"random" hangs during boot, ACPI related ? Workaround problem by disabling HTT

[Kevin Oberman]

On Sun, Dec 28, 2014 at 5:49 AM, Thomas Steen Rasmussen <thomas at gibfest.dk>
wrote:

>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hello list,
>
> I've had a problem a couple of times now that I wanted to share with
> the list to see what you have to say. The problem is that FreeBSD
> sometimes hangs during boot, the last line of output from the kernel
> is "ACPI APIC Table: <INTEL  DENLOW>" and then nothing more happens.
>
> This seems to be related to something not getting initialized correctly
> after a soft reboot, but only sometimes. The first time I tried this it
> was on
> a server that has been in production for months, and suddently after a
> reboot during an installworld run it froze. Even a verbose boot doesn't
> reveal anything that jumps at me: http://i.imgur.com/y04ACyP.png
>
> Hard rebooting the server appears to solve it, until next reboot where
> it may or may not happen again.
>
> The same problem seems to affect others. This is the thread where I
> found suggestions for workarounds: https://bugs.pcbsd.org/issues/4024
>
> Workarounds include disabling C3 or disabling HTT in bios. Disabling C3
> did nothing for me, but disabling HTT made the server boot successfully
> again.
>
> The FreeBSD versions involved are 9 and 10, various versions. I've
> attached /var/run/dmesg.boot from one of the servers that had the
> problem a few weeks ago. On this server I've disabled HTT in the bios
> which worked around the problem.
>
> I just bought a new server at Hetzner (the hosting provider where I am
> experienzing this problem) and the server is not in production yet. I
> wanted to hold off on disabling HTT to see if anyone wants me to
> check something so we can get this fixed.
>
> Thoughts ? More info available on request.
>
>
> Best regards,
>
> /Thomas Steen Rasmussen
>

I know that C3 (and higher) and TCC don't get along on some platforms. I
thought that throttling was now off by default, but I'm not sure as I have
it disabled in /boot/loader.conf:
# Disable CPU throttling
hint.p4tcc.0.disabled=1
hint.acpi_throttle.0.disabled=1

This will not show up until ACPI and EST come up, so it sort of fits.
Whether it causes a lockup or not, disabling it is a good idea. It was
intended for use as thermal management, not power management. (The 'T' in
TCC (or P4TCC) is "Thermal", after all.)
--
R. Kevin Oberman, Network Engineer, Retired
E-mail: rkoberman at gmail.com