-CURRENT barfing on UDMA on Via 8233

[Damian Gerow]

Thus spake Kirk Strauser (kirk at strauser.com) [05/02/04 14:36]:
: > The PR (at quick glance) seems to bring up issue with 5.2-RC.  It also
: > states that PIO4 mode doesn't work.
: 
: At the end, PIO4 was in fact working and solid.  It was pretty flaky at the
: time I filed the PR, but stabilized somewhat afterward.

I had -RC on this machine, installed from JPSNAP the day before -RELEASE
was, well, released.

: > While my problem is still definitely with ATAng, I'm seeing very different
: > behaviour, and at different points in the tree, than you.  I can't even boot
: > in UDMA, but booting in PIO4 works just fine, even in a heavy load -- load
: > average at about 6, compiling jdk14, X11-4-libraries, and world at the same
: > time.
: 
: My gold standard was running Amanda backup.  During regular operation, when
: in a UDMA mode, I'd see a few intermittent warnings and a rare crash.  When
: I ran Amanda, my system would freeze - every time, guaranteed.  The same
: operations in PIO4 were slow but completely successfully.

I'm just finishing up a level 0 dump of all filesystems, and I'm running
pretty stable.  This isn't surprising, and it's in line with what you saw --
running under PIO4 is stable, UDMA100 is not.

Would it be worth it to try to force the system into UDMA66 or UDMA33?