Crash after trying powerd on FreeBSD 7.0BETA-2

JoaoBR joao at matik.com.br
Wed Nov 21 06:18:38 PST 2007


On Tuesday 20 November 2007 22:51:27 Arno J. Klaassen wrote:
> Kris Kennaway <kris at FreeBSD.org> writes:
> > Arno J. Klaassen wrote:
> > > Gregory Wright <gwright at antiope.com> writes:
> > >> Hi,
> > >>
> > >> I tried enabling powerd on FreeBSd 7.0BETA-2 by adding
> > >>
> > >> powerd_enable="YES"
> > >>
> > >> to /etc/rc.conf.  The mobo is a Tyan 2882-D with a single Opteron
> > >> 270 HE processor (dual core), 2 GB RAM.  AMD powernow was
> > >> enabled in the BIOS.
> > >
> > > I never succeeded in using cpufreq on any 288x MB. If freezes
> > > box even with latest bios.
> >
> > Have you tried disabling acpi_throttle?
>
> yes; but quite a long time ago; didn't make a difference
>

no change at all, doesn't matter if ULE or 4BSD

> those boards go rather low in their minimal frequency (about
> 200-300Mhz if not even lower) and in at least 8 steps; acpi_ppc, which
> works for me, seems to take into account only the four most powerfull
> states (about 1Ghz being minimal then)
>


are you saying that your setup there is changing the cpu freq levels by itself 
or acpi?

here, as long as I do not set it manually nothing happens, the averages are 
coming from top: 

Wed Nov 21 10:55:00 BRST 2007 Freq:2189   av: 0.00, 0.00, 0.00
Wed Nov 21 11:00:00 BRST 2007 Freq:2189   av: 0.00, 0.00, 0.00
Wed Nov 21 11:05:01 BRST 2007 Freq:2189   av: 0.00, 0.00, 0.00
Wed Nov 21 11:10:00 BRST 2007 Freq:2189   av: 0.00, 0.00, 0.00

I can run this for hours and days and the cpu freq stays where it is, machine 
idle

for the records powernow is enabled in the BIOS and cpufreq in the kernel:
Timecounter "ACPI-fast" frequency 3579545 Hz quality 1000
acpi_timer0: <24-bit timer at 3.579545MHz> port 0x1008-0x100b on acpi0
cpu0: <ACPI CPU> on acpi0
acpi_throttle0: <ACPI CPU Throttling> on cpu0
powernow0: <Cool`n'Quiet K8> on cpu0
cpu1: <ACPI CPU> on acpi0
powernow1: <Cool`n'Quiet K8> on cpu1
cpu2: <ACPI CPU> on acpi0
powernow2: <Cool`n'Quiet K8> on cpu2
cpu3: <ACPI CPU> on acpi0
powernow3: <Cool`n'Quiet K8> on cpu3



setting the cpu freq manually to a lower value seems to work because compling 
the kernel needs more time but running something the level do not switch up, 
it is locked

Wed Nov 21 11:15:00 BRST 2007 Freq:995   av: 0.00, 0.00, 0.00
Wed Nov 21 11:20:00 BRST 2007 Freq:995   av: 0.46, 0.13, 0.04
Wed Nov 21 11:25:00 BRST 2007 Freq:995   av: 0.96, 0.63, 0.29 
Wed Nov 21 11:30:00 BRST 2007 Freq:995   av: 1.01, 0.88, 0.51 

the cpu freq itself as you can see does not switch up as it also does not 
switch down

then here is a larger sheet with 3 temps and 3 fans measured with healthd 
*/5min and even if this values might not be accurate you can see they do not 
change I can put more load ont the CPUs and the fan speed grows up to 6200 
(first two) and the temps grow a little bit but whatever I try the cpu freq 
stays where it is


Freq:995   av: 0.96, 0.63, 0.29  35.0 36.5 43.5 4017 4115 20454
Freq:995   av: 1.01, 0.88, 0.51  35.0 36.5 43.5 4017 4115 20454
Freq:995   av: 1.03, 0.98, 0.65  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 1.05, 0.99, 0.75  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 1.02, 1.02, 0.83  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 2.25, 1.19, 0.90  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 5.80, 3.58, 2.06  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 7.14, 6.24, 3.80  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 9.68, 8.12, 5.33  35.0 36.5 43.5 4017 4115 20454
Freq:2189   av: 3.82, 6.73, 5.62  35.0 36.5 43.5 4017 4115 20454

so I guess that powernow is certainly working in means of temperature reading 
and fan control but cpu freq unfortunatly is not

to complete:
FreeBSD wco-nov.matik.com.br 7.0-BETA3 FreeBSD 7.0-BETA3 #0: Tue Nov 20 
14:49:30 BRST 2007     jm at wco-nov.matik.com.br:/usr/obj/usr/src/sys/GENERIC  
amd64




> might be a "piste"
>
> Arno
>
> PS, while here, and totaly unrelated, rev 1.252.2.7 (MFC "un-neede
> singl-threading code") of kern_fork.c to releng_6, seems to break
> things for me in a 4-way SMP (twice dual-core) amd64_releng6 box when
> forking procs in java. I'll triple-check before once again sending a
> bogus problem email but I got plenty of strange things when upgrading
> a production box to recent releng_6 (for sake of testing ata-promise
> TX4 patch), which disappeared after reverting to 1.252.2.6. fyi
> _______________________________________________
> freebsd-current at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe at freebsd.org"
>
>
>
>
>
>
>
> A mensagem foi scaneada pelo sistema de e-mail e pode ser considerada
> segura. Service fornecido pelo Datacenter Matik 
> https://datacenter.matik.com.br

-- 

João







A mensagem foi scaneada pelo sistema de e-mail e pode ser considerada segura.
Service fornecido pelo Datacenter Matik  https://datacenter.matik.com.br


More information about the freebsd-current mailing list