Correction: It *says* it's using udma5 and udma6, but running hdparm -t /dev/hd[a|b] gives around 5-6 MB/s whereas another drive that is not reporting these issues gives 21-22 MB/s. Quite a leap I'd say... Or am I reading too much into the hdparm tests? On Sat, 6 Nov 2004 02:17:04 +0100, Martin Moeller wrote:

Hi all.

Recently my harddisks in an older system have seemed to fail or be close to failing and I decided to splash on a new disk, just in case.

However, now I get some very annoying timeout messages during startup of the above mentioned variety (dma_timer_expiry) typically followed by either 0x21, 0x61 or 0x41 as the 'error'. Nothing much of relevance shows up in google searches.

I tried booting PartitionMagic just to see if this is Linux only and there were no delays for Partition Magic (DOS version).

I then tried a 2.4 kernel KNOPPIX boot. Also no problems (or maybe it's just not smart enough to know there should be any?).

The weird thing is that afterwards the drives function well enough and at full DMA speed (typically UDMA5 or UDMA6).

Can anyone explain to me why 2.6 kernels (at least 2.6.8/2.6.8.1) does this and if I can prevent it in some way? I've tried turning APIC on and off, using idewait=50 and such. I have not tried ide=nodma, since I *want* DMA :).

PS: It's the exact same thing with Yoper.

The new disk is a Hitachi DeskStar 7K250 160 GB disk. After setting both this and the previous disk on 80 connector cables, both exhibit this behaviour. Jumpers are set correctly. Cabling is correct, according to tech docs at Hitachi Storage homepage....

Anyone?

/Martin.