Skip to contentSkip to conversation

W83627EHF driver update

PATCH New Rudolf Marek 5 participants 30 messages Jun 6, 2006 – Jun 26, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 W83627EHF driver update

Hello

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>

Some notes: The patch should be OK. It was checked/tested many times by me and David. This patch is based on ehf_patch_rc1 and ehf_patch_fix_fan45. More than 80 columns were fixed and on 4 places mutex lock/unlock was placed.

David, please can you check/test this final patch before Jean takes it? Please have a look to store_pwm_enable which has non-trivial locking. I belive it is OK and I tested briefly, but better to be sure.

Thanks,
regards
Rudolf

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_u.patch
Type: text/x-patch
Size: 18050 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060606/5822ece9/attachment-0001.bin

#1 lore raw

PATCH 1/1 W83627EHF driver update

Hi Rudolf,

Okay, I will test this and get back to you tomorrow.

Thanks,
David

Quoted 21 lines · Rudolf Marek

Hello

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>

Some notes: The patch should be OK. It was checked/tested many times by me and David. This patch is based on ehf_patch_rc1 and ehf_patch_fix_fan45. More than 80 columns were fixed and on 4 places mutex lock/unlock was placed.

David, please can you check/test this final patch before Jean takes it? Please have a look to store_pwm_enable which has non-trivial locking. I belive it is OK and I tested briefly, but better to be sure.

Thanks,
regards
Rudolf

#2 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Review

Quoted 12 lines · Rudolf Marek

Please have a look to store_pwm_enable which has non-trivial locking. I belive it is OK and I tested briefly, but better to be sure.

(quoted diff) +5 -0
+static ssize_t
+store_pwm_enable(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
 
+	if (data->pwm_enable[nr] != val) {

I guess *here* is where you mean non-trivial.

why not take the lock here -> instead of in both branches of if ?

Review

Quoted 2 lines
(quoted diff) +1 -0
+		if (!val) {

you specifically and knowingly dont protect this call,
and advised us of it in your msg. It certainly warrants a comment here too.

Review

Quoted 2 lines
(quoted diff) +1 -0
+			store_pwm(dev, attr, "255", 3);

and now take the lock.

Review

Quoted 5 lines
(quoted diff) +4 -0
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+		}
+		else {

repeating, this lock could be hoisted, but for the missing/unstated reason.

Quoted 4 lines
(quoted diff) +3 -0
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+			val--;

Review

Quoted 10 lines
(quoted diff) +10 -0
+		}
+		reg = (reg & ~(11 << W83627EHF_PWM_ENABLE_SHIFT[nr]));
+		reg |= val << W83627EHF_PWM_ENABLE_SHIFT[nr];
+		w83627ehf_write_value(client, W83627EHF_REG_PWM_ENABLE[nr],
+					reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}
+

:-)

#3 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

I like the added mutex locking. It's a good idea. It sounds like you'll be modifying it a little?

Here is the output of the regression test on my machine:

# ./w83627ehf_regression.sh
* WARNING: This will run your system fans through many possible
           combinations. There is a possibility your system will
           overheat. Use this script at your own risk!
Mapping fans...setting all PWMs to 100%
    fan 1 0 RPM: ignored. (Fan not installed?)
    fan 2 0 RPM: ignored. (Fan not installed?)
    fan 3 3479 RPM:
        pwm1 3443 RPM
        pwm2 3443 RPM
        pwm3 1654 RPM: connected.
Testing fan 3 - pwm3 ... div OK, PWM/DC OK, enable OK
test_smartfan1 $1="3" $2="pwm3"
, SF1 OK

Obviously, I'm right in the middle of building SmartFan 1 testing. It doesn't test that. Everything else works fine. I'm attaching the regression test which is updated for this patch. So, sign me off on it.

Thanks,
David
-------------- next part --------------
A non-text attachment was scrubbed...
Name: w83627ehf_regression.sh
Type: application/x-sh
Size: 15366 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060607/ba81aba3/attachment.sh

#4 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Jim,

Quoted 9 lines

Review

Quoted 7 lines
(quoted diff) +5 -0
+static ssize_t
+store_pwm_enable(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
 
+	if (data->pwm_enable[nr] != val) {

I guess *here* is where you mean non-trivial.

Quoted 6 lines

why not take the lock here -> instead of in both branches of if ?

Review

Quoted 2 lines
(quoted diff) +1 -0
+		if (!val) {

you specifically and knowingly dont protect this call,
and advised us of it in your msg. It certainly warrants a comment here too.

The call protect itself because it locks on its own. If the lock would be before, it tries to lock 2 times and will deadlock.

Quoted 10 lines

Review

Quoted 3 lines
(quoted diff) +1 -0
+			store_pwm(dev, attr, "255", 3);

and now take the lock.

Review

Quoted 5 lines
(quoted diff) +4 -0
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+		}
+		else {

repeating, this lock could be hoisted, but for the missing/unstated reason.

I know this code is not perfect, but how can it be done without more locking or more confusing temporary variables?

Thanks,

regards
Rudolf

#5 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi David,

Quoted 2 lines

I like the added mutex locking. It's a good idea. It sounds like you'll be modifying it a little?

I hope not.

There is just a place that Jim mentioned with the lock in both branches of if. It is not elegant, but without more temporary variables or multiple lock/unlock it wont simply work as expected.

Thanks for the testing. If you have better idea how to handle the particular place it would be good.

Otherwise if Jim has not any ideas I would like Jean to accept it as it is.

Thanks,
Regards
Rudolf

#6 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi David,

I tested the module with your script. Here's the output :
* WARNING: This will run your system fans through many possible combinations. There is a possibility your system will overheat. Use this script at your own risk!
Mapping fans...setting all PWMs to 100%

     fan 1 2083 RPM:
         pwm1 RPM highly variable? ignored.
         pwm2 2136 RPM
         pwm3 2109 RPM
     fan 2 3668 RPM:
         pwm1 3668 RPM
         pwm2 1140 RPM: connected.
         pwm3 3668 RPM
     fan 3 0 RPM: ignored. (Fan not installed?)

Testing fan 2 - pwm2 ... div OK, PWM/DC OK
Failure: 50% 3479 RPM, disabled (100%) 3515 RPM.

Fan 1 is my PSU fan, reporting (sometimes crazy) values, but not powered my the motherboard. Thus, fan control won't affect it. It's perfectly normal.

pwm2 seem to perfectly control my CPU fan (fan2). I was very surprised when my CPU fan (normaly running at 20%) started to run at full power !

I'll receive another fan tomorrow, and will connect it to fan1 (and this time, it may be controlled by the w83627ehf).
I'll keep you informed. Thanks a lot for your work ! I was _really_ waiting for this feature !

Cheers,
Sylvain

Note : to pass the files verification phase, I had to change ls -l | cut -c-28,42- | (
into
/bin/ls -l | cut -c-28,46- | cut -c-11,13- | (
to make it look like what you expected. Looks like ls hasn't a standard output among distros.

Quoted 29 lines · David Hubbard

Hi Rudolf,

I like the added mutex locking. It's a good idea. It sounds like you'll be modifying it a little?

Here is the output of the regression test on my machine:

# ./w83627ehf_regression.sh
* WARNING: This will run your system fans through many possible
         combinations. There is a possibility your system will
         overheat. Use this script at your own risk!
Mapping fans...setting all PWMs to 100%
  fan 1 0 RPM: ignored. (Fan not installed?)
  fan 2 0 RPM: ignored. (Fan not installed?)
  fan 3 3479 RPM:
      pwm1 3443 RPM
      pwm2 3443 RPM
      pwm3 1654 RPM: connected.
Testing fan 3 - pwm3 ... div OK, PWM/DC OK, enable OK
test_smartfan1 $1="3" $2="pwm3"
, SF1 OK

Obviously, I'm right in the middle of building SmartFan 1 testing. It doesn't test that. Everything else works fine. I'm attaching the regression test which is updated for this patch. So, sign me off on it.

Thanks,
David

#7 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Sylvain,

Thanks for the feedback! The ls | cut stuff is broken, I know. But it's a quick way to check that stuff. Maybe I should just remove it--the sysfs files wouldn't ever have wrong permissions, would they?

Quoted 9 lines

I tested the module with your script. Here's the output :
* WARNING: This will run your system fans through many possible combinations. There is a possibility your system will overheat. Use this script at your own risk!
Mapping fans...setting all PWMs to 100%

     fan 1 2083 RPM:
         pwm1 RPM highly variable? ignored.
         pwm2 2136 RPM
         pwm3 2109 RPM

The highly variable message and the fact that it didn't find anything here just means that the fan1 input isn't connected to any of the pwms. That's what you said here, right?

Quoted 3 lines

Fan 1 is my PSU fan, reporting (sometimes crazy) values, but not powered my the motherboard. Thus, fan control won't affect it. It's perfectly normal.

Okay, so it starts testing the fan 2 - pwm2 stuff:

Quoted 2 lines

Testing fan 2 - pwm2 ... div OK, PWM/DC OK
Failure: 50% 3479 RPM, disabled (100%) 3515 RPM.

What that means is that it set pwm2 to 50, then measured the RPM (3479 RPM). Then it disabled pwm2, which should set pwm2 to 100. It got 3515 RPM. That's not a significant change. However, I wouldn't say the disabled mode is broken. I'd say that the regression test script isn't robust enough, and should maybe take more samples at 50% before disabling and sampling at 100%. Anyway, I'll tweak it and see if the problem goes away. One thing is--the more samples it takes, the longer the script takes to run. :-(

I'm still thinking about the best way to test all this stuff. But I like getting test results!

David

#8 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi David,

Quoted 9 lines · David Hubbard
Quoted 5 lines

Mapping fans...setting all PWMs to 100%

     fan 1 2083 RPM:
         pwm1 RPM highly variable? ignored.
         pwm2 2136 RPM
         pwm3 2109 RPM

The highly variable message and the fact that it didn't find anything here just means that the fan1 input isn't connected to any of the pwms. That's what you said here, right?

Not exactly. My PSU has got a fan connector, so that I can monitor my PSU's RPMs. But on this fan connector, only the RPM wire is connected (and the ground I believe). Plus, the reported RPMs are quite .. variable (some glitches at 20000-30000 RPMs sometimes). So maybe pwm1 does affect fan1, but I'll need using a standard fan to test that (I just received it !).

About fan3/ pwm3, I only have 2 fans.

Quoted 5 lines

Okay, so it starts testing the fan 2 - pwm2 stuff:

Quoted 2 lines

Testing fan 2 - pwm2 ... div OK, PWM/DC OK
Failure: 50% 3479 RPM, disabled (100%) 3515 RPM.

Quoted 8 lines

What that means is that it set pwm2 to 50, then measured the RPM (3479 RPM). Then it disabled pwm2, which should set pwm2 to 100. It got 3515 RPM. That's not a significant change. However, I wouldn't say the disabled mode is broken. I'd say that the regression test script isn't robust enough, and should maybe take more samples at 50% before disabling and sampling at 100%. Anyway, I'll tweak it and see if the problem goes away. One thing is--the more samples it takes, the longer the script takes to run. :-(

Mmm. When the script ended, it left my fan speed at 100% (huge noise). So I had to rapidly figure out how to set fan speed down to 20%. I tried doing

$ echo 27 > pwm2
It didn't work. I messed a little with pwm2_enable / pwm2_mode (setting 
their values to other ones, then setting them back to original) and 
retried setting pwm to 27. This time, the fan slowed down.
I'll need to do further testing to understand what's going on.

After it worked, I could set my fan speed from 40 to 255 and hear the fan accelerate slowly.

Quoted 2 lines

I'm still thinking about the best way to test all this stuff. But I like getting test results!

No problem, I like controling my fan so precisely :)

Sylvain

#9 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Sorry for the late answer.

Quoted 8 lines

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>
Quoted 7 lines

Some notes: The patch should be OK. It was checked/tested many times by me and David. This patch is based on ehf_patch_rc1 and ehf_patch_fix_fan45. More than 80 columns were fixed and on 4 places mutex lock/unlock was placed.

David, please can you check/test this final patch before Jean takes it? Please have a look to store_pwm_enable which has non-trivial locking. I belive it is OK and I tested briefly, but better to be sure.

Here's my review:

Review

Quoted 8 lines
(quoted diff) +8 -0
+/* SmartFan registers */
+/* DC or PWM output fan configuration */
+static const u8 W83627EHF_REG_PWM_ENABLE[] = {
+	0x04,			/* SYS FAN0 output mode and PWM mode */
+	0x04,			/* CPU FAN0 output mode and PWM mode */
+	0x12,			/* AUX FAN mode all fan MIN */
+	0x62			/* CPU fan1 mode */
+};

Missing comma after the last entry.

Also the comments lack consistency, in particular we don't care that register 0x12 contains "fan min" bits here.

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+/* maps the user modes to chip modes */

I don't see how this comment is related to the following data.

Review

Quoted 6 lines
(quoted diff) +6 -0
+static const u8 W83627EHF_PWM_MODE_SHIFT[] = { 0, 1, 0, 6 };
+static const u8 W83627EHF_PWM_ENABLE_SHIFT[] = { 2, 4, 1, 4 };
+
+/* FAN Duty Cycle, be used to control */
+static const u8 W83627EHF_REG_PWM[] = { 0x01, 0x03, 0x11, 0x61 };
+static const u8 W83627EHF_REG_TARGET_TEMP[] = { 0x05, 0x06, 0x13, 0x63 };

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

Review

Quoted 10 lines
(quoted diff) +10 -0
+static const u8 W83627EHF_REG_TOLERANCE[] = { 0x07, 0x07, 0x14, 0x62 };
+
+
+/* Advanced Fan control */
+static const u8 W83627EHF_REG_FAN_MIN_OUTPUT[] = { 0x08, 0x09, 0x15, 0x64 };
+static const u8 W83627EHF_REG_FAN_MAX_OUTPUT[] = { 0x67, 0x69 };
+static const u8 W83627EHF_REG_FAN_STEP[] = { 0x68, 0x6A };
+static const u8 W83627EHF_REG_FAN_STOP_TIME[] = { 0x0C, 0x0D, 0x17, 0x66 };
+static const u8 W83627EHF_REG_FAN_STEP_TIME[] = { 0x0E, 0x0F };
+

For arrays with four values it's quite clear that there's one value for each fan, but for the arrays with only two values, it's not so clear. Some comments would be welcome.

Review

Quoted 9 lines

/*
* Conversions
*/

(quoted diff) +5 -0
+/* 0 is PWM mode, output in ms */
+static inline unsigned int step_time_from_reg(u8 reg, u8 mode)
+{
+	return mode ? 100 * reg : 400 * reg; 
+}

Comment is wrong, PWM mode is 1 not 0.

Review

Quoted 6 lines
(quoted diff) +6 -0
+
+static inline u8 step_time_to_reg(unsigned int msec, u8 mode)
+{
+	return mode ? (msec + 50) / 100 : (msec + 200) / 400;
+}
+

Missing check for overflow!

Quoted 26 lines

static inline unsigned int
fan_from_reg(u8 reg, unsigned int div)
{

(quoted diff) +15 -0
@@ -223,6 +262,21 @@ struct w83627ehf_data {
 	s16 temp_max[2];
 	s16 temp_max_hyst[2];
 	u32 alarms;
+ 
+	u8 pwm_mode[4]; /* 0->DC variable voltage, 1->PWM variable duty cycle */
+	u8 pwm_enable[4]; /* 0->disabled
+			   * 1->manual
+			   * 2->thermal cruise (also called SmartFan I)
+			   */
+	u8 pwm[4];
+	u8 target_temp[4];
+	u8 tolerance[4];
+
+	u8 fan_min_output[4]; /* minimum fan speed */
+	u8 fan_max_output[2];
+	u8 fan_step[2];
+	u8 fan_stop_time[4];
+	u8 fan_step_time[2]; /* 0 Down time, 1 up time */
 };
 
 static inline int is_word_sized(u16 reg)

Review static struct w83627ehf_data *w83627ehf_ · line 407

Quoted 4 lines
(quoted diff) +1 -0
@@ -350,6 +404,7 @@ static struct w83627ehf_data *w83627ehf_
 	struct i2c_client *client = to_i2c_client(dev);
 	struct w83627ehf_data *data = i2c_get_clientdata(client);
 	int i;
+	int tmp = 0;

Initialization not needed.

Quoted 8 lines
(quoted diff) +1 -0
 	mutex_lock(&data->update_lock);
 
@@ -416,6 +471,43 @@ static struct w83627ehf_data *w83627ehf_
 			}
 		}
 
+		for (i = 0; i < 4; i++) {

Review static struct w83627ehf_data *w83627ehf_ · line 484

Quoted 10 lines
(quoted diff) +10 -0
+			if (i != 1)
+				tmp = w83627ehf_read_value(client,
+						W83627EHF_REG_PWM_ENABLE[i]);
+			data->pwm_mode[i] > +				((tmp >> W83627EHF_PWM_MODE_SHIFT[i]) & 1)
+				? 0 : 1;
+			if (data->pwm_enable[i] != 0 ||
+			    ((tmp >> W83627EHF_PWM_ENABLE_SHIFT[i]) & 3) != 0)
+				data->pwm_enable[i] = 
+					((tmp >> W83627EHF_PWM_ENABLE_SHIFT[i])
+						& 3) + 1;

I don't understand this test, can you please explain?

Review static struct w83627ehf_data *w83627ehf_ · line 497

Quoted 13 lines
(quoted diff) +13 -0
+			data->pwm[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_PWM[i]);
+			data->fan_min_output[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_MIN_OUTPUT[i]);
+			data->fan_stop_time[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STOP_TIME[i]);
+			data->target_temp[i] > +				w83627ehf_read_value(client,
+					W83627EHF_REG_TARGET_TEMP[i]) &
+					(data->pwm_mode[i] = 1 ? 0x7f : 0xff);
+			data->tolerance[i] > +				(w83627ehf_read_value(client,
+					W83627EHF_REG_TOLERANCE[i]) >>
+					(i = 1 ? 4 : 0)) & 0x0f;
+		}

Register 0x07 is read twice, this could be avoided with the same trick you use for register 0x04. I would also suggest that you make the temporary variable(s) local as you only use it (them) here. What about the following?

		for (i = 0; i < 4; i++) {
			int pwmcfg, tolerance;

			if (i != 1) {
				pwmcfg = w83627ehf_read_value(client,
						W83627EHF_REG_PWM_ENABLE[i]);
				tolerance = w83627ehf_read_value(client,
						W83627EHF_REG_TOLERANCE[i]);
			}

			data->pwm_mode[i] 				((pwmcfg >> W83627EHF_PWM_MODE_SHIFT[i]) & 1)
				? 0 : 1;

			(... other register reads ...)

			data->tolerance[i] = (tolerance >> (i = 1 ? 4 : 0)) & 0x0f;
		}
Quoted 15 lines
(quoted diff) +10 -0
+
+		for (i = 0; i < 2; i++) {
+			data->fan_max_output[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_MAX_OUTPUT[i]);
+			data->fan_step[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STEP[i]);
+			data->fan_step_time[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STEP_TIME[i]);
+		}
+
 		/* Measured temperatures and limits */
 		data->temp1 = w83627ehf_read_value(client,
 			      W83627EHF_REG_TEMP1);
@@ -777,6 +869,321 @@ static struct sensor_device_attribute sd
 	SENSOR_ATTR(temp3_alarm, S_IRUGO, show_alarm, NULL, 13),

Review static struct sensor_device_attribute sd · line 879

Quoted 10 lines
(quoted diff) +8 -0
 };
 
+#define show_pwm_reg(reg) \
+static ssize_t show_##reg (struct device *dev, struct device_attribute *attr, \
+				char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);\
+	int nr = sensor_attr->index;\
+	return sprintf(buf,"%d\n", data->reg[nr]); \

Missing space after comma. Your spaces before backslashes are also inconsistent.

Review static struct sensor_device_attribute sd · line 890

Quoted 11 lines
(quoted diff) +11 -0
+}
+
+show_pwm_reg(pwm_mode)
+show_pwm_reg(pwm_enable)
+show_pwm_reg(pwm)
+
+static ssize_t
+store_pwm_mode(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);

No, you don't need to call update_device in a "store" sysfs callback.

Review static struct sensor_device_attribute sd · line 894

Quoted 4 lines
(quoted diff) +4 -0
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 1);

Using SENSOR_LIMIT here doesn't make sense, as we're not dealing with a range. Better return -EINVAL if value is neither 0 nor 1.

Review static struct sensor_device_attribute sd · line 900

Quoted 6 lines
(quoted diff) +6 -0
+	u16 reg = w83627ehf_read_value(client, W83627EHF_REG_PWM_ENABLE[nr]);
+
+	if (data->pwm_mode[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->pwm_mode[nr] = val;
+		reg = (reg & ~(1 << W83627EHF_PWM_MODE_SHIFT[nr]));

reg &= ..., or at least drop the extra pair of parentheses.

Review static struct sensor_device_attribute sd · line 908

Quoted 8 lines
(quoted diff) +8 -0
+		if (!val)
+			reg |= 1 << W83627EHF_PWM_MODE_SHIFT[nr];
+		w83627ehf_write_value(client, W83627EHF_REG_PWM_ENABLE[nr],
+					reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}

Locking is broken. You're doing a read-modify-write cycle, you must hold the lock before the read. You also must hold the lock when testing data->pwm_mode[nr] - even though I don't think you should test it at all, I see no reason to make the register write conditional.

Same comments apply to all your store callback functions, it seems.

Quoted 34 lines
(quoted diff) +34 -0
+
+static ssize_t
+store_pwm(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 255);
+
+	if (data->pwm_enable[nr] != 0 && data->pwm[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->pwm[nr] = val;
+		w83627ehf_write_value(client, W83627EHF_REG_PWM[nr],
+					val);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}
+
+static ssize_t
+store_pwm_enable(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 3);
+	u16 reg = w83627ehf_read_value(client, W83627EHF_REG_PWM_ENABLE[nr]);
+
+	if (data->pwm_enable[nr] != val) {
+		if (!val) {

Review static struct sensor_device_attribute sd · line 952

Quoted 10 lines
(quoted diff) +10 -0
+			store_pwm(dev, attr, "255", 3);
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+		}
+		else {
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+			val--;
+		}
+		reg = (reg & ~(11 << W83627EHF_PWM_ENABLE_SHIFT[nr]));

What is this "11" falling from the sky?

Review static struct sensor_device_attribute sd · line 959

Quoted 7 lines
(quoted diff) +7 -0
+		reg |= val << W83627EHF_PWM_ENABLE_SHIFT[nr];
+		w83627ehf_write_value(client, W83627EHF_REG_PWM_ENABLE[nr],
+					reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}

I just realized that you are emulating pwm_enable = 0. I guess the chip doesn't support it? Then you don't want to implement it in the driver. This emulation makes your code much more complex with no gain at all. Drivers must implement what the chip support, they must NOT emulate features.

I'm stopping my review here. Please fix:
* use of w83627ehf_update_device
* use of SENSORS_LIMIT
* locking
* conditional register writes
all around the place, then test the patch again, and then I'll review it again.

Quoted 294 lines
(quoted diff) +255 -3
+
+
+#define show_tol_temp(reg) \
+static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \
+				char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	return sprintf(buf,"%d\n", temp1_from_reg(data->reg[nr])); \
+}
+
+show_tol_temp(tolerance)
+show_tol_temp(target_temp)
+
+static ssize_t
+store_target_temp(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u8 val = temp1_to_reg(
+			SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 127000));
+
+	if (data->target_temp[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->target_temp[nr] = val;
+		w83627ehf_write_value(client, W83627EHF_REG_TARGET_TEMP[nr],
+					val);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}
+
+static ssize_t
+store_tolerance(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	/* Limit the temp to 0C - 15C */
+	u8 val = temp1_to_reg(
+			SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 15000));
+
+	u16 reg = w83627ehf_read_value(client, W83627EHF_REG_TOLERANCE[nr]);
+
+	if (data->tolerance[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->tolerance[nr] = val;
+		if (nr = 1)
+			reg = (reg & 0x0f) | (val << 4);
+		else
+			reg = (reg & 0xf0) | val;
+		w83627ehf_write_value(client,
+				W83627EHF_REG_TOLERANCE[nr], reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}
+
+static struct sensor_device_attribute sda_pwm[] = {
+	SENSOR_ATTR(pwm1, S_IWUSR | S_IRUGO, show_pwm, store_pwm, 0),
+	SENSOR_ATTR(pwm2, S_IWUSR | S_IRUGO, show_pwm, store_pwm, 1),
+	SENSOR_ATTR(pwm3, S_IWUSR | S_IRUGO, show_pwm, store_pwm, 2),
+	SENSOR_ATTR(pwm4, S_IWUSR | S_IRUGO, show_pwm, store_pwm, 3),
+};
+
+static struct sensor_device_attribute sda_pwm_mode[] = {
+	SENSOR_ATTR(pwm1_mode, S_IWUSR | S_IRUGO, show_pwm_mode,
+		    store_pwm_mode, 0),
+	SENSOR_ATTR(pwm2_mode, S_IWUSR | S_IRUGO, show_pwm_mode,
+		    store_pwm_mode, 1),
+	SENSOR_ATTR(pwm3_mode, S_IWUSR | S_IRUGO, show_pwm_mode,
+		    store_pwm_mode, 2),
+	SENSOR_ATTR(pwm4_mode, S_IWUSR | S_IRUGO, show_pwm_mode,
+		    store_pwm_mode, 3),
+};
+
+static struct sensor_device_attribute sda_pwm_enable[] = {
+	SENSOR_ATTR(pwm1_enable, S_IWUSR | S_IRUGO, show_pwm_enable,
+		    store_pwm_enable, 0),
+	SENSOR_ATTR(pwm2_enable, S_IWUSR | S_IRUGO, show_pwm_enable,
+		    store_pwm_enable, 1),
+	SENSOR_ATTR(pwm3_enable, S_IWUSR | S_IRUGO, show_pwm_enable,
+		    store_pwm_enable, 2),
+	SENSOR_ATTR(pwm4_enable, S_IWUSR | S_IRUGO, show_pwm_enable,
+		    store_pwm_enable, 3),
+};
+
+static struct sensor_device_attribute sda_target_temp[] = {
+	SENSOR_ATTR(temp1_target, S_IWUSR | S_IRUGO, show_target_temp,
+		    store_target_temp, 0),
+	SENSOR_ATTR(temp2_target, S_IWUSR | S_IRUGO, show_target_temp,
+		    store_target_temp, 1),
+	SENSOR_ATTR(temp3_target, S_IWUSR | S_IRUGO, show_target_temp,
+		    store_target_temp, 2),
+	SENSOR_ATTR(temp4_target, S_IWUSR | S_IRUGO, show_target_temp,
+		    store_target_temp, 3),
+};
+
+static struct sensor_device_attribute sda_tolerance[] = {
+	SENSOR_ATTR(temp1_tolerance, S_IWUSR | S_IRUGO, show_tolerance,
+		    store_tolerance, 0),
+	SENSOR_ATTR(temp2_tolerance, S_IWUSR | S_IRUGO, show_tolerance,
+		    store_tolerance, 1),
+	SENSOR_ATTR(temp3_tolerance, S_IWUSR | S_IRUGO, show_tolerance,
+		    store_tolerance, 2),
+	SENSOR_ATTR(temp4_tolerance, S_IWUSR | S_IRUGO, show_tolerance,
+		    store_tolerance, 3),
+};
+
+static void device_create_file_pwm(struct device *dev, int i)
+{
+	device_create_file(dev, &sda_pwm[i].dev_attr);
+	device_create_file(dev, &sda_pwm_mode[i].dev_attr);
+	device_create_file(dev, &sda_pwm_enable[i].dev_attr);
+	device_create_file(dev, &sda_target_temp[i].dev_attr);
+	device_create_file(dev, &sda_tolerance[i].dev_attr);
+}
+
+/* Smart Fan registers */
+
+#define fan_functions(reg, REG) \
+static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \
+		       char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	return sprintf(buf,"%d\n", data->reg[nr]); \
+}\
+static ssize_t \
+store_##reg(struct device *dev, struct device_attribute *attr, \
+			    const char *buf, size_t count) \
+{\
+	struct i2c_client *client = to_i2c_client(dev); \
+	struct w83627ehf_data *data = i2c_get_clientdata(client); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 255);\
+	if ( data->reg[nr] != val ) { \
+		mutex_lock(&data->update_lock); \
+		data->reg[nr] = val; \
+		w83627ehf_write_value(client, W83627EHF_REG_##REG[nr], \
+					 val); \
+		mutex_unlock(&data->update_lock); \
+	} \
+	return count; \
+}
+
+fan_functions(fan_min_output, FAN_MIN_OUTPUT)
+fan_functions(fan_max_output, FAN_MAX_OUTPUT)
+fan_functions(fan_step, FAN_STEP)
+
+#define fan_time_functions(reg, REG) \
+static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \
+				char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	return sprintf(buf,"%d\n", \
+			step_time_from_reg(data->reg[nr], data->pwm_mode[nr])); \
+} \
+\
+static ssize_t \
+store_##reg(struct device *dev, struct device_attribute *attr, \
+			const char *buf, size_t count) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct i2c_client *client = to_i2c_client(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	u8 val = step_time_to_reg(simple_strtoul(buf, NULL, 10), \
+					data->pwm_mode[nr]); \
+	if ( data->reg[nr] != val ) { \
+		mutex_lock(&data->update_lock); \
+		data->reg[nr] = val; \
+		w83627ehf_write_value(client, W83627EHF_REG_##REG[nr], \
+					 val); \
+		mutex_unlock(&data->update_lock); \
+	} \
+	return count; \
+} \
+
+fan_time_functions(fan_step_time, FAN_STEP_TIME)
+fan_time_functions(fan_stop_time, FAN_STOP_TIME)
+
+
+static struct sensor_device_attribute sda_sf3_arrays_fan4[] = {
+	SENSOR_ATTR(fan4_stop_time, S_IWUSR | S_IRUGO, show_fan_stop_time,
+		    store_fan_stop_time, 3),
+	SENSOR_ATTR(fan4_min_output, S_IWUSR | S_IRUGO, show_fan_min_output,
+		    store_fan_min_output, 3),
+	SENSOR_ATTR(fan4_max_output, S_IWUSR | S_IRUGO, show_fan_max_output,
+		    store_fan_max_output, 1),
+	SENSOR_ATTR(fan4_step, S_IWUSR | S_IRUGO, show_fan_step,
+		    store_fan_step, 1),
+};
+
+static struct sensor_device_attribute sda_sf3_arrays[] = {
+	SENSOR_ATTR(fan_step_up_time, S_IWUSR | S_IRUGO, show_fan_step_time,
+		    store_fan_step_time, 1),
+	SENSOR_ATTR(fan_step_down_time, S_IWUSR | S_IRUGO, show_fan_step_time,
+		    store_fan_step_time, 0),
+	SENSOR_ATTR(fan1_stop_time, S_IWUSR | S_IRUGO, show_fan_stop_time,
+		    store_fan_stop_time, 0),
+	SENSOR_ATTR(fan2_stop_time, S_IWUSR | S_IRUGO, show_fan_stop_time,
+		    store_fan_stop_time, 1),
+	SENSOR_ATTR(fan3_stop_time, S_IWUSR | S_IRUGO, show_fan_stop_time,
+		    store_fan_stop_time, 2),
+	SENSOR_ATTR(fan1_min_output, S_IWUSR | S_IRUGO, show_fan_min_output,
+		    store_fan_min_output, 0),
+	SENSOR_ATTR(fan2_min_output, S_IWUSR | S_IRUGO, show_fan_min_output,
+		    store_fan_min_output, 1),
+	SENSOR_ATTR(fan3_min_output, S_IWUSR | S_IRUGO, show_fan_min_output,
+		    store_fan_min_output, 2),
+	SENSOR_ATTR(fan2_max_output, S_IWUSR | S_IRUGO, show_fan_max_output,
+		    store_fan_max_output, 0),
+	SENSOR_ATTR(fan2_step, S_IWUSR | S_IRUGO, show_fan_step,
+		    store_fan_step, 0),
+};
+
 /*
  * Driver and client management
  */
@@ -810,6 +1217,7 @@ static int w83627ehf_detect(struct i2c_a
 	struct i2c_client *client;
 	struct w83627ehf_data *data;
 	struct device *dev;
+	u8 cr24, cr29;
 	int i, err = 0;
 
 	if (!request_region(address + REGION_OFFSET, REGION_LENGTH,
@@ -848,13 +1256,22 @@ static int w83627ehf_detect(struct i2c_a
 		data->fan_min[i] = w83627ehf_read_value(client,
 				   W83627EHF_REG_FAN_MIN[i]);
 
+	/* fan4 and fan5 share some pins with the GPIO and serial flash */
+
+	superio_enter();
+	/* flash needs to be disabled for AUXFANIN1/fan5 -> 00 */
+	cr24 = superio_inb(0x24) & 0x2; 
+	cr29 = superio_inb(0x29) & 0x6; /* CPUFANIN1/fan4 ->00 */
+	superio_exit();
+	
 	/* It looks like fan4 and fan5 pins can be alternatively used
 	   as fan on/off switches */
+	
 	data->has_fan = 0x07; /* fan1, fan2 and fan3 */
 	i = w83627ehf_read_value(client, W83627EHF_REG_FANDIV1);
-	if (i & (1 << 2))
+	if ((i & (1 << 2)) && (!cr29))
 		data->has_fan |= (1 << 3);
-	if (i & (1 << 0))
+	if ((i & (1 << 0)) && (!cr24))
 		data->has_fan |= (1 << 4);
 
 	/* Register sysfs hooks */
@@ -864,13 +1281,28 @@ static int w83627ehf_detect(struct i2c_a
 		goto exit_detach;
 	}
 
+  	for (i = 0; i < ARRAY_SIZE(sda_sf3_arrays); i++)
+  		device_create_file(dev, &sda_sf3_arrays[i].dev_attr);
+
+	/* if fan4 is enabled create the sf3 files for it */
+	if (data->has_fan & (1 << 3))
+		for (i = 0; i < ARRAY_SIZE(sda_sf3_arrays_fan4); i++)
+			device_create_file(dev, &sda_sf3_arrays_fan4[i].dev_attr);
+
 	for (i = 0; i < 10; i++)
 		device_create_file_in(dev, i);
 
-	for (i = 0; i < 5; i++) {
+	for (i = 0; i < 5; i++)
 		if (data->has_fan & (1 << i))
 			device_create_file_fan(dev, i);
+
+	for (i = 0; i < 4; i++) {
+		if (data->has_fan & (1 << i)) {
+			device_create_file_pwm(dev, i);
+			data->pwm_mode[i] = 1; /* initialize in PWM mode */
+		}
 	}
+
 	for (i = 0; i < ARRAY_SIZE(sda_temp); i++)
 		device_create_file(dev, &sda_temp[i].dev_attr);
 
Signature · 1 line
Jean Delvare
#10 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Jean, Rudolf,

Quoted 7 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+/* maps the user modes to chip modes */

I don't see how this comment is related to the following data.

Review

Quoted 2 lines
(quoted diff) +2 -0
+static const u8 W83627EHF_PWM_MODE_SHIFT[] = { 0, 1, 0, 6 };
+static const u8 W83627EHF_PWM_ENABLE_SHIFT[] = { 2, 4, 1, 4 };

The PWM modes are located at different bit positions in the registers. the PWM_MODE_SHIFT values give the bit positions. Remember that the pwmN_mode sets PWM or DC.

The PWM_ENABLE_SHIFT values give the bit positions from the PWM enable settings. After the emulated mode 0 (disabled) is accounted for, then they go into the register.

So I guess we should replace the comment with a better explanation. I think that comment is a remnant from some code that was deleted.

Quoted 8 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+/* FAN Duty Cycle, be used to control */
+static const u8 W83627EHF_REG_PWM[] = { 0x01, 0x03, 0x11, 0x61 };
+static const u8 W83627EHF_REG_TARGET_TEMP[] = { 0x05, 0x06, 0x13, 0x63 };

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

This is a good suggestion. In the near future, the registers will hold one or the other value, and the driver will support both modes. Rudolf and I talked about storing both values in the driver, with separate sysfs files so libsensors could read either value, the one on the chip or the one that will replace the value on the chip if the mode switched. Advantages to doing it this way: we can set a reasonable default for the value that's not on the chip, so when the mdoe switches, the value in the register is not left unchanged (and probably not reasonable); also, from userspace, it looks like there are two registers, even though in hardware there's only one.

Disadvantages: well, emulation is bad, isn't it? But here I would argue for emulating separate registers in the driver because the register sharing in the hardware is impossible to map directly to a sysfs file. If we're going to have two sysfs files with different meanings, then can we just emulate separate registers?

Quoted 6 lines

I just realized that you are emulating pwm_enable = 0. I guess the chip doesn't support it? Then you don't want to implement it in the driver. This emulation makes your code much more complex with no gain at all. Drivers must implement what the chip support, they must NOT emulate features.

Quoted 7 lines

I'm stopping my review here. Please fix:
* use of w83627ehf_update_device
* use of SENSORS_LIMIT
* locking
* conditional register writes
all around the place, then test the patch again, and then I'll review it again.

So it looks like there are some comments to fix, a few minor things, plus what you've listed here. Rudolf, I'll be able to spend some time on this over the weekend. That's a way off, if you feel like working on it before then.

Thanks for the review, Jean.

David

#11 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi David,

Quoted 16 lines
Quoted 4 lines

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

This is a good suggestion. In the near future, the registers will hold one or the other value, and the driver will support both modes. Rudolf and I talked about storing both values in the driver, with separate sysfs files so libsensors could read either value, the one on the chip or the one that will replace the value on the chip if the mode switched. Advantages to doing it this way: we can set a reasonable default for the value that's not on the chip, so when the mdoe switches, the value in the register is not left unchanged (and probably not reasonable); also, from userspace, it looks like there are two registers, even though in hardware there's only one.

Quoted 5 lines

Disadvantages: well, emulation is bad, isn't it? But here I would argue for emulating separate registers in the driver because the register sharing in the hardware is impossible to map directly to a sysfs file. If we're going to have two sysfs files with different meanings, then can we just emulate separate registers?

I'm fine with having two sets of internal variables, the reasons you give in that sense are good ones. I don't think doing so qualifies as "emulation". The hardware designers took an unfortunate shortcut by using the same registers for different functions, we need to deal with this.

Thanks,

Signature · 1 line
Jean Delvare
#12 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 24 lines · Jean Delvare

Hi David,

Quoted 16 lines
Quoted 4 lines

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

This is a good suggestion. In the near future, the registers will hold one or the other value, and the driver will support both modes. Rudolf and I talked about storing both values in the driver, with separate sysfs files so libsensors could read either value, the one on the chip or the one that will replace the value on the chip if the mode switched. Advantages to doing it this way: we can set a reasonable default for the value that's not on the chip, so when the mdoe switches, the value in the register is not left unchanged (and probably not reasonable); also, from userspace, it looks like there are two registers, even though in hardware there's only one.

Quoted 5 lines

Disadvantages: well, emulation is bad, isn't it? But here I would argue for emulating separate registers in the driver because the register sharing in the hardware is impossible to map directly to a sysfs file. If we're going to have two sysfs files with different meanings, then can we just emulate separate registers?

Quoted 5 lines

I'm fine with having two sets of internal variables, the reasons you give in that sense are good ones. I don't think doing so qualifies as "emulation". The hardware designers took an unfortunate shortcut by using the same registers for different functions, we need to deal with this.

Okay, that won't show up in the code this week. I'd like to see the new features we have working signed off. I think that's Rudolf opinion too. Then we'll add the other features, including that one.

David

#13 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi all,

Quoted 5 lines

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

I think this is in the plan for future driver upgrade.

Quoted 15 lines

Review

Quoted 11 lines
(quoted diff) +10 -0
+static const u8 W83627EHF_REG_TOLERANCE[] = { 0x07, 0x07, 0x14, 0x62 };
+
+
+/* Advanced Fan control */
+static const u8 W83627EHF_REG_FAN_MIN_OUTPUT[] = { 0x08, 0x09, 0x15, 0x64 };
+static const u8 W83627EHF_REG_FAN_MAX_OUTPUT[] = { 0x67, 0x69 };
+static const u8 W83627EHF_REG_FAN_STEP[] = { 0x68, 0x6A };
+static const u8 W83627EHF_REG_FAN_STOP_TIME[] = { 0x0C, 0x0D, 0x17, 0x66 };
+static const u8 W83627EHF_REG_FAN_STEP_TIME[] = { 0x0E, 0x0F };
+

For arrays with four values it's quite clear that there's one value for each fan, but for the arrays with only two values, it's not so clear. Some comments would be welcome.

OK

Quoted 5 lines
Quoted 5 lines

/*
* Conversions
*/

Quoted 7 lines

Review

Quoted 5 lines
(quoted diff) +5 -0
+/* 0 is PWM mode, output in ms */
+static inline unsigned int step_time_from_reg(u8 reg, u8 mode)
+{
+	return mode ? 100 * reg : 400 * reg; 
+}

Comment is wrong, PWM mode is 1 not 0.

Good catch.

Quoted 9 lines

Review

Quoted 7 lines
(quoted diff) +6 -0
+
+static inline u8 step_time_to_reg(unsigned int msec, u8 mode)
+{
+	return mode ? (msec + 50) / 100 : (msec + 200) / 400;
+}
+

Missing check for overflow!

When it is bigger than 255... Will fix...

=>

Quoted 1 line

Initialization not needed.

OK

Quoted 8 lines
Quoted 8 lines
(quoted diff) +0 -0
 	mutex_lock(&data->update_lock);
@@ -416,6 +471,43 @@ static struct w83627ehf_data *w83627ehf_
 			}
 		}
 
Quoted 13 lines

Review static struct w83627ehf_data *w83627ehf_ · line 484

Quoted 11 lines
(quoted diff) +11 -0
+		for (i = 0; i < 4; i++) {
+			if (i != 1)
+				tmp = w83627ehf_read_value(client,
+						W83627EHF_REG_PWM_ENABLE[i]);
+			data->pwm_mode[i] >> +				((tmp >> W83627EHF_PWM_MODE_SHIFT[i]) & 1)
+				? 0 : 1;
+			if (data->pwm_enable[i] != 0 ||
+			    ((tmp >> W83627EHF_PWM_ENABLE_SHIFT[i]) & 3) != 0)
+				data->pwm_enable[i] = 
+					((tmp >> W83627EHF_PWM_ENABLE_SHIFT[i])
+						& 3) + 1;

I don't understand this test, can you please explain?

Well this is for that MODE 0 emulation. I think we agreed we can ignore the mode.

Quoted 20 lines
Quoted 4 lines
(quoted diff) +3 -0
+			data->pwm[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_PWM[i]);
+			data->fan_min_output[i] = w83627ehf_read_value(client,

Review static struct w83627ehf_data *w83627ehf_ · line 497

Quoted 10 lines
(quoted diff) +10 -0
+						W83627EHF_REG_FAN_MIN_OUTPUT[i]);
+			data->fan_stop_time[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STOP_TIME[i]);
+			data->target_temp[i] >> +				w83627ehf_read_value(client,
+					W83627EHF_REG_TARGET_TEMP[i]) &
+					(data->pwm_mode[i] = 1 ? 0x7f : 0xff);
+			data->tolerance[i] >> +				(w83627ehf_read_value(client,
+					W83627EHF_REG_TOLERANCE[i]) >>
+					(i = 1 ? 4 : 0)) & 0x0f;
+		}

Register 0x07 is read twice, this could be avoided with the same trick you use for register 0x04. I would also suggest that you make the temporary variable(s) local as you only use it (them) here. What about the following?

Quoted 10 lines
		for (i = 0; i < 4; i++) {
			int pwmcfg, tolerance;

			if (i != 1) {
				pwmcfg = w83627ehf_read_value(client,
						W83627EHF_REG_PWM_ENABLE[i]);
				tolerance = w83627ehf_read_value(client,
						W83627EHF_REG_TOLERANCE[i]);
			}

Ok looks good.

Quoted 25 lines
			data->pwm_mode[i] > 				((pwmcfg >> W83627EHF_PWM_MODE_SHIFT[i]) & 1)
				? 0 : 1;

			(... other register reads ...)

			data->tolerance[i] = (tolerance >> (i = 1 ? 4 : 0)) & 0x0f;
		}
Quoted 17 lines
(quoted diff) +10 -0
+
+		for (i = 0; i < 2; i++) {
+			data->fan_max_output[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_MAX_OUTPUT[i]);
+			data->fan_step[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STEP[i]);
+			data->fan_step_time[i] = w83627ehf_read_value(client,
+						W83627EHF_REG_FAN_STEP_TIME[i]);
+		}
+
 		/* Measured temperatures and limits */
 		data->temp1 = w83627ehf_read_value(client,
 			      W83627EHF_REG_TEMP1);
@@ -777,6 +869,321 @@ static struct sensor_device_attribute sd
 	SENSOR_ATTR(temp3_alarm, S_IRUGO, show_alarm, NULL, 13),
 };
 
Quoted 11 lines

Review static struct sensor_device_attribute sd · line 879

Quoted 8 lines
(quoted diff) +8 -0
+#define show_pwm_reg(reg) \
+static ssize_t show_##reg (struct device *dev, struct device_attribute *attr, \
+				char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);\
+	int nr = sensor_attr->index;\
+	return sprintf(buf,"%d\n", data->reg[nr]); \

Missing space after comma. Your spaces before backslashes are also inconsistent.

Ok I will try to unify this. There are three authors of this patch so sometimes it is quite hard to get all stuff right.

Quoted 14 lines

Review static struct sensor_device_attribute sd · line 890

Quoted 12 lines
(quoted diff) +11 -0
+}
+
+show_pwm_reg(pwm_mode)
+show_pwm_reg(pwm_enable)
+show_pwm_reg(pwm)
+
+static ssize_t
+store_pwm_mode(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);

No, you don't need to call update_device in a "store" sysfs callback.

Yes good catch.

Quoted 7 lines

Review static struct sensor_device_attribute sd · line 894

Quoted 4 lines
(quoted diff) +4 -0
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 1);

Using SENSOR_LIMIT here doesn't make sense, as we're not dealing with a range. Better return -EINVAL if value is neither 0 nor 1.

Well yes why not, maybe we should write a note to a docs?

Quoted 9 lines

Review static struct sensor_device_attribute sd · line 900

Quoted 7 lines
(quoted diff) +6 -0
+	u16 reg = w83627ehf_read_value(client, W83627EHF_REG_PWM_ENABLE[nr]);
+
+	if (data->pwm_mode[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->pwm_mode[nr] = val;
+		reg = (reg & ~(1 << W83627EHF_PWM_MODE_SHIFT[nr]));

reg &= ..., or at least drop the extra pair of parentheses.

oOK

Quoted 15 lines

Review static struct sensor_device_attribute sd · line 908

Quoted 9 lines
(quoted diff) +8 -0
+		if (!val)
+			reg |= 1 << W83627EHF_PWM_MODE_SHIFT[nr];
+		w83627ehf_write_value(client, W83627EHF_REG_PWM_ENABLE[nr],
+					reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}

Locking is broken. You're doing a read-modify-write cycle, you must hold the lock before the read.
You also must hold the lock when testing data->pwm_mode[nr] - even though I don't think you should test it at all, I see no reason to make the register write conditional.

The register writes are just time "savers" I think it Yuan did it also for the
new driver...

Quoted 2 lines

Same comments apply to all your store callback functions, it seems.

I'm not a author of this code, I just did not catch it ...

Quoted 47 lines
Quoted 34 lines
(quoted diff) +34 -0
+
+static ssize_t
+store_pwm(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 255);
+
+	if (data->pwm_enable[nr] != 0 && data->pwm[nr] != val) {
+		mutex_lock(&data->update_lock);
+		data->pwm[nr] = val;
+		w83627ehf_write_value(client, W83627EHF_REG_PWM[nr],
+					val);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}
+
+static ssize_t
+store_pwm_enable(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct w83627ehf_data *data = w83627ehf_update_device(dev);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 3);
+	u16 reg = w83627ehf_read_value(client, W83627EHF_REG_PWM_ENABLE[nr]);
+
+	if (data->pwm_enable[nr] != val) {
+		if (!val) {

Review static struct sensor_device_attribute sd · line 952

Quoted 10 lines
(quoted diff) +10 -0
+			store_pwm(dev, attr, "255", 3);
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+		}
+		else {
+			mutex_lock(&data->update_lock);
+			data->pwm_enable[nr] = val;
+			val--;
+		}
+		reg = (reg & ~(11 << W83627EHF_PWM_ENABLE_SHIFT[nr]));

What is this "11" falling from the sky?

Quoted 13 lines

Review static struct sensor_device_attribute sd · line 959

Quoted 7 lines
(quoted diff) +7 -0
+		reg |= val << W83627EHF_PWM_ENABLE_SHIFT[nr];
+		w83627ehf_write_value(client, W83627EHF_REG_PWM_ENABLE[nr],
+					reg);
+		mutex_unlock(&data->update_lock);
+	}
+	return count;
+}

I just realized that you are emulating pwm_enable = 0. I guess the chip doesn't support it? Then you don't want to implement it in the driver. This emulation makes your code much more complex with no gain at all. Drivers must implement what the chip support, they must NOT emulate features.

On the other hand we want consistent interface... I thought that either the file could be present or not, but if present it MUST support 0 and 1 at least.

I think you told me that that -EINVAL should do it and I can revert the emulation stuff back. But I think we should FIX this in docs so it cannot be a matter of interpretation.

I will fix the minor stuff too.

Quoted 4 lines

* use of w83627ehf_update_device
* use of SENSORS_LIMIT
* locking
* conditional register writes

As for the conditional register writes. You just told me to save one read and you dont want the conditional writes? hmmm?

Quoted 2 lines

all around the place, then test the patch again, and then I'll review it again.

OK

Many thanks for the review. As I already told you we need some kind of wiki page describing all those pitfalls so I can easily check the stuff and dont forget about anything..

Anyone will help me?

Regards
Rudolf

#14 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 6 lines
Quoted 4 lines

The name is not very well chosen, these registers can hold both a target temperature or a target fan speed depending on the control mode (even if we don't support the latter at the moment.) What about just W83627EHF_REG_TARGET?

I think this is in the plan for future driver upgrade.

Let's get the name right now, so that this future driver upgrade doesn't have to change it everywhere.

Quoted 9 lines
Quoted 7 lines

Review

Quoted 4 lines
(quoted diff) +4 -0
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 1);

Using SENSOR_LIMIT here doesn't make sense, as we're not dealing with a range. Better return -EINVAL if value is neither 0 nor 1.

Well yes why not, maybe we should write a note to a docs?

I think the best we can do is to make it right in our drivers, because this is where people look when writing new drivers. If you want to extend the sysfs-interface documentation file to describe how invalid values should be handled (on write), feel free to do so.

Quoted 6 lines
Quoted 5 lines

I just realized that you are emulating pwm_enable = 0. I guess the chip doesn't support it? Then you don't want to implement it in the driver. This emulation makes your code much more complex with no gain at all. Drivers must implement what the chip support, they must NOT emulate features.

Quoted 6 lines

On the other hand we want consistent interface... I thought that either the file could be present or not, but if present it MUST support 0 and 1 at least.

I think you told me that that -EINVAL should do it and I can revert the emulation stuff back. But I think we should FIX this in docs so it cannot be a matter of interpretation.

Patch welcome :)

Quoted 2 lines

As for the conditional register writes. You just told me to save one read and you dont want the conditional writes? hmmm?

Saving one read in a safe way, in a function which will be called hundreds of times in the driver lifetime, is worth it. Saving a write in a function which will be called only a few times in the driver lifetime, in an unsafe way, is not good.

Quoted 5 lines

Many thanks for the review. As I already told you we need some kind of wiki page describing all those pitfalls so I can easily check the stuff and dont forget about anything..

Anyone will help me?

I think I already helped much by reviewing so much code in the past. Everything I said there is a candidate for your page if you want to write it.

Signature · 1 line
Jean Delvare
#15 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi,

Quoted 2 lines

Let's get the name right now, so that this future driver upgrade doesn't have to change it everywhere.

OK I changed it. Three places total ;)

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

Review

Quoted 4 lines
(quoted diff) +4 -0
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 1);

Using SENSOR_LIMIT here doesn't make sense, as we're not dealing with a range. Better return -EINVAL if value is neither 0 nor 1.

Quoted 6 lines
Quoted 1 line

Well yes why not, maybe we should write a note to a docs?

I think the best we can do is to make it right in our drivers, because this is where people look when writing new drivers. If you want to extend the sysfs-interface documentation file to describe how invalid values should be handled (on write), feel free to do so.

Ok

Well they are handled at least with two different ways.

1) -EINVAL
2) if out of bounds - nearest bound is written

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

I just realized that you are emulating pwm_enable = 0. I guess the chip doesn't support it? Then you don't want to implement it in the driver. This emulation makes your code much more complex with no gain at all. Drivers must implement what the chip support, they must NOT emulate features.

Quoted 8 lines
Quoted 6 lines

On the other hand we want consistent interface... I thought that either the file could be present or not, but if present it MUST support 0 and 1 at least.

I think you told me that that -EINVAL should do it and I can revert the emulation stuff back. But I think we should FIX this in docs so it cannot be a matter of interpretation.

Patch welcome :)

OK

Quoted 5 lines
Quoted 1 line

Anyone will help me?

I think I already helped much by reviewing so much code in the past. Everything I said there is a candidate for your page if you want to write it.

Yes I had it in mind. But this was not addressed to you. But to others round here...

Regards
Rudolf

#16 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Rudolf,

Quoted 10 lines
Quoted 9 lines
Quoted 4 lines
Quoted 2 lines

Using SENSOR_LIMIT here doesn't make sense, as we're not dealing with a range. Better return -EINVAL if value is neither 0 nor 1.

Well yes why not, maybe we should write a note to a docs?

I think the best we can do is to make it right in our drivers, because this is where people look when writing new drivers. If you want to extend the sysfs-interface documentation file to describe how invalid values should be handled (on write), feel free to do so.

Quoted 6 lines

Ok

Well they are handled at least with two different ways.

1) -EINVAL
2) if out of bounds - nearest bound is written

First for discrete values, second for ranges. This makes sense, and is still simple enough so everyone should be able to get it.

Signature · 1 line
Jean Delvare
#17 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi all,

Here is some updated version for the test. It fixes the stuff Jean found. I want to read it once again I'm not falling asleep while reading it. It works for me and I think it should work for you all too.

Regards
Rudolf

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_next_try4.patch
Type: text/x-patch
Size: 16859 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060614/cc694273/attachment-0001.bin

#18 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

+ for (i = 0; i < 4; i++) {
+ int pwmcfg, tolerance;
+ if (i != 1) {
+ pwmcfg = w83627ehf_read_value(client,
+ W83627EHF_REG_PWM_ENABLE[i]);

Can we put a comment here explaining the if condition? Suggestion:
/* W83627EHF_REG_PWM_ENABLE[1] = W83627EHF_REG_PWM_ENABLE[0]. The values of pwmcfg and tolerance are not changed from i=0 to i=1 */

+store_pwm_enable(struct device *dev, struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct i2c_client *client = to_i2c_client(dev);
+ struct w83627ehf_data *data = i2c_get_clientdata(client);
+ struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+ int nr = sensor_attr->index;
+ u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 3);
+ u16 reg;
+
+ if (!val) /* pwm off not supported */
+ return -EINVAL;

We return -EINVAL in store_pwm_mode() if val is invalid. We shouldn't use SENSORS_LIMIT here either. Can we return -EINVAL instead? Also, these are the PWM modes the driver currently supports:
1 - manual PWM control
2 - thermal cruise

These PWM modes will soon be supported:
3 - RPM cruise
4 - Smart Fan III

So should val be limited to 1, 2, 3, or should val be limited to 1, 2?

Suggested change:

	u32 val = simple_strtoul(buf, NULL, 10);
	if (!val || val > 3)	/* mode 0 (disabled) not supported */
		return -EINVAL;

I'm testing it on my system right now. I will let you know if there are problems.

Thanks,
David

Quoted 5 lines · Rudolf Marek

Hi all,

Here is some updated version for the test. It fixes the stuff Jean found. I want to read it once again I'm not falling asleep while reading it. It works for me and I think it should work for you all too.

#19 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek
Quoted 2 lines · David Hubbard

I'm testing it on my system right now. I will let you know if there are problems.

Everything's OK here.

Manual fan control works fine.
David, is there anything else I should test ?
[ pwm2 is controling fan2. Other pwm don't have any effect (since the beginning). ]

Sylvain

Quoted 12 lines · Rudolf Marek
Quoted 5 lines

Hi all,

Here is some updated version for the test. It fixes the stuff Jean found. I want to read it once again I'm not falling asleep while reading it. It works for me and I think it should work for you all too.

_______________________________________________
lm-sensors mailing list
lm-sensors at lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#20 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Sylvain,

Quoted 4 lines

Everything's OK here.

Manual fan control works fine.
David, is there anything else I should test ?

I think that's all.

Quoted 4 lines

[ pwm2 is controling fan2. Other pwm don't have any effect (since the beginning). ]

Sylvain

David

#21 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi all,

Here comes the updated patch

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>

David, I fix the comment you had, thanks! Also I ripped out two more register exports (sf3 fan_max and fan_step). This will be neccessary to put back next round.

Regards
Rudolf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_next_try5.patch
Type: text/x-patch
Size: 16043 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060616/f0e6212b/attachment.bin

#22 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 12 lines · Rudolf Marek

Hi all,

Here comes the updated patch

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>
Quoted 5 lines

David, I fix the comment you had, thanks! Also I ripped out two more register exports (sf3 fan_max and fan_step). This will be neccessary to put back next round.

Regards
Rudolf

Just signing off. It looks fine to me.

David

#23 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 5 lines

Here comes the updated patch

This patch adds long-awaited suport for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documenation update will follow.

I'd rather take this and the documentation patch together. We all know what happens to patches which "will follow" ;)

Quoted 2 lines

David, I fix the comment you had, thanks! Also I ripped out two more register exports (sf3 fan_max and fan_step). This will be neccessary to put back next round.

It's in a much better shape now, thanks to both of you for the work. I would still have some comments though:

Review static struct w83627ehf_data *w83627ehf_ · line 473

Quoted 13 lines
(quoted diff) +9 -0
@@ -416,6 +462,35 @@ static struct w83627ehf_data *w83627ehf_
 			}
 		}
 
+		for (i = 0; i < 4; i++) {
+			int pwmcfg, tolerance;
+			/* pwmcfg, tolarance mapped for i=0, i=1 to same reg */
+			if (i != 1) {
+				pwmcfg = w83627ehf_read_value(client,
+						W83627EHF_REG_PWM_ENABLE[i]);
+				tolerance = w83627ehf_read_value(client,
+						W83627EHF_REG_TOLERANCE[i]);
+			}

Unfortunately my compiler (gcc 3.3.6) complains:

drivers/hwmon/w83627ehf.c: In function `w83627ehf_update_device':
drivers/hwmon/w83627ehf.c:466: warning: pwmcfg' might be used uninitialized in this function drivers/hwmon/w83627ehf.c:466: warning: tolerance' might be used uninitialized in this function

The compiler is wrong, but we can't leave warnings in the code. I guess this is the reason why you where initializing your temporary variable in the first place, I'm sorry for asking you to change that.

At this point I'm not too sure what is worst, two redundant inb_p, or two initializations and four comparisons. I let you decide, as long as there is no warning. If you go for initialization, please add a comment explaining it's there to prevent a compiler warning.

Quoted 32 lines
(quoted diff) +32 -0
+static ssize_t
+store_target_temp(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct w83627ehf_data *data = i2c_get_clientdata(client);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u8 val = temp1_to_reg(
+			SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 127000));
+
+	mutex_lock(&data->update_lock);
+	data->target_temp[nr] = val;
+	w83627ehf_write_value(client, W83627EHF_REG_TARGET[nr], val);
+	mutex_unlock(&data->update_lock);
+	return count;
+}
+
+static ssize_t
+store_tolerance(struct device *dev, struct device_attribute *attr,
+			const char *buf, size_t count)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct w83627ehf_data *data = i2c_get_clientdata(client);
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr);
+	int nr = sensor_attr->index;
+	u16 reg;
+	/* Limit the temp to 0C - 15C */
+	u8 val = temp1_to_reg(
+			SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 15000));
+
+	mutex_lock(&data->update_lock);

Review static struct w83627ehf_data *w83627ehf_

Quoted 10 lines
(quoted diff) +10 -0
+	reg = w83627ehf_read_value(client, W83627EHF_REG_TOLERANCE[nr]);
+	data->tolerance[nr] = val;
+	if (nr = 1)
+		reg = (reg & 0x0f) | (val << 4);
+	else
+		reg = (reg & 0xf0) | val;
+	w83627ehf_write_value(client, W83627EHF_REG_TOLERANCE[nr], reg);
+	mutex_unlock(&data->update_lock);
+	return count;
+}

These combinations of temp1_to_reg and SENSORS_LIMIT are inefficient. I suggest that you instead add two parameters (min and max values) to temp1_to_reg(). As it is inlined there will be no performance penalty and the code will look better too.

Quoted 48 lines
(quoted diff) +48 -0
+/* Smart Fan registers */
+
+#define fan_functions(reg, REG) \
+static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \
+		       char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	return sprintf(buf, "%d\n", data->reg[nr]); \
+}\
+static ssize_t \
+store_##reg(struct device *dev, struct device_attribute *attr, \
+			    const char *buf, size_t count) \
+{\
+	struct i2c_client *client = to_i2c_client(dev); \
+	struct w83627ehf_data *data = i2c_get_clientdata(client); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	u32 val = SENSORS_LIMIT(simple_strtoul(buf, NULL, 10), 0, 255); \
+	mutex_lock(&data->update_lock); \
+	data->reg[nr] = val; \
+	w83627ehf_write_value(client, W83627EHF_REG_##REG[nr],  val); \
+	mutex_unlock(&data->update_lock); \
+	return count; \
+}
+
+fan_functions(fan_min_output, FAN_MIN_OUTPUT)
+
+#define fan_time_functions(reg, REG) \
+static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \
+				char *buf) \
+{ \
+	struct w83627ehf_data *data = w83627ehf_update_device(dev); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \
+	return sprintf(buf, "%d\n", \
+			step_time_from_reg(data->reg[nr], data->pwm_mode[nr])); \
+} \
+\
+static ssize_t \
+store_##reg(struct device *dev, struct device_attribute *attr, \
+			const char *buf, size_t count) \
+{ \
+	struct i2c_client *client = to_i2c_client(dev); \
+	struct w83627ehf_data *data = i2c_get_clientdata(client); \
+	struct sensor_device_attribute *sensor_attr = to_sensor_dev_attr(attr); \
+	int nr = sensor_attr->index; \

Review

Quoted 10 lines
(quoted diff) +10 -0
+	u8 val = step_time_to_reg(simple_strtoul(buf, NULL, 10), \
+					data->pwm_mode[nr]); \
+	mutex_lock(&data->update_lock); \
+	data->reg[nr] = val; \
+	w83627ehf_write_value(client, W83627EHF_REG_##REG[nr], val); \
+	mutex_unlock(&data->update_lock); \
+	return count; \
+} \
+
+fan_time_functions(fan_stop_time, FAN_STOP_TIME)

Why do you define these two macros, if you then use them only once?

Quoted 12 lines

/*
* Driver and client management
*/

(quoted diff) +1 -0
@@ -810,6 +1162,7 @@ static int w83627ehf_detect(struct i2c_a
 	struct i2c_client *client;
 	struct w83627ehf_data *data;
 	struct device *dev;
+	u8 cr24, cr29;
 	int i, err = 0;
 
 	if (!request_region(address + REGION_OFFSET, REGION_LENGTH,

Review static int w83627ehf_detect(struct i2c_a · line 1210

Quoted 10 lines
(quoted diff) +7 -0
@@ -848,13 +1201,22 @@ static int w83627ehf_detect(struct i2c_a
 		data->fan_min[i] = w83627ehf_read_value(client,
 				   W83627EHF_REG_FAN_MIN[i]);
 
+	/* fan4 and fan5 share some pins with the GPIO and serial flash */
+
+	superio_enter();
+	/* flash needs to be disabled for AUXFANIN1/fan5 -> 00 */
+	cr24 = superio_inb(0x24) & 0x2; 
+	cr29 = superio_inb(0x29) & 0x6; /* CPUFANIN1/fan4 ->00 */
+	superio_exit();

What does "-> 00" mean? These comments could be improved (maybe merged with the one right below?) I would also prefer that the masking is done when testing, rather than when reading the registers, else your variable names don't make much sense.

Quoted 14 lines
(quoted diff) +4 -2
+	
 	/* It looks like fan4 and fan5 pins can be alternatively used
 	   as fan on/off switches */
+	
 	data->has_fan = 0x07; /* fan1, fan2 and fan3 */
 	i = w83627ehf_read_value(client, W83627EHF_REG_FANDIV1);
-	if (i & (1 << 2))
+	if ((i & (1 << 2)) && (!cr29))
 		data->has_fan |= (1 << 3);
-	if (i & (1 << 0))
+	if ((i & (1 << 0)) && (!cr24))
 		data->has_fan |= (1 << 4);
 
 	for (i = 0; i < 10; i++)

Review

Quoted 6 lines
(quoted diff) +1 -1
 		device_create_file_in(dev, i);
 
-	for (i = 0; i < 5; i++) {
+	for (i = 0; i < 5; i++)
 		if (data->has_fan & (1 << i))
 			device_create_file_fan(dev, i);

Please don't change this.

Review

Quoted 7 lines
(quoted diff) +6 -0
+
+	for (i = 0; i < 4; i++) {
+		if (data->has_fan & (1 << i)) {
+			device_create_file_pwm(dev, i);
+			data->pwm_mode[i] = 1; /* initialize in PWM mode */
+		}
 	}

This initialization doesn't make much sense, it is arbitrary and the value will be overwritten on first update anyway. What's the purpose?

Thanks,

Signature · 1 line
Jean Delvare
#24 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Jean, Rudolf,

Quoted 2 lines · Jean Delvare

This initialization doesn't make much sense, it is arbitrary and the value will be overwritten on first update anyway. What's the purpose?

The value can't be zero, so I initialized it to 1. This is also a holdover from emulating 0 (disabled). Let's take it out and test it that way. What do you think, Rudolf?

David

#25 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi all,

Here comes fixed patch, I left the macro there because it will be used in future update. Rest was fixed.

I'm attaching the documentation patch too. David, please can you read it and check it? (I will sign that off when we know that there are no mistakes)

This patch adds long-awaited support for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. The documentation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>

Regards
Rudolf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_docs.patch
Type: text/x-patch
Size: 3393 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060624/922b5044/attachment.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_next_try7.patch
Type: text/x-patch
Size: 16868 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060624/922b5044/attachment-0001.bin

#26 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 2 lines

Here comes fixed patch, I left the macro there because it will be used in future update. Rest was fixed.

I am happy with the code now, but there is a problem with the sysfs file names, see below.

Quoted 2 lines

I'm attaching the documentation patch too. David, please can you read it and check it? (I will sign that off when we know that there are no mistakes)

My review of it:

Documentation/hwmon/w83627ehf line 4

Quoted 8 lines

diff -uprN a/Documentation/hwmon/w83627ehf linux-2.6.17-rc6/Documentation/hwmon/w83627ehf

Documentation/hwmon/w83627ehf +4 -0
@@ -0,0 +1,75 @@
+Kernel driver w83627ehf
+===========> +
+Supported chips:
+  * Winbond W83627EHF/EHG/EF/EG (ISA accesses ONLY)

s/accesses/access/

EF and EG verions are not actually supported, see below.

Documentation/hwmon/w83627ehf line 8

Quoted 4 lines
Documentation/hwmon/w83627ehf +4 -0
+    Prefix: 'w83627ehf'
+    Addresses scanned: ISA address retrieved from Super I/O registers
+    Datasheet: http://www.winbond-usa.com/products/winbond_products/pdfs/
+                      /PCIC/W83627EHF_%20W83627EHGb.pdf

Please don't split the URL.

Weird file name, BTW :(

Documentation/hwmon/w83627ehf line 20

Quoted 12 lines
Documentation/hwmon/w83627ehf +12 -0
+
+Authors:
+        Jean Delvare <khali at linux-fr.org>,
+        Yuan Mu <Ymu at Winbond.com.tw>,
+        Rudolf Marek <r.marek at sh.cvut.cz>
+
+Description
+-----------
+
+This driver implements support for the Winbond W83627EF, W83627EHF,
+W83627EHG and W83627EG super I/O chips. We will refer to them
+collectively as Winbond chips.

Not correct. The W83627EF and W83627EG do not have the hardware monitoring features so the driver is useless for them.

Documentation/hwmon/w83627ehf line 24

Quoted 4 lines
Documentation/hwmon/w83627ehf +4 -0
+
+The chips implement three temperature sensors, five fan rotation 
+speed sensors, ten analog voltage sensors, alarms with beep warnings (control
+unimplemented), and some automatic fan regulation strategies.

And manual fan speed control too.

Documentation/hwmon/w83627ehf line 29

Quoted 5 lines
Documentation/hwmon/w83627ehf +5 -0
+
+Temperatures are measured in degrees Celsius and measurement resolution is 1
+degC for temp1 and 0.5 degC for temp2 and temp3. An alarm is triggered when
+the temperature gets higher than the Overtemperature Shutdown value; it stays
+on until the temperature falls below the Hysteresis value.

"Overtemperature Shutdown" is an old term, I think it was used in the LM75 datasheet, but doesn't make much sense here. The system is usually NOT going to shutdown if this limit is crossed.

Documentation/hwmon/w83627ehf line 35

Quoted 6 lines
Documentation/hwmon/w83627ehf +6 -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. Some fans might not
+be present because they share pins with other funtions.

This driver implements automatic fan clock divider switching, so the divider values are hidden from the user.

Documentation/hwmon/w83627ehf line 43

Quoted 8 lines
Documentation/hwmon/w83627ehf +8 -0
+
+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.
+This mode works for fan1-fan4. Programmable mapping of the temperatures
+to fans is not implemented, BIOS should have done it.

We should at least have read-only files to let the user know the mapping.

It would be nice to describe what is the "thermal cruise" mode.

Documentation/hwmon/w83627ehf line 52

Quoted 9 lines
Documentation/hwmon/w83627ehf +9 -0
+
+/sys files
+----------
+
+pwm[1-4] - this file stores PWM duty cycle or DC value (fan speed) in range:
+	0 (stop) to 255 (full)
+
+pwm[1-4]_enable - this file controls mode of fan/temperature control:
+	* 1 Manual Mode, write to pwm file any value 0-255 (fullspeed)

s/fullspeed/full speed/

Documentation/hwmon/w83627ehf line 53

Quoted 1 line
Documentation/hwmon/w83627ehf +1 -0
+        * 2 Thermal Cruise

Broken indentation.

Documentation/hwmon/w83627ehf line 58

Quoted 5 lines
Documentation/hwmon/w83627ehf +5 -0
+
+Thermal Cruise mode
+--------------------
+
+If the temperature is in its range, defined with:

s/its/the/

Documentation/hwmon/w83627ehf line 61

Quoted 3 lines
Documentation/hwmon/w83627ehf +3 -0
+
+temp[1-4]_target    - set target temperature, unit miliC (range 0 - 127000)
+temp[1-4]_tolerance - tolerance, unit miliC (range 0 - 15000)

Unknown unit "miliC".

I'm just realizing that these names are quite wrong. For example temp1_target may not be related to temp1 but to a different temperature sensor depending on the temp/fan mapping. The "1" here really means pwm1, not temp1. So these files should be named pwm[1-4]_target_temp and pwm[1-4]_tolerance_temp, respectively.

BTW, these file names should also be added to sysfs-interface.

Documentation/hwmon/w83627ehf line 68

Quoted 7 lines
Documentation/hwmon/w83627ehf +7 -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. 
+There are defined steps and times, but not exported by the driver yet.
+
+fan[1-4]_min_output - minimum fan speed (range 1 - 255), when the temperature
+                      is bellow defined range.

Should be named pwm[1-4]_min.

s/bellow/below/

Documentation/hwmon/w83627ehf line 72

Quoted 4 lines
Documentation/hwmon/w83627ehf +4 -0
+fan[1-4]_stop_time  - how many milliseconds [ms] must elapse to switch
+                      corresponding fan off. (when the temperature was bellow
+                      defined range).
+

Should be named pwm[1-4]_stop_time.

Documentation/hwmon/w83627ehf line 74

Quoted 3 lines
Documentation/hwmon/w83627ehf +2 -0
+It seems last two funtions are exclusive, and are controlled with register 0x12.
+(Should be programed by BIOS)

But the driver uncondionally creates both files? This needs to be investigated, and fixed (either the driver or the documentation.)

Thanks,

Signature · 1 line
Jean Delvare
#27 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi all,

Quoted 2 lines

I am happy with the code now, but there is a problem with the sysfs file names, see below.

Ok. I'm attaching fixed versions of both.

Quoted 1 line

And manual fan speed control too.

Ok I will mention, although I considered this a special case of automatic regulation

Quoted 2 lines

We should at least have read-only files to let the user know the mapping.

The mapping is trivial and programmable for fan4 only. I fixed the docs.

Quoted 5 lines

I'm just realizing that these names are quite wrong. For example temp1_target may not be related to temp1 but to a different temperature sensor depending on the temp/fan mapping. The "1" here really means pwm1, not temp1. So these files should be named pwm[1-4]_target_temp and pwm[1-4]_tolerance_temp, respectively.

Yes correct. Fixed.

Quoted 1 line

BTW, these file names should also be added to sysfs-interface.

If we consider this as standard for some chips. We would need some mapping standard too (Yuan did some file for w83793)

Quoted 2 lines

But the driver uncondionally creates both files? This needs to be investigated, and fixed (either the driver or the documentation.)

I fixed the note. The files are used in SF3 iirc, so we might have them.

This patch adds long-awaited support for automatic fan modes. Based on the work of Yuan Mu from Winbond, I finished the support with the great help of David Hubbard. Many thanks goes to Jean Delvare, for his reviews and patience. The documentation update will follow.

Signed-off-by Yuan Mu <Ymu at Winbond.com.tw>
Signed-off-by Rudolf Marek <r.marek at sh.cvut.cz>
Signed-off-by David Hubbard <david.c.hubbard at gmail.com>

Regards
Rudolf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_next_try8.patch
Type: text/x-patch
Size: 16860 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060625/878aa562/attachment.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ehf_docs1.patch
Type: text/x-patch
Size: 3706 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060625/878aa562/attachment-0001.bin

#28 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Rudolf,

Quoted 16 lines
Quoted 2 lines

I am happy with the code now, but there is a problem with the sysfs file names, see below.

Ok. I'm attaching fixed versions of both.

Quoted 1 line

And manual fan speed control too.

Ok I will mention, although I considered this a special case of automatic regulation

Quoted 2 lines

We should at least have read-only files to let the user know the mapping.

The mapping is trivial and programmable for fan4 only. I fixed the docs.

Quoted 7 lines
Quoted 5 lines

I'm just realizing that these names are quite wrong. For example temp1_target may not be related to temp1 but to a different temperature sensor depending on the temp/fan mapping. The "1" here really means pwm1, not temp1. So these files should be named pwm[1-4]_target_temp and pwm[1-4]_tolerance_temp, respectively.

Yes correct. Fixed.

Thanks.

Quoted 2 lines

If we consider this as standard for some chips. We would need some mapping standard too (Yuan did some file for w83793)

The temp <-> pwm mapping standard already exists.

Quoted 4 lines
Quoted 2 lines

But the driver uncondionally creates both files? This needs to be investigated, and fixed (either the driver or the documentation.)

I fixed the note. The files are used in SF3 iirc, so we might have them.

I'm still not exactly happy. Documenting that we implemented things in a possibly non-working way is better than not documenting it, agreed, but even better would be to have things working in the first place.

Also on the to do list is the export of the fan <-> temperature mapping, at least read only, as I asked before already. For now pwm4 is useless in Temperature Cruise mode, as you don't know what temperature channel it is working with.

Anyway, I will apply these patches now, as we have gone trough many review cycles already. I made some fixes to the documentation patch to fix typos and reword one sentence. Any future change to the w83627ehf driver or its documentation should be incremental on top of these patches.

Thanks,

Signature · 1 line
Jean Delvare
#29 lore raw

PATCH 1/1 W83627EHF driver update

In reply to Rudolf Marek

Hi Jean,

Quoted 1 line

The temp <-> pwm mapping standard already exists.

Aha ok. pwm[1-*]_auto_channels_temp I guess.

Quoted 7 lines
Quoted 3 lines
Quoted 2 lines

But the driver uncondionally creates both files? This needs to be investigated, and fixed (either the driver or the documentation.)

I fixed the note. The files are used in SF3 iirc, so we might have them.

I'm still not exactly happy. Documenting that we implemented things in a possibly non-working way is better than not documenting it, agreed, but even better would be to have things working in the first place.

This stuff is so complex that even if you think you got it right, then some bits will make even bigger mess than it was.

Quoted 4 lines

Also on the to do list is the export of the fan <-> temperature mapping, at least read only, as I asked before already. For now pwm4 is useless in Temperature Cruise mode, as you don't know what temperature channel it is working with.

pwm4 is disabled on my board so I cannot tell which is which. I remmber that on david board it is disabled too.

I will put this into my queue, do it soon.

Quoted 5 lines

Anyway, I will apply these patches now, as we have gone trough many review cycles already. I made some fixes to the documentation patch to fix typos and reword one sentence. Any future change to the w83627ehf driver or its documentation should be incremental on top of these patches.

Thank you very much. As you know I'm not in exactly good situation so I hope I will do better next time when things are sorted out.

Regards
Rudolf