ACPI / SATA / FFS - truncated directories, etc.

Chris Turner c.turner at
Fri Oct 26 16:34:24 PDT 2007

Hello all -

I've been trying to do some performance tests on a dual-core opteron
machine using the tyan tomcat H1000E board [1]_ - when I boot 1.10.1 UP
with acpi disabled, things run fine, but running UP with ACPI yields
strange dld problems and appears to cause filesystem corruption after
less than an hour or so of operation (afterwords, booting from cd +=
fsck yields truncated directory inodes, etc, so it appears that the
system is corrupting the disk).

Booting w/ACPI results in a dmesg with SATA, whereas without ACPI
results in the system booting in regular ATA mode.

Currently, the machine has been running UP for about 24H without ACPI
and managed to crunch it's way through a subset of a pkgsrc bulk build,
so I'm thinking there's something in either the ACPI or the SATA driver
for the onboard H1000 controller thats causing the problem - this has
been the second occurrence of similar as well.

I've got another identical machine I can use for testing - what's the
best way to proceed? Thinking to boot a -HEAD cdrom and do some
filesystem benchmarking inside a chroot.. I'll probably do that and
respond to thread but my cynical nature makes me think it might still be
an issue despite all the great ACPI work going on, so hoping to get a
'leg up' so to speak on what people think the right method might be for
troubleshooting this.. ACPI is a new concept for me so I'm trying to get
up to speed on things.. Also, the bios supports runtime setting of ACPI
v1, 2, 3 - what is best (aside from experimenting, of course)

in any case - thanks in advance.. looking forward to getting back in to
the swing of things in DF land!


- Chris


More information about the Kernel mailing list