[PATCH dev-5.0 1/2] hwmon (occ): Store error condition for rate-limited polls

[PATCH dev-5.0 1/2] hwmon (occ): Store error condition for rate-limited polls

[Eddie James]

The OCC driver limits the rate of sending poll commands to the OCC. If a
user reads a hwmon entry after a poll response resulted in an error and
is rate-limited, the error is invisible to the user. Fix this by storing
the last error and returning that in the rate-limited case.

Signed-off-by: Eddie James <eajames@linux.ibm.com>
---
 drivers/hwmon/occ/common.c | 4 ++++
 drivers/hwmon/occ/common.h | 3 ++-
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/hwmon/occ/common.c b/drivers/hwmon/occ/common.c
index c888f4a..0e812ef 100644
--- a/drivers/hwmon/occ/common.c
+++ b/drivers/hwmon/occ/common.c
@@ -138,6 +138,7 @@ static int occ_poll(struct occ *occ)
 	/* mutex should already be locked if necessary */
 	rc = occ->send_cmd(occ, cmd);
 	if (rc) {
+		occ->last_error = rc;
 		if (occ->error_count++ > OCC_ERROR_COUNT_THRESHOLD)
 			occ->error = rc;
 
@@ -146,6 +147,7 @@ static int occ_poll(struct occ *occ)
 
 	/* clear error since communication was successful */
 	occ->error_count = 0;
+	occ->last_error = 0;
 	occ->error = 0;
 
 	/* check for safe state */
@@ -207,6 +209,8 @@ int occ_update_response(struct occ *occ)
 	if (time_after(jiffies, occ->last_update + OCC_UPDATE_FREQUENCY)) {
 		rc = occ_poll(occ);
 		occ->last_update = jiffies;
+	} else {
+		rc = occ->last_error;
 	}
 
 	mutex_unlock(&occ->lock);
diff --git a/drivers/hwmon/occ/common.h b/drivers/hwmon/occ/common.h
index 7c44df3..c676e48 100644
--- a/drivers/hwmon/occ/common.h
+++ b/drivers/hwmon/occ/common.h
@@ -105,7 +105,8 @@ struct occ {
 	struct attribute_group group;
 	const struct attribute_group *groups[2];
 
-	int error;                      /* latest transfer error */
+	int error;                      /* final transfer error after retry */
+	int last_error;			/* latest transfer error */
 	unsigned int error_count;       /* number of xfr errors observed */
 	unsigned long last_safe;        /* time OCC entered "safe" state */
 
-- 
2.7.4

[PATCH dev-5.0 2/2] hwmon (occ): Prevent sysfs error attribute from returning error

[Eddie James]

The error sysfs attribute returns the stored error state of the OCC and
doesn't depend on the OCC poll response. Therefore, split the error
attribute into it's own function to avoid failing out of the function if
the poll response fails.

Signed-off-by: Eddie James <eajames@linux.ibm.com>
---
 drivers/hwmon/occ/sysfs.c | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)

diff --git a/drivers/hwmon/occ/sysfs.c b/drivers/hwmon/occ/sysfs.c
index 743b26ec..9bf1526 100644
--- a/drivers/hwmon/occ/sysfs.c
+++ b/drivers/hwmon/occ/sysfs.c
@@ -71,9 +71,6 @@ static ssize_t occ_sysfs_show(struct device *dev,
 		else
 			val = 1;
 		break;
-	case 8:
-		val = occ->error;
-		break;
 	default:
 		return -EINVAL;
 	}
@@ -81,6 +78,16 @@ static ssize_t occ_sysfs_show(struct device *dev,
 	return snprintf(buf, PAGE_SIZE - 1, "%d\n", val);
 }
 
+static ssize_t occ_sysfs_error(struct device *dev,
+			       struct device_attribute *attr, char *buf)
+{
+	struct occ *occ = dev_get_drvdata(dev);
+
+	occ_update_response(occ);
+
+	return snprintf(buf, PAGE_SIZE - 1, "%d\n", occ->error);
+}
+
 static SENSOR_DEVICE_ATTR(occ_master, 0444, occ_sysfs_show, NULL, 0);
 static SENSOR_DEVICE_ATTR(occ_active, 0444, occ_sysfs_show, NULL, 1);
 static SENSOR_DEVICE_ATTR(occ_dvfs_overtemp, 0444, occ_sysfs_show, NULL, 2);
@@ -89,7 +96,7 @@ static SENSOR_DEVICE_ATTR(occ_mem_throttle, 0444, occ_sysfs_show, NULL, 4);
 static SENSOR_DEVICE_ATTR(occ_quick_pwr_drop, 0444, occ_sysfs_show, NULL, 5);
 static SENSOR_DEVICE_ATTR(occ_state, 0444, occ_sysfs_show, NULL, 6);
 static SENSOR_DEVICE_ATTR(occs_present, 0444, occ_sysfs_show, NULL, 7);
-static SENSOR_DEVICE_ATTR(occ_error, 0444, occ_sysfs_show, NULL, 8);
+static DEVICE_ATTR(occ_error, 0444, occ_sysfs_error, NULL);
 
 static struct attribute *occ_attributes[] = {
 	&sensor_dev_attr_occ_master.dev_attr.attr,
@@ -100,7 +107,7 @@ static struct attribute *occ_attributes[] = {
 	&sensor_dev_attr_occ_quick_pwr_drop.dev_attr.attr,
 	&sensor_dev_attr_occ_state.dev_attr.attr,
 	&sensor_dev_attr_occs_present.dev_attr.attr,
-	&sensor_dev_attr_occ_error.dev_attr.attr,
+	&dev_attr_occ_error.attr,
 	NULL
 };
 
@@ -164,7 +171,7 @@ void occ_sysfs_poll_done(struct occ *occ)
 	}
 
 	if (occ->error && occ->error != occ->prev_error) {
-		name = sensor_dev_attr_occ_error.dev_attr.attr.name;
+		name = dev_attr_occ_error.attr.name;
 		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
 	}
 
-- 
2.7.4

Re: [PATCH dev-5.0 2/2] hwmon (occ): Prevent sysfs error attribute from returning error

[Andrew Jeffery]

On Tue, 16 Apr 2019, at 07:58, Eddie James wrote:
> The error sysfs attribute returns the stored error state of the OCC and
> doesn't depend on the OCC poll response. Therefore, split the error
> attribute into it's own function to avoid failing out of the function if
> the poll response fails.
> 
> Signed-off-by: Eddie James <eajames@linux.ibm.com>

Should this be sent upstream?

> ---
>  drivers/hwmon/occ/sysfs.c | 19 +++++++++++++------
>  1 file changed, 13 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/hwmon/occ/sysfs.c b/drivers/hwmon/occ/sysfs.c
> index 743b26ec..9bf1526 100644
> --- a/drivers/hwmon/occ/sysfs.c
> +++ b/drivers/hwmon/occ/sysfs.c
> @@ -71,9 +71,6 @@ static ssize_t occ_sysfs_show(struct device *dev,
>  		else
>  			val = 1;
>  		break;
> -	case 8:
> -		val = occ->error;
> -		break;
>  	default:
>  		return -EINVAL;
>  	}
> @@ -81,6 +78,16 @@ static ssize_t occ_sysfs_show(struct device *dev,
>  	return snprintf(buf, PAGE_SIZE - 1, "%d\n", val);
>  }
>  
> +static ssize_t occ_sysfs_error(struct device *dev,
> +			       struct device_attribute *attr, char *buf)
> +{
> +	struct occ *occ = dev_get_drvdata(dev);
> +
> +	occ_update_response(occ);
> +
> +	return snprintf(buf, PAGE_SIZE - 1, "%d\n", occ->error);
> +}
> +
>  static SENSOR_DEVICE_ATTR(occ_master, 0444, occ_sysfs_show, NULL, 0);
>  static SENSOR_DEVICE_ATTR(occ_active, 0444, occ_sysfs_show, NULL, 1);
>  static SENSOR_DEVICE_ATTR(occ_dvfs_overtemp, 0444, occ_sysfs_show, 
> NULL, 2);
> @@ -89,7 +96,7 @@ static SENSOR_DEVICE_ATTR(occ_mem_throttle, 0444, 
> occ_sysfs_show, NULL, 4);
>  static SENSOR_DEVICE_ATTR(occ_quick_pwr_drop, 0444, occ_sysfs_show, 
> NULL, 5);
>  static SENSOR_DEVICE_ATTR(occ_state, 0444, occ_sysfs_show, NULL, 6);
>  static SENSOR_DEVICE_ATTR(occs_present, 0444, occ_sysfs_show, NULL, 7);
> -static SENSOR_DEVICE_ATTR(occ_error, 0444, occ_sysfs_show, NULL, 8);
> +static DEVICE_ATTR(occ_error, 0444, occ_sysfs_error, NULL);
>  
>  static struct attribute *occ_attributes[] = {
>  	&sensor_dev_attr_occ_master.dev_attr.attr,
> @@ -100,7 +107,7 @@ static struct attribute *occ_attributes[] = {
>  	&sensor_dev_attr_occ_quick_pwr_drop.dev_attr.attr,
>  	&sensor_dev_attr_occ_state.dev_attr.attr,
>  	&sensor_dev_attr_occs_present.dev_attr.attr,
> -	&sensor_dev_attr_occ_error.dev_attr.attr,
> +	&dev_attr_occ_error.attr,
>  	NULL
>  };
>  
> @@ -164,7 +171,7 @@ void occ_sysfs_poll_done(struct occ *occ)
>  	}
>  
>  	if (occ->error && occ->error != occ->prev_error) {
> -		name = sensor_dev_attr_occ_error.dev_attr.attr.name;
> +		name = dev_attr_occ_error.attr.name;
>  		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
>  	}
>  
> -- 
> 2.7.4
> 
>

Re: [PATCH dev-5.0 1/2] hwmon (occ): Store error condition for rate-limited polls

[Andrew Jeffery]

On Tue, 16 Apr 2019, at 07:59, Eddie James wrote:
> The OCC driver limits the rate of sending poll commands to the OCC. If a
> user reads a hwmon entry after a poll response resulted in an error and
> is rate-limited, the error is invisible to the user. Fix this by storing
> the last error and returning that in the rate-limited case.
> 
> Signed-off-by: Eddie James <eajames@linux.ibm.com>

Should this be sent upstream?

> ---
>  drivers/hwmon/occ/common.c | 4 ++++
>  drivers/hwmon/occ/common.h | 3 ++-
>  2 files changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/hwmon/occ/common.c b/drivers/hwmon/occ/common.c
> index c888f4a..0e812ef 100644
> --- a/drivers/hwmon/occ/common.c
> +++ b/drivers/hwmon/occ/common.c
> @@ -138,6 +138,7 @@ static int occ_poll(struct occ *occ)
>  	/* mutex should already be locked if necessary */
>  	rc = occ->send_cmd(occ, cmd);
>  	if (rc) {
> +		occ->last_error = rc;
>  		if (occ->error_count++ > OCC_ERROR_COUNT_THRESHOLD)
>  			occ->error = rc;
>  
> @@ -146,6 +147,7 @@ static int occ_poll(struct occ *occ)
>  
>  	/* clear error since communication was successful */
>  	occ->error_count = 0;
> +	occ->last_error = 0;
>  	occ->error = 0;
>  
>  	/* check for safe state */
> @@ -207,6 +209,8 @@ int occ_update_response(struct occ *occ)
>  	if (time_after(jiffies, occ->last_update + OCC_UPDATE_FREQUENCY)) {
>  		rc = occ_poll(occ);
>  		occ->last_update = jiffies;
> +	} else {
> +		rc = occ->last_error;
>  	}
>  
>  	mutex_unlock(&occ->lock);
> diff --git a/drivers/hwmon/occ/common.h b/drivers/hwmon/occ/common.h
> index 7c44df3..c676e48 100644
> --- a/drivers/hwmon/occ/common.h
> +++ b/drivers/hwmon/occ/common.h
> @@ -105,7 +105,8 @@ struct occ {
>  	struct attribute_group group;
>  	const struct attribute_group *groups[2];
>  
> -	int error;                      /* latest transfer error */
> +	int error;                      /* final transfer error after retry */
> +	int last_error;			/* latest transfer error */
>  	unsigned int error_count;       /* number of xfr errors observed */
>  	unsigned long last_safe;        /* time OCC entered "safe" state */
>  
> -- 
> 2.7.4
> 
>