Re: AMD Bulldozer FX-8150 Powers off during kernel build

[Sid Boyce]

# uname -r
3.6.0-rc5-u1-smp+

I built a new 3.6-rc5 kernel (3.6.0-rc5-u2) using 3.6.0-rc5-u1 with 8 
cores and power off didn't ocur.
slipstream:/usr/src/linux-3.6.0-rc5-u1 # grep POWER .config
# CONFIG_ACPI_PROCFS_POWER is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_X86_POWERNOW_K8=m
# CONFIG_PCIEASPM_POWERSAVE is not set
CONFIG_INPUT_POWERMATE=m
CONFIG_IPMI_POWEROFF=m
CONFIG_POWER_SUPPLY=y
# CONFIG_POWER_SUPPLY_DEBUG is not set
CONFIG_PDA_POWER=m
CONFIG_TEST_POWER=m
CONFIG_POWER_AVS=y
CONFIG_SENSORS_FAM15H_POWER=m
CONFIG_SENSORS_ACPI_POWER=m
CONFIG_SND_AC97_POWER_SAVE=y
CONFIG_SND_AC97_POWER_SAVE_DEFAULT=0
# CONFIG_SND_HDA_POWER_SAVE is not set
# CONFIG_HID_LCPOWER is not set
CONFIG_DEVFREQ_GOV_POWERSAVE=y
CONFIG_EVENT_POWER_TRACING_DEPRECATED=y
# CONFIG_XZ_DEC_POWERPC is not set

When it was powering off "CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y" was 
set.
slipstream:/usr/src/linux-3.6.0-rc5-u1 # grep PERFORMANCE .config
# CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_PCIEASPM_PERFORMANCE=y
CONFIG_DEVFREQ_GOV_PERFORMANCE=y

slipstream:/usr/src/linux-3.6.0-rc5-u1 # grep MCE .config
CONFIG_X86_MCE=y
# CONFIG_X86_MCE_INTEL is not set
CONFIG_X86_MCE_AMD=y
CONFIG_X86_MCE_THRESHOLD=y
# CONFIG_X86_MCE_INJECT is not set
CONFIG_EDAC_DECODE_MCE=y
# CONFIG_EDAC_MCE_INJ is not set

During the build temperature and power was around these values
-------------------------------------------------------------------------------------
fam15h_power-pci-00c4
Adapter: PCI adapter
power1:      133.30 W  (crit = 124.77 W)

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +61.9°C  (high = +70.0°C)
                       (crit = +90.0°C, hyst = +87.0°C)

Immediately after the build the values are much lower than what it was 
with the kernel and config that caused the power off.
----------------------------------------
fam15h_power-pci-00c4
Adapter: PCI adapter
power1:       31.10 W  (crit = 124.77 W)

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +33.2°C  (high = +70.0°C)
                       (crit = +90.0°C, hyst = +87.0°C)
------------------------------------------

If needed I can go back to the earlier 3.6.0-rc5 kernel and config to 
recreate the power off situation.
With the kernel that powered off, MCE was not set and 
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

For the 3.6.0-rc5-u1 kernel only those 2 were changed.
Regards
Sid.

On 13/09/12 10:44, Borislav Petkov wrote:
> On Thu, Sep 13, 2012 at 02:30:27AM +0100, Sid Boyce wrote:
> > I have a huge heatsink and large CPU fan plus lots of cooling fans
> > in the case and nothing gets hot.
> > If I build e.g 3.6-rc5 with 8 or 6 cores, part way through it
> > suddenly powers off.
> Ok, can you catch the whole dmesg when you boot the machine _after_ the
> sudden poweroff? You can send it to me and Andreas (on CC) privately if
> you prefer.
>
> Important: make sure the kernel has CONFIG_X86_MCE and
> CONFIG_EDAC_DECODE_MCE built-in.
>
> Please make sure to use a recent kernel, i.e. 3.4, 3.5 is fine.
>
> Thanks.
>
> (Leaving in the rest for reference)
>
> > I have checked hwmon/k10temp.c to see if I could see where these
> > values were defined.
> >
> > k10temp.h is 0 bytes.
> > -rw-r--r-- 1 root root 0 Sep  9 01:59
> > /usr/src/linux-3.6.0-rc5/include/config/sensors/k10temp.h
> >
> > Currently I build with "make -j 1" and temperature and power values
> > are around those below.
> > # sensors
> > k10temp-pci-00c3
> > Adapter: PCI adapter
> > temp1:        +60.4°C  (high = +70.0°C)
> >                         (crit = +90.0°C, hyst = +87.0°C)
> >
> > fam15h_power-pci-00c4
> > Adapter: PCI adapter
> > power1:      127.49 W  (crit = 124.77 W)
> >
> > # cat /proc/cpuinfo
> > processor       : 0
> > vendor_id       : AuthenticAMD
> > cpu family      : 21
> > model           : 1
> > model name      : AMD FX(tm)-8150 Eight-Core Processor
> > stepping        : 2
> > microcode       : 0x6000626
> > cpu MHz         : 3600.000
> > cache size      : 2048 KB
> >
> > from .config:-
> > # grep HWMON .config
> > CONFIG_IXGBE_HWMON=y
> > CONFIG_HWMON=y
> > CONFIG_HWMON_VID=m
> > # CONFIG_HWMON_DEBUG_CHIP is not set
> > CONFIG_THERMAL_HWMON=y
> >
> > # grep POWERSAVE .config
> > # CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
> > CONFIG_CPU_FREQ_GOV_POWERSAVE=m
> > # CONFIG_PCIEASPM_POWERSAVE is not set
> > CONFIG_DEVFREQ_GOV_POWERSAVE=y
> >
> > On another 6-core box I can build kernels with "make -j 6" without problems.
> > # cat /proc/cpuinfo
> > processor       : 0
> > vendor_id       : AuthenticAMD
> > cpu family      : 21
> > model           : 1
> > model name      : AMD FX(tm)-6100 Six-Core Processor
> > stepping        : 2
> > microcode       : 0x6000623
> > cpu MHz         : 3300.000
> > cache size      : 2048 KB
> >
> > With a kernel build going on six core box, temperature and power
> > hover around the values below.
> > sabre:~ # sensors
> > k10temp-pci-00c3
> > Adapter: PCI adapter
> > temp1:        +50.2°C  (high = +70.0°C)
> >                         (crit = +90.0°C, hyst = +87.0°C)
> >
> > fam15h_power-pci-00c4
> > Adapter: PCI adapter
> > power1:       94.40 W  (crit =  95.01 W)
> >
> > 73 ... Sid.
> >
> > --


--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/