Missing: hw.acpi.thermal.tz%d._HOT

Ian Smith smithi at nimnet.asn.au
Sat Jun 14 06:06:41 UTC 2014


On Fri, 13 Jun 2014 10:08:57 -0700, hiren panchasara wrote:
 > On Fri, Jun 13, 2014 at 9:22 AM, Ian Smith <smithi at nimnet.asn.au> wrote:
 > > On Thu, 12 Jun 2014 14:28:33 -0700, hiren panchasara wrote:
 > >  > On Tue, Jun 10, 2014 at 8:40 AM, Eric Neblock <cen5848 at louisiana.edu> wrote:
 > >  > > On Wed, 2014-06-11 at 01:33 +1000, Ian Smith wrote:
 > >  > >> On Tue, 10 Jun 2014 09:54:14 -0500, Eric Neblock wrote:
 > >  > >>  > Hello all,
 > >  > >>  >   I'm trying to figure out what is the _HOT temperature on my particular
 > >  > >>  > processor. I'm running FreeBSD 10 GENERIC on a Sunfire X2200.
 > >  > >>  >
 > >  > >>  > The processor is an Dual Core AMD Opteron 2218.
 > >  > >>  >
 > >  > >>  > In the GENERIC kernel, acpi is built in; so, kldload acpi fails. I've
 > >  > >>  > also loaded the amdtemp module at boot time to figure out what the
 > >  > >>  > current temp of the processor is.
[..]
 > >  > > sysctl: Unknown oid 'hw.acpi.thermal' : No such file or directory
 > >  >
 > >  > Similar thing here at home desktop running -CURRENT:
 > >  >
 > >  > CPU: AMD FX(tm)-8350 Eight-Core Processor            (4000.24-MHz K8-class CPU)
 > >  >   Origin="AuthenticAMD"  Id=0x600f20  Family=0x15  Model=0x2  Stepping=0

So looking at /sys/dev/amdtemp/amdtemp.c .. here on stable/9 from a few 
weeks ago, whic appears to be an MFC of this one on head: 
http://svnweb.freebsd.org/base/head/sys/dev/amdtemp/amdtemp.c?view=log

"Driver for the AMD CPU on-die thermal sensors for Family 0Fh/10h/11h 
procs." with support added recently also for the 0x16h family, but no 
mention of 0x15 .. going by Eric's report, his would appear suoported.

Looking at amdtemp_gettemp() there, I suspect the 0x15 family uses yet 
another number or placement of register bits; your ~13C to 15C range of 
temps shown seems much more likely to be in the ~52C to 60C range ..

cc'ing jkim@, although others have messed with amdtemp more recently.

 > >  > acpi0: <7596MS A7596100> on motherboard

 > >  > # sysctl dev.amdtemp
 > >  > dev.amdtemp.0.%desc: AMD CPU On-Die Thermal Sensors
 > >  > dev.amdtemp.0.%driver: amdtemp
 > >  > dev.amdtemp.0.%parent: hostb4
 > >  > dev.amdtemp.0.sensor_offset: 0
 > >  > dev.amdtemp.0.core0.sensor0: 15.3C
 > >  >
 > >  > # sysctl -a dev.cpu | grep temp
 > >  > dev.cpu.0.temperature: 15.2C
[..]
 > >  > dev.cpu.7.temperature: 15.2C
 > >  >
 > >  > I am not sure how this ^ relates to what acpi reports under thermal.

I assume dev.amdtemp.0.core0.sensor0 is the source for all of those.

 > I am not sure how correct these numbers are but I've enabled AMD's
 > Cool'n'Quiet thingi in BIOS.

Looks like you need to ask someone to add support for family 0x15.  As 
for Eric, with no _TZ support, I don't know how you'd handle overtemps.

 > > And neither of these are reporting hw.acpi.thermal .. is it because the
 > > BIOS / ACPI doesn't present thermal zone information?
 > I'd believe so.
 > 
 > >  Or there aren't
 > > suitable drivers to interpret it?  I've no idea, but does seem curious.
 > >
 > > Any output from?
 > > # acpidump -dt | egrep -i 'TZ|thermal'
 > nothing.
 > 
 > # acpidump -dt | egrep -i 'TZ|thermal'
 > acpidump: RSDT entry 3 (sig OEMB) is corrupt
 > 
 > Now this ^^ error might also suggest something is wrong.

Don't know, but probably not related to the sensor temperatures.

 > > If so, you might want to put your full ASL up somewhere.
 > # acpidump -dt | gzip -c9 >  amd_fx8350.asl.gz
 > 
 > amd_fx8350.asl.gz is attached.

It's no use to me and the list swallowed it; you'd need to put it up at 
an URL somewhere .. but as it has no Thermal Zone section it can't help 
with this issue anyway.

 > By the time I collected everything,
 > 
 > # sysctl dev.cpu | grep temp
 > dev.cpu.0.temperature: 14.0C
 > dev.cpu.1.temperature: 14.0C
 > dev.cpu.2.temperature: 14.0C
 > dev.cpu.3.temperature: 14.0C
 > dev.cpu.4.temperature: 14.0C
 > dev.cpu.5.temperature: 14.0C
 > dev.cpu.6.temperature: 14.0C
 > dev.cpu.7.temperature: 14.0C

56C most likely, unless there's also an offset.  I'm out of clues ..

cheers, Ian


More information about the freebsd-acpi mailing list