Am 17.02.2014 22:50, schrieb Gilles Chanteperdrix: > On 02/17/2014 10:14 PM, Christian Göbeler wrote: >> Am 17.02.2014 18:52, schrieb Gilles Chanteperdrix: >>> On 02/17/2014 06:47 PM, Christian Göbeler wrote: >>>> also disabled SMI by using the workaround described in the wiki >>>> (xeno_hal.smi=1) >>> Most importantly, after you have booted doing that run >>> >>> dmesg | grep -i smi >>> >>> to See if the SMI workaround actually worked. >>> >>> >> dmesg shows the following smi messages: >> >> [ 0.000000] Command line: >> BOOT_IMAGE=/boot/vmlinuz-3.8.13-xenomai-2.6.3 >> root=UUID=c01a5565-8f04-4959-a6a0-1bc4e920a017 ro text xeno_hal.smi=1 >> pci=nomsi nomodeset >> [ 0.000000] Kernel command line: >> BOOT_IMAGE=/boot/vmlinuz-3.8.13-xenomai-2.6.3 >> root=UUID=c01a5565-8f04-4959-a6a0-1bc4e920a017 ro text xeno_hal.smi=1 >> pci=nomsi nomodeset >> [ 0.000908] CPU0: Thermal monitoring handled by SMI >> [ 0.149371] CPU1: Thermal monitoring handled by SMI >> [ 0.170683] CPU2: Thermal monitoring handled by SMI >> [ 0.191977] CPU3: Thermal monitoring handled by SMI >> [ 0.213274] CPU4: Thermal monitoring handled by SMI >> [ 0.234576] CPU5: Thermal monitoring handled by SMI >> [ 0.255876] CPU6: Thermal monitoring handled by SMI >> [ 0.277184] CPU7: Thermal monitoring handled by SMI >> [ 0.814047] Xenomai: SMI-enabled chipset found >> [ 0.814093] Xenomai: SMI workaround enabled >> >> I'm not sure how to read that. The workaround is enabled but not working? >> >> Unfortunatly I couldn't resist and changed more of the kernel config >> than probably necessary, now things are getting really slow. >> >> None of my latencies are below 4000 and some are even beyond 170000. >> Additionally the posix mutex torture script crashes every once in a >> while with varying messages (i.e. device or resource busy, lock_stealing >> not supported, access violations, and so on) > You could be overheating, that could be the effect of disabling SMI > globally, if thermal monitoring relies on SMI. So, you should not try to > run with the smi workaround enabled. > > I am afraid if there is no option in BIOS to let the OS do the thermal > management instead of the SMI, and if the thermal monitoring is what is > causing the high latencies, you are out of luck. > I doubt that it's an overheating problem ... I've had the workaround enabled all day and most of my measurements were "OK" besides those occasional latency peaks.
I just installed lm-sensors to check the temperature: 65 to 70 degrees celsius while compiling a kernel on all cores. Quite warm, but still ok I guess. _______________________________________________ Xenomai mailing list [email protected] http://www.xenomai.org/mailman/listinfo/xenomai
