Review your favorite Linux distribution.
Go Back > Forums > Linux Forums > Linux - Hardware
User Name
Linux - Hardware This forum is for Hardware issues.
Having trouble installing a piece of hardware? Want to know if that peripheral is compatible with Linux?


  Search this Thread
Old 11-23-2004, 02:28 PM   #1
Registered: Jul 2004
Location: Boston, MA, USA
Distribution: FreeBSD, Debian/AMD64, Ubuntu/i386
Posts: 59

Rep: Reputation: 15
Problem - IDE channels not independent

Short story:

my IDE channels on ICH5 are obviously not independent. In /proc/ide they are pointing to each other as "mates". How do I make them operate independently?

Long story:

I have a Asus P4C800-E Deluxe, which has onboard ICH5 (i875 chipset) and an onboard Promise controller. When I was using this combo with SATA drives, everything was fine and all drive independent, RAID-0 performance was about the sum of the drives.

Now I dropped the SATA stuff and got parallel ATA drives (and stuffed in another P-ATA promise PCI controller) and now obviously the ICH5 channels are not independent anymore. The performance in RAID-0 is very clear that any combination of one drive on ICH5 and one drive on the promise is full speed, but two drives on the ICH5 is dog slow.

- drive alone:
4294967296 B 4.0 GB 68.50 s 62698102 B/s 59.79 MB/s
- RAID0 one drive on ICH5 and one on Promise:
8589934592 B 8.0 GB 75.84 s 113269421 B/s 108.02 MB/s
- RAID0 with two drives on ICH5:
8589934592 B 8.0 GB 135.02 s 63620366 B/s 60.67 MB/s

Further remarks:
- the drives are on different channels on the ICH5
- in SATA mode everything worked as expected, only P-ATA seems to marry them

I run Fedora Core 2 2.6.8-1.521custom. I didn't change the configuration, it is only marked "custom" because I recompiled and installed it to make software like vmware happy that the source matches the running kernel.


1) I looked at the kernel source for IDE and didn't see anything resembling a knob to turn this on or off. Any ideas?

2) is it possible that one of Fedora's patches caused this? Maybe this is one of the Redhat/Fedora safety belts like not turning on DMA in the past? I diffed the IDE driver to the Linus driver and couldn't spot anything that could have caused this.


Here's dmesg for IDE. The full dmesg is on

EDIT: the forum didn't let me include the dmesg cut'n'paste, complaining about too many images. Yeah, right

Last edited by cracauer; 11-23-2004 at 03:34 PM.
Old 11-23-2004, 04:41 PM   #2
LQ Guru
Registered: Feb 2003
Location: Blue Ridge Mountain
Distribution: Linux Mint 17, Debian 8
Posts: 7,863

Rep: Reputation: 311Reputation: 311Reputation: 311Reputation: 311
"my IDE channels on ICH5 are obviously not independent. In /proc/ide they are pointing to each other as "mates". How do I make them operate independently?"

"onboard Promise controller"

Promise has a history of making IDE chipsets in which the two channels are not independent. Promise gets around this problem by writing a Windows driver which single threads the IDE chipset so that they never are accessed simultaneously. If you access them simultaneously you get hardware errors. Promise refuses to release the specs for their IDE chipsets and so the kernel developers have to reverse engineer the chipsets and write a Linux driver that handles the Promise design flaws the same way that the Promise Windows driver does.

I don't know whether your chipset falls into this catagory or not because 3 years ago I threw away the Promise chipset that taught me all this and have made it a point not to buy any more Promise chipsets. I have never used your Promise chipset. But what it sounds like to me is that Promise has design flaws (not necessarily the same one that I ran into) in your chipset which the kernel developers manage to accomodate at the price of a severe performance hit. I did not find any relevant error messages in your dmesg so maybe the kernel is handling the Promise chipset the way that it is supposed too.

I finally figured out what my problem with Promise was by looking at the kernel changelog for the 2 or 3 kernel releases before my kernel version and scanning for my Promise chipset. I suggest that you scan recent kernel changelogs for problems with your chipset that the developers had to write a workaround for.

Steve Stites

Last edited by jailbait; 11-23-2004 at 04:50 PM.
Old 11-23-2004, 04:54 PM   #3
Registered: Jul 2004
Location: Boston, MA, USA
Distribution: FreeBSD, Debian/AMD64, Ubuntu/i386
Posts: 59

Original Poster
Rep: Reputation: 15
Sorry, Steve, but the channels not independent are on the onboard Intel controller. You see the bad striping performance I posted is when I benchmark two disks on the Intel controller.

When I did SATA, the onboard Intel and the Promise were working together (independetly) like a charm.

Another person just pointed out to me that I have one drive configured as slave with no master. I'll try changing that when I come home.


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
4 independent sound channels player kubokubik Linux - Software 0 12-02-2004 04:34 AM
Independent sound channels? Phantomas Linux - General 3 08-21-2003 02:43 PM
how2 make the kernel scan both PCI IDE and Mboard IDE channels? carboncopy Slackware 1 07-23-2003 03:26 PM
IDE channels michap Linux - Hardware 3 11-04-2002 06:49 PM
swappnig ide channels acid_kewpie Linux - General 1 09-08-2001 02:10 PM > Forums > Linux Forums > Linux - Hardware

All times are GMT -5. The time now is 01:31 AM.

Main Menu
Write for LQ is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
Facebook: linuxquestions Google+: linuxquestions
Open Source Consulting | Domain Registration