on the second (Maxtor 6Y080P0).
> 
> I found out this is for good reasons because after enabling UDMA100 on 
> both disks, via hdparm, filesystem corruption quickly resulted.

That all suprises me a great deal, although the modes are normally BIOS
selected by the Serverworks and then read by the kernel. CBS5/CSB6 IDE
should be very reliable indeed (OSB4 is a much older chipset and the
problem it has is OSB4 specific)

I'd also suggest you raise it with your RHEL support and/or
bugzilla.redhat.com

Alan