Message-ID: <53E31694.8050107@gmail.com>
Date:	Thu, 07 Aug 2014 08:03:00 +0200
From:	Goffredo Baroncelli <kreijack@...il.com>
To:	Guenter Roeck <linux@...ck-us.net>
CC:	Benjamin Herrenschmidt <benh@...nel.crashing.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Jean Delvare <jdelvare@...e.de>
Subject: Re: [PATCH 5/5] Export the temperatures via hwmon

On 08/07/2014 01:18 AM, Guenter Roeck wrote:
> On Wed, Aug 06, 2014 at 09:05:03PM +0000, Goffredo Baroncelli wrote:
>> From: Goffredo Baroncelli <kreijack@...wnd.it>
>>
>> Export the temperature via the hwmon subsystem.
>> See the list below for the sensors exported:
>>
>> $ cd /sys/devices/temperature/hwmon/hwmon0
>> $ echo "name: $(cat name)"; for i in temp*; do echo "$i: $(cat $i)"; done
>> name: therm_windtunnel
>> temp1_input: 59312
>> temp1_label: CPU
>> temp2_input: 36750
>> temp2_label: Case
>> temp3_input: 37750
>> temp3_label: Case2
>>
> Makes me wonder why you don't report the fan speed through hwmon.

See the first email. Basically when I start to read the fan speed,
it became irregular, so I stopped the test. I don't know
if it is a my faulted hardware, or an hw design problem.
The fan is a 2 wire fan.

> 
>> The Case2 temperature is the sensor temperature inside the adm1030.
>>
> There are standard hwmon drivers for lm75, ds1775, and adm1030.
> Personally I think it would make more sense to use those directly.
> But I guess that would be for another day.
> 
>> Signed-off-by: Goffredo Baroncelli <kreijack@...ind.it>
>> ---
>>  drivers/macintosh/therm_windtunnel.c | 103 +++++++++++++++++++++++++++++++++--
>>  1 file changed, 99 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/macintosh/therm_windtunnel.c b/drivers/macintosh/therm_windtunnel.c
>> index b6cba98..a6757d7 100644
>> --- a/drivers/macintosh/therm_windtunnel.c
>> +++ b/drivers/macintosh/therm_windtunnel.c
>> @@ -37,6 +37,8 @@
>>  #include <linux/init.h>
>>  #include <linux/kthread.h>
>>  #include <linux/of_platform.h>
>> +#include <linux/hwmon.h>
>> +#include <linux/hwmon-sysfs.h>
>>  
>>  #include <asm/prom.h>
>>  #include <asm/machdep.h>
>> @@ -58,9 +60,12 @@ static struct {
>>  	struct i2c_client	*thermostat;
>>  	struct i2c_client	*fan;
>>  
>> +	struct device		*hwmon;
>> +
>>  	int			overheat_temp;		/* 100% fan at this temp */
>>  	int			overheat_hyst;
>>  	int			temp;
>> +	int			casetemp2;
>>  	int			casetemp;
>>  	int			fan_level;		/* active fan_table setting */
>>  
>> @@ -120,6 +125,75 @@ static DEVICE_ATTR(case_temperature, S_IRUGO, show_case_temperature, NULL );
>>  static DEVICE_ATTR(fan_level, S_IRUGO, show_fan_level, NULL);
>>  
>>  
>> +static ssize_t
>> +show_temp1(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "%d%03d\n", x.temp>>8, (x.temp & 0xff)*1000>>8);
> 
> I personally prefer if code follows coding style, with spaces around operands,
> and if calculations are less cryptic.
> 
>> +}
>> +
>> +static ssize_t
>> +show_temp1_label(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "CPU\n");
>> +}
>> +
>> +static ssize_t
>> +show_temp2(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "%d%03d\n", x.casetemp>>8,
>> +				(x.casetemp & 0xff)*1000>>8);
>> +}
>> +
>> +static ssize_t
>> +show_temp2_label(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "Case\n");
>> +}
>> +
>> +
>> +static ssize_t
>> +show_temp3(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "%d%03d\n", x.casetemp2>>8,
>> +				(x.casetemp2 & 0xff)*1000>>8);
> 
> ... and aligned second lines.
> 
>> +}
>> +
>> +static ssize_t
>> +show_temp3_label(struct device *dev, struct device_attribute *attr, char *buf)
>> +{
>> +	return sprintf(buf, "Case2\n");
>> +}
>> +
>> +#define TWT_ATTRIB(name, func) \
>> +	static SENSOR_DEVICE_ATTR(name##_input, S_IRUGO, func, NULL, 0); \
>> +	static SENSOR_DEVICE_ATTR(name##_label, S_IRUGO, func##_label, \
>> +					NULL, 0)
>> +
>> +TWT_ATTRIB(temp1, show_temp1);
>> +TWT_ATTRIB(temp2, show_temp2);
>> +TWT_ATTRIB(temp3, show_temp3);
>> +
>> +
> A single empty line should be sufficient. Personally I am also not
> a fan of such macros (and neither is checkpatch), and prefer the use
> of direct macros as less confusing.

True, I put a three line macro to avoid to write three line... it
doesn't make sense.

> 
> Also, using three different functions for three different attributes
> where the only difference is the variable name defeats the purpose
> of using SENSOR_DEVICE_ATTR, which has an index variable for exactly 
> that reason. I would suggest to either use an indexed array to access
> the temperatures, or use DEVICE_ATTR.
> 
>> +static struct attribute  *therm_windtunnel_attributes[] = {
>> +	&sensor_dev_attr_temp1_input.dev_attr.attr,
>> +	&sensor_dev_attr_temp1_label.dev_attr.attr,
>> +	&sensor_dev_attr_temp2_input.dev_attr.attr,
>> +	&sensor_dev_attr_temp2_label.dev_attr.attr,
>> +	&sensor_dev_attr_temp3_input.dev_attr.attr,
>> +	&sensor_dev_attr_temp3_label.dev_attr.attr,
>> +
>> +	NULL,
>> +};
>> +
>> +static const struct attribute_group therm_windtunnel_attr_group = {
>> +	.attrs  = therm_windtunnel_attributes,
>> +};
>> +
>> +static const struct attribute_group *therm_windtunnel_attr_groups[] = {
>> +	&therm_windtunnel_attr_group,
>> +	NULL,
>> +};
> 
> ATTRIBUTE_GROUPS() is a nice and useful macro.
I will give a look to this macro.
> 
>> +
>>  /************************************************************************/
>>  /*	controller thread						*/
>>  /************************************************************************/
>> @@ -172,16 +246,23 @@ tune_fan( int fan_setting )
>>  static void
>>  poll_temp( void )
>>  {
>> -	int temp, i, level, casetemp, tempchanged;
>> +	int temp, i, level, casetemp, tempchanged, casetemp2, reg06;
>>  
>> +	/* temperature read from ds1775 */
>>  	temp = read_reg( x.thermostat, 0, 2 );
>>  
>>  	/* this actually occurs when the computer is loaded */
>>  	if( temp < 0 )
>>  		return;
>>  
>> -	casetemp = read_reg(x.fan, 0x0b, 1) << 8;
>> -	casetemp |= (read_reg(x.fan, 0x06, 1) & 0x7) << 5;
>> +	/*
>> +	 * temperatures read from the adm1030
>> +	 * casetemp is the external temperature sensor
>> +	 * casetemp2 is the internal temperature sensor
>> +	 */
> 
> What if there is no adm1030 ? Or is that always there ?
> Just wondering.

poll_temp() is called by the daemon, which is started only after 
the discovering of both the adm1030 and the ds1775: (see patch #1
at the function try_start_control_loop())

> 
>> +	reg06 = read_reg(x.fan, 0x06, 1);
>> +	casetemp = (read_reg(x.fan, 0x0b, 1) << 8) | (reg06 & 0x07 << 5);
>> +	casetemp2 = (read_reg(x.fan, 0x0a, 1) << 8) | (reg06 & 0xc0);
>>  
>>  	level = -1;
>>  	for( i=0; (temp & 0xffff) > fan_table[i].temp ; i++ )
>> @@ -200,13 +281,16 @@ poll_temp( void )
>>  	 * if verbose >0 log each fan tuning
>>  	 * if verbose >1 log each cpu temperature change
>>  	 */
>> -	tempchanged = x.temp != temp || x.casetemp != casetemp;
>> +	tempchanged = x.temp != temp || x.casetemp != casetemp ||
>> +		x.casetemp2 != casetemp2;
>>  	if ((verbose > 1 && tempchanged) ||
>>  	    (verbose > 0 && level >= 0)) {
>>  		printk(KERN_INFO);
>>  		print_temp("CPU-temp: ", temp);
>>  		if (casetemp)
>>  			print_temp(", Case: ", casetemp);
>> +		if (casetemp2)
>> +			print_temp(", Case2: ", casetemp2);
>>  		if (level >= 0)
>>  			printk(", Fan: %d (tuned %+d)\n", 11-level,
>>  				x.fan_level-level);
> 
> Wow, this logging must clutter the kernel log quite substantially
> if enabled.

It is  disable by default.

> 
>> @@ -216,6 +300,7 @@ poll_temp( void )
>>  
>>  	x.temp = temp;
>>  	x.casetemp = casetemp;
>> +	x.casetemp2 = casetemp2;
>>  
>>  	if( level >= 0 )
>>  		tune_fan( level );
>> @@ -275,11 +360,21 @@ setup_hardware( void )
>>  	if (err)
>>  		printk(KERN_WARNING
>>  			"Failed to create temperature attribute file(s).\n");
>> +
>> +	x.hwmon = hwmon_device_register_with_groups(&x.of_dev->dev,
>> +			"therm_windtunnel", NULL,
>> +			therm_windtunnel_attr_groups);
>> +	if (!x.hwmon)
>> +		dev_warn(&x.of_dev->dev, "Failed to create the hwmon device\n");
>>  }
>>  
>>  static void
>>  restore_regs( void )
>>  {
>> +	if (x.hwmon)
>> +		hwmon_device_unregister(x.hwmon);
>> +	x.hwmon = NULL;
>> +
>>  	device_remove_file( &x.of_dev->dev, &dev_attr_cpu_temperature );
>>  	device_remove_file( &x.of_dev->dev, &dev_attr_case_temperature );
>>  	device_remove_file(&x.of_dev->dev, &dev_attr_fan_level);
>> -- 
>> 2.0.1
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@...r.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
>>
>>
>>
> 


-- 
gpg @keyserver.linux.it: Goffredo Baroncelli (kreijackATinwind.it>
Key fingerprint BBF5 1610 0B64 DAC6 5F7D  17B2 0EDA 9B37 8B82 E0B5
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists