ATA still broken, but now differently

[sos at deepcore.dk]

It seems Kirk Strauser wrote:
-- Start of PGP signed section.
> Well, Thursday's world seems to have broken ATA in a new and interesting
> way.  Before updating world in mid-September, my drive ran perfectly in
> UDMA66 mode.  Since that time, the system is absolutely guaranteed to freeze
> if I run in any mode higher than BIOSPIO.  Before two days ago, that meant
> getting errors like:
> 
>    ad0: TIMEOUT - WRITE_DMA retrying (2 reties left)
>    ata0: resetting devices
>    ad0: FAILURE - already active DMA on this device
>    ad0: setting up DMA failed
> 
> before infinite, unbreakable hangs.  As of today, when running in any DMA
> mode and exercising the ATA system, I get an endless stream of messages:
> 
>    swap pager: indefinite wait buffer
> 
> and I can only break out of the error cycle with a reset.
> 
> For the record, I've checked the cabling multiple times.  "smartctl" gives
> good results.  The machine runs perfectly under FreeBSD 4.9 or versions of
> 5.x prior to mid-September.
> 
> Any thoughts before I file (another) PR?

Hmm, please try a stock GENERIC kernel from todays sources, and boot verbosely
then get me the output of dmesg. 
I known of no issue with that chipset, so it might be something special 
about your setup that makes things fail.

-Søren