Tag snooping points to cache again.
]]>Tag Snoop Error
]]>echo 0 >/sys/devices/system/cpu/cpu1/online
This doesn't help, as the failure occurs before. Also, dmesg reports this:
microcode: AMD CPU family 0xf not supported
so the last hope is gone. For now it works if I disable Cool'n'quiet (which I can live with) and if I down clock the cpu, which is not acceptable.
Any other ideas?
Edit: I've already updated BIOS, still the same.
]]>Another way:
echo 0 >/sys/devices/system/cpu/cpu1/online
Run this before the system crashes, e.g. in /etc/rc.local or whatever is the systemd equivalent if you use this.
Yet another: add nosmp to kernel command line if it's a dual-core CPU.
]]>Another option is disabling the faulting core. This should help with K10 which has separate L2 caches for each core. Not sure about K8, though.
]]>Install lm_sensors. Run sensors-detect and load driver for the SuperIO chip it finds. Run sensors.
If it reports temperatures above 60°C, it's possible that you have insufficient cooling and the CPU goes crazy due to high temperature.
Otherwise, it's core1 L2 cache damage or some extremely weird bug.
I've also installed amd-ucode, but I honestly don't know what to do with it.
]]>Did you perform any kernel updates recently?
Don't overclock (in case you do).
Clean CPU cooler from dust.
Try disabling SpeedStep/Cool'n'Quiet - somebody recently reported that CPUFReq appears to overclock his CPU to 5GHz (???) and crashes the system.
Check if you can boot some live cd, preferably one known to work on this computer in the past.
Check CPU temperature.
As a last resort, downclocking the CPU may increase probability that it will boot.
]]>if don't, activate "microcode updation" on BIOS, put clocksource=hpet and try again. Once booted, install microcode to your processor "pacman -Ss microcode" choose one and install
]]>