ATA and SATA problems (timeouts/resetting)

[Josh Endries]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hey everyone,

I hope this goes through. For some reason I get bounces saying it
can't reverse my IP, though I see no problems.

I'm having major issues getting FreeBSD to install on a server. It's
been a couple weeks now and nothing I've tried has helped. The
server in question used to be running 4-STABLE until I upgraded it
to 5-STABLE, which is when I started getting ATAPI errors:

ata1-master FAILURE - ATAPI_IDENTIFY timed out

The only ATA/IDE device plugged in is a CD-ROM, which was in
secondary master position when this error happened. I've moved it
around and nothing helps, it just changes the source of the problem
(ata0-master, etc.). The system also has a 3ware SATA RAID PCI card
in it, twa0, which it is booting from. Both the system BIOS and
3ware firmware is fresh. After three of the above errors it gives up
I guess and then I get these:

twa0: Request timed out!
twa0: Resetting controller
twa0: INFO: 0x04 0x005e Cache synchronized after power fail
twa0: INFO: 0x04 0x0001 Controller reset occurred
twa0: Controller reset done!

I get the same thing with the latest 6 ISO (beta 1?).

I have an almost identical system that is working just fine with
5-STABLE. The only difference is that machine has a LSI MegaRAID
SCSI card also. I had these problems initially with that machine,
but they just disappeared and it's running/rebooting fine, which
worries me a bit. I think I booted into safe mode and cvsuped,
custom kernel, and it started working, but I tried that with the
new machine (same kernel config file) and it didn't have the same
effect. I've scoured through the BIOSes and they're set up
identically. It isn't sporadic either, I get the errors every single
time, just after the "timecounters tick at 1msec" line (or whatever
it is, I forget).

Anyway I found some into online about mkIII patches and applied
those and now I just get different errors. I don't remember
specifically what they were, I can reinstall again and get them, but
it was similar, timeout setting transferrate (or tranfer mode), then
it said "danger will robinson" and started mixing in the above twa0
errors.

Booting normally doesn't work at all, neither does single-user mode.
The only way I can get in (to use and/or initially install) is using
safe mode. I added an option to the menu Safer Mode to try and find
out what difference was causing it but tried with/without the
ATA/DMA, APIC, and ACPI lines individually and it didn't change
anything. I've tried GENERIC and SMP (they are DP machines) and
various kernel changes, stripping it bare, disabling DMS and ACPI in
/boot/loader.conf...nothing helped. I turned off DMA in the BIOS,
changed the transfer speed (PIO, standard, etc.) and just about
every other thing I could think of.

I just successfully installed 4.11 and it boots fine, no errors
whatsoever. I was wondering if it's a hardware problem, but
everything seems to run fine on the other 5.x machine (after the
problems went away :/) and this 4.x one, so I'm not sure.

Anyone have any ideas what I can do to troubleshoot or (hopefully)
fix this? I'd much rather run 5 on it than 4, but if all else fails
I guess I'm stuck with what works.

Thanks,
Josh
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (FreeBSD)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFC4QBeV/+PyAj2L+IRAtmgAJ4s68SSJQjQtxQTzL+/gi2FN4Qm1gCeM0oN
2LBqpERB6cOpZCbWMG2+crQ=
=wTZf
-----END PGP SIGNATURE-----