Re: ADT746X: logical-bitwise & confusion in set_max_duty_at_crit()

Benjamin Herrenschmidt wrote:

On Mon, 2008-03-10 at 08:46 +0100, Colin Leroy wrote:

quoted

On Mon, 10 Mar 2008 01:04:33 +0100, Roel Kluin wrote:

quoted

logical-bitwise & confusion

Looks good to me, but I'm not really maintaining that anymore :-)
I'm not sure who does, Cc:ing Benjamin as he'll probably know.

Nobody is U suspect...

Send it to the list linuxppc-dev@ozlabs.org, it should be picked up
anyway.

(linuxppc-dev CC'd)
---
logical-bitwise & confusion

Signed-off-by: Roel Kluin <redacted>
---

diff --git a/drivers/hwmon/adt7473.c b/drivers/hwmon/adt7473.c
index 9587869..8ea7da2 100644
--- a/drivers/hwmon/adt7473.c
+++ b/drivers/hwmon/adt7473.c

@@ -570,7 +570,7 @@ static ssize_t set_max_duty_at_crit(struct device *dev,
 	struct i2c_client *client = to_i2c_client(dev);
 	struct adt7473_data *data = i2c_get_clientdata(client);
 	int temp = simple_strtol(buf, NULL, 10);
-	temp = temp && 0xFF;
+	temp &= 0xFF;
 
 	mutex_lock(&data->lock);
 	data->max_duty_at_overheat = temp;

The & 0xff here is bogus anyway; temp is only ever used as an u8,
so just declare it as that, or do proper overflow/underflow checking
on it.  The patch will need testing on hardware too, since it changes
behaviour (it should be a bugfix, but who knows).


Segher

Segher Boessenkool wrote:

quoted

diff --git a/drivers/hwmon/adt7473.c b/drivers/hwmon/adt7473.c
index 9587869..8ea7da2 100644
--- a/drivers/hwmon/adt7473.c
+++ b/drivers/hwmon/adt7473.c

@@ -570,7 +570,7 @@ static ssize_t set_max_duty_at_crit(struct device

*dev,
     struct i2c_client *client = to_i2c_client(dev);
     struct adt7473_data *data = i2c_get_clientdata(client);
     int temp = simple_strtol(buf, NULL, 10);
-    temp = temp && 0xFF;
+    temp &= 0xFF;

     mutex_lock(&data->lock);
     data->max_duty_at_overheat = temp;

The & 0xff here is bogus anyway; temp is only ever used as an u8,
so just declare it as that, or do proper overflow/underflow checking
on it.  The patch will need testing on hardware too, since it changes
behaviour (it should be a bugfix, but who knows).

Maybe someone can test this?
---
logical-bitwise & confusion

Signed-off-by: Roel Kluin <redacted>
---

diff --git a/drivers/hwmon/adt7473.c b/drivers/hwmon/adt7473.c
index 9587869..2a2de73 100644
--- a/drivers/hwmon/adt7473.c
+++ b/drivers/hwmon/adt7473.c

@@ -566,11 +566,11 @@ static ssize_t set_max_duty_at_crit(struct device *dev,
 				    const char *buf,
 				    size_t count)
 {
-	u8 reg;
+	u8 reg, temp;
 	struct i2c_client *client = to_i2c_client(dev);
 	struct adt7473_data *data = i2c_get_clientdata(client);
-	int temp = simple_strtol(buf, NULL, 10);
-	temp = temp && 0xFF;
+
+	temp = simple_strtol(buf, NULL, 10) & 0xFF;
 
 	mutex_lock(&data->lock);
 	data->max_duty_at_overheat = temp;

On Mon, Mar 10, 2008 at 10:59:43AM +0100, Roel Kluin wrote:

quoted

The & 0xff here is bogus anyway; temp is only ever used as an u8,
so just declare it as that, or do proper overflow/underflow checking
on it.  The patch will need testing on hardware too, since it changes
behaviour (it should be a bugfix, but who knows).

Maybe someone can test this?

I did.  No regressions observed and it fixes that bug as well.  Sorry I
didn't catch it earlier... :/

Acked-by: Darrick J. Wong <redacted>

--D

Andrew, I think you may want this patch instead of the other
adt746x-logical-bitwise-confusion-in-set_max_duty_at_crit.patch

It includes suggested changes by Segher Boessenkool and I think this
version was tested by Darrick J. Wong
---
logical-bitwise & confusion

Signed-off-by: Roel Kluin <redacted>
---

diff --git a/drivers/hwmon/adt7473.c b/drivers/hwmon/adt7473.c
index 9587869..2a2de73 100644
--- a/drivers/hwmon/adt7473.c
+++ b/drivers/hwmon/adt7473.c

@@ -566,11 +566,11 @@ static ssize_t set_max_duty_at_crit(struct device *dev,
 				    const char *buf,
 				    size_t count)
 {
-	u8 reg;
+	u8 reg, temp;
 	struct i2c_client *client = to_i2c_client(dev);
 	struct adt7473_data *data = i2c_get_clientdata(client);
-	int temp = simple_strtol(buf, NULL, 10);
-	temp = temp && 0xFF;
+
+	temp = simple_strtol(buf, NULL, 10) & 0xFF;
 
 	mutex_lock(&data->lock);
 	data->max_duty_at_overheat = temp;

It includes suggested changes by Segher Boessenkool and I think this
version was tested by Darrick J. Wong

It still does this superfluous "& 0xff", which hides the lack of
range checking.


Segher

Segher Boessenkool wrote:

quoted

It includes suggested changes by Segher Boessenkool and I think this
version was tested by Darrick J. Wong

quoted

-    u8 reg;
+    u8 reg, temp;
     struct i2c_client *client = to_i2c_client(dev);
     struct adt7473_data *data = i2c_get_clientdata(client);
-    int temp = simple_strtol(buf, NULL, 10);
-    temp = temp && 0xFF;
+
+    temp = simple_strtol(buf, NULL, 10) & 0xFF;

It still does this superfluous "& 0xff", which hides the lack of
range checking.

Sorry didn't quite grep that
---
logical-bitwise & confusion

Signed-off-by: Roel Kluin <redacted>
---

diff --git a/drivers/hwmon/adt7473.c b/drivers/hwmon/adt7473.c
index 9587869..98937d3 100644
--- a/drivers/hwmon/adt7473.c
+++ b/drivers/hwmon/adt7473.c

@@ -566,11 +566,11 @@ static ssize_t set_max_duty_at_crit(struct device *dev,
 				    const char *buf,
 				    size_t count)
 {
-	u8 reg;
+	u8 reg, temp;
 	struct i2c_client *client = to_i2c_client(dev);
 	struct adt7473_data *data = i2c_get_clientdata(client);
-	int temp = simple_strtol(buf, NULL, 10);
-	temp = temp && 0xFF;
+
+	temp = simple_strtol(buf, NULL, 10);
 
 	mutex_lock(&data->lock);
 	data->max_duty_at_overheat = temp;