Skip to contentSkip to conversation

hwmon: Driver for Nuvoton NCT6775F, NCT6776F, and NCT6779D

PATCH Superseded Guenter Roeck 2 participants 4 messages Jul 27, 2012 – Aug 7, 2012

Subsystems (2)

v4 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Guenter Roeck

author maintainer
#0 lore raw

PATCH v4 1/1 hwmon: Driver for Nuvoton NCT6775F, NCT6776F, and NCT6779D

This driver will replace the w83627ehf driver for NCT6775F and NCT6776F, and provides support for NCT6779D.

---
v4:
- Fix "temp" array size (reported by smatch)
- Fix whitespace errors reported by "git am" (but not by checkpatch).
- Order register definition data by chip type
- Improve parameter alignment

v3:
- Variables pwm_temp, REG_TEMP_MON, and NCT6775_REG_TEMP_MON are not used, drop it.

v2:
- superio region access protection
- Add support for pwmX_step_enable attribute
- Improved documentation

 Documentation/hwmon/nct6775 |  187 +++
 drivers/hwmon/Kconfig       |   13 +
 drivers/hwmon/Makefile      |    1 +
 drivers/hwmon/nct6775.c     | 3887 +++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 4088 insertions(+)
 create mode 100644 Documentation/hwmon/nct6775
 create mode 100644 drivers/hwmon/nct6775.c
Documentation/hwmon/nct6775 new file +186 -0
new file mode 100644
index 0000000..09139a6
@@ -0,0 +1,187 @@
+Note
+==
+
+This driver supercedes the NCT6775F and NCT6776F support in the W83627EHF
+driver.
+
+Kernel driver NCT6775
+===========+
+Supported chips:
+  * Nuvoton NCT6775F/W83667HG-I
+    Prefix: 'nct6775'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: Available from Nuvoton upon request
+  * Nuvoton NCT6776F
+    Prefix: 'nct6776'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: Available from Nuvoton upon request
+  * Nuvoton NCT6779D
+    Prefix: 'nct6779'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: Available from Nuvoton upon request
+
+Authors:
+        Guenter Roeck <[email protected]>
+
+Description
+-----------
+
+This driver implements support for the Nuvoton NCT6775F, NCT6776F, and NCT6779D
+super I/O chips. We will refer to them collectively as Nuvoton chips.
+
+The chips implement up to 8 temperature sensors depending on the chip type and
+configuration. Temperatures used to control fan speed are reported separately.
+There are 4 to 5 fan rotation speed sensors, 8 to 15 analog voltage sensors,
+one VID, alarms with beep warnings (control unimplemented), and some automatic
+fan regulation strategies (plus manual fan control mode).
+
+The temperature sensor sources on all chips are configurable. temp4 and higher
+attributes are only reported if its temperature source differs from the
+temperature sources of the already reported temperature sensors.
+The configured source for each of the temperature sensors is provided
+in tempX_label.
+
+Temperatures are measured in degrees Celsius and measurement resolution is
+either 1 degC or 0.5 degC, depending on the temperature source and
+configuration. An alarm is triggered when the temperature gets higher than
+the high limit; it stays on until the temperature falls below the hysteresis
+value. Alarms are only supported for temp1, temp2, and temp3.
+
+Fan rotation speeds are reported in RPM (rotations per minute). An alarm is
+triggered if the rotation speed has dropped below a programmable limit. Fan
+readings can be divided by a programmable divider (1, 2, 4, 8, 16, 32, 64 or
+128) to give the readings more range or accuracy. The driver sets the most
+suitable fan divisor itself. Some fans might not be present because they
+share pins with other functions.
+
+Voltage sensors (also known as IN sensors) report their values in millivolts.
+An alarm is triggered if the voltage has crossed a programmable minimum
+or maximum limit.
+
+The driver supports automatic fan control mode known as Thermal Cruise.
+In this mode, the chip attempts to keep the measured temperature in a
+predefined temperature range. If the temperature goes out of range, fan
+is driven slower/faster to reach the predefined range again.
+
+The mode works for fan1-fan5.
+
+sysfs attributes
+----------------
+
+name - this is a standard hwmon device entry, it contains the name of
+       the device (see the prefix in the list of supported devices at
+       the top of this file)
+
+pwm[1-5] - this file stores PWM duty cycle or DC value (fan speed) in range:
+	   0 (stop) to 255 (full)
+
+pwm[1-5]_enable - this file controls mode of fan/temperature control:
+	* 0 Fan control disabled (fans set to maximum speed)
+	* 1 Manual mode, write to pwm file any value 0-255 (full speed)
+	* 2 "Thermal Cruise" mode
+	* 3 "Fan Speed Cruise" mode
+	* 4 "Smart Fan III" mode (NCT6775F only)
+	* 5 "Smart Fan IV" mode
+
+pwm[1-5]_mode - controls if output is PWM or DC level
+        * 0 DC output (0 - 12v)
+        * 1 PWM output
+
+Common fan control attributes
+-----------------------------
+
+pwm[1-5]_temp_sel	Temperature source. Value is temperature sensor index.
+			For example, select '1' for temp1_input.
+pwm[1-5]_weight_temp_sel
+			Secondary temperature source. Value is temperature
+			sensor index. For example, select '1' for temp1_input.
+pwm[1-5]_weight_enable	Set to 1 to enable secondary temperature control.
+
+If secondary temperature functionality is enabled, it is controlled with the
+following attributes.
+
+pwm[1-5]_weight_duty_step
+			Duty step size.
+pwm[1-5]_weight_temp_step
+			Temperature step size. With each step over
+			temp_step_base, the value of weight_duty_step is added
+			to the current pwm value.
+pwm[1-5]_weight_temp_step_base
+			Temperature at which secondary temperature control kicks
+			in.
+pwm[1-5]_weight_temp_step_tol
+			Temperature step tolerance or hysteresis. This is a
+			relative value.
+
+
+Thermal Cruise mode (2)
+-----------------------
+
+If the temperature is in the range defined by:
+
+pwm[1-5]_target		Target temperature, unit millidegree Celsius
+			(range 0 - 127000)
+pwm[1-5]_auto_temp1_hyst
+			Hysteresis, unit millidegree Celsius
+			Hysteresis value is relative to pwm[1-5]_auto_temp1.
+
+there are no changes to fan speed. Once the temperature leaves the interval,
+fan speed increases (temp is higher) or decreases if lower than desired,
+using the following steps and times.
+
+pwm[1-5]_start_output	fan pwm start value (range 1 - 255), to start fan
+			when the temperature is above defined range.
+pwm[1-5]_stop_output	lowest fan pwm (range 1 - 255) if temperature is below
+			the defined range.
+pwm[1-5]_stop_output_enable
+			Set to 1 to enable pwm[1-5]_stop_output. If disabled
+			(set to 0), the fan will stop if the temperature is
+			below the defined range.
+pwm[1-5]_step_up_time	milliseconds [ms] before fan speed is increased
+pwm[1-5]_step_down_time	milliseconds [ms] before fan speed is decreased
+pwm[1-5]_stop_time	how many milliseconds [ms] must elapse to switch
+			corresponding fan off (when the temperature was below
+			defined range).
+
+Speed Cruise mode (3)
+---------------------
+
+This modes tries to keep the fan speed constant.
+Untested; use at your own risk.
+
+Smart Fan IV mode (5)
+---------------------
+
+The fan is regulated to maintain a target temperature. There are five data
+points. Subsequent data points should be set to higher temperatures and higher
+pwm values to achieve higher fan speeds with increasing temperature. The last
+data point reflects critical temperature mode, in which the fans should run at
+full speed.
+
+pwm[1-5]_auto_point[1-5]_pwm
+			pwm value to be set if temperature reaches matching
+			temperature range.
+pwm[1-5]_auto_point[1-5]_temp
+			Temperature at which the matching pwm is enabled.
+pwm[1-5]_step_up_time	milliseconds [ms] before fan speed is increased
+pwm[1-5]_step_down_time	milliseconds [ms] before fan speed is decreased
+pwm[1-5]_step_enable	Set to 1 to enable fine grain speed control.
+			If disabled, pwm values will increase or decrease to
+			the values configured in the auto_point_pwm attributes.
+			If enabled, pwm values will be interpolated, resulting
+			in a more fine-grained fan control, resulting in a more
+			fine-grained fan control
+
+
+Usage Notes
+-----------
+
+On various ASUS boards with NCT6776F, it appears that CPUTIN is not really
+connected to anything and floats, or that it is connected to some non-standard
+temperature measurement device. As a result, the temperature reported on CPUTIN
+will not reflect a usable value. It often reports unreasonablyy high
+temperatures, and in some cases the reported temperature declines if the actual
+temperature increases (similar to the raw PECI temperature value - see PECI
+specification for details). CPUTIN should therefore be be ignored on ASUS
+boards. The CPU temperature on ASUS boards is reported from PECI 0.
drivers/hwmon/Kconfig +13 -0
index b0a2e4c..131a7a1 100644
@@ -844,6 +844,19 @@ config SENSORS_MCP3021
 	  This driver can also be built as a module.  If so, the module
 	  will be called mcp3021.
 
+config SENSORS_NCT6775
+	tristate "Nuvoton NCT6775F, NCT6776F, NCT6779D"
+	depends on !PPC && EXPERIMENTAL
+	select HWMON_VID
+	help
+	  If you say yes here you get support for the hardware monitoring
+	  functionality of the Nuvoton NCT6775F, NCT6776F, and NCT6779D
+	  Super-I/O chips. It replaces the w83627ehf driver for NCT6775F
+	  and NCT6776F.
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called nct6775.
+
 config SENSORS_NTC_THERMISTOR
 	tristate "NTC thermistor support"
 	depends on EXPERIMENTAL
drivers/hwmon/Makefile +1 -0
index 7aa9811..7856141 100644
@@ -99,6 +99,7 @@ obj-$(CONFIG_SENSORS_MAX6642)	+= max6642.o
 obj-$(CONFIG_SENSORS_MAX6650)	+= max6650.o
 obj-$(CONFIG_SENSORS_MC13783_ADC)+= mc13783-adc.o
 obj-$(CONFIG_SENSORS_MCP3021)	+= mcp3021.o
+obj-$(CONFIG_SENSORS_NCT6775)	+= nct6775.o
 obj-$(CONFIG_SENSORS_NTC_THERMISTOR)	+= ntc_thermistor.o
 obj-$(CONFIG_SENSORS_PC87360)	+= pc87360.o
 obj-$(CONFIG_SENSORS_PC87427)	+= pc87427.o
drivers/hwmon/nct6775.c new file +3875 -0

2,788 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 7 lines
1.7.9.7


_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Part of the diff is not shown here. View the raw message.

#1 lore raw

Hi Guenter,

Quoted 2 lines · Guenter Roeck

This driver will replace the w83627ehf driver for NCT6775F and NCT6776F, and provides support for NCT6779D.

Wow, what a huge driver. 1000 lines larger than the largest hwmon drivers we had so far.

This is actually too much for a complete review from me. For today I'll only comment on the documentation file and driver integration. I'll review the code itself another day, possibly in several parts.

This is already the 4th submission of this driver, and it did not receive any review so far. This is certainly related to the size of the driver. It would probably have helped to first submit a minimal version of the driver, and then add more features (in particular automatic fan speed control support) later.

Quoted 32 lines

---
v4:
- Fix "temp" array size (reported by smatch)
- Fix whitespace errors reported by "git am" (but not by checkpatch).
- Order register definition data by chip type
- Improve parameter alignment

v3:
- Variables pwm_temp, REG_TEMP_MON, and NCT6775_REG_TEMP_MON are not used, drop it.

v2:
- superio region access protection
- Add support for pwmX_step_enable attribute
- Improved documentation

 Documentation/hwmon/nct6775 |  187 +++
 drivers/hwmon/Kconfig       |   13 +
 drivers/hwmon/Makefile      |    1 +
 drivers/hwmon/nct6775.c     | 3887 +++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 4088 insertions(+)
 create mode 100644 Documentation/hwmon/nct6775
 create mode 100644 drivers/hwmon/nct6775.c
Documentation/hwmon/nct6775 new file +0 -0
new file mode 100644
index 0000000..09139a6

Documentation/hwmon/nct6775 line 5

Quoted 5 lines
Documentation/hwmon/nct6775 new file +5 -0
@@ -0,0 +1,187 @@
+Note
+==
+
+This driver supercedes the NCT6775F and NCT6776F support in the W83627EHF
+driver.

Spelling: supersedes. I'd write "w83627ehf driver" in lower-case. I'd also move this note at the end of the following section, as it is really only a note and not an intrinsic property.

Documentation/hwmon/nct6775

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+
+Kernel driver NCT6775
+===========

Remove two "=" for alignment.

Quoted 13 lines
Documentation/hwmon/nct6775 new file +13 -0
+
+Supported chips:
+  * Nuvoton NCT6775F/W83667HG-I
+    Prefix: 'nct6775'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: Available from Nuvoton upon request
+  * Nuvoton NCT6776F
+    Prefix: 'nct6776'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: Available from Nuvoton upon request
+  * Nuvoton NCT6779D
+    Prefix: 'nct6779'
+    Addresses scanned: ISA address retrieved from Super I/O registers

Documentation/hwmon/nct6775 line 30

Quoted 10 lines
Documentation/hwmon/nct6775 new file +10 -0
+    Datasheet: Available from Nuvoton upon request
+
+Authors:
+        Guenter Roeck <[email protected]>
+
+Description
+-----------
+
+This driver implements support for the Nuvoton NCT6775F, NCT6776F, and NCT6779D
+super I/O chips. We will refer to them collectively as Nuvoton chips.

You don't actually refer to them collectively anywhere in the document.

Documentation/hwmon/nct6775 line 33

Quoted 2 lines
Documentation/hwmon/nct6775 new file +2 -0
+
+The chips implement up to 8 temperature sensors depending on the chip type and

The comment at the beginning of the driver suggests up to 9 temperature sensors, and the code itself appears to support up to 10.

Quoted 6 lines
Documentation/hwmon/nct6775 new file +6 -0
+configuration. Temperatures used to control fan speed are reported separately.
+There are 4 to 5 fan rotation speed sensors, 8 to 15 analog voltage sensors,
+one VID, alarms with beep warnings (control unimplemented), and some automatic
+fan regulation strategies (plus manual fan control mode).
+
+The temperature sensor sources on all chips are configurable. temp4 and higher

Documentation/hwmon/nct6775 line 48

Quoted 10 lines
Documentation/hwmon/nct6775 new file +10 -0
+attributes are only reported if its temperature source differs from the
+temperature sources of the already reported temperature sensors.
+The configured source for each of the temperature sensors is provided
+in tempX_label.
+
+Temperatures are measured in degrees Celsius and measurement resolution is
+either 1 degC or 0.5 degC, depending on the temperature source and
+configuration. An alarm is triggered when the temperature gets higher than
+the high limit; it stays on until the temperature falls below the hysteresis
+value. Alarms are only supported for temp1, temp2, and temp3.

The code seems to support them on temp4, temp5 and temp6 too?

Documentation/hwmon/nct6775

Quoted 5 lines
Documentation/hwmon/nct6775 new file +5 -0
+
+Fan rotation speeds are reported in RPM (rotations per minute). An alarm is
+triggered if the rotation speed has dropped below a programmable limit. Fan
+readings can be divided by a programmable divider (1, 2, 4, 8, 16, 32, 64 or
+128) to give the readings more range or accuracy.

This sentence does only seem to apply to the NCT6775F.

Quoted 2 lines

The driver sets the most
+suitable fan divisor itself.

Please explain when and how it decides.

Quoted 10 lines

Some fans might not be present because they

Documentation/hwmon/nct6775 new file +9 -0
+share pins with other functions.
+
+Voltage sensors (also known as IN sensors) report their values in millivolts.
+An alarm is triggered if the voltage has crossed a programmable minimum
+or maximum limit.
+
+The driver supports automatic fan control mode known as Thermal Cruise.
+In this mode, the chip attempts to keep the measured temperature in a
+predefined temperature range. If the temperature goes out of range, fan

Documentation/hwmon/nct6775 line 73

Quoted 10 lines
Documentation/hwmon/nct6775 new file +10 -0
+is driven slower/faster to reach the predefined range again.
+
+The mode works for fan1-fan5.
+
+sysfs attributes
+----------------
+
+name - this is a standard hwmon device entry, it contains the name of
+       the device (see the prefix in the list of supported devices at
+       the top of this file)

As this is a completely standard attribute and not even related to fan speed control, I'm not sure why you list it here.

Documentation/hwmon/nct6775 line 76

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+
+pwm[1-5] - this file stores PWM duty cycle or DC value (fan speed) in range:
+	   0 (stop) to 255 (full)

0 is lowest speed for the given fan. It will not necessarily stop a 4-pin fan.

Documentation/hwmon/nct6775 line 80

Quoted 4 lines
Documentation/hwmon/nct6775 new file +4 -0
+
+pwm[1-5]_enable - this file controls mode of fan/temperature control:
+	* 0 Fan control disabled (fans set to maximum speed)
+	* 1 Manual mode, write to pwm file any value 0-255 (full speed)

"write to pwm[1-5]" would be clearer. Also the "(full speed)" is a little confusing and probably not needed, as the meaning of the pwm[1-5] is already explained above.

Documentation/hwmon/nct6775 line 87

Quoted 7 lines
Documentation/hwmon/nct6775 new file +7 -0
+	* 2 "Thermal Cruise" mode
+	* 3 "Fan Speed Cruise" mode
+	* 4 "Smart Fan III" mode (NCT6775F only)
+	* 5 "Smart Fan IV" mode
+
+pwm[1-5]_mode - controls if output is PWM or DC level
+        * 0 DC output (0 - 12v)

Capital "V", if you really want to mention a voltage value. I wouldn't, as I doubt the chip itself can output 12 V, being powered with 3.3 V. As a matter of fact, the DC output voltage scale tops at 3.3 V (NCT6775F datasheet version 1.7, PDF page 95.) I presume that an amplifier is needed between the chip's output pin and the fan supply pin in DC output mode.

Quoted 17 lines
Documentation/hwmon/nct6775 new file +17 -0
+        * 1 PWM output
+
+Common fan control attributes
+-----------------------------
+
+pwm[1-5]_temp_sel	Temperature source. Value is temperature sensor index.
+			For example, select '1' for temp1_input.
+pwm[1-5]_weight_temp_sel
+			Secondary temperature source. Value is temperature
+			sensor index. For example, select '1' for temp1_input.
+pwm[1-5]_weight_enable	Set to 1 to enable secondary temperature control.
+
+If secondary temperature functionality is enabled, it is controlled with the
+following attributes.
+
+pwm[1-5]_weight_duty_step
+			Duty step size.

Documentation/hwmon/nct6775 line 114

Quoted 10 lines
Documentation/hwmon/nct6775 new file +10 -0
+pwm[1-5]_weight_temp_step
+			Temperature step size. With each step over
+			temp_step_base, the value of weight_duty_step is added
+			to the current pwm value.
+pwm[1-5]_weight_temp_step_base
+			Temperature at which secondary temperature control kicks
+			in.
+pwm[1-5]_weight_temp_step_tol
+			Temperature step tolerance or hysteresis. This is a
+			relative value.

Wow, this is incredibly complex. Took me two reads to understand what it is for and a third to understand how I would program the settings. The standard interface has a single file for roughly the same:
pwm[1-*]_auto_channels_temp.

Are the above settings really that useful? Do the BIOS out there actually make use of these?

Documentation/hwmon/nct6775

Quoted 2 lines
Documentation/hwmon/nct6775 new file +2 -0
+
+

Your spacing between sections is inconsistent, sometimes you have a single blank line and sometimes you have two.

Documentation/hwmon/nct6775 line 123

Quoted 7 lines
Documentation/hwmon/nct6775 new file +7 -0
+Thermal Cruise mode (2)
+-----------------------
+
+If the temperature is in the range defined by:
+
+pwm[1-5]_target		Target temperature, unit millidegree Celsius
+			(range 0 - 127000)

This looks wrong. The name should reflect the fact that the target is a temperature value. As a matter of fact we use fan[1-*]_target when the target is a fan speed value.

Documentation/hwmon/nct6775 line 126

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+pwm[1-5]_auto_temp1_hyst
+			Hysteresis, unit millidegree Celsius
+			Hysteresis value is relative to pwm[1-5]_auto_temp1.

I don't see these two attributes in the driver, instead I see pwm[1-5]_auto_point1_temp and pwm[1-5]_auto_point1_temp_hyst. Which I prefer because they are part of the standard interface, while the documented ones aren't.

I am still confused by the fact that the datasheet mentions a temperature target and a tolerance, not a high limit with an hysteresis. While the concepts are similar, there are implementation differences. Most notably, tolerance is reported as a relative value to user-space, while hysteresis must always be reported as an absolute value. Also, hysteresis typically applies on one side of a given limit, while tolerance applies on both. Here you cleanly want to expose a tolerance, exactly as the w83627ehf driver was doing.

Documentation/hwmon/nct6775 line 130

Quoted 4 lines
Documentation/hwmon/nct6775 new file +4 -0
+
+there are no changes to fan speed. Once the temperature leaves the interval,
+fan speed increases (temp is higher) or decreases if lower than desired,
+using the following steps and times.

The compact wording hurts readability somewhat. What about the more verbose:

"Once the temperature leaves the interval, fan speed increases (if temperature is higher that desired) or decreases (if temperature is lower than desired), using the following steps and times."

"steps and time" is a strange formulation, BTW, I see time settings but no step settings.

Documentation/hwmon/nct6775 line 139

Quoted 9 lines
Documentation/hwmon/nct6775 new file +9 -0
+
+pwm[1-5]_start_output	fan pwm start value (range 1 - 255), to start fan
+			when the temperature is above defined range.
+pwm[1-5]_stop_output	lowest fan pwm (range 1 - 255) if temperature is below
+			the defined range.
+pwm[1-5]_stop_output_enable
+			Set to 1 to enable pwm[1-5]_stop_output. If disabled
+			(set to 0), the fan will stop if the temperature is
+			below the defined range.

I don't much like the name pwm[1-5]_stop_output, even though I know this is the name found in several datasheets. Using the word "stop" to say that the fan should actually keep spinning is quite confusing. What the value really describes is a minimum PWM output. Maybe pwm[1-5]_floor would be a better name.

Also the "output" part of the names seems redundant.

These attribute are relatively common, several hwmon drivers already implement them. Maybe this would be the right time to come up with standard names and add these to Documentation/hwmon/sysfs-interface?

Documentation/hwmon/nct6775 line 144

Quoted 5 lines
Documentation/hwmon/nct6775 new file +5 -0
+pwm[1-5]_step_up_time	milliseconds [ms] before fan speed is increased
+pwm[1-5]_step_down_time	milliseconds [ms] before fan speed is decreased
+pwm[1-5]_stop_time	how many milliseconds [ms] must elapse to switch
+			corresponding fan off (when the temperature was below
+			defined range).

I don't think the [ms] are needed.

The step_up_time and step_down_time attributes don't actually seem to affect thermal cruise mode, according to the NCT6775F datasheet. I think it only affects Smart Fan III and/or IV.

Documentation/hwmon/nct6775 line 155

Quoted 11 lines
Documentation/hwmon/nct6775 new file +11 -0
+
+Speed Cruise mode (3)
+---------------------
+
+This modes tries to keep the fan speed constant.
+Untested; use at your own risk.
+
+Smart Fan IV mode (5)
+---------------------
+
+The fan is regulated to maintain a target temperature. There are five data

This definition makes it look like thermal cruise mode, while it is different. Smart Fan IV is based on a PWM vs. temperature response curve (or trip points.)

Documentation/hwmon/nct6775 line 163

Quoted 8 lines
Documentation/hwmon/nct6775 new file +8 -0
+points. Subsequent data points should be set to higher temperatures and higher
+pwm values to achieve higher fan speeds with increasing temperature. The last
+data point reflects critical temperature mode, in which the fans should run at
+full speed.
+
+pwm[1-5]_auto_point[1-5]_pwm
+			pwm value to be set if temperature reaches matching
+			temperature range.

The driver actually exposes up to 7 trip points.

Documentation/hwmon/nct6775 line 165

Quoted 2 lines
Documentation/hwmon/nct6775 new file +2 -0
+pwm[1-5]_auto_point[1-5]_temp
+			Temperature at which the matching pwm is enabled.

I suggest "over which" instead of "at which", for clarity.

Documentation/hwmon/nct6775

Quoted 2 lines
Documentation/hwmon/nct6775 new file +2 -0
+pwm[1-5]_step_up_time	milliseconds [ms] before fan speed is increased
+pwm[1-5]_step_down_time	milliseconds [ms] before fan speed is decreased

Here again, [ms] is redundant.

Quoted 1 line

+pwm[1-5]_step_enable Set to 1 to enable fine grain speed control.

The attribute name doesn't match the function. It's about interpolation / smoothing, the attribute name should reflect that. This attribute is also a candidate for standardization.

Documentation/hwmon/nct6775 line 171

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+			If disabled, pwm values will increase or decrease to
+			the values configured in the auto_point_pwm attributes.
+			If enabled, pwm values will be interpolated, resulting

I presume temperature values are interpolated too.

Documentation/hwmon/nct6775 line 173

Quoted 2 lines
Documentation/hwmon/nct6775 new file +2 -0
+			in a more fine-grained fan control, resulting in a more
+			fine-grained fan control

Duplicate end of sentence.

Documentation/hwmon/nct6775 line 182

Quoted 9 lines
Documentation/hwmon/nct6775 new file +9 -0
+
+
+Usage Notes
+-----------
+
+On various ASUS boards with NCT6776F, it appears that CPUTIN is not really
+connected to anything and floats, or that it is connected to some non-standard
+temperature measurement device. As a result, the temperature reported on CPUTIN
+will not reflect a usable value. It often reports unreasonablyy high

Spelling: unreasonably.

Quoted 11 lines
Documentation/hwmon/nct6775 new file +4 -0
+temperatures, and in some cases the reported temperature declines if the actual
+temperature increases (similar to the raw PECI temperature value - see PECI
+specification for details). CPUTIN should therefore be be ignored on ASUS
+boards. The CPU temperature on ASUS boards is reported from PECI 0.
drivers/hwmon/Kconfig +0 -0
index b0a2e4c..131a7a1 100644
@@ -844,6 +844,19 @@ config SENSORS_MCP3021
 	  This driver can also be built as a module.  If so, the module
 	  will be called mcp3021.

drivers/hwmon/Kconfig config SENSORS_MCP3021 · line 855

Quoted 10 lines
drivers/hwmon/Kconfig +9 -0
 
+config SENSORS_NCT6775
+	tristate "Nuvoton NCT6775F, NCT6776F, NCT6779D"
+	depends on !PPC && EXPERIMENTAL
+	select HWMON_VID
+	help
+	  If you say yes here you get support for the hardware monitoring
+	  functionality of the Nuvoton NCT6775F, NCT6776F, and NCT6779D
+	  Super-I/O chips. It replaces the w83627ehf driver for NCT6775F
+	  and NCT6776F.

It isn't clear what "It" refers to in the last sentence.

Quoted 19 lines
drivers/hwmon/Kconfig +4 -0
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called nct6775.
+
 config SENSORS_NTC_THERMISTOR
 	tristate "NTC thermistor support"
 	depends on EXPERIMENTAL
drivers/hwmon/Makefile +1 -0
index 7aa9811..7856141 100644
@@ -99,6 +99,7 @@ obj-$(CONFIG_SENSORS_MAX6642)	+= max6642.o
 obj-$(CONFIG_SENSORS_MAX6650)	+= max6650.o
 obj-$(CONFIG_SENSORS_MC13783_ADC)+= mc13783-adc.o
 obj-$(CONFIG_SENSORS_MCP3021)	+= mcp3021.o
+obj-$(CONFIG_SENSORS_NCT6775)	+= nct6775.o
 obj-$(CONFIG_SENSORS_NTC_THERMISTOR)	+= ntc_thermistor.o
 obj-$(CONFIG_SENSORS_PC87360)	+= pc87360.o
 obj-$(CONFIG_SENSORS_PC87427)	+= pc87427.o
Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
author maintainer
#2 lore raw
In reply to Guenter Roeck
Quoted 13 lines · Jean Delvare

Hi Guenter,

Quoted 2 lines · Guenter Roeck

This driver will replace the w83627ehf driver for NCT6775F and NCT6776F, and provides support for NCT6779D.

Wow, what a huge driver. 1000 lines larger than the largest hwmon drivers we had so far.

This is actually too much for a complete review from me. For today I'll only comment on the documentation file and driver integration. I'll review the code itself another day, possibly in several parts.

Quoted 6 lines

This is already the 4th submission of this driver, and it did not receive any review so far. This is certainly related to the size of the driver. It would probably have helped to first submit a minimal version of the driver, and then add more features (in particular automatic fan speed control support) later.

Hi Jean,

thanks a lot for the review and your time.

Yes, I know, it has been growing. I can split it into multiple pieces if you prefer / if you think it helps (I would also split out NCT6776 and NCT6779 support in that case).

Much of your comments are straightforward, so I won't comment on those. Some of it was due to cut-and-paste from the w83627ehf documentation, so maybe I should address it there as well.

Quoted 7 lines

Review

Quoted 3 lines
(quoted diff) +2 -0
+
+The chips implement up to 8 temperature sensors depending on the chip type and

The comment at the beginning of the driver suggests up to 9 temperature sensors, and the code itself appears to support up to 10.

Yes, that is a bit tricky with those chips, since they have so many registers and temperature sources. I'll go through the specifications and make sure the numbers are at least consistent.

Quoted 22 lines
Quoted 22 lines
Documentation/hwmon/nct6775 new file +22 -0
+Common fan control attributes
+-----------------------------
+
+pwm[1-5]_temp_sel	Temperature source. Value is temperature sensor index.
+			For example, select '1' for temp1_input.
+pwm[1-5]_weight_temp_sel
+			Secondary temperature source. Value is temperature
+			sensor index. For example, select '1' for temp1_input.
+pwm[1-5]_weight_enable	Set to 1 to enable secondary temperature control.
+
+If secondary temperature functionality is enabled, it is controlled with the
+following attributes.
+
+pwm[1-5]_weight_duty_step
+			Duty step size.
+pwm[1-5]_weight_temp_step
+			Temperature step size. With each step over
+			temp_step_base, the value of weight_duty_step is added
+			to the current pwm value.
+pwm[1-5]_weight_temp_step_base
+			Temperature at which secondary temperature control kicks
+			in.
Quoted 9 lines

Documentation/hwmon/nct6775 line 114

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+pwm[1-5]_weight_temp_step_tol
+			Temperature step tolerance or hysteresis. This is a
+			relative value.

Wow, this is incredibly complex. Took me two reads to understand what it is for and a third to understand how I would program the settings. The standard interface has a single file for roughly the same:
pwm[1-*]_auto_channels_temp.

Problem with pwm[1-*]_auto_channels_temp is that it is a non-weighted bit mask. I could use it for pwm[1-5]_temp_sel.

Quoted 3 lines

Are the above settings really that useful? Do the BIOS out there actually make use of these?

Yes, my Intel board configures and enables the weight settings.

/sys/class/hwmon/hwmon1/device/pwm1_weight_duty_step:36 /sys/class/hwmon/hwmon1/device/pwm1_weight_enable:1 /sys/class/hwmon/hwmon1/device/pwm1_weight_temp_sel:1 /sys/class/hwmon/hwmon1/device/pwm1_weight_temp_step:2000 /sys/class/hwmon/hwmon1/device/pwm1_weight_temp_step_base:70000 /sys/class/hwmon/hwmon1/device/pwm1_weight_temp_step_tol:1000

This is why I added it in.

Quoted 12 lines

Documentation/hwmon/nct6775 line 123

Quoted 7 lines
Documentation/hwmon/nct6775 new file +7 -0
+Thermal Cruise mode (2)
+-----------------------
+
+If the temperature is in the range defined by:
+
+pwm[1-5]_target		Target temperature, unit millidegree Celsius
+			(range 0 - 127000)

This looks wrong. The name should reflect the fact that the target is a temperature value. As a matter of fact we use fan[1-*]_target when the target is a fan speed value.

Ah yes, and worse, the register is overloaded and means the target speed in speed cruise mode. I'll have to address that. What should I use ?
pwm[1-5]_target_temp for thermal cruise mode
fan[1-*]_target for speed cruise mode

Quoted 9 lines

Documentation/hwmon/nct6775 line 126

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+pwm[1-5]_auto_temp1_hyst
+			Hysteresis, unit millidegree Celsius
+			Hysteresis value is relative to pwm[1-5]_auto_temp1.

I don't see these two attributes in the driver, instead I see pwm[1-5]_auto_point1_temp and pwm[1-5]_auto_point1_temp_hyst. Which I prefer because they are part of the standard interface, while the documented ones aren't.

I changed that in the code to follow the ABI and forgot to update Documentation.

Quoted 9 lines

I am still confused by the fact that the datasheet mentions a temperature target and a tolerance, not a high limit with an hysteresis. While the concepts are similar, there are implementation differences. Most notably, tolerance is reported as a relative value to user-space, while hysteresis must always be reported as an absolute value. Also, hysteresis typically applies on one side of a given limit, while tolerance applies on both. Here you cleanly want to expose a tolerance, exactly as the w83627ehf driver was doing.

I went back and forth about that. Yes, tolerance is much better. I started to use hyst only because that is the standard attribute. I'll go back to tolerance; that makes the code simpler anyway.

Quoted 8 lines

Documentation/hwmon/nct6775 line 130

Quoted 4 lines
Documentation/hwmon/nct6775 new file +4 -0
+
+there are no changes to fan speed. Once the temperature leaves the interval,
+fan speed increases (temp is higher) or decreases if lower than desired,
+using the following steps and times.

The compact wording hurts readability somewhat. What about the more verbose:

Quoted 7 lines

"Once the temperature leaves the interval, fan speed increases (if temperature is higher that desired) or decreases (if temperature is lower than desired), using the following steps and times."

"steps and time" is a strange formulation, BTW, I see time settings but no step settings.

Correct, there are no step settings.

How about that:
"Once the temperature leaves the interval, fan speed increases (if temperature is higher that desired) or decreases (if temperature is lower than desired), using the following limits and time intervals."

Quoted 16 lines

Documentation/hwmon/nct6775 line 139

Quoted 9 lines
Documentation/hwmon/nct6775 new file +9 -0
+
+pwm[1-5]_start_output	fan pwm start value (range 1 - 255), to start fan
+			when the temperature is above defined range.
+pwm[1-5]_stop_output	lowest fan pwm (range 1 - 255) if temperature is below
+			the defined range.
+pwm[1-5]_stop_output_enable
+			Set to 1 to enable pwm[1-5]_stop_output. If disabled
+			(set to 0), the fan will stop if the temperature is
+			below the defined range.

I don't much like the name pwm[1-5]_stop_output, even though I know this is the name found in several datasheets. Using the word "stop" to say that the fan should actually keep spinning is quite confusing. What the value really describes is a minimum PWM output. Maybe pwm[1-5]_floor would be a better name.

Quoted 2 lines

Also the "output" part of the names seems redundant.

How about
pwm[1-5]_start fan pwm start value (range 1 - 255), to start fan
when the temperature is above defined range.
pwm[1-5]_floor lowest fan pwm (range 1 - 255) if temperature is below
the defined range.
pwm[1-5]_floor_enable Set to 1 to enable pwm[1-5]_floor_output.
If disabled (set to 0), the fan will stop if the temperature is below the defined range.

Quoted 4 lines

These attribute are relatively common, several hwmon drivers already implement them. Maybe this would be the right time to come up with standard names and add these to Documentation/hwmon/sysfs-interface?

Sure.

Quoted 8 lines

Documentation/hwmon/nct6775 line 144

Quoted 5 lines
Documentation/hwmon/nct6775 new file +5 -0
+pwm[1-5]_step_up_time	milliseconds [ms] before fan speed is increased
+pwm[1-5]_step_down_time	milliseconds [ms] before fan speed is decreased
+pwm[1-5]_stop_time	how many milliseconds [ms] must elapse to switch
+			corresponding fan off (when the temperature was below
+			defined range).

I don't think the [ms] are needed.

Quoted 4 lines

The step_up_time and step_down_time attributes don't actually seem to affect thermal cruise mode, according to the NCT6775F datasheet. I think it only affects Smart Fan III and/or IV.

Table 8-7 ?

Quoted 9 lines

Documentation/hwmon/nct6775 line 155

Quoted 4 lines
Documentation/hwmon/nct6775 new file +4 -0
+Smart Fan IV mode (5)
+---------------------
+
+The fan is regulated to maintain a target temperature. There are five data

This definition makes it look like thermal cruise mode, while it is different. Smart Fan IV is based on a PWM vs. temperature response curve (or trip points.)

Is this better ?

"This mode offers multiple slopes to control the fan speed. The slopes can be controlled by setting the pwm and temperature attributes. When the temperature rises, the chip will calculate the DC/PWM output based on the current slope. There are up to seven data points depending on the chip type. Subsequent data points should be set to higher temperatures and higher pwm values to achieve higher fan speeds with increasing temperature. The last data point reflects critical temperature mode, in which the fans should run at full speed."

Quoted 7 lines
Quoted 2 lines

+pwm[1-5]_step_enable Set to 1 to enable fine grain speed control.

The attribute name doesn't match the function. It's about interpolation / smoothing, the attribute name should reflect that. This attribute is also a candidate for standardization.

pwm[1-5]_interpolate ?

Quoted 5 lines

Documentation/hwmon/nct6775 line 171

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+			If disabled, pwm values will increase or decrease to
+			the values configured in the auto_point_pwm attributes.
+			If enabled, pwm values will be interpolated, resulting

I presume temperature values are interpolated too.

pwm values are interpolated based on the actual temperature. For example, if two steps are

	{ pwm = 100, temp = 50 }
	{ pwm = 200, temp = 60 }

and the temperature is 55 degrees C, the pwm value will be set to 150. Not sure how to explain that. Any idea ?

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#3 lore raw
In reply to Guenter Roeck

Fast reply as I am short on time...

Quoted 3 lines · Guenter Roeck

Yes, I know, it has been growing. I can split it into multiple pieces if you prefer / if you think it helps (I would also split out NCT6776 and NCT6779 support in that case).

Smaller chunks are easier to review, but it is also easier if you don't have to know what's coming next in order to review every patch. OTOH the goal is not to force you to write intermediate code just to make reviewers happy. So I presume there's a balance to be found.

Quoted 3 lines

Much of your comments are straightforward, so I won't comment on those. Some of it was due to cut-and-paste from the w83627ehf documentation, so maybe I should address it there as well.

I noticed this, yes.

Quoted 12 lines
Quoted 11 lines

Documentation/hwmon/nct6775 line 123

Quoted 7 lines
Documentation/hwmon/nct6775 new file +7 -0
+Thermal Cruise mode (2)
+-----------------------
+
+If the temperature is in the range defined by:
+
+pwm[1-5]_target		Target temperature, unit millidegree Celsius
+			(range 0 - 127000)

This looks wrong. The name should reflect the fact that the target is a temperature value. As a matter of fact we use fan[1-*]_target when the target is a fan speed value.

Quoted 4 lines

Ah yes, and worse, the register is overloaded and means the target speed in speed cruise mode. I'll have to address that. What should I use ?
pwm[1-5]_target_temp for thermal cruise mode
fan[1-*]_target for speed cruise mode

The recurrent problem with naming is that it all depends on what the chip attaches the attribute to. The same attribute can be attached to a temperature channel, a fan input or PWM output, depending on the hardware implementation, so the name that seems better because it is more generic may not actually always fit.

My first choice for thermal cruise mode would be temp[1-*]_target, however it turns out that the Nuvoton chips attach the temperature target to PWM outputs, not temperature channels directly, so it won't work. Your proposal makes more sense.

Unfortunately these implementation differences between chips make it very difficult to define a standard sysfs interface for automatic fan speed control :(

Quoted 14 lines
Quoted 7 lines

Documentation/hwmon/nct6775 line 130

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+there are no changes to fan speed. Once the temperature leaves the interval,
+fan speed increases (temp is higher) or decreases if lower than desired,
+using the following steps and times.

The compact wording hurts readability somewhat. What about the more verbose:

Quoted 6 lines

"Once the temperature leaves the interval, fan speed increases (if temperature is higher that desired) or decreases (if temperature is lower than desired), using the following steps and times."

"steps and time" is a strange formulation, BTW, I see time settings but no step settings.

Quoted 6 lines

Correct, there are no step settings.

How about that:
"Once the temperature leaves the interval, fan speed increases (if temperature is higher that desired) or decreases (if temperature is lower than desired), using the following limits and time intervals."

Looks good.

Quoted 16 lines
Quoted 16 lines

Documentation/hwmon/nct6775 line 139

Quoted 9 lines
Documentation/hwmon/nct6775 new file +9 -0
+
+pwm[1-5]_start_output	fan pwm start value (range 1 - 255), to start fan
+			when the temperature is above defined range.
+pwm[1-5]_stop_output	lowest fan pwm (range 1 - 255) if temperature is below
+			the defined range.
+pwm[1-5]_stop_output_enable
+			Set to 1 to enable pwm[1-5]_stop_output. If disabled
+			(set to 0), the fan will stop if the temperature is
+			below the defined range.

I don't much like the name pwm[1-5]_stop_output, even though I know this is the name found in several datasheets. Using the word "stop" to say that the fan should actually keep spinning is quite confusing. What the value really describes is a minimum PWM output. Maybe pwm[1-5]_floor would be a better name.

Quoted 8 lines
Quoted 1 line

Also the "output" part of the names seems redundant.

How about
pwm[1-5]_start fan pwm start value (range 1 - 255), to start fan
when the temperature is above defined range.
pwm[1-5]_floor lowest fan pwm (range 1 - 255) if temperature is below
the defined range.
pwm[1-5]_floor_enable Set to 1 to enable pwm[1-5]_floor_output.

s/_output//

Quoted 2 lines

If disabled (set to 0), the fan will stop if the temperature is below the defined range.

The problem is that I can actually think of 3 behaviors, not 2:
* If fan speed should be below floor, it is set to floor value.
* If fan speed should be below floor, it is set to 0.
* No special handling for low speed values.

One could argue that the third behavior is better described by not implementing the pwm[1-*]_floor* attributes at all, but what about chips which support all behaviors?

Quoted 5 lines
Quoted 3 lines

The step_up_time and step_down_time attributes don't actually seem to affect thermal cruise mode, according to the NCT6775F datasheet. I think it only affects Smart Fan III and/or IV.

Table 8-7 ?

You're right, not sure how I missed that.

Quoted 12 lines
Quoted 10 lines

Documentation/hwmon/nct6775 line 155

Quoted 5 lines
Documentation/hwmon/nct6775 new file +4 -0
+Smart Fan IV mode (5)
+---------------------
+
+The fan is regulated to maintain a target temperature. There are five data

This definition makes it look like thermal cruise mode, while it is different. Smart Fan IV is based on a PWM vs. temperature response curve (or trip points.)

Is this better ?

Quoted 7 lines

"This mode offers multiple slopes to control the fan speed. The slopes can be controlled by setting the pwm and temperature attributes. When the temperature rises, the chip will calculate the DC/PWM output based on the current slope. There are up to seven data points depending on the chip type. Subsequent data points should be set to higher temperatures and higher pwm values to achieve higher fan speeds with increasing temperature. The last data point reflects critical temperature mode, in which the fans should run at full speed."

Sounds good, although the notion of slope only applies in interpolation mode. Without interpolation, it's really only a set of trip points.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

+pwm[1-5]_step_enable Set to 1 to enable fine grain speed control.

The attribute name doesn't match the function. It's about interpolation / smoothing, the attribute name should reflect that. This attribute is also a candidate for standardization.

pwm[1-5]_interpolate ?

Agreed.

Quoted 6 lines
Quoted 5 lines

Documentation/hwmon/nct6775 line 171

Quoted 3 lines
Documentation/hwmon/nct6775 new file +3 -0
+			If disabled, pwm values will increase or decrease to
+			the values configured in the auto_point_pwm attributes.
+			If enabled, pwm values will be interpolated, resulting

I presume temperature values are interpolated too.

Quoted 6 lines

pwm values are interpolated based on the actual temperature. For example, if two steps are

	{ pwm = 100, temp = 50 }
	{ pwm = 200, temp = 60 }

and the temperature is 55 degrees C, the pwm value will be set to 150. Not sure how to explain that. Any idea ?

Your explanation was actually just fine, I was probably just tired.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors