OK, I just realized this isn't specifically related to 2.6.19-rc2 as
it's been around since 2.6.16 w/ Alan's PATA patches for me. I'm
currently running 2.6.17-ck kernel due to a couple hard-locks I
experienced in 2.6.19-rc2, which I haven't yet had time to mess with.

On bootup, the pata_amd driver mis-detects the cable connected to the
2nd port on my system as 40 wire and sets UDMA/33 for this drive. Prior
to using the libata patches, the system did correctly setup this drive
for UDMA/100.  The cable is a standard 80 conductor, flat IDE cable.
I've tried switching the cables around to see if it truly was a cable
issue, but it makes no difference.  The 2nd port always gets detected as
a non-80-wire cable.

For kicks, I commented out the ap->cbl = ATA_CBL_PATA40 definition in
pata_amd.c and set it to ATA_CBL_PATA80 and it's been running fine for
~24 hrs (2.6.17-ck).

Any ideas?


dmesg, lspci & config follows: