| Juerg Haefliger | 61d0b53 | 2006-09-24 20:54:46 +0200 | [diff] [blame] | 1 | Kernel driver vt1211 | 
|  | 2 | ==================== | 
|  | 3 |  | 
|  | 4 | Supported chips: | 
|  | 5 | * VIA VT1211 | 
|  | 6 | Prefix: 'vt1211' | 
|  | 7 | Addresses scanned: none, address read from Super-I/O config space | 
|  | 8 | Datasheet: Provided by VIA upon request and under NDA | 
|  | 9 |  | 
|  | 10 | Authors: Juerg Haefliger <juergh@gmail.com> | 
|  | 11 |  | 
|  | 12 | This driver is based on the driver for kernel 2.4 by Mark D. Studebaker and | 
|  | 13 | its port to kernel 2.6 by Lars Ekman. | 
|  | 14 |  | 
|  | 15 | Thanks to Joseph Chan and Fiona Gatt from VIA for providing documentation and | 
|  | 16 | technical support. | 
|  | 17 |  | 
|  | 18 |  | 
| Juerg Haefliger | a1fdcb9 | 2006-09-24 20:55:34 +0200 | [diff] [blame] | 19 | Module Parameters | 
|  | 20 | ----------------- | 
|  | 21 |  | 
|  | 22 | * uch_config: int	Override the BIOS default universal channel (UCH) | 
|  | 23 | configuration for channels 1-5. | 
|  | 24 | Legal values are in the range of 0-31. Bit 0 maps to | 
|  | 25 | UCH1, bit 1 maps to UCH2 and so on. Setting a bit to 1 | 
|  | 26 | enables the thermal input of that particular UCH and | 
|  | 27 | setting a bit to 0 enables the voltage input. | 
|  | 28 |  | 
|  | 29 | * int_mode: int		Override the BIOS default temperature interrupt mode. | 
|  | 30 | The only possible value is 0 which forces interrupt | 
|  | 31 | mode 0. In this mode, any pending interrupt is cleared | 
|  | 32 | when the status register is read but is regenerated as | 
|  | 33 | long as the temperature stays above the hysteresis | 
|  | 34 | limit. | 
|  | 35 |  | 
|  | 36 | Be aware that overriding BIOS defaults might cause some unwanted side effects! | 
|  | 37 |  | 
|  | 38 |  | 
| Juerg Haefliger | 61d0b53 | 2006-09-24 20:54:46 +0200 | [diff] [blame] | 39 | Description | 
|  | 40 | ----------- | 
|  | 41 |  | 
|  | 42 | The VIA VT1211 Super-I/O chip includes complete hardware monitoring | 
|  | 43 | capabilities. It monitors 2 dedicated temperature sensor inputs (temp1 and | 
|  | 44 | temp2), 1 dedicated voltage (in5) and 2 fans. Additionally, the chip | 
|  | 45 | implements 5 universal input channels (UCH1-5) that can be individually | 
|  | 46 | programmed to either monitor a voltage or a temperature. | 
|  | 47 |  | 
|  | 48 | This chip also provides manual and automatic control of fan speeds (according | 
|  | 49 | to the datasheet). The driver only supports automatic control since the manual | 
|  | 50 | mode doesn't seem to work as advertised in the datasheet. In fact I couldn't | 
|  | 51 | get manual mode to work at all! Be aware that automatic mode hasn't been | 
|  | 52 | tested very well (due to the fact that my EPIA M10000 doesn't have the fans | 
|  | 53 | connected to the PWM outputs of the VT1211 :-(). | 
|  | 54 |  | 
|  | 55 | The following table shows the relationship between the vt1211 inputs and the | 
|  | 56 | sysfs nodes. | 
|  | 57 |  | 
|  | 58 | Sensor          Voltage Mode   Temp Mode   Default Use (from the datasheet) | 
|  | 59 | ------          ------------   ---------   -------------------------------- | 
|  | 60 | Reading 1                      temp1       Intel thermal diode | 
|  | 61 | Reading 3                      temp2       Internal thermal diode | 
|  | 62 | UCH1/Reading2   in0            temp3       NTC type thermistor | 
|  | 63 | UCH2            in1            temp4       +2.5V | 
|  | 64 | UCH3            in2            temp5       VccP (processor core) | 
|  | 65 | UCH4            in3            temp6       +5V | 
|  | 66 | UCH5            in4            temp7       +12V | 
|  | 67 | +3.3V           in5                        Internal VCC (+3.3V) | 
|  | 68 |  | 
|  | 69 |  | 
|  | 70 | Voltage Monitoring | 
|  | 71 | ------------------ | 
|  | 72 |  | 
|  | 73 | Voltages are sampled by an 8-bit ADC with a LSB of ~10mV. The supported input | 
|  | 74 | range is thus from 0 to 2.60V. Voltage values outside of this range need | 
|  | 75 | external scaling resistors. This external scaling needs to be compensated for | 
|  | 76 | via compute lines in sensors.conf, like: | 
|  | 77 |  | 
|  | 78 | compute inx @*(1+R1/R2), @/(1+R1/R2) | 
|  | 79 |  | 
|  | 80 | The board level scaling resistors according to VIA's recommendation are as | 
|  | 81 | follows. And this is of course totally dependent on the actual board | 
|  | 82 | implementation :-) You will have to find documentation for your own | 
|  | 83 | motherboard and edit sensors.conf accordingly. | 
|  | 84 |  | 
|  | 85 | Expected | 
|  | 86 | Voltage       R1     R2     Divider   Raw Value | 
|  | 87 | ----------------------------------------------- | 
|  | 88 | +2.5V         2K     10K    1.2       2083 mV | 
|  | 89 | VccP          ---    ---    1.0       1400 mV (1) | 
|  | 90 | +5V           14K    10K    2.4       2083 mV | 
|  | 91 | +12V          47K    10K    5.7       2105 mV | 
|  | 92 | +3.3V (int)   2K     3.4K   1.588     3300 mV (2) | 
|  | 93 | +3.3V (ext)   6.8K   10K    1.68      1964 mV | 
|  | 94 |  | 
|  | 95 | (1) Depending on the CPU (1.4V is for a VIA C3 Nehemiah). | 
|  | 96 | (2) R1 and R2 for 3.3V (int) are internal to the VT1211 chip and the driver | 
|  | 97 | performs the scaling and returns the properly scaled voltage value. | 
|  | 98 |  | 
|  | 99 | Each measured voltage has an associated low and high limit which triggers an | 
|  | 100 | alarm when crossed. | 
|  | 101 |  | 
|  | 102 |  | 
|  | 103 | Temperature Monitoring | 
|  | 104 | ---------------------- | 
|  | 105 |  | 
|  | 106 | Temperatures are reported in millidegree Celsius. Each measured temperature | 
|  | 107 | has a high limit which triggers an alarm if crossed. There is an associated | 
|  | 108 | hysteresis value with each temperature below which the temperature has to drop | 
|  | 109 | before the alarm is cleared (this is only true for interrupt mode 0). The | 
|  | 110 | interrupt mode can be forced to 0 in case the BIOS doesn't do it | 
| Juerg Haefliger | a1fdcb9 | 2006-09-24 20:55:34 +0200 | [diff] [blame] | 111 | automatically. See the 'Module Parameters' section for details. | 
| Juerg Haefliger | 61d0b53 | 2006-09-24 20:54:46 +0200 | [diff] [blame] | 112 |  | 
|  | 113 | All temperature channels except temp2 are external. Temp2 is the VT1211 | 
|  | 114 | internal thermal diode and the driver does all the scaling for temp2 and | 
|  | 115 | returns the temperature in millidegree Celsius. For the external channels | 
|  | 116 | temp1 and temp3-temp7, scaling depends on the board implementation and needs | 
|  | 117 | to be performed in userspace via sensors.conf. | 
|  | 118 |  | 
|  | 119 | Temp1 is an Intel-type thermal diode which requires the following formula to | 
|  | 120 | convert between sysfs readings and real temperatures: | 
|  | 121 |  | 
|  | 122 | compute temp1 (@-Offset)/Gain, (@*Gain)+Offset | 
|  | 123 |  | 
|  | 124 | According to the VIA VT1211 BIOS porting guide, the following gain and offset | 
|  | 125 | values should be used: | 
|  | 126 |  | 
|  | 127 | Diode Type      Offset   Gain | 
|  | 128 | ----------      ------   ---- | 
|  | 129 | Intel CPU       88.638   0.9528 | 
|  | 130 | 65.000   0.9686   *) | 
|  | 131 | VIA C3 Ezra     83.869   0.9528 | 
|  | 132 | VIA C3 Ezra-T   73.869   0.9528 | 
|  | 133 |  | 
|  | 134 | *) This is the formula from the lm_sensors 2.10.0 sensors.conf file. I don't | 
|  | 135 | know where it comes from or how it was derived, it's just listed here for | 
|  | 136 | completeness. | 
|  | 137 |  | 
|  | 138 | Temp3-temp7 support NTC thermistors. For these channels, the driver returns | 
|  | 139 | the voltages as seen at the individual pins of UCH1-UCH5. The voltage at the | 
|  | 140 | pin (Vpin) is formed by a voltage divider made of the thermistor (Rth) and a | 
|  | 141 | scaling resistor (Rs): | 
|  | 142 |  | 
|  | 143 | Vpin = 2200 * Rth / (Rs + Rth)   (2200 is the ADC max limit of 2200 mV) | 
|  | 144 |  | 
|  | 145 | The equation for the thermistor is as follows (google it if you want to know | 
|  | 146 | more about it): | 
|  | 147 |  | 
|  | 148 | Rth = Ro * exp(B * (1 / T - 1 / To))   (To is 298.15K (25C) and Ro is the | 
|  | 149 | nominal resistance at 25C) | 
|  | 150 |  | 
|  | 151 | Mingling the above two equations and assuming Rs = Ro and B = 3435 yields the | 
|  | 152 | following formula for sensors.conf: | 
|  | 153 |  | 
|  | 154 | compute tempx 1 / (1 / 298.15 - (` (2200 / @ - 1)) / 3435) - 273.15, | 
|  | 155 | 2200 / (1 + (^ (3435 / 298.15 - 3435 / (273.15 + @)))) | 
|  | 156 |  | 
|  | 157 |  | 
|  | 158 | Fan Speed Control | 
|  | 159 | ----------------- | 
|  | 160 |  | 
|  | 161 | The VT1211 provides 2 programmable PWM outputs to control the speeds of 2 | 
|  | 162 | fans. Writing a 2 to any of the two pwm[1-2]_enable sysfs nodes will put the | 
|  | 163 | PWM controller in automatic mode. There is only a single controller that | 
|  | 164 | controls both PWM outputs but each PWM output can be individually enabled and | 
|  | 165 | disabled. | 
|  | 166 |  | 
|  | 167 | Each PWM has 4 associated distinct output duty-cycles: full, high, low and | 
|  | 168 | off. Full and off are internally hard-wired to 255 (100%) and 0 (0%), | 
|  | 169 | respectively. High and low can be programmed via | 
|  | 170 | pwm[1-2]_auto_point[2-3]_pwm. Each PWM output can be associated with a | 
|  | 171 | different thermal input but - and here's the weird part - only one set of | 
|  | 172 | thermal thresholds exist that controls both PWMs output duty-cycles. The | 
|  | 173 | thermal thresholds are accessible via pwm[1-2]_auto_point[1-4]_temp. Note | 
|  | 174 | that even though there are 2 sets of 4 auto points each, they map to the same | 
|  | 175 | registers in the VT1211 and programming one set is sufficient (actually only | 
|  | 176 | the first set pwm1_auto_point[1-4]_temp is writable, the second set is | 
|  | 177 | read-only). | 
|  | 178 |  | 
|  | 179 | PWM Auto Point             PWM Output Duty-Cycle | 
|  | 180 | ------------------------------------------------ | 
|  | 181 | pwm[1-2]_auto_point4_pwm   full speed duty-cycle (hard-wired to 255) | 
|  | 182 | pwm[1-2]_auto_point3_pwm   high speed duty-cycle | 
|  | 183 | pwm[1-2]_auto_point2_pwm   low speed duty-cycle | 
|  | 184 | pwm[1-2]_auto_point1_pwm   off duty-cycle (hard-wired to 0) | 
|  | 185 |  | 
|  | 186 | Temp Auto Point             Thermal Threshold | 
|  | 187 | --------------------------------------------- | 
|  | 188 | pwm[1-2]_auto_point4_temp   full speed temp | 
|  | 189 | pwm[1-2]_auto_point3_temp   high speed temp | 
|  | 190 | pwm[1-2]_auto_point2_temp   low speed temp | 
|  | 191 | pwm[1-2]_auto_point1_temp   off temp | 
|  | 192 |  | 
|  | 193 | Long story short, the controller implements the following algorithm to set the | 
|  | 194 | PWM output duty-cycle based on the input temperature: | 
|  | 195 |  | 
|  | 196 | Thermal Threshold             Output Duty-Cycle | 
|  | 197 | (Rising Temp)           (Falling Temp) | 
|  | 198 | ---------------------------------------------------------- | 
|  | 199 | full speed duty-cycle   full speed duty-cycle | 
|  | 200 | full speed temp | 
|  | 201 | high speed duty-cycle   full speed duty-cycle | 
|  | 202 | high speed temp | 
|  | 203 | low speed duty-cycle    high speed duty-cycle | 
|  | 204 | low speed temp | 
|  | 205 | off duty-cycle          low speed duty-cycle | 
|  | 206 | off temp |