On Fri June 12 2009, Harald Welte wrote: > On Fri, Jun 12, 2009 at 06:46:45AM -0500, Michael S. Zick wrote: > > > Temperature values are supposed to be expressed in millidegrees C, not > > > degrees C as it seems to be doing (although 25 degrees C seems pretty > > > low for a CPU temperature?) The drivers needs to multiply values by > > > 1000 before exporting them to sysfs. Then "sensors" will report the > > > correct temperature value. > > > > > > > Ah, 25 degrees C is room temperature - real hard for the junction temperature > > to be 25 degrees C with power applied; lacking an infinitely perfect heatsink. > > > > Look for an "off by one" error in shifting or masking the value. > > there is no shifting and the masking is 0xffffffff :) > > it might be that the BIOS is doing something wrong when programming the > calibration MSR's at early botoup. I would need the contents of MSR > 0x1160 ... 0x116C as well as 0x1152 and 0x1153 to be able to determine that. > I was commenting on someone else's observation - - Will build with your patch and poke at it a bit. I should be able to read those registers from /dev/cpu/0/msr, correct? Or is there a "safe" mask in the msr driver that prevents reading random MSR registers? Mike
