Very slow disk IO on SATA RAID1 Jetway NC92-330-LF

[Guest]

[DFen]

I have installed endian on a Jetway NC92-330-LF MB with two SATA HDD in software RAID1

Apart from the install being very slow, after installation I am seeing extremely slow IO on the hard disks.

 hdparm -tT /dev/hda

/dev/hda:
 Timing cached reads:   2392 MB in  2.00 seconds = 1195.39 MB/sec
 Timing buffered disk reads:   10 MB in  3.08 seconds =   3.25 MB/sec

I was also surprised to see the disks using device hdX and not sdX!

Am I missing something here or does endian CE not fully support SATA?

[DFen]

Just a follow up on my problems with 2.4 IO speed

I installed 2.3 on the same machine with no speed poblems at all
hdparm was showing 100MB/s Buffered read speed.

I also note that 2.3 was using kernel module libata and ata_piix as well as some scsi modules.

libata is NOT present on my copy of endian CF 2.4
Can anybody suggest where I can get additional kernel modules?

Further follow-up
============

I have upgraded using efw-upgrade and the system has reverted to being very slow!!

The disk devices are now showing as hda and hdb instead of sda and sdb (I guess having raid 1 md devices allows it to boot)
kernel mosule ata_piix is loaded but no sign of libata.

This looks to me similar to the problem found at:
h**p://bugs.endian.com/view.php?id=208


Yet more follow up 3/July/2010:
I am not at all sure where the problem lies however it appears that libata is built in to the kernel and not created as a module.
I have tried ordinary IDE disks and the speed problem remains the same.

I suspect now that this problem relates to the removal of Combined mode in kernel 2.6.22
I understand this may cause problems when the OLD ide drivers are included in the kernel along with libata drivers.

I will be the first to admit I don't understand all this, but see the following references:

Combined mode
h**p://www.mail-archive.com/linux-ide@vger.kernel.org/msg04681.html

Performance issues after combined mode removed:
h**p://forum.soft32.com/linux/Combined-mode-quirk-removal-kills-performance-ftopict345203.html

Related stuff I don't understand
h**p://forum.soft32.com/linux/PATCH-22-remove-Intel-combined-mode-quirk-ftopict341159.html

A possible fix in kernel config
h**p://linux.derkeiler.com/Mailing-Lists/Kernel/2007-07/msg08563.html


Now all I need is to try a kernel built to endian standards but without the old ide driver!!