Skip to contentSkip to conversation

Export 1-wire thermal sensors as hwmon device

PATCH Superseded Jaghathiswari Rankappagounder Natarajan 4 participants 10 messages Jul 12, 2017 – Jul 17, 2017

Subsystems (2)

v2 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v2 cover Export 1-wire thermal sensors as hwmon device

Our board has 4 DS18B20 1-wire temperature sensors. Each 1-wire bus and the sensor under it is already configured against the Linux 1-wire driver (called w1). They have a sysfs file(e.g. /sys/bus/w1/devices/w1_bus_master1/28-000007704f4c/w1_slave) which, when read, runs code to read the temperature. We'd like the temperatures to show up in hwmon, so that the BMC IPMI sensor plumbing can forward those to host.

The changes in version 2 are mentioned in the individual patches.

This patchset is based on linux mainline version v4.10.

Tested:
Yes; On a board with 4 DS18B20 1-wire temperature sensors:

root@zaius:/sys/class/hwmon# ls
hwmon0  hwmon1  hwmon2  hwmon3  hwmon4  hwmon5
root@zaius:/sys/class/hwmon# cd hwmon0
root@zaius:/sys/class/hwmon/hwmon0# ls
device       name         subsystem    temp1_input  uevent
root@zaius:/sys/class/hwmon/hwmon0# cat temp1_input
24500
root@zaius:/sys/class/hwmon/hwmon0# cd ..
root@zaius:/sys/class/hwmon# cd hwmon1
root@zaius:/sys/class/hwmon/hwmon1# cat temp1_input
26562
root@zaius:/sys/class/hwmon/hwmon1# cd ..
root@zaius:/sys/class/hwmon# cd hwmon2
root@zaius:/sys/class/hwmon/hwmon2# cat temp1_input
27250
root@zaius:/sys/class/hwmon/hwmon2# cd ..
root@zaius:/sys/class/hwmon# cd hwmon3
root@zaius:/sys/class/hwmon/hwmon3# cat temp1_input
22250
root@zaius:/sys/class/hwmon/hwmon3#

Jaghathiswari Rankappagounder Natarajan (3):
drivers: w1: add hwmon support structures
drivers: w1: refactor w1_slave_show to make the temp reading
functionality separate
drivers: w1: add hwmon temp support for w1_therm

 drivers/w1/slaves/w1_therm.c | 197 ++++++++++++++++++++++++++++++++++++-------
 drivers/w1/w1.c              |  21 ++++-
 drivers/w1/w1.h              |   2 +
 drivers/w1/w1_family.h       |   2 +
 4 files changed, 191 insertions(+), 31 deletions(-)

--
2.13.2.932.g7449e964c-goog

#1 lore raw

PATCH v2 1/3 drivers: w1: add hwmon support structures

This patch has changes to w1.h/w1.c/w1_family.h generic files to add (optional) hwmon support structures.

---
v2
- made changes to support hwmon_device_register_with_info mentioned by Guenter.

 drivers/w1/w1.c        | 21 ++++++++++++++++++++-
 drivers/w1/w1.h        |  2 ++
 drivers/w1/w1_family.h |  2 ++
 3 files changed, 24 insertions(+), 1 deletion(-)
drivers/w1/w1.c +20 -1
index e213c678bbfe..c37e21597c5c 100644
@@ -32,6 +32,7 @@
 #include <linux/sched.h>
 #include <linux/kthread.h>
 #include <linux/freezer.h>
+#include <linux/hwmon.h>
 
 #include <linux/atomic.h>
 
@@ -659,13 +660,31 @@ static int w1_family_notify(unsigned long action, struct w1_slave *sl)
 				return err;
 			}
 		}
-
+#ifdef CONFIG_HWMON
+		if (fops->chip_info) {
+			struct device *hwmon
+				= hwmon_device_register_with_info(&sl->dev,
+						"w1_slave_temp", sl,
+						fops->chip_info,
+						NULL);
+			if (IS_ERR(hwmon)) {
+				dev_warn(&sl->dev,
+					 "could not create hwmon device\n");
+			} else {
+				sl->hwmon = hwmon;
+			}
+		}
+#endif
 		break;
 	case BUS_NOTIFY_DEL_DEVICE:
 		if (fops->remove_slave)
 			sl->family->fops->remove_slave(sl);
 		if (fops->groups)
 			sysfs_remove_groups(&sl->dev.kobj, fops->groups);
+#ifdef CONFIG_HWMON
+		if (fops->chip_info && sl->hwmon)
+			hwmon_device_unregister(sl->hwmon);
+#endif
 		break;
 	}
 	return 0;
drivers/w1/w1.h +2 -0
index 129895f562b0..74ef1a543856 100644
@@ -83,6 +83,7 @@ struct w1_reg_num
  * @family: module for device family type
  * @family_data: pointer for use by the family module
  * @dev: kernel device identifier
+ * @hwmon: pointer to hwmon device
  *
  */
 struct w1_slave
@@ -99,6 +100,7 @@ struct w1_slave
 	struct w1_family	*family;
 	void			*family_data;
 	struct device		dev;
+	struct device		*hwmon;
 };
 
 typedef void (*w1_slave_found_callback)(struct w1_master *, u64);
drivers/w1/w1_family.h +2 -0
index 10a7a0767187..9001851e1316 100644
@@ -55,12 +55,14 @@ struct w1_slave;
  * @add_slave: add_slave
  * @remove_slave: remove_slave
  * @groups: sysfs group
+ * @chip_info: pointer to struct hwmon_chip_info
  */
 struct w1_family_ops
 {
 	int  (* add_slave)(struct w1_slave *);
 	void (* remove_slave)(struct w1_slave *);
 	const struct attribute_group **groups;
+	const struct hwmon_chip_info *chip_info;
 };
 
 /**

--
2.13.2.932.g7449e964c-goog

maintainer
#2 lore raw

Re: [PATCH linux v2 1/3] drivers: w1: add hwmon support structures

Quoted 27 lines · Jaghathiswari Rankappagounder Natarajan

This patch has changes to w1.h/w1.c/w1_family.h generic files to add (optional) hwmon support structures.

---
v2
- made changes to support hwmon_device_register_with_info mentioned by Guenter.

  drivers/w1/w1.c        | 21 ++++++++++++++++++++-
  drivers/w1/w1.h        |  2 ++
  drivers/w1/w1_family.h |  2 ++
  3 files changed, 24 insertions(+), 1 deletion(-)
drivers/w1/w1.c +1 -0
index e213c678bbfe..c37e21597c5c 100644
@@ -32,6 +32,7 @@
  #include <linux/sched.h>
  #include <linux/kthread.h>
  #include <linux/freezer.h>
+#include <linux/hwmon.h>
 
  #include <linux/atomic.h>
 

drivers/w1/w1.c static int w1_family_notify(unsigned long action, struct w1_slave *sl) · line 663

Quoted 5 lines
drivers/w1/w1.c +1 -1
@@ -659,13 +660,31 @@ static int w1_family_notify(unsigned long action, struct w1_slave *sl)
  				return err;
  			}
  		}
-
+#ifdef CONFIG_HWMON

HWMON can be built as module. If so, this won't be enabled, even if W1 is also built as module. Better would be

#if IS_REACHABLE(CONFIG_HWMON)

I personally would probably use

	if (IS_REACHABLE(CONFIG_HWMON) && fops->chip_info) {

to avoid the ifdef and to ensure that the code always compiles. But that is a personal preference.

Quoted 14 lines
drivers/w1/w1.c +14 -0
+		if (fops->chip_info) {
+			struct device *hwmon
+				= hwmon_device_register_with_info(&sl->dev,
+						"w1_slave_temp", sl,
+						fops->chip_info,
+						NULL);
+			if (IS_ERR(hwmon)) {
+				dev_warn(&sl->dev,
+					 "could not create hwmon device\n");
+			} else {
+				sl->hwmon = hwmon;
+			}
+		}
+#endif

drivers/w1/w1.c static int w1_family_notify(unsigned long action, struct w1_slave *sl) · line 687

Quoted 10 lines
drivers/w1/w1.c +4 -0
  		break;
  	case BUS_NOTIFY_DEL_DEVICE:
  		if (fops->remove_slave)
  			sl->family->fops->remove_slave(sl);
  		if (fops->groups)
  			sysfs_remove_groups(&sl->dev.kobj, fops->groups);
+#ifdef CONFIG_HWMON
+		if (fops->chip_info && sl->hwmon)
+			hwmon_device_unregister(sl->hwmon);
+#endif

Not completely understanding the code, I may be wrong, but wouldn't it be better to remove the sysfs group and the hwmon device before calling remove_slave() ?

Quoted 50 lines
drivers/w1/w1.c +0 -0
  		break;
  	}
  	return 0;
drivers/w1/w1.h +2 -0
index 129895f562b0..74ef1a543856 100644
@@ -83,6 +83,7 @@ struct w1_reg_num
   * @family: module for device family type
   * @family_data: pointer for use by the family module
   * @dev: kernel device identifier
+ * @hwmon: pointer to hwmon device
   *
   */
  struct w1_slave
@@ -99,6 +100,7 @@ struct w1_slave
  	struct w1_family	*family;
  	void			*family_data;
  	struct device		dev;
+	struct device		*hwmon;
  };
 
  typedef void (*w1_slave_found_callback)(struct w1_master *, u64);
drivers/w1/w1_family.h +2 -1
index 10a7a0767187..9001851e1316 100644
@@ -55,12 +55,14 @@ struct w1_slave;
   * @add_slave: add_slave
   * @remove_slave: remove_slave
   * @groups: sysfs group
+ * @chip_info: pointer to struct hwmon_chip_info
   */
  struct w1_family_ops
  {
  	int  (* add_slave)(struct w1_slave *);
  	void (* remove_slave)(struct w1_slave *);
  	const struct attribute_group **groups;
+	const struct hwmon_chip_info *chip_info;
  };
 
  /**
--

2.13.2.932.g7449e964c-goog

--
To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw

PATCH v2 2/3 drivers: w1: refactor w1_slave_show to make the temp reading functionality separate

In reply to Jaghathiswari Rankappagounder Natarajan

Inside the w1_slave_show function refactor the code to read the temp into a separate function.

---
v2
- no changes

 drivers/w1/slaves/w1_therm.c | 82 +++++++++++++++++++++++++++-----------------
 1 file changed, 51 insertions(+), 31 deletions(-)
drivers/w1/slaves/w1_therm.c +51 -31
index 82611f197b0a..88f4cc8c3908 100644
@@ -64,6 +64,12 @@ struct w1_therm_family_data {
 	atomic_t refcnt;
 };
 
+struct therm_info {
+	u8 rom[9];
+	u8 crc;
+	u8 verdict;
+};
+
 /* return the address of the refcnt in the family data */
 #define THERM_REFCNT(family_data) \
 	(&((struct w1_therm_family_data *)family_data)->refcnt)
@@ -427,33 +433,23 @@ static ssize_t w1_slave_store(struct device *device,
 	return ret ? : size;
 }
 
-static ssize_t w1_slave_show(struct device *device,
-	struct device_attribute *attr, char *buf)
+static ssize_t read_therm(struct device *device,
+			  struct w1_slave *sl, struct therm_info *info)
 {
-	struct w1_slave *sl = dev_to_w1_slave(device);
 	struct w1_master *dev = sl->master;
-	u8 rom[9], crc, verdict, external_power;
-	int i, ret, max_trying = 10;
-	ssize_t c = PAGE_SIZE;
-	u8 *family_data = sl->family_data;
+	u8 external_power;
+	int ret, max_trying = 10;
 
 	ret = mutex_lock_interruptible(&dev->bus_mutex);
 	if (ret != 0)
 		goto post_unlock;
 
-	if (!sl->family_data) {
-		ret = -ENODEV;
-		goto pre_unlock;
-	}
-
-	/* prevent the slave from going away in sleep */
-	atomic_inc(THERM_REFCNT(family_data));
-	memset(rom, 0, sizeof(rom));
+	memset(info->rom, 0, sizeof(info->rom));
 
 	while (max_trying--) {
 
-		verdict = 0;
-		crc = 0;
+		info->verdict = 0;
+		info->crc = 0;
 
 		if (!w1_reset_select_slave(sl)) {
 			int count = 0;
@@ -496,30 +492,57 @@ static ssize_t w1_slave_show(struct device *device,
 			if (!w1_reset_select_slave(sl)) {
 
 				w1_write_8(dev, W1_READ_SCRATCHPAD);
-				count = w1_read_block(dev, rom, 9);
+				count = w1_read_block(dev, info->rom, 9);
 				if (count != 9) {
 					dev_warn(device, "w1_read_block() "
 						"returned %u instead of 9.\n",
 						count);
 				}
 
-				crc = w1_calc_crc8(rom, 8);
+				info->crc = w1_calc_crc8(info->rom, 8);
 
-				if (rom[8] == crc)
-					verdict = 1;
+				if (info->rom[8] == info->crc)
+					info->verdict = 1;
 			}
 		}
 
-		if (verdict)
+		if (info->verdict)
 			break;
 	}
 
+pre_unlock:
+	mutex_unlock(&dev->bus_mutex);
+
+post_unlock:
+	return ret;
+}
+
+static ssize_t w1_slave_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
+{
+	struct w1_slave *sl = dev_to_w1_slave(device);
+	u8 *family_data = sl->family_data;
+	struct therm_info info;
+	int ret;
+	ssize_t c = PAGE_SIZE;
+	int i;
+
+	if (!sl->family_data)
+		return -ENODEV;
+
+	/* prevent the slave from going away in sleep */
+	atomic_inc(THERM_REFCNT(family_data));
+
+	ret = read_therm(device, sl, &info);
+	if (ret)
+		goto dec_refcnt;
+
 	for (i = 0; i < 9; ++i)
-		c -= snprintf(buf + PAGE_SIZE - c, c, "%02x ", rom[i]);
+		c -= snprintf(buf + PAGE_SIZE - c, c, "%02x ", info.rom[i]);
 	c -= snprintf(buf + PAGE_SIZE - c, c, ": crc=%02x %s\n",
-		      crc, (verdict) ? "YES" : "NO");
-	if (verdict)
-		memcpy(family_data, rom, sizeof(rom));
+		      info.crc, (info.verdict) ? "YES" : "NO");
+	if (info.verdict)
+		memcpy(family_data, info.rom, sizeof(info.rom));
 	else
 		dev_warn(device, "Read failed CRC check\n");
 
@@ -528,13 +551,10 @@ static ssize_t w1_slave_show(struct device *device,
 			      ((u8 *)family_data)[i]);
 
 	c -= snprintf(buf + PAGE_SIZE - c, c, "t=%d\n",
-		w1_convert_temp(rom, sl->family->fid));
+			w1_convert_temp(info.rom, sl->family->fid));
 	ret = PAGE_SIZE - c;
 
-pre_unlock:
-	mutex_unlock(&dev->bus_mutex);
-
-post_unlock:
+dec_refcnt:
 	atomic_dec(THERM_REFCNT(family_data));
 	return ret;
 }

--
2.13.2.932.g7449e964c-goog

#4 lore raw

PATCH v2 3/3 drivers: w1: add hwmon temp support for w1_therm

In reply to Jaghathiswari Rankappagounder Natarajan

This change adds hwmon temp support for w1_therm.

---
v2
- make changes to support hwmon_device_register_with_info mentioned by Guenter.

 drivers/w1/slaves/w1_therm.c | 117 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 117 insertions(+)
drivers/w1/slaves/w1_therm.c +117 -0
index 88f4cc8c3908..147f9dfa13de 100644
@@ -29,6 +29,7 @@
 #include <linux/types.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/hwmon.h>
 
 #include "../w1.h"
 #include "../w1_int.h"
@@ -118,19 +119,61 @@ static struct attribute *w1_ds28ea00_attrs[] = {
 	&dev_attr_w1_seq.attr,
 	NULL,
 };
+
 ATTRIBUTE_GROUPS(w1_therm);
 ATTRIBUTE_GROUPS(w1_ds28ea00);
 
+static umode_t w1_is_visible(const void *_data,
+			     enum hwmon_sensor_types type,
+			     u32 attr, int channel);
+
+static int w1_read_temp(struct device *dev, u32 attr, int channel,
+			long *val);
+
+static int w1_read(struct device *dev, enum hwmon_sensor_types type,
+		   u32 attr, int channel, long *val);
+
+static int w1_write(struct device *dev, enum hwmon_sensor_types type,
+		    u32 attr, int channel, long val);
+
+static const u32 w1_temp_config[] = {
+	HWMON_T_INPUT,
+	0
+};
+
+static const struct hwmon_channel_info w1_temp = {
+	.type = hwmon_temp,
+	.config = w1_temp_config,
+};
+
+static const struct hwmon_channel_info *w1_info[] = {
+	&w1_temp,
+	NULL
+};
+
+static const struct hwmon_ops w1_hwmon_ops = {
+	.is_visible = w1_is_visible,
+	.read = w1_read,
+	.write = w1_write,
+};
+
+static const struct hwmon_chip_info w1_chip_info = {
+	.ops = &w1_hwmon_ops,
+	.info = w1_info,
+};
+
 static struct w1_family_ops w1_therm_fops = {
 	.add_slave	= w1_therm_add_slave,
 	.remove_slave	= w1_therm_remove_slave,
 	.groups		= w1_therm_groups,
+	.chip_info	= &w1_chip_info,
 };
 
 static struct w1_family_ops w1_ds28ea00_fops = {
 	.add_slave	= w1_therm_add_slave,
 	.remove_slave	= w1_therm_remove_slave,
 	.groups		= w1_ds28ea00_groups,
+	.chip_info	= &w1_chip_info,
 };
 
 static struct w1_family w1_therm_family_DS18S20 = {
@@ -559,6 +602,80 @@ static ssize_t w1_slave_show(struct device *device,
 	return ret;
 }
 
+static int w1_read_temp(struct device *device, u32 attr, int channel,
+			long *val)
+{
+	struct w1_slave *sl = dev_get_drvdata(device);
+	u8 *family_data = sl->family_data;
+	struct therm_info info;
+	int ret;
+
+	if (!sl->family_data)
+		return -ENODEV;
+
+	/* prevent the slave from going away in sleep */
+	atomic_inc(THERM_REFCNT(family_data));
+
+	switch (attr) {
+	case hwmon_temp_input:
+	ret = read_therm(device, sl, &info);
+	if (ret)
+		goto dec_refcnt;
+
+	if (!info.verdict) {
+		dev_warn(device, "Read failed CRC check\n");
+		ret = -EIO;
+		goto dec_refcnt;
+	}
+
+	*val = w1_convert_temp(info.rom, sl->family->fid);
+	ret = 0;
+	break;
+	default:
+		ret = -EOPNOTSUPP;
+		break;
+	}
+
+dec_refcnt:
+	atomic_dec(THERM_REFCNT(family_data));
+	return ret;
+}
+
+static umode_t w1_is_visible(const void *_data, enum hwmon_sensor_types type,
+			     u32 attr, int channel)
+{
+	umode_t mode = 0444;
+
+	switch (attr) {
+	case hwmon_temp_input:
+		break;
+	default:
+		mode = 0;
+		break;
+	}
+	return mode;
+}
+
+static int w1_read(struct device *dev, enum hwmon_sensor_types type,
+		   u32 attr, int channel, long *val)
+{
+	switch (type) {
+	case hwmon_temp:
+		return w1_read_temp(dev, attr, channel, val);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static int w1_write(struct device *dev, enum hwmon_sensor_types type,
+		    u32 attr, int channel, long val)
+{
+	switch (type) {
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 #define W1_42_CHAIN	0x99
 #define W1_42_CHAIN_OFF	0x3C
 #define W1_42_CHAIN_OFF_INV	0xC3

--
2.13.2.932.g7449e964c-goog

maintainer
#5 lore raw

Re: [linux,v2,3/3] drivers: w1: add hwmon temp support for w1_therm

Quoted 32 lines · Jaghathiswari Rankappagounder Natarajan

This change adds hwmon temp support for w1_therm.

---
v2
- make changes to support hwmon_device_register_with_info mentioned by Guenter.

 drivers/w1/slaves/w1_therm.c | 117 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 117 insertions(+)

--
2.13.2.932.g7449e964c-goog

--
To unsubscribe from this list: send the line "unsubscribe linux-hwmon" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

drivers/w1/slaves/w1_therm.c +1 -0
index 88f4cc8c3908..147f9dfa13de 100644
@@ -29,6 +29,7 @@
 #include <linux/types.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/hwmon.h>
 
 #include "../w1.h"
 #include "../w1_int.h"

drivers/w1/slaves/w1_therm.c static struct attribute *w1_ds28ea00_attrs[] = { · line 125

Quoted 7 lines
drivers/w1/slaves/w1_therm.c +1 -0
@@ -118,19 +119,61 @@ static struct attribute *w1_ds28ea00_attrs[] = {
 	&dev_attr_w1_seq.attr,
 	NULL,
 };
+
 ATTRIBUTE_GROUPS(w1_therm);
 ATTRIBUTE_GROUPS(w1_ds28ea00);
 

Do you want the following code all the time, or enclose it in IS_REACHABLE() ?

drivers/w1/slaves/w1_therm.c static struct attribute *w1_ds28ea00_attrs[] = { · line 138

Quoted 13 lines
drivers/w1/slaves/w1_therm.c +13 -0
+static umode_t w1_is_visible(const void *_data,
+			     enum hwmon_sensor_types type,
+			     u32 attr, int channel);
+
+static int w1_read_temp(struct device *dev, u32 attr, int channel,
+			long *val);
+
+static int w1_read(struct device *dev, enum hwmon_sensor_types type,
+		   u32 attr, int channel, long *val);
+
+static int w1_write(struct device *dev, enum hwmon_sensor_types type,
+		    u32 attr, int channel, long val);
+

Any chance you can rearrange the code to not require forward declarations ?

Quoted 44 lines
drivers/w1/slaves/w1_therm.c +28 -0
+static const u32 w1_temp_config[] = {
+	HWMON_T_INPUT,
+	0
+};
+
+static const struct hwmon_channel_info w1_temp = {
+	.type = hwmon_temp,
+	.config = w1_temp_config,
+};
+
+static const struct hwmon_channel_info *w1_info[] = {
+	&w1_temp,
+	NULL
+};
+
+static const struct hwmon_ops w1_hwmon_ops = {
+	.is_visible = w1_is_visible,
+	.read = w1_read,
+	.write = w1_write,
+};
+
+static const struct hwmon_chip_info w1_chip_info = {
+	.ops = &w1_hwmon_ops,
+	.info = w1_info,
+};
+
 static struct w1_family_ops w1_therm_fops = {
 	.add_slave	= w1_therm_add_slave,
 	.remove_slave	= w1_therm_remove_slave,
 	.groups		= w1_therm_groups,
+	.chip_info	= &w1_chip_info,
 };
 
 static struct w1_family_ops w1_ds28ea00_fops = {
 	.add_slave	= w1_therm_add_slave,
 	.remove_slave	= w1_therm_remove_slave,
 	.groups		= w1_ds28ea00_groups,
+	.chip_info	= &w1_chip_info,
 };
 
 static struct w1_family w1_therm_family_DS18S20 = {
@@ -559,6 +602,80 @@ static ssize_t w1_slave_show(struct device *device,
 	return ret;
 }

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 613

Quoted 10 lines
drivers/w1/slaves/w1_therm.c +9 -0
 
+static int w1_read_temp(struct device *device, u32 attr, int channel,
+			long *val)
+{
+	struct w1_slave *sl = dev_get_drvdata(device);
+	u8 *family_data = sl->family_data;
+	struct therm_info info;
+	int ret;
+
+	if (!sl->family_data)

Why not use family_data here ?

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 617

Quoted 4 lines
drivers/w1/slaves/w1_therm.c +4 -0
+		return -ENODEV;
+
+	/* prevent the slave from going away in sleep */
+	atomic_inc(THERM_REFCNT(family_data));

The original code increments the reference count with the bus mutex locked. Similar, the original code only checks if family_data is NULL after locking the bus mutex.

Are you sure this change in sequence doesn't cause race conditions ? If this function can be called with family_data == NULL, what guarantees that it isn't set to NULL after being checked but before the reference count is increased (or, in other words, that the atomic_inc() doesn't increment a reference counter in freed memory) ?

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 620

Quoted 3 lines
drivers/w1/slaves/w1_therm.c +3 -0
+
+	switch (attr) {
+	case hwmon_temp_input:

indentation below is off.

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 626

Quoted 6 lines
drivers/w1/slaves/w1_therm.c +6 -0
+	ret = read_therm(device, sl, &info);
+	if (ret)
+		goto dec_refcnt;
+
+	if (!info.verdict) {
+		dev_warn(device, "Read failed CRC check\n");

I personally dislike such noise. Personal preference, of course.

Quoted 20 lines
drivers/w1/slaves/w1_therm.c +20 -0
+		ret = -EIO;
+		goto dec_refcnt;
+	}
+
+	*val = w1_convert_temp(info.rom, sl->family->fid);
+	ret = 0;
+	break;
+	default:
+		ret = -EOPNOTSUPP;
+		break;
+	}
+
+dec_refcnt:
+	atomic_dec(THERM_REFCNT(family_data));
+	return ret;
+}
+
+static umode_t w1_is_visible(const void *_data, enum hwmon_sensor_types type,
+			     u32 attr, int channel)
+{

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 656

Quoted 10 lines
drivers/w1/slaves/w1_therm.c +10 -0
+	umode_t mode = 0444;
+
+	switch (attr) {
+	case hwmon_temp_input:
+		break;
+	default:
+		mode = 0;
+		break;
+	}
+	return mode;
	return attr == hwmon_temp_input ? 0444 : 0;

would be a bit simpler here.

Quoted 5 lines
drivers/w1/slaves/w1_therm.c +5 -0
+}
+
+static int w1_read(struct device *dev, enum hwmon_sensor_types type,
+		   u32 attr, int channel, long *val)
+{

drivers/w1/slaves/w1_therm.c static ssize_t w1_slave_show(struct device *device, · line 671

Quoted 10 lines
drivers/w1/slaves/w1_therm.c +10 -0
+	switch (type) {
+	case hwmon_temp:
+		return w1_read_temp(dev, attr, channel, val);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static int w1_write(struct device *dev, enum hwmon_sensor_types type,
+		    u32 attr, int channel, long val)

You don't need a write function, since attributes are never writeable.

Quoted 10 lines
drivers/w1/slaves/w1_therm.c +7 -0
+{
+	switch (type) {
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 #define W1_42_CHAIN	0x99
 #define W1_42_CHAIN_OFF	0x3C
 #define W1_42_CHAIN_OFF_INV	0xC3
#6 lore raw
In reply to Jaghathiswari Rankappagounder Natarajan

Hi

13.07.2017, 00:41, "Jaghathiswari Rankappagounder Natarajan" <[email protected]>:

Quoted 10 lines

Our board has 4 DS18B20 1-wire temperature sensors. Each 1-wire bus and the sensor under it is already configured against the Linux 1-wire driver (called w1). They have a sysfs file(e.g. /sys/bus/w1/devices/w1_bus_master1/28-000007704f4c/w1_slave) which, when read, runs code to read the temperature. We'd like the temperatures to show up in hwmon, so that the BMC IPMI sensor plumbing can forward those to host.

The changes in version 2 are mentioned in the individual patches.

This patchset is based on linux mainline version v4.10.

I believe this is a resend of your previous patchet, isn't it? Greg, if you hadn't yet, please pull it into the tree.

Thank you.

#7 lore raw

Hi

13.07.2017, 18:36, "Evgeniy Polyakov" <[email protected]>:

Quoted 2 lines

I believe this is a resend of your previous patchet, isn't it? Greg, if you hadn't yet, please pull it into the tree.

Aaand it looks like we have a discussion now, so please postpone it for a while :)

#8 lore raw

Hi Evgeniy/Greg,

Please pull in version 4 patchset into the tree.

Quoted 7 lines · Evgeniy Polyakov

Hi

13.07.2017, 18:36, "Evgeniy Polyakov" <[email protected]>:

Quoted 2 lines

I believe this is a resend of your previous patchet, isn't it? Greg, if you hadn't yet, please pull it into the tree.

Aaand it looks like we have a discussion now, so please postpone it for a while :)

#9 lore raw
Quoted 3 lines · Jaghathiswari Rankappagounder Natarajan

Hi Evgeniy/Greg,

Please pull in version 4 patchset into the tree.

I have no idea what that means :(

Please resend the correct ones so I know what to do here.

thanks,

greg k-h