I've been running a kernel I compiled from the 19981001 snapshot
for a while now, and have recently upgraded my user-land to -current
(to get egcs, binutils, and so forth), so I thought it would be
a good idea to upgrade the kernel to go along with it.

Unfortunately, when I rebooted with the new kernel (compiled from
to-day's source), I get a bunch of errors like this:

wd0: soft error (corrected)
wd0a: DMA error reading fsbn 130096 of 130096-130111 (wd0 bn 130096; cn 129 tn 1
 sn 1), retrying
wd0: soft error (corrected)
wd0a: DMA error reading fsbn 130176 of 130176-130191 (wd0 bn 130176; cn 129 tn 2
 sn 18), retrying
wd0: soft error (corrected)
wd0a: DMA error reading fsbn 130192 of 130192-130207 (wd0 bn 130192; cn 129 tn 2
 sn 34), retrying
wd0e: DMA error reading fsbn 16320 of 16320-16335 (wd0 bn 672528; cn 667 tn 3 sn
 3), retrying

Now, this didn't show up at all in the 19981001 kernel, so I'm thinking
something in the IDE subsystem has changed since then.

FWIW, I'm using an IBM 9 GB drive (can't remember the model name,
DataStore?) on the on-board PCI IDE controller on my CATS motherboard.
Both claim to support PIO modes up to 4 and UDMA. However, in previous
kernels, the DMA mode had been disabled because the kernel thought
the controller didn't support it. From the message above, though, one
can gather that DMA mode is now working (sort of) on the system.

My best guess is that the controller doesn't actually support UDMA,
and that's what's causing these errors. Is there an option I can specify
to turn it off?

-bjc

-- 
Brian Cully						<shmit@rcn.com>
``I'm not surprised,'' said I. ``You created God in your own image,
and when you found out he was no good you abolished him. It's quite a
common form of psychological suicide.''
				-- Robertson Davies, Fifth Buisiness