[ale] IDE problems with Software RAID-1

Danny Cox danscox at mindspring.com
Fri Oct 4 07:14:21 EDT 2002


Jeff,

On Fri, 2002-10-04 at 06:53, Jeff Layton wrote:
>    I've got a box at home that I'm experimenting with. I've got 3
> IDE HDs and an IDE CDRW. The first IDE drive has everything
> but /home on it. It's the master on ide 0 (/dev/hda). This drive is
> a 20 Gig WD drive. The remaining IDE HDS are setup as
> Software RAID-1. The first drive is the slave on ide 0 (/dev/hdb)
> and the second drive is the master on ide 1 (/dev/hdc). These
> drives are 40 Gig Seagate drives (I forgot the model number
> right now). The CDRW is the slave on ide 1 (/dev/hdd).
>    Anyway, I'm getting errors on /dev/hdb (the first drive in the
> RAID-1 set). I will sometimes get the usual errors:
> 
> kernel: hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error }
> kernel: hdb: dma_intr: error=0x84 { DriveStatusError BadCRC }
> 
> for a while. I'm also getting errors on /dev/hda. I can't remember
> the exact errors messages, but it keeps saying,
> 
> hda: lost interrupt
> 
> I also will get a few messages about DMA resets (or something
> like that) and 'ide 0 reset'.
>    I'm using a 2.4.19 kernel with -aa patches and -jam0 patches.
> I'm sorry I don't have concrete error messages (the machine was
> having real trouble last night and I couldn't get any of the logs out
> before I fell asleep :).
>    One last thing. All of the HDs are in removable racks. I've checked
> the cables to the racks. Their less than 18" and the cables tested fine
> in another box (not extensively tested though).

	It may be the cables, but I suspect the HDs more.  You might try
throttling 'em back some.  Sometimes DMA is just too much for some
chipsets/drives/cables.  Also, if you're using UDMA Mode 4 or higher,
you'll need the 80 conductor cables instead of the usual 40.  Try this:

	hdparm -d0 -c1 /dev/hdb

That turns off DMA, and enables 32bit transfers.

	If unmask_irq is set (1), you might try turning that off (hdparm -u0). 
That *might* be the cause of your lost interrupt.  What chipset is in
this beastie?  The bootup log will say, when the IDE driver discovers
it.

	See man hdparm for more info.

-- 
kernel, n.: A part of an operating system that preserves the
medieval traditions of sorcery and black art.

Danny


---
This message has been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.






More information about the Ale mailing list