Skip to contentSkip to conversation

sysfs mount_matrix for st_lsm6dsx gyro

DISCUSSION Philippe De Muyter 3 participants 8 messages Jan 11, 2023 – Jan 13, 2023

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Some more info :

I have backported drivers/iio/imu/st_lsm6dsx to linux-4.9 in order to drive a ism330dlc imu on a custom board. The chip is correctly detected and two devices are created in /sys/bus/iio/devices/

the first one (where name is 'ism330dlc_gyro') has the following entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device1/
 buffer                      in_anglvel_x_raw  sampling_frequency
 current_timestamp_clock     in_anglvel_y_raw  sampling_frequency_available
 dev                         in_anglvel_z_raw  scan_elements
 in_anglvel_scale            name              subsystem
 in_anglvel_scale_available  power             uevent
 me@proto4:~$

the second one (where name is 'ism330dlc_accel') has those entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device2
 buffer                    in_accel_x_raw  sampling_frequency
 current_timestamp_clock   in_accel_y_raw  sampling_frequency_available
 dev                       in_accel_z_raw  scan_elements
 events                    mount_matrix    subsystem
 in_accel_scale            name            uevent
 in_accel_scale_available  power
 me@proto4:~$

The 'mount_matrix' entry is only present in the 'ism330dlc_accel' device but not in the 'ism330dlc_gyro' device.

On a similar board, but with mpu9250 imu, I get only one iio:deviceX entry but with two *mount_matrix entries :

in_accel_mount_matrix
in_anglvel_mount_matrix

In both cases, I would have expected only one 'iio:deviceX' entry with only one 'mount_matrix' entry.

Best regards

Philippe

Signature · 1 line
Philippe De Muyter +32 2 6101532 Macq SA rue de l'Aeronef 2 B-1140 Bruxelles
#1 lore raw
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

Quoted 31 lines

Some more info :

I have backported drivers/iio/imu/st_lsm6dsx to linux-4.9 in order to drive a ism330dlc imu on a custom board. The chip is correctly detected and two devices are created in /sys/bus/iio/devices/

the first one (where name is 'ism330dlc_gyro') has the following entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device1/
 buffer                      in_anglvel_x_raw  sampling_frequency
 current_timestamp_clock     in_anglvel_y_raw  sampling_frequency_available
 dev                         in_anglvel_z_raw  scan_elements
 in_anglvel_scale            name              subsystem
 in_anglvel_scale_available  power             uevent
 me@proto4:~$

the second one (where name is 'ism330dlc_accel') has those entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device2
 buffer                    in_accel_x_raw  sampling_frequency
 current_timestamp_clock   in_accel_y_raw  sampling_frequency_available
 dev                       in_accel_z_raw  scan_elements
 events                    mount_matrix    subsystem
 in_accel_scale            name            uevent
 in_accel_scale_available  power
 me@proto4:~$

The 'mount_matrix' entry is only present in the 'ism330dlc_accel' device but not in the 'ism330dlc_gyro' device.

Quoted 8 lines

On a similar board, but with mpu9250 imu, I get only one iio:deviceX entry but with two *mount_matrix entries :

in_accel_mount_matrix
in_anglvel_mount_matrix

In both cases, I would have expected only one 'iio:deviceX' entry with only one 'mount_matrix' entry.

There are multiple devices because the driver predates the addition of multiple buffer support to IIO and IIRC is capable of producing data at different sampling rates for the accelerometer and the gyros. Hence when it was implemented the only choice was to register multiple devices in order to get the multiple buffers. It's ABI now so we can't fix it in an old driver unfortunately. We'd do this differently today..

The double mount_matrix for the mpu9250 is a little odd and I can't immediately spot why that one is happening.

Quoted 5 lines

Best regards

Philippe

maintainer
#2 lore raw
Quoted 10 lines
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Quoted 5 lines

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

I am not sure if accel and gyro can be mounted with a different orientation. Do you think we should have a per-sensor mount_matrix?

Regards,
Lorenzo

Quoted 57 lines
Quoted 31 lines

Some more info :

I have backported drivers/iio/imu/st_lsm6dsx to linux-4.9 in order to drive a ism330dlc imu on a custom board. The chip is correctly detected and two devices are created in /sys/bus/iio/devices/

the first one (where name is 'ism330dlc_gyro') has the following entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device1/
 buffer                      in_anglvel_x_raw  sampling_frequency
 current_timestamp_clock     in_anglvel_y_raw  sampling_frequency_available
 dev                         in_anglvel_z_raw  scan_elements
 in_anglvel_scale            name              subsystem
 in_anglvel_scale_available  power             uevent
 me@proto4:~$

the second one (where name is 'ism330dlc_accel') has those entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device2
 buffer                    in_accel_x_raw  sampling_frequency
 current_timestamp_clock   in_accel_y_raw  sampling_frequency_available
 dev                       in_accel_z_raw  scan_elements
 events                    mount_matrix    subsystem
 in_accel_scale            name            uevent
 in_accel_scale_available  power
 me@proto4:~$

The 'mount_matrix' entry is only present in the 'ism330dlc_accel' device but not in the 'ism330dlc_gyro' device.

Quoted 8 lines

On a similar board, but with mpu9250 imu, I get only one iio:deviceX entry but with two *mount_matrix entries :

in_accel_mount_matrix
in_anglvel_mount_matrix

In both cases, I would have expected only one 'iio:deviceX' entry with only one 'mount_matrix' entry.

There are multiple devices because the driver predates the addition of multiple buffer support to IIO and IIRC is capable of producing data at different sampling rates for the accelerometer and the gyros. Hence when it was implemented the only choice was to register multiple devices in order to get the multiple buffers. It's ABI now so we can't fix it in an old driver unfortunately. We'd do this differently today..

The double mount_matrix for the mpu9250 is a little odd and I can't immediately spot why that one is happening.

Quoted 5 lines

Best regards

Philippe

#3 lore raw

Hello Lorenzo,

Quoted 16 lines · Lorenzo Bianconi

Date: Thu, 12 Jan 2023 10:51:03 +0100
From: Lorenzo Bianconi <[email protected]>
To: Jonathan Cameron <[email protected]>
Cc: Philippe De Muyter <[email protected]>, [email protected]
Subject: Re: sysfs mount_matrix for st_lsm6dsx gyro

Quoted 10 lines
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Quoted 8 lines
Quoted 5 lines

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

I am not sure if accel and gyro can be mounted with a different orientation. Do you think we should have a per-sensor mount_matrix?

My chip is a 'ism330dlc'. It's one chip containing an accel and a gyro, so the mount_matrix should be the same for the accel and the gyro. Unfortunately the accel and the gyro are presented in /sysfs as two separate devices, and only the accel one has a 'mount_matrix' entry. So a user looking for any gyro will find the gyro entry, but without 'mount_matrix'.

I have followed Jonathan's proposal and added this simple patch in the definition of the ST_LSM6DSX_CHANNEL macro :

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +0 -0
index d80ba2e688ed..9d18145d5041 100644
@@ -96,6 +96,7 @@ enum st_lsm6dsx_hw_id {
 		.storagebits = 16,                                      \
 		.endianness = IIO_LE,                                   \
	},                                                              \

+ .ext_info = st_lsm6dsx_accel_ext_info, \
}

struct st_lsm6dsx_reg {

Another fix would be to create only one chip entry in sysfs, but that's above my knowledge of this driver.

Best regards

Philippe

maintainer
#4 lore raw
Quoted 43 lines

Hello Lorenzo,

Quoted 16 lines · Lorenzo Bianconi

Date: Thu, 12 Jan 2023 10:51:03 +0100
From: Lorenzo Bianconi <[email protected]>
To: Jonathan Cameron <[email protected]>
Cc: Philippe De Muyter <[email protected]>, [email protected]
Subject: Re: sysfs mount_matrix for st_lsm6dsx gyro

Quoted 10 lines
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Quoted 8 lines
Quoted 5 lines

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

I am not sure if accel and gyro can be mounted with a different orientation. Do you think we should have a per-sensor mount_matrix?

My chip is a 'ism330dlc'. It's one chip containing an accel and a gyro, so the mount_matrix should be the same for the accel and the gyro. Unfortunately the accel and the gyro are presented in /sysfs as two separate devices, and only the accel one has a 'mount_matrix' entry. So a user looking for any gyro will find the gyro entry, but without 'mount_matrix'.

I have followed Jonathan's proposal and added this simple patch in the definition of the ST_LSM6DSX_CHANNEL macro :

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +0 -0
index d80ba2e688ed..9d18145d5041 100644

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h enum st_lsm6dsx_hw_id { · line 100

Quoted 5 lines
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +1 -0
@@ -96,6 +96,7 @@ enum st_lsm6dsx_hw_id {
 		.storagebits = 16,                                      \
 		.endianness = IIO_LE,                                   \
 	},                                                              \
+	.ext_info = st_lsm6dsx_accel_ext_info,                          \
 }

I am fine with this approach, probably having a per-sensor mount_matrix is unnecessary. Can you please just rename st_lsm6dsx_accel_ext_info in st_lsm6dsx_ext_info?

Regards,
Lorenzo

Quoted 9 lines
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +0 -0
 struct st_lsm6dsx_reg {
 

Another fix would be to create only one chip entry in sysfs, but that's above my knowledge of this driver.

Best regards

Philippe

#5 lore raw
Quoted 53 lines · On Thu, 12 Jan 2023 12:32:11 +0100 Lorenzo Bianconi <[email protected]> wrote:
Quoted 43 lines

Hello Lorenzo,

Quoted 16 lines · Lorenzo Bianconi

Date: Thu, 12 Jan 2023 10:51:03 +0100
From: Lorenzo Bianconi <[email protected]>
To: Jonathan Cameron <[email protected]>
Cc: Philippe De Muyter <[email protected]>, [email protected]
Subject: Re: sysfs mount_matrix for st_lsm6dsx gyro

Quoted 10 lines
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Quoted 8 lines
Quoted 5 lines

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

I am not sure if accel and gyro can be mounted with a different orientation. Do you think we should have a per-sensor mount_matrix?

My chip is a 'ism330dlc'. It's one chip containing an accel and a gyro, so the mount_matrix should be the same for the accel and the gyro. Unfortunately the accel and the gyro are presented in /sysfs as two separate devices, and only the accel one has a 'mount_matrix' entry. So a user looking for any gyro will find the gyro entry, but without 'mount_matrix'.

I have followed Jonathan's proposal and added this simple patch in the definition of the ST_LSM6DSX_CHANNEL macro :

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +0 -0
index d80ba2e688ed..9d18145d5041 100644

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h enum st_lsm6dsx_hw_id { · line 100

Quoted 5 lines
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +1 -0
@@ -96,6 +96,7 @@ enum st_lsm6dsx_hw_id {
 		.storagebits = 16,                                      \
 		.endianness = IIO_LE,                                   \
 	},                                                              \
+	.ext_info = st_lsm6dsx_accel_ext_info,                          \
 }  

I am fine with this approach, probably having a per-sensor mount_matrix is unnecessary. Can you please just rename st_lsm6dsx_accel_ext_info in st_lsm6dsx_ext_info?

Quoted 8 lines

Regards,
Lorenzo

Quoted 5 lines
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +0 -0
 struct st_lsm6dsx_reg {
 

Another fix would be to create only one chip entry in sysfs, but that's above my knowledge of this driver.

We can't make that change. It would be ABI breakage - userspace code would stop working.

Jonathan

Quoted 5 lines
Quoted 4 lines

Best regards

Philippe

#6 lore raw
In reply to Lorenzo Bianconi
Quoted 18 lines · Lorenzo Bianconi

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h enum st_lsm6dsx_hw_id { · line 100

Quoted 13 lines

I have followed Jonathan's proposal and added this simple patch in the definition of the ST_LSM6DSX_CHANNEL macro :

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +1 -0
index d80ba2e688ed..9d18145d5041 100644
@@ -96,6 +96,7 @@ enum st_lsm6dsx_hw_id {
 		.storagebits = 16,                                      \
 		.endianness = IIO_LE,                                   \
 	},                                                              \
+	.ext_info = st_lsm6dsx_accel_ext_info,                          \
 }

I am fine with this approach, probably having a per-sensor mount_matrix is unnecessary. Can you please just rename st_lsm6dsx_accel_ext_info in st_lsm6dsx_ext_info?

The compilation fails because there is already something different with that same name in drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_shub.c

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_shub.c:715:30: error: conflicting type qualifiers for ‘st_lsm6dsx_ext_info’

  static const struct iio_info st_lsm6dsx_ext_info = {
                               ^~~~~~~~~~~~~~~~~~~

In file included from drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_shub.c:31:0:
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h:498:46: note: previous definition of ‘st_lsm6dsx_ext_info’ was here struct iio_chan_spec_ext_info __maybe_unused st_lsm6dsx_ext_info[] = {

Best regards

Philippe

maintainer
#7 lore raw
In reply to Jonathan Cameron
Quoted 10 lines
Quoted 6 lines · On Wed, 11 Jan 2023 13:09:40 +0100 Philippe De Muyter <[email protected]> wrote:

Hello Lorenzo and list,

I do not find a "*mount_matrix" entry in sysfs for a 'ism330dlc_gyro' iio device.
Is that normal ?
Is a fix available ?

Quoted 5 lines

Looks like the channel definition for the gyro does not include an appropriate ext_info entry unlike the accelerometer channels which have one with mount_matrix support.

From a quick glance looks like a simple fix. Add that entry.

something like (per-sensor mount_matrix):

drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h +12 -4
index 5b6f195748fc..5e0c184e1055 100644
@@ -113,6 +113,8 @@ enum st_lsm6dsx_hw_id {
 		.storagebits = 16,					\
 		.endianness = IIO_LE,					\
 	},								\
+	.ext_info = st_lsm6dsx_gyro_ext_info,				\
+	.num_event_specs = 1,						\
 }
 
 struct st_lsm6dsx_reg {
@@ -356,6 +358,7 @@ enum st_lsm6dsx_fifo_mode {
  * @decimator: Sensor decimation factor.
  * @sip: Number of samples in a given pattern.
  * @ts_ref: Sensor timestamp reference for hw one.
+ * @orientation: sensor chip orientation relative to main hardware.
  * @ext_info: Sensor settings if it is connected to i2c controller
  */
 struct st_lsm6dsx_sensor {
@@ -371,6 +374,8 @@ struct st_lsm6dsx_sensor {
 	u8 sip;
 	s64 ts_ref;
 
+	struct iio_mount_matrix orientation;
+
 	struct {
 		const struct st_lsm6dsx_ext_dev_settings *settings;
 		u32 slv_odr;
@@ -398,7 +403,6 @@ struct st_lsm6dsx_sensor {
  * @enable_event: enabled event bitmask.
  * @iio_devs: Pointers to acc/gyro iio_dev instances.
  * @settings: Pointer to the specific sensor settings in use.
- * @orientation: sensor chip orientation relative to main hardware.
  * @scan: Temporary buffers used to align data before iio_push_to_buffers()
  */
 struct st_lsm6dsx_hw {
@@ -427,7 +431,6 @@ struct st_lsm6dsx_hw {
 
 	const struct st_lsm6dsx_settings *settings;
 
-	struct iio_mount_matrix orientation;
 	/* Ensure natural alignment of buffer elements */
 	struct {
 		__le16 channels[3];
@@ -511,9 +514,8 @@ st_lsm6dsx_get_mount_matrix(const struct iio_dev *iio_dev,
 			    const struct iio_chan_spec *chan)
 {
 	struct st_lsm6dsx_sensor *sensor = iio_priv(iio_dev);
-	struct st_lsm6dsx_hw *hw = sensor->hw;
 
-	return &hw->orientation;
+	return &sensor->orientation;
 }
 
 static inline int
@@ -533,4 +535,10 @@ struct iio_chan_spec_ext_info __maybe_unused st_lsm6dsx_accel_ext_info[] = {
 	{ }
 };
 
+static const
+struct iio_chan_spec_ext_info __maybe_unused st_lsm6dsx_gyro_ext_info[] = {
+	IIO_MOUNT_MATRIX(IIO_SHARED_BY_ALL, st_lsm6dsx_get_mount_matrix),
+	{ }
+};
+
 #endif /* ST_LSM6DSX_H */
drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c +3 -4
index 3f6060c64f32..15c7184a1895 100644
@@ -2359,6 +2359,9 @@ static struct iio_dev *st_lsm6dsx_alloc_iiodev(struct st_lsm6dsx_hw *hw,
 	sensor->gain = hw->settings->fs_table[id].fs_avl[0].gain;
 	sensor->watermark = 1;
 
+	if (iio_read_mount_matrix(hw->dev, &sensor->orientation))
+		return NULL;
+
 	switch (id) {
 	case ST_LSM6DSX_ID_ACC:
 		iio_dev->info = &st_lsm6dsx_acc_info;
@@ -2676,10 +2679,6 @@ int st_lsm6dsx_probe(struct device *dev, int irq, int hw_id,
 			return err;
 	}
 
-	err = iio_read_mount_matrix(hw->dev, &hw->orientation);
-	if (err)
-		return err;
-
 	for (i = 0; i < ST_LSM6DSX_ID_MAX; i++) {
 		if (!hw->iio_devs[i])
 			continue;
Quoted 57 lines
Quoted 31 lines

Some more info :

I have backported drivers/iio/imu/st_lsm6dsx to linux-4.9 in order to drive a ism330dlc imu on a custom board. The chip is correctly detected and two devices are created in /sys/bus/iio/devices/

the first one (where name is 'ism330dlc_gyro') has the following entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device1/
 buffer                      in_anglvel_x_raw  sampling_frequency
 current_timestamp_clock     in_anglvel_y_raw  sampling_frequency_available
 dev                         in_anglvel_z_raw  scan_elements
 in_anglvel_scale            name              subsystem
 in_anglvel_scale_available  power             uevent
 me@proto4:~$

the second one (where name is 'ism330dlc_accel') has those entries :

 me@proto4:~$ ls /sys/bus/iio/devices/iio\:device2
 buffer                    in_accel_x_raw  sampling_frequency
 current_timestamp_clock   in_accel_y_raw  sampling_frequency_available
 dev                       in_accel_z_raw  scan_elements
 events                    mount_matrix    subsystem
 in_accel_scale            name            uevent
 in_accel_scale_available  power
 me@proto4:~$

The 'mount_matrix' entry is only present in the 'ism330dlc_accel' device but not in the 'ism330dlc_gyro' device.

Quoted 8 lines

On a similar board, but with mpu9250 imu, I get only one iio:deviceX entry but with two *mount_matrix entries :

in_accel_mount_matrix
in_anglvel_mount_matrix

In both cases, I would have expected only one 'iio:deviceX' entry with only one 'mount_matrix' entry.

There are multiple devices because the driver predates the addition of multiple buffer support to IIO and IIRC is capable of producing data at different sampling rates for the accelerometer and the gyros. Hence when it was implemented the only choice was to register multiple devices in order to get the multiple buffers. It's ABI now so we can't fix it in an old driver unfortunately. We'd do this differently today..

The double mount_matrix for the mpu9250 is a little odd and I can't immediately spot why that one is happening.

Quoted 5 lines

Best regards

Philippe