Hi Nicolas, > FWIW, I have a Gigabyte motherboard with an it87 chip too. Reading > about this it87 polarity thing I'm suspecting something is really > wrong here: > > When system is idle, the sensors report shows: > CPU temp = +25?C and CPU fan = 2136 RPM (and rather noisy) > > When system is 100% busy (with dd if=/dev/urandom of=/dev/null): > CPU temp = +41?C and CPU fan = 1288 RPM (and obviously much quieter) > > I'm running a 2.6.10 kernel (not -mm) so I guess the BIOS settings for > fan control are not altered. And incidentally the BIOS has a setting > called "smart fan control" set to "enabled" which maps to the ITxxF > automatic PWM control mode I suppose. So if the BIOS actually set the > fan polarity wrong then the fan would slow down when the temperature > rises and vice versa, right? That's right, what you describe really sounds like a wrong polarity setting. Could you please tell us what model it is, with what BIOS revision? I would also appreciate a dump of the chip (isadump 0x295 0x296 unless it lives at some uncommon address) to confirm the guess. The code Jonas and I have been adding to the it87 driver recently will probably not work for you if you leave the "smart fan control" enabled in your BIOS setup screen (because we supposed that no reponsible motherboard maker would enable this mode without properly configuring the polarity beforehand - wrong guess in your case - and are not allowing a polarity inversion in this case). However, by disabling this mode in the BIOS setup screen, you should be able to use the new fix_pwm_polarity module register to get the polarity fixed, with manual PWM control (no auto mode yet). You might also search for a BIOS update for your board. I consider the behavior your describe a major problem and would expect Gigabyte to fix it at some point in time. Thanks, -- Jean Delvare http://khali.linux-fr.org/
