Skip to contentSkip to conversation

IIO: Alignment fixes part 2 - struct used to ensure alignment

PATCH Applied Jonathan Cameron 5 participants 35 messages May 1, 2021 – May 13, 2021

Subsystems (5)
lore
Share & tools
lore

Jonathan Cameron

author maintainer
#0 lore raw

cover IIO: Alignment fixes part 2 - struct used to ensure alignment

From: Jonathan Cameron <[email protected]>

I finally got around to do a manual audit of all the calls to iio_push_to_buffers_with_timestamp() which has the somewhat odd requirements of:
1. 8 byte alignment of the provided buffer.
2. space for an 8 byte naturally aligned timestamp to be inserted at the end.

Unfortuantely there were rather a lot of these left, but time to bite the bullet and clean them up.

As discussed previous in
https://lore.kernel.org/linux-iio/[email protected]/
it is not easy to fix the alignment issue without requiring a bounce buffer (see part 4 of the alignment fixes for a proposal for that where it is absolutely necessary).

Part 2 is the simple cases where the timestamp position is always the same (as not more than 8 bytes of other scan elements) and where it is either in a structure that is kzalloc'd or there was already an explicit memset that would clear any holes in the structure (and hence avoid a potential kernel data leak).

Cc Peter Meerwald <[email protected]>
Cc Andy Shevchenko <[email protected]>
Cc Srinivas Pandruvada <[email protected]>
Cc Daniel Baluta <[email protected]>
Cc Andy Shevchenko <[email protected]>
Cc Stefan-Gabriel Mirea <[email protected]>
Cc Sanchayan Maity <[email protected]>
Cc Stephan Gerhold <[email protected]>
Cc Andreas Klinger <[email protected]>
Cc Matt Ranostay <[email protected]>
Cc Linus Walleij <[email protected]>

Jonathan Cameron (19):
iio: accel: bma180: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: bma220: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: hid: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: kxcjk-1013: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: mxc4005: Fix overread of data and alignment issue.
iio: accel: stk8312: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: stk8ba50: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: ti-ads1015: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: vf610: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: gyro: bmg160: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: humidity: am2315: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: srf08: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: pulsed-light: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: as3935: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: hmc5843: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: bmc150: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: isl29125: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3414: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3472: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()

 drivers/iio/accel/bma180.c                    | 10 +++++---
 drivers/iio/accel/bma220_spi.c                | 10 +++++---
 drivers/iio/accel/hid-sensor-accel-3d.c       | 13 ++++++----
 drivers/iio/accel/kxcjk-1013.c                | 24 +++++++++++--------
 drivers/iio/accel/mxc4005.c                   | 10 +++++---
 drivers/iio/accel/stk8312.c                   | 12 ++++++----
 drivers/iio/accel/stk8ba50.c                  | 17 +++++++------
 drivers/iio/adc/ti-ads1015.c                  | 12 ++++++----
 drivers/iio/adc/vf610_adc.c                   | 10 +++++---
 drivers/iio/gyro/bmg160_core.c                | 10 +++++---
 drivers/iio/humidity/am2315.c                 | 16 ++++++++-----
 drivers/iio/light/isl29125.c                  | 10 +++++---
 drivers/iio/light/tcs3414.c                   | 10 +++++---
 drivers/iio/light/tcs3472.c                   | 10 +++++---
 drivers/iio/magnetometer/bmc150_magn.c        | 11 +++++----
 drivers/iio/magnetometer/hmc5843.h            |  8 +++++--
 drivers/iio/magnetometer/hmc5843_core.c       |  4 ++--
 drivers/iio/proximity/as3935.c                | 10 +++++---
 .../iio/proximity/pulsedlight-lidar-lite-v2.c | 10 +++++---
 drivers/iio/proximity/srf08.c                 | 14 +++++------
 20 files changed, 148 insertions(+), 83 deletions(-)
Signature · 1 line
2.31.1
author maintainer
#1 lore raw

PATCH 1/19 iio: accel: bma180: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes b9a6a237ffc9 ("iio:bma180: Drop _update_scan_mode()")
Cc Peter Meerwald <[email protected]>
---
 drivers/iio/accel/bma180.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/bma180.c +7 -3
index b8a7469cdae4..2e901f900ea7 100644
@@ -162,7 +162,11 @@ struct bma180_data {
 	int scale;
 	int bw;
 	bool pmode;
-	u8 buff[16]; /* 3x 16-bit + 8-bit + padding + timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s16 chan[3];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 enum bma180_chan {
@@ -938,12 +942,12 @@ static irqreturn_t bma180_trigger_handler(int irq, void *p)
 			mutex_unlock(&data->mutex);
 			goto err;
 		}
-		((s16 *)data->buff)[i++] = ret;
+		data->scan.chan[i++] = ret;
 	}
 
 	mutex_unlock(&data->mutex);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buff, time_ns);
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, time_ns);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
 
Signature · 1 line
2.31.1
author maintainer Applied
#2 lore raw

Re: [PATCH 01/19] iio: accel: bma180: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 10 lines · On Sat, 1 May 2021 18:01:03 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes b9a6a237ffc9 ("iio:bma180: Drop _update_scan_mode()")
Cc Peter Meerwald <[email protected]>

Applied to the togreg branch of iio.git with a minor fix.

Note I'm taking these the slow way and nor marking them for stable as they've been there a very long time.

Quoted 9 lines
---
 drivers/iio/accel/bma180.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/bma180.c +0 -0
index b8a7469cdae4..2e901f900ea7 100644

drivers/iio/accel/bma180.c struct bma180_data { · line 167

Quoted 7 lines
drivers/iio/accel/bma180.c +3 -1
@@ -162,7 +162,11 @@ struct bma180_data {
 	int scale;
 	int bw;
 	bool pmode;
-	u8 buff[16]; /* 3x 16-bit + 8-bit + padding + timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s16 chan[3];

The 8 bit in the comment above is misleading, but made me take another
look at this. There are actually 4 channels on some devices, the last
of which (temp) is 8 bits in a 16bit field.

I changed this to be s16 chan[4] to reflect that.

Thanks,

Jonathan

Quoted 20 lines
drivers/iio/accel/bma180.c +4 -2
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 enum bma180_chan {
@@ -938,12 +942,12 @@ static irqreturn_t bma180_trigger_handler(int irq, void *p)
 			mutex_unlock(&data->mutex);
 			goto err;
 		}
-		((s16 *)data->buff)[i++] = ret;
+		data->scan.chan[i++] = ret;
 	}
 
 	mutex_unlock(&data->mutex);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buff, time_ns);
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, time_ns);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
 
author maintainer
#3 lore raw

PATCH 2/19 iio: accel: bma220: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 194dc4c71413 ("iio: accel: Add triggered buffer support for BMA220")
Cc Andy Shevchenko <[email protected]>
---
 drivers/iio/accel/bma220_spi.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/bma220_spi.c +7 -3
index 36fc9876dbca..2802ce1852ef 100644
@@ -63,7 +63,11 @@ static const int bma220_scale_table[][2] = {
 struct bma220_data {
 	struct spi_device *spi_device;
 	struct mutex lock;
-	s8 buffer[16]; /* 3x8-bit channels + 5x8 padding + 8x8 timestamp */
+	struct {
+		s8 chans[2];
+		/* Ensure timestamp is naturally aligned. */
+		s64 timestamp __aligned(8);
+	} scan;
 	u8 tx_buf[2] ____cacheline_aligned;
 };
 
@@ -94,12 +98,12 @@ static irqreturn_t bma220_trigger_handler(int irq, void *p)
 
 	mutex_lock(&data->lock);
 	data->tx_buf[0] = BMA220_REG_ACCEL_X | BMA220_READ_MASK;
-	ret = spi_write_then_read(spi, data->tx_buf, 1, data->buffer,
+	ret = spi_write_then_read(spi, data->tx_buf, 1, &data->scan.chans,
 				  ARRAY_SIZE(bma220_channels) - 1);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	mutex_unlock(&data->lock);
Signature · 1 line
2.31.1
author maintainer
#4 lore raw

Re: [PATCH 02/19] iio: accel: bma220: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 6 lines · On Sat, 1 May 2021 18:01:04 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

~Gah. Clearly wasn't on my best form when I originally wrote these. Number of chans in this one is wrong as well. Fixed to 3 as per the comment that got dropped because it should have been obvious.

As you might imagine, the error in patch 1 made me look closer at the rest of these whilst applying.

Quoted 4 lines
Fixes 194dc4c71413 ("iio: accel: Add triggered buffer support for BMA220")
Cc Andy Shevchenko <[email protected]>

Applied to the togreg branch of iio.git and pushed out as testing for the autobuilders to poke at.

Thanks,

Jonathan

Quoted 36 lines
---
 drivers/iio/accel/bma220_spi.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/bma220_spi.c +7 -3
index 36fc9876dbca..2802ce1852ef 100644
@@ -63,7 +63,11 @@ static const int bma220_scale_table[][2] = {
 struct bma220_data {
 	struct spi_device *spi_device;
 	struct mutex lock;
-	s8 buffer[16]; /* 3x8-bit channels + 5x8 padding + 8x8 timestamp */
+	struct {
+		s8 chans[2];
+		/* Ensure timestamp is naturally aligned. */
+		s64 timestamp __aligned(8);
+	} scan;
 	u8 tx_buf[2] ____cacheline_aligned;
 };
 
@@ -94,12 +98,12 @@ static irqreturn_t bma220_trigger_handler(int irq, void *p)
 
 	mutex_lock(&data->lock);
 	data->tx_buf[0] = BMA220_REG_ACCEL_X | BMA220_READ_MASK;
-	ret = spi_write_then_read(spi, data->tx_buf, 1, data->buffer,
+	ret = spi_write_then_read(spi, data->tx_buf, 1, &data->scan.chans,
 				  ARRAY_SIZE(bma220_channels) - 1);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	mutex_unlock(&data->lock);
author maintainer
#5 lore raw

PATCH 3/19 iio: accel: hid: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.
Note this matches what was done in all the other hid sensor drivers. This one was missed previously due to an extra level of indirection.

Found during an audit of all calls of this function.

Fixes a96cd0f901ee ("iio: accel: hid-sensor-accel-3d: Add timestamp")
Cc Srinivas Pandruvada <[email protected]>
---
 drivers/iio/accel/hid-sensor-accel-3d.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)
drivers/iio/accel/hid-sensor-accel-3d.c +8 -5
index 2f9465cb382f..27f47e1c251e 100644
@@ -28,8 +28,11 @@ struct accel_3d_state {
 	struct hid_sensor_hub_callbacks callbacks;
 	struct hid_sensor_common common_attributes;
 	struct hid_sensor_hub_attribute_info accel[ACCEL_3D_CHANNEL_MAX];
-	/* Reserve for 3 channels + padding + timestamp */
-	u32 accel_val[ACCEL_3D_CHANNEL_MAX + 3];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u32 accel_val[3];
+		s64 timestamp __aligned(8);
+	} scan;
 	int scale_pre_decml;
 	int scale_post_decml;
 	int scale_precision;
@@ -245,8 +248,8 @@ static int accel_3d_proc_event(struct hid_sensor_hub_device *hsdev,
 			accel_state->timestamp = iio_get_time_ns(indio_dev);
 
 		hid_sensor_push_data(indio_dev,
-				     accel_state->accel_val,
-				     sizeof(accel_state->accel_val),
+				     &accel_state->scan,
+				     sizeof(accel_state->scan),
 				     accel_state->timestamp);
 
 		accel_state->timestamp = 0;
@@ -271,7 +274,7 @@ static int accel_3d_capture_sample(struct hid_sensor_hub_device *hsdev,
 	case HID_USAGE_SENSOR_ACCEL_Y_AXIS:
 	case HID_USAGE_SENSOR_ACCEL_Z_AXIS:
 		offset = usage_id - HID_USAGE_SENSOR_ACCEL_X_AXIS;
-		accel_state->accel_val[CHANNEL_SCAN_INDEX_X + offset] =
+		accel_state->scan.accel_val[CHANNEL_SCAN_INDEX_X + offset] =
 						*(u32 *)raw_data;
 		ret = 0;
 	break;
Signature · 1 line
2.31.1
author maintainer
#6 lore raw

Re: [PATCH 03/19] iio: accel: hid: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 7 lines · On Sat, 1 May 2021 18:01:05 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.
Note this matches what was done in all the other hid sensor drivers. This one was missed previously due to an extra level of indirection.

Quoted 5 lines

Found during an audit of all calls of this function.

Fixes a96cd0f901ee ("iio: accel: hid-sensor-accel-3d: Add timestamp")
Cc Srinivas Pandruvada <[email protected]>

Applied to the togreg branch of iio.git and pushed out as testing for the autobuilders to poke at.

Thanks,

Jonathan

Quoted 42 lines
---
 drivers/iio/accel/hid-sensor-accel-3d.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)
drivers/iio/accel/hid-sensor-accel-3d.c +8 -5
index 2f9465cb382f..27f47e1c251e 100644
@@ -28,8 +28,11 @@ struct accel_3d_state {
 	struct hid_sensor_hub_callbacks callbacks;
 	struct hid_sensor_common common_attributes;
 	struct hid_sensor_hub_attribute_info accel[ACCEL_3D_CHANNEL_MAX];
-	/* Reserve for 3 channels + padding + timestamp */
-	u32 accel_val[ACCEL_3D_CHANNEL_MAX + 3];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u32 accel_val[3];
+		s64 timestamp __aligned(8);
+	} scan;
 	int scale_pre_decml;
 	int scale_post_decml;
 	int scale_precision;
@@ -245,8 +248,8 @@ static int accel_3d_proc_event(struct hid_sensor_hub_device *hsdev,
 			accel_state->timestamp = iio_get_time_ns(indio_dev);
 
 		hid_sensor_push_data(indio_dev,
-				     accel_state->accel_val,
-				     sizeof(accel_state->accel_val),
+				     &accel_state->scan,
+				     sizeof(accel_state->scan),
 				     accel_state->timestamp);
 
 		accel_state->timestamp = 0;
@@ -271,7 +274,7 @@ static int accel_3d_capture_sample(struct hid_sensor_hub_device *hsdev,
 	case HID_USAGE_SENSOR_ACCEL_Y_AXIS:
 	case HID_USAGE_SENSOR_ACCEL_Z_AXIS:
 		offset = usage_id - HID_USAGE_SENSOR_ACCEL_X_AXIS;
-		accel_state->accel_val[CHANNEL_SCAN_INDEX_X + offset] =
+		accel_state->scan.accel_val[CHANNEL_SCAN_INDEX_X + offset] =
 						*(u32 *)raw_data;
 		ret = 0;
 	break;
author maintainer
#7 lore raw

PATCH 4/19 iio: accel: kxcjk-1013: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 1a4fbf6a9286 ("iio: accel: kxcjk1013 3-axis accelerometer driver")
Cc Srinivas Pandruvada <[email protected]>
---
 drivers/iio/accel/kxcjk-1013.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
drivers/iio/accel/kxcjk-1013.c +14 -10
index ff724bc17a45..96ab247f17b3 100644
@@ -133,6 +133,13 @@ enum kx_acpi_type {
 	ACPI_KIOX010A,
 };
 
+enum kxcjk1013_axis {
+	AXIS_X,
+	AXIS_Y,
+	AXIS_Z,
+	AXIS_MAX,
+};
+
 struct kxcjk1013_data {
 	struct regulator_bulk_data regulators[2];
 	struct i2c_client *client;
@@ -140,7 +147,11 @@ struct kxcjk1013_data {
 	struct iio_trigger *motion_trig;
 	struct iio_mount_matrix orientation;
 	struct mutex mutex;
-	s16 buffer[8];
+	/* Ensure timestamp naturally aligned */
+	struct {
+		s16 chans[AXIS_MAX];
+		s64 timestamp __aligned(8);
+	} scan;
 	u8 odr_bits;
 	u8 range;
 	int wake_thres;
@@ -154,13 +165,6 @@ struct kxcjk1013_data {
 	enum kx_acpi_type acpi_type;
 };
 
-enum kxcjk1013_axis {
-	AXIS_X,
-	AXIS_Y,
-	AXIS_Z,
-	AXIS_MAX,
-};
-
 enum kxcjk1013_mode {
 	STANDBY,
 	OPERATION,
@@ -1094,12 +1098,12 @@ static irqreturn_t kxcjk1013_trigger_handler(int irq, void *p)
 	ret = i2c_smbus_read_i2c_block_data_or_emulated(data->client,
 							KXCJK1013_REG_XOUT_L,
 							AXIS_MAX * 2,
-							(u8 *)data->buffer);
+							(u8 *)data->scan.chans);
 	mutex_unlock(&data->mutex);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   data->timestamp);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
Signature · 1 line
2.31.1
#8 lore raw

Re: [PATCH 04/19] iio: accel: kxcjk-1013: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 20 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 1a4fbf6a9286 ("iio: accel: kxcjk1013 3-axis accelerometer driver")
Cc Srinivas Pandruvada <[email protected]>
---
 drivers/iio/accel/kxcjk-1013.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
drivers/iio/accel/kxcjk-1013.c +0 -0
index ff724bc17a45..96ab247f17b3 100644

drivers/iio/accel/kxcjk-1013.c enum kx_acpi_type { · line 141

Quoted 9 lines
drivers/iio/accel/kxcjk-1013.c +5 -0
@@ -133,6 +133,13 @@ enum kx_acpi_type {
        ACPI_KIOX010A,
 };
 
+enum kxcjk1013_axis {
+       AXIS_X,
+       AXIS_Y,
+       AXIS_Z,
 
+       AXIS_MAX,

I see that this is from original code, though I think you have a chance to remove unneeded comma here. Let's make the terminator line terminate.

Quoted 50 lines
drivers/iio/accel/kxcjk-1013.c +2 -0
+};
+
 struct kxcjk1013_data {
        struct regulator_bulk_data regulators[2];
        struct i2c_client *client;
drivers/iio/accel/kxcjk-1013.c +7 -11
@@ -140,7 +147,11 @@ struct kxcjk1013_data {
        struct iio_trigger *motion_trig;
        struct iio_mount_matrix orientation;
        struct mutex mutex;
-       s16 buffer[8];
+       /* Ensure timestamp naturally aligned */
+       struct {
+               s16 chans[AXIS_MAX];
+               s64 timestamp __aligned(8);
+       } scan;
        u8 odr_bits;
        u8 range;
        int wake_thres;
@@ -154,13 +165,6 @@ struct kxcjk1013_data {
        enum kx_acpi_type acpi_type;
 };
 
-enum kxcjk1013_axis {
-       AXIS_X,
-       AXIS_Y,
-       AXIS_Z,
-       AXIS_MAX,
-};
-
 enum kxcjk1013_mode {
        STANDBY,
        OPERATION,
@@ -1094,12 +1098,12 @@ static irqreturn_t kxcjk1013_trigger_handler(int irq, void *p)
        ret = i2c_smbus_read_i2c_block_data_or_emulated(data->client,
                                                        KXCJK1013_REG_XOUT_L,
                                                        AXIS_MAX * 2,
-                                                       (u8 *)data->buffer);
+                                                       (u8 *)data->scan.chans);
        mutex_unlock(&data->mutex);
        if (ret < 0)
                goto err;
 
-       iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+       iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
                                           data->timestamp);
 err:
        iio_trigger_notify_done(indio_dev->trig);
--

2.31.1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author maintainer
#9 lore raw

Re: [PATCH 04/19] iio: accel: kxcjk-1013: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 29 lines · On Sat, 1 May 2021 22:10:48 +0300 Andy Shevchenko <[email protected]>…
Quoted 27 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 1a4fbf6a9286 ("iio: accel: kxcjk1013 3-axis accelerometer driver")
Cc Srinivas Pandruvada <[email protected]>
---
 drivers/iio/accel/kxcjk-1013.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
drivers/iio/accel/kxcjk-1013.c +4 -0
index ff724bc17a45..96ab247f17b3 100644
@@ -133,6 +133,13 @@ enum kx_acpi_type {
        ACPI_KIOX010A,
 };
 
+enum kxcjk1013_axis {
+       AXIS_X,
+       AXIS_Y,
+       AXIS_Z,  
Quoted 5 lines

drivers/iio/accel/kxcjk-1013.c enum kx_acpi_type { · line 140

Quoted 1 line
drivers/iio/accel/kxcjk-1013.c +1 -0
+       AXIS_MAX,  

I see that this is from original code, though I think you have a chance to remove unneeded comma here. Let's make the terminator line terminate.

Good idea. Did that whilst applying. This one had a lot of fuzz due to support of new parts added to the driver in the meantime, but the patch seems to have applied cleanly.

+CC various people involved in that series so they know this change is going in as well (and can perhaps give it a spin!:)

Thanks,

Jonathan

Quoted 53 lines
Quoted 51 lines
drivers/iio/accel/kxcjk-1013.c +9 -11
+};
+
 struct kxcjk1013_data {
        struct regulator_bulk_data regulators[2];
        struct i2c_client *client;
@@ -140,7 +147,11 @@ struct kxcjk1013_data {
        struct iio_trigger *motion_trig;
        struct iio_mount_matrix orientation;
        struct mutex mutex;
-       s16 buffer[8];
+       /* Ensure timestamp naturally aligned */
+       struct {
+               s16 chans[AXIS_MAX];
+               s64 timestamp __aligned(8);
+       } scan;
        u8 odr_bits;
        u8 range;
        int wake_thres;
@@ -154,13 +165,6 @@ struct kxcjk1013_data {
        enum kx_acpi_type acpi_type;
 };
 
-enum kxcjk1013_axis {
-       AXIS_X,
-       AXIS_Y,
-       AXIS_Z,
-       AXIS_MAX,
-};
-
 enum kxcjk1013_mode {
        STANDBY,
        OPERATION,
@@ -1094,12 +1098,12 @@ static irqreturn_t kxcjk1013_trigger_handler(int irq, void *p)
        ret = i2c_smbus_read_i2c_block_data_or_emulated(data->client,
                                                        KXCJK1013_REG_XOUT_L,
                                                        AXIS_MAX * 2,
-                                                       (u8 *)data->buffer);
+                                                       (u8 *)data->scan.chans);
        mutex_unlock(&data->mutex);
        if (ret < 0)
                goto err;
 
-       iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+       iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
                                           data->timestamp);
 err:
        iio_trigger_notify_done(indio_dev->trig);
--

2.31.1

author maintainer
#10 lore raw

PATCH 5/19 iio: accel: mxc4005: Fix overread of data and alignment issue.

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

The bulk read size is based on the size of an array that also has space for the timestamp alongside the channels.
Fix that and also fix alignment of the buffer passed
to iio_push_to_buffers_with_timestamp.

Found during an audit of all calls to this function.

Fixes 1ce0eda0f757 ("iio: mxc4005: add triggered buffer mode for mxc4005")
---
 drivers/iio/accel/mxc4005.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/mxc4005.c +7 -3
index fb3cbaa62bd8..9e6066c6a2df 100644
@@ -56,7 +56,11 @@ struct mxc4005_data {
 	struct mutex mutex;
 	struct regmap *regmap;
 	struct iio_trigger *dready_trig;
-	__be16 buffer[8];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		__be16 chans[3];
+		s64 timestamp;
+	} scan;
 	bool trigger_enabled;
 };
 
@@ -135,7 +139,7 @@ static int mxc4005_read_xyz(struct mxc4005_data *data)
 	int ret;
 
 	ret = regmap_bulk_read(data->regmap, MXC4005_REG_XOUT_UPPER,
-			       data->buffer, sizeof(data->buffer));
+			       data->scan.chans, sizeof(data->scan.chans));
 	if (ret < 0) {
 		dev_err(data->dev, "failed to read axes\n");
 		return ret;
@@ -301,7 +305,7 @@ static irqreturn_t mxc4005_trigger_handler(int irq, void *private)
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 
 err:
Signature · 1 line
2.31.1
author maintainer
#11 lore raw

Re: [PATCH 05/19] iio: accel: mxc4005: Fix overread of data and alignment issue.

Quoted 11 lines · On Sat, 1 May 2021 18:01:07 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

The bulk read size is based on the size of an array that also has space for the timestamp alongside the channels.
Fix that and also fix alignment of the buffer passed
to iio_push_to_buffers_with_timestamp.

Found during an audit of all calls to this function.

Fixes 1ce0eda0f757 ("iio: mxc4005: add triggered buffer mode for mxc4005")

Applied to the togreg branch of iio.git and pushed out as testing for the autobuilders to see if they can find anything. Minor tweak to enforce the 8 byte alignment as done in other drivers.

Quoted 9 lines
---
 drivers/iio/accel/mxc4005.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/accel/mxc4005.c +0 -0
index fb3cbaa62bd8..9e6066c6a2df 100644

drivers/iio/accel/mxc4005.c struct mxc4005_data { · line 62

Quoted 8 lines
drivers/iio/accel/mxc4005.c +4 -1
@@ -56,7 +56,11 @@ struct mxc4005_data {
 	struct mutex mutex;
 	struct regmap *regmap;
 	struct iio_trigger *dready_trig;
-	__be16 buffer[8];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		__be16 chans[3];
+		s64 timestamp;

Added __aligned(8) here to be consistent and ensure natural alignment.

Quoted 22 lines
drivers/iio/accel/mxc4005.c +3 -2
+	} scan;
 	bool trigger_enabled;
 };
 
@@ -135,7 +139,7 @@ static int mxc4005_read_xyz(struct mxc4005_data *data)
 	int ret;
 
 	ret = regmap_bulk_read(data->regmap, MXC4005_REG_XOUT_UPPER,
-			       data->buffer, sizeof(data->buffer));
+			       data->scan.chans, sizeof(data->scan.chans));
 	if (ret < 0) {
 		dev_err(data->dev, "failed to read axes\n");
 		return ret;
@@ -301,7 +305,7 @@ static irqreturn_t mxc4005_trigger_handler(int irq, void *private)
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 
 err:
author maintainer
#12 lore raw

PATCH 6/19 iio: accel: stk8312: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 95c12bba51c3 ("iio: accel: Add buffer mode for Sensortek STK8312")
---
 drivers/iio/accel/stk8312.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/iio/accel/stk8312.c +8 -4
index 157d8faefb9e..c8b6915a757c 100644
@@ -103,7 +103,11 @@ struct stk8312_data {
 	u8 mode;
 	struct iio_trigger *dready_trig;
 	bool dready_trigger_on;
-	s8 buffer[16]; /* 3x8-bit channels + 5x8 padding + 64-bit timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s8 chans[3];
+		s64 timestamp;
+	} scan;
 };
 
 static IIO_CONST_ATTR(in_accel_scale_available, STK8312_SCALE_AVAIL);
@@ -438,7 +442,7 @@ static irqreturn_t stk8312_trigger_handler(int irq, void *p)
 		ret = i2c_smbus_read_i2c_block_data(data->client,
 						    STK8312_REG_XOUT,
 						    STK8312_ALL_CHANNEL_SIZE,
-						    data->buffer);
+						    data->scan.chans);
 		if (ret < STK8312_ALL_CHANNEL_SIZE) {
 			dev_err(&data->client->dev, "register read failed\n");
 			mutex_unlock(&data->lock);
@@ -452,12 +456,12 @@ static irqreturn_t stk8312_trigger_handler(int irq, void *p)
 				mutex_unlock(&data->lock);
 				goto err;
 			}
-			data->buffer[i++] = ret;
+			data->scan.chans[i++] = ret;
 		}
 	}
 	mutex_unlock(&data->lock);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
Signature · 1 line
2.31.1
author maintainer
#13 lore raw

Re: [PATCH 06/19] iio: accel: stk8312: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 9 lines · On Sat, 1 May 2021 18:01:08 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes 95c12bba51c3 ("iio: accel: Add buffer mode for Sensortek STK8312")

Also missing the __aligned(8) for the timestamp. Added it.

thanks,

Jonathan

Quoted 45 lines
---
 drivers/iio/accel/stk8312.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/iio/accel/stk8312.c +8 -4
index 157d8faefb9e..c8b6915a757c 100644
@@ -103,7 +103,11 @@ struct stk8312_data {
 	u8 mode;
 	struct iio_trigger *dready_trig;
 	bool dready_trigger_on;
-	s8 buffer[16]; /* 3x8-bit channels + 5x8 padding + 64-bit timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s8 chans[3];
+		s64 timestamp;
+	} scan;
 };
 
 static IIO_CONST_ATTR(in_accel_scale_available, STK8312_SCALE_AVAIL);
@@ -438,7 +442,7 @@ static irqreturn_t stk8312_trigger_handler(int irq, void *p)
 		ret = i2c_smbus_read_i2c_block_data(data->client,
 						    STK8312_REG_XOUT,
 						    STK8312_ALL_CHANNEL_SIZE,
-						    data->buffer);
+						    data->scan.chans);
 		if (ret < STK8312_ALL_CHANNEL_SIZE) {
 			dev_err(&data->client->dev, "register read failed\n");
 			mutex_unlock(&data->lock);
@@ -452,12 +456,12 @@ static irqreturn_t stk8312_trigger_handler(int irq, void *p)
 				mutex_unlock(&data->lock);
 				goto err;
 			}
-			data->buffer[i++] = ret;
+			data->scan.chans[i++] = ret;
 		}
 	}
 	mutex_unlock(&data->lock);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
author maintainer
#14 lore raw

PATCH 7/19 iio: accel: stk8ba50: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes db6a19b8251f ("iio: accel: Add trigger support for STK8BA50")
---
 drivers/iio/accel/stk8ba50.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
drivers/iio/accel/stk8ba50.c +8 -9
index 7cf9cb7e8666..eb9daa4e623a 100644
@@ -91,12 +91,11 @@ struct stk8ba50_data {
 	u8 sample_rate_idx;
 	struct iio_trigger *dready_trig;
 	bool dready_trigger_on;
-	/*
-	 * 3 x 16-bit channels (10-bit data, 6-bit padding) +
-	 * 1 x 16 padding +
-	 * 4 x 16 64-bit timestamp
-	 */
-	s16 buffer[8];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s16 chans[3];
+		s64 timetamp __aligned(8);
+	} scan;
 };
 
 #define STK8BA50_ACCEL_CHANNEL(index, reg, axis) {			\
@@ -324,7 +323,7 @@ static irqreturn_t stk8ba50_trigger_handler(int irq, void *p)
 		ret = i2c_smbus_read_i2c_block_data(data->client,
 						    STK8BA50_REG_XOUT,
 						    STK8BA50_ALL_CHANNEL_SIZE,
-						    (u8 *)data->buffer);
+						    (u8 *)data->scan.chans);
 		if (ret < STK8BA50_ALL_CHANNEL_SIZE) {
 			dev_err(&data->client->dev, "register read failed\n");
 			goto err;
@@ -337,10 +336,10 @@ static irqreturn_t stk8ba50_trigger_handler(int irq, void *p)
 			if (ret < 0)
 				goto err;
 
-			data->buffer[i++] = ret;
+			data->scan.chans[i++] = ret;
 		}
 	}
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	mutex_unlock(&data->lock);
Signature · 1 line
2.31.1
author maintainer
#15 lore raw

PATCH 8/19 iio: adc: ti-ads1015: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of this function.

Fixes ecc24e72f437 ("iio: adc: Add TI ADS1015 ADC driver support")
Cc Daniel Baluta <[email protected]>
Cc Andy Shevchenko <[email protected]>
---
 drivers/iio/adc/ti-ads1015.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/iio/adc/ti-ads1015.c +8 -4
index 9fef39bcf997..5b828428be77 100644
@@ -395,10 +395,14 @@ static irqreturn_t ads1015_trigger_handler(int irq, void *p)
 	struct iio_poll_func *pf = p;
 	struct iio_dev *indio_dev = pf->indio_dev;
 	struct ads1015_data *data = iio_priv(indio_dev);
-	s16 buf[8]; /* 1x s16 ADC val + 3x s16 padding +  4x s16 timestamp */
+	/* Ensure natural alignment of timestamp */
+	struct {
+		s16 chan;
+		s64 timestamp __aligned(8);
+	} scan;
 	int chan, ret, res;
 
-	memset(buf, 0, sizeof(buf));
+	memset(&scan, 0, sizeof(scan));
 
 	mutex_lock(&data->lock);
 	chan = find_first_bit(indio_dev->active_scan_mask,
@@ -409,10 +413,10 @@ static irqreturn_t ads1015_trigger_handler(int irq, void *p)
 		goto err;
 	}
 
-	buf[0] = res;
+	scan.chan = res;
 	mutex_unlock(&data->lock);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, buf,
+	iio_push_to_buffers_with_timestamp(indio_dev, &scan,
 					   iio_get_time_ns(indio_dev));
 
 err:
Signature · 1 line
2.31.1
author maintainer
#16 lore raw

PATCH 9/19 iio: adc: vf610: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 0010d6b44406 ("iio: adc: vf610: Add IIO buffer support for Vybrid ADC")
Cc Stefan-Gabriel Mirea <[email protected]>
Cc Sanchayan Maity <[email protected]>
---
 drivers/iio/adc/vf610_adc.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/adc/vf610_adc.c +7 -3
index 1d794cf3e3f1..fd57fc43e8e5 100644
@@ -167,7 +167,11 @@ struct vf610_adc {
 	u32 sample_freq_avail[5];
 
 	struct completion completion;
-	u16 buffer[8];
+	/* Ensure the timestamp is naturally aligned */
+	struct {
+		u16 chan;
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 static const u32 vf610_hw_avgs[] = { 1, 4, 8, 16, 32 };
@@ -579,9 +583,9 @@ static irqreturn_t vf610_adc_isr(int irq, void *dev_id)
 	if (coco & VF610_ADC_HS_COCO0) {
 		info->value = vf610_adc_read_data(info);
 		if (iio_buffer_enabled(indio_dev)) {
-			info->buffer[0] = info->value;
+			info->scan.chan = info->value;
 			iio_push_to_buffers_with_timestamp(indio_dev,
-					info->buffer,
+					&info->scan,
 					iio_get_time_ns(indio_dev));
 			iio_trigger_notify_done(indio_dev->trig);
 		} else
Signature · 1 line
2.31.1
author maintainer
#17 lore raw

PATCH 10/19 iio: gyro: bmg160: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 13426454b649 ("iio: bmg160: Separate i2c and core driver")
Cc Stephan Gerhold <[email protected]>
---
 drivers/iio/gyro/bmg160_core.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/gyro/bmg160_core.c +7 -3
index b11ebd9bb7a4..7bc13ff2c3ac 100644
@@ -98,7 +98,11 @@ struct bmg160_data {
 	struct iio_trigger *motion_trig;
 	struct iio_mount_matrix orientation;
 	struct mutex mutex;
-	s16 buffer[8];
+	/* Ensure naturally aligned timestamp */
+	struct {
+		s16 chans[3];
+		s64 timestamp __aligned(8);
+	} scan;
 	u32 dps_range;
 	int ev_enable_state;
 	int slope_thres;
@@ -882,12 +886,12 @@ static irqreturn_t bmg160_trigger_handler(int irq, void *p)
 
 	mutex_lock(&data->mutex);
 	ret = regmap_bulk_read(data->regmap, BMG160_REG_XOUT_L,
-			       data->buffer, AXIS_MAX * 2);
+			       data->scan.chans, AXIS_MAX * 2);
 	mutex_unlock(&data->mutex);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
Signature · 1 line
2.31.1
author maintainer
#18 lore raw

PATCH 11/19 iio: humidity: am2315: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 0d96d5ead3f7 ("iio: humidity: Add triggered buffer support for AM2315")
---
 drivers/iio/humidity/am2315.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)
drivers/iio/humidity/am2315.c +10 -6
index 23bc9c784ef4..248d0f262d60 100644
@@ -33,7 +33,11 @@
 struct am2315_data {
 	struct i2c_client *client;
 	struct mutex lock;
-	s16 buffer[8]; /* 2x16-bit channels + 2x16 padding + 4x16 timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s16 chans[2];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 struct am2315_sensor_data {
@@ -167,20 +171,20 @@ static irqreturn_t am2315_trigger_handler(int irq, void *p)
 
 	mutex_lock(&data->lock);
 	if (*(indio_dev->active_scan_mask) == AM2315_ALL_CHANNEL_MASK) {
-		data->buffer[0] = sensor_data.hum_data;
-		data->buffer[1] = sensor_data.temp_data;
+		data->scan.chans[0] = sensor_data.hum_data;
+		data->scan.chans[1] = sensor_data.temp_data;
 	} else {
 		i = 0;
 		for_each_set_bit(bit, indio_dev->active_scan_mask,
 				 indio_dev->masklength) {
-			data->buffer[i] = (bit ? sensor_data.temp_data :
-						 sensor_data.hum_data);
+			data->scan.chans[i] = (bit ? sensor_data.temp_data :
+					       sensor_data.hum_data);
 			i++;
 		}
 	}
 	mutex_unlock(&data->lock);
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 err:
 	iio_trigger_notify_done(indio_dev->trig);
Signature · 1 line
2.31.1
author maintainer
#19 lore raw

PATCH 12/19 iio: prox: srf08: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 78f839029e1d ("iio: distance: srf08: add IIO driver for us ranger")
Cc Andreas Klinger <[email protected]>
---
 drivers/iio/proximity/srf08.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
drivers/iio/proximity/srf08.c +7 -7
index 70beac5c9c1d..9b0886760f76 100644
@@ -63,11 +63,11 @@ struct srf08_data {
 	int			range_mm;
 	struct mutex		lock;
 
-	/*
-	 * triggered buffer
-	 * 1x16-bit channel + 3x16 padding + 4x16 timestamp
-	 */
-	s16			buffer[8];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s16 chan;
+		s64 timestamp __aligned(8);
+	} scan;
 
 	/* Sensor-Type */
 	enum srf08_sensor_type	sensor_type;
@@ -190,9 +190,9 @@ static irqreturn_t srf08_trigger_handler(int irq, void *p)
 
 	mutex_lock(&data->lock);
 
-	data->buffer[0] = sensor_data;
+	data->scan.chan = sensor_data;
 	iio_push_to_buffers_with_timestamp(indio_dev,
-						data->buffer, pf->timestamp);
+					   &data->scan, pf->timestamp);
 
 	mutex_unlock(&data->lock);
 err:
Signature · 1 line
2.31.1
author maintainer
#20 lore raw

PATCH 13/19 iio: prox: pulsed-light: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes cb119d535083 ("iio: proximity: add support for PulsedLight LIDAR")
Cc Matt Ranostay <[email protected]>
---
 drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +7 -3
index c685f10b5ae4..822a68ae5e03 100644
@@ -44,7 +44,11 @@ struct lidar_data {
 	int (*xfer)(struct lidar_data *data, u8 reg, u8 *val, int len);
 	int i2c_enabled;
 
-	u16 buffer[8]; /* 2 byte distance + 8 byte timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u16 chan;
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 static const struct iio_chan_spec lidar_channels[] = {
@@ -229,9 +233,9 @@ static irqreturn_t lidar_trigger_handler(int irq, void *private)
 	struct lidar_data *data = iio_priv(indio_dev);
 	int ret;
 
-	ret = lidar_get_measurement(data, data->buffer);
+	ret = lidar_get_measurement(data, &data->scan.chan);
 	if (!ret) {
-		iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+		iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 						   iio_get_time_ns(indio_dev));
 	} else if (ret != -EINVAL) {
 		dev_err(&data->client->dev, "cannot read LIDAR measurement");
Signature · 1 line
2.31.1
Acked-by on v1 13/19
#21 lore raw

Re: [PATCH 13/19] iio: prox: pulsed-light: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 11 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes cb119d535083 ("iio: proximity: add support for PulsedLight LIDAR")
Quoted 37 lines
Cc Matt Ranostay <[email protected]>
---
 drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +7 -4
index c685f10b5ae4..822a68ae5e03 100644
@@ -44,7 +44,11 @@ struct lidar_data {
        int (*xfer)(struct lidar_data *data, u8 reg, u8 *val, int len);
        int i2c_enabled;
 
-       u16 buffer[8]; /* 2 byte distance + 8 byte timestamp */
+       /* Ensure timestamp is naturally aligned */
+       struct {
+               u16 chan;
+               s64 timestamp __aligned(8);
+       } scan;
 };
 
 static const struct iio_chan_spec lidar_channels[] = {
@@ -229,9 +233,9 @@ static irqreturn_t lidar_trigger_handler(int irq, void *private)
        struct lidar_data *data = iio_priv(indio_dev);
        int ret;
 
-       ret = lidar_get_measurement(data, data->buffer);
+       ret = lidar_get_measurement(data, &data->scan.chan);
        if (!ret) {
-               iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+               iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
                                                   iio_get_time_ns(indio_dev));
        } else if (ret != -EINVAL) {
                dev_err(&data->client->dev, "cannot read LIDAR measurement");
--

2.31.1

author maintainer
#22 lore raw

PATCH 14/19 iio: prox: as3935: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 37b1ba2c68cf ("iio: proximity: as3935: fix buffer stack trashing")
Cc Matt Ranostay <[email protected]>
---
 drivers/iio/proximity/as3935.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/proximity/as3935.c +7 -3
index edc4a35ae66d..1d5ace2bde44 100644
@@ -59,7 +59,11 @@ struct as3935_state {
 	unsigned long noise_tripped;
 	u32 tune_cap;
 	u32 nflwdth_reg;
-	u8 buffer[16]; /* 8-bit data + 56-bit padding + 64-bit timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u8 chan;
+		s64 timestamp __aligned(8);
+	} scan;
 	u8 buf[2] ____cacheline_aligned;
 };
 
@@ -225,8 +229,8 @@ static irqreturn_t as3935_trigger_handler(int irq, void *private)
 	if (ret)
 		goto err_read;
 
-	st->buffer[0] = val & AS3935_DATA_MASK;
-	iio_push_to_buffers_with_timestamp(indio_dev, &st->buffer,
+	st->scan.chan = val & AS3935_DATA_MASK;
+	iio_push_to_buffers_with_timestamp(indio_dev, &st->scan,
 					   iio_get_time_ns(indio_dev));
 err_read:
 	iio_trigger_notify_done(indio_dev->trig);
Signature · 1 line
2.31.1
Acked-by on v1 14/19
#23 lore raw

Re: [PATCH 14/19] iio: prox: as3935: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 11 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 37b1ba2c68cf ("iio: proximity: as3935: fix buffer stack trashing")
Quoted 36 lines
Cc Matt Ranostay <[email protected]>
---
 drivers/iio/proximity/as3935.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/proximity/as3935.c +7 -4
index edc4a35ae66d..1d5ace2bde44 100644
@@ -59,7 +59,11 @@ struct as3935_state {
        unsigned long noise_tripped;
        u32 tune_cap;
        u32 nflwdth_reg;
-       u8 buffer[16]; /* 8-bit data + 56-bit padding + 64-bit timestamp */
+       /* Ensure timestamp is naturally aligned */
+       struct {
+               u8 chan;
+               s64 timestamp __aligned(8);
+       } scan;
        u8 buf[2] ____cacheline_aligned;
 };
 
@@ -225,8 +229,8 @@ static irqreturn_t as3935_trigger_handler(int irq, void *private)
        if (ret)
                goto err_read;
 
-       st->buffer[0] = val & AS3935_DATA_MASK;
-       iio_push_to_buffers_with_timestamp(indio_dev, &st->buffer,
+       st->scan.chan = val & AS3935_DATA_MASK;
+       iio_push_to_buffers_with_timestamp(indio_dev, &st->scan,
                                           iio_get_time_ns(indio_dev));
 err_read:
        iio_trigger_notify_done(indio_dev->trig);
--

2.31.1

author maintainer
#24 lore raw

PATCH 15/19 iio: magn: hmc5843: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 7247645f6865 ("iio: hmc5843: Move hmc5843 out of staging")
---
 drivers/iio/magnetometer/hmc5843.h      | 8 ++++++--
 drivers/iio/magnetometer/hmc5843_core.c | 4 ++--
 2 files changed, 8 insertions(+), 4 deletions(-)
drivers/iio/magnetometer/hmc5843.h +6 -2
index 3f6c0b662941..242f742f2643 100644
@@ -33,7 +33,8 @@ enum hmc5843_ids {
  * @lock:		update and read regmap data
  * @regmap:		hardware access register maps
  * @variant:		describe chip variants
- * @buffer:		3x 16-bit channels + padding + 64-bit timestamp
+ * @scan:		buffer to pack data for passing to
+ *			iio_push_to_buffers_with_timestamp()
  */
 struct hmc5843_data {
 	struct device *dev;
@@ -41,7 +42,10 @@ struct hmc5843_data {
 	struct regmap *regmap;
 	const struct hmc5843_chip_info *variant;
 	struct iio_mount_matrix orientation;
-	__be16 buffer[8];
+	struct {
+		__be16 chans[3];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 int hmc5843_common_probe(struct device *dev, struct regmap *regmap,
drivers/iio/magnetometer/hmc5843_core.c +2 -2
index 780faea61d82..221563e0c18f 100644
@@ -446,13 +446,13 @@ static irqreturn_t hmc5843_trigger_handler(int irq, void *p)
 	}
 
 	ret = regmap_bulk_read(data->regmap, HMC5843_DATA_OUT_MSB_REGS,
-			       data->buffer, 3 * sizeof(__be16));
+			       data->scan.chans, sizeof(data->scan.chans));
 
 	mutex_unlock(&data->lock);
 	if (ret < 0)
 		goto done;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   iio_get_time_ns(indio_dev));
 
 done:
Signature · 1 line
2.31.1
author maintainer
#25 lore raw

PATCH 16/19 iio: magn: bmc150: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes c91746a2361d ("iio: magn: Add support for BMC150 magnetometer")
Cc Stephan Gerhold <[email protected]>
Cc Linus Walleij <[email protected]>
---
 drivers/iio/magnetometer/bmc150_magn.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iio/magnetometer/bmc150_magn.c +7 -4
index 00f9766bad5c..dd5f80093a18 100644
@@ -138,8 +138,11 @@ struct bmc150_magn_data {
 	struct regmap *regmap;
 	struct regulator_bulk_data regulators[2];
 	struct iio_mount_matrix orientation;
-	/* 4 x 32 bits for x, y z, 4 bytes align, 64 bits timestamp */
-	s32 buffer[6];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s32 chans[4];
+		s64 timestamp __aligned(8);
+	} scan;
 	struct iio_trigger *dready_trig;
 	bool dready_trigger_on;
 	int max_odr;
@@ -675,11 +678,11 @@ static irqreturn_t bmc150_magn_trigger_handler(int irq, void *p)
 	int ret;
 
 	mutex_lock(&data->mutex);
-	ret = bmc150_magn_read_xyz(data, data->buffer);
+	ret = bmc150_magn_read_xyz(data, data->scan.chans);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 
 err:
Signature · 1 line
2.31.1
Reviewed-by on v1 16/19
#26 lore raw

Re: [PATCH 16/19] iio: magn: bmc150: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 5 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Quoted 7 lines

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes c91746a2361d ("iio: magn: Add support for BMC150 magnetometer")
Cc Stephan Gerhold <[email protected]>
Cc Linus Walleij <[email protected]>

Excellent work Jonathan.

I wonder if there is some way for us to abstract this into the core so we can't get it wrong.

Yours,
Linus Walleij

#27 lore raw

Re: [PATCH 16/19] iio: magn: bmc150: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 15 lines · On Wed, 5 May 2021 14:57:12 +0200 Linus Walleij <[email protected]>…
Quoted 5 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Quoted 7 lines

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes c91746a2361d ("iio: magn: Add support for BMC150 magnetometer")
Cc Stephan Gerhold <[email protected]>
Cc Linus Walleij <[email protected]>
Quoted 5 lines

Excellent work Jonathan.

I wonder if there is some way for us to abstract this into the core so we can't get it wrong.

Abstracting is a bit of a pain, because we end up creating unnecessary limitations on what can be done. Often having buffer a lot larger than it needs to be is sensible for example...

However, I'm definitely thinking we should look at what checks we can add once all these cases are fixed and there might be a nice pattern to use for those cases where we currently have a bounce buffer anyway due to hardware restrictions. In most others, moving to the pattern where the timestamp is explicit in the structure makes it obvious (subject to the fun question of x86_32 alignment and whether that matters - we don't know of any bugs as a result but it's possible some buffer consumer will assume 8 byte alignment - hence the hardening in these cases).

The size being too small case for example should be easy - we just augment iio_push_to_buffers_with_timestamp() to take the size and check it against scan_bytes. Alignment is trickier...

Jonathan

Quoted 2 lines

Yours,
Linus Walleij

author maintainer Applied
#28 lore raw

Re: [PATCH 16/19] iio: magn: bmc150: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron
Quoted 5 lines · On Sat, 1 May 2021 18:01:18 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Quoted 7 lines

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes c91746a2361d ("iio: magn: Add support for BMC150 magnetometer")
Cc Stephan Gerhold <[email protected]>
Cc Linus Walleij <[email protected]>

Applied, but with a tiny tweak to make this more obviously correct.

Quoted 9 lines
---
 drivers/iio/magnetometer/bmc150_magn.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iio/magnetometer/bmc150_magn.c +0 -0
index 00f9766bad5c..dd5f80093a18 100644

drivers/iio/magnetometer/bmc150_magn.c struct bmc150_magn_data { · line 143

Quoted 8 lines
drivers/iio/magnetometer/bmc150_magn.c +3 -2
@@ -138,8 +138,11 @@ struct bmc150_magn_data {
 	struct regmap *regmap;
 	struct regulator_bulk_data regulators[2];
 	struct iio_mount_matrix orientation;
-	/* 4 x 32 bits for x, y z, 4 bytes align, 64 bits timestamp */
-	s32 buffer[6];
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		s32 chans[4];

The whole point of the structure is to enforce the alignment without needing
to specify padding. There are only 3 actual chans, so I've reduced this to
s32 chans[3]

Result is exactly the same in practice, but it's more readable.

Thanks,

Jonathan

Quoted 19 lines
drivers/iio/magnetometer/bmc150_magn.c +4 -2
+		s64 timestamp __aligned(8);
+	} scan;
 	struct iio_trigger *dready_trig;
 	bool dready_trigger_on;
 	int max_odr;
@@ -675,11 +678,11 @@ static irqreturn_t bmc150_magn_trigger_handler(int irq, void *p)
 	int ret;
 
 	mutex_lock(&data->mutex);
-	ret = bmc150_magn_read_xyz(data, data->buffer);
+	ret = bmc150_magn_read_xyz(data, data->scan.chans);
 	if (ret < 0)
 		goto err;
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 					   pf->timestamp);
 
 err:
author maintainer
#29 lore raw

PATCH 17/19 iio: light: isl29125: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes 6c25539cbc46 ("iio: Add Intersil isl29125 digital color light sensor driver")
---
 drivers/iio/light/isl29125.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/light/isl29125.c +7 -3
index b93b85dbc3a6..ba53b50d711a 100644
@@ -51,7 +51,11 @@
 struct isl29125_data {
 	struct i2c_client *client;
 	u8 conf1;
-	u16 buffer[8]; /* 3x 16-bit, padding, 8 bytes timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u16 chans[3];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 #define ISL29125_CHANNEL(_color, _si) { \
@@ -184,10 +188,10 @@ static irqreturn_t isl29125_trigger_handler(int irq, void *p)
 		if (ret < 0)
 			goto done;
 
-		data->buffer[j++] = ret;
+		data->scan.chans[j++] = ret;
 	}
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 		iio_get_time_ns(indio_dev));
 
 done:
Signature · 1 line
2.31.1
author maintainer
#30 lore raw

PATCH 18/19 iio: light: tcs3414: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes a244e7b57f0f ("iio: Add driver for AMS/TAOS tcs3414 digital color sensor")
---
 drivers/iio/light/tcs3414.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/light/tcs3414.c +7 -3
index 6fe5d46f80d4..0593abd600ec 100644
@@ -53,7 +53,11 @@ struct tcs3414_data {
 	u8 control;
 	u8 gain;
 	u8 timing;
-	u16 buffer[8]; /* 4x 16-bit + 8 bytes timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u16 chans[4];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 #define TCS3414_CHANNEL(_color, _si, _addr) { \
@@ -209,10 +213,10 @@ static irqreturn_t tcs3414_trigger_handler(int irq, void *p)
 		if (ret < 0)
 			goto done;
 
-		data->buffer[j++] = ret;
+		data->scan.chans[j++] = ret;
 	}
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 		iio_get_time_ns(indio_dev));
 
 done:
Signature · 1 line
2.31.1
author maintainer
#31 lore raw

PATCH 19/19 iio: light: tcs3472: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Fixes tag is not strictly accurate as prior to that patch there was potentially an unaligned write. However, any backport past there will need to be done manually.

Fixes 0624bf847dd0 ("iio:tcs3472: Use iio_push_to_buffers_with_timestamp()")
---
 drivers/iio/light/tcs3472.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/light/tcs3472.c +7 -3
index a0dc447aeb68..90dc3fef59e6 100644
@@ -64,7 +64,11 @@ struct tcs3472_data {
 	u8 control;
 	u8 atime;
 	u8 apers;
-	u16 buffer[8]; /* 4 16-bit channels + 64-bit timestamp */
+	/* Ensure timestamp is naturally aligned */
+	struct {
+		u16 chans[4];
+		s64 timestamp __aligned(8);
+	} scan;
 };
 
 static const struct iio_event_spec tcs3472_events[] = {
@@ -386,10 +390,10 @@ static irqreturn_t tcs3472_trigger_handler(int irq, void *p)
 		if (ret < 0)
 			goto done;
 
-		data->buffer[j++] = ret;
+		data->scan.chans[j++] = ret;
 	}
 
-	iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
 		iio_get_time_ns(indio_dev));
 
 done:
Signature · 1 line
2.31.1
#32 lore raw

Re: [PATCH 19/19] iio: light: tcs3472: Fix buffer alignment in iio_push_to_buffers_with_timestamp()

Quoted 8 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

To make code more readable, use a structure to express the channel layout and ensure the timestamp is 8 byte aligned.

Found during an audit of all calls of uses of
iio_push_to_buffers_with_timestamp()

Missed period at the end of the line.

Quoted 43 lines

Fixes tag is not strictly accurate as prior to that patch there was potentially an unaligned write. However, any backport past there will need to be done manually.

Fixes 0624bf847dd0 ("iio:tcs3472: Use iio_push_to_buffers_with_timestamp()")
---
 drivers/iio/light/tcs3472.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/iio/light/tcs3472.c +7 -4
index a0dc447aeb68..90dc3fef59e6 100644
@@ -64,7 +64,11 @@ struct tcs3472_data {
        u8 control;
        u8 atime;
        u8 apers;
-       u16 buffer[8]; /* 4 16-bit channels + 64-bit timestamp */
+       /* Ensure timestamp is naturally aligned */
+       struct {
+               u16 chans[4];
+               s64 timestamp __aligned(8);
+       } scan;
 };
 
 static const struct iio_event_spec tcs3472_events[] = {
@@ -386,10 +390,10 @@ static irqreturn_t tcs3472_trigger_handler(int irq, void *p)
                if (ret < 0)
                        goto done;
 
-               data->buffer[j++] = ret;
+               data->scan.chans[j++] = ret;
        }
 
-       iio_push_to_buffers_with_timestamp(indio_dev, data->buffer,
+       iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
                iio_get_time_ns(indio_dev));
 
 done:
--

2.31.1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
Reviewed-by on the v1 series
#33 lore raw
In reply to Jonathan Cameron
Quoted 11 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

I finally got around to do a manual audit of all the calls to iio_push_to_buffers_with_timestamp() which has the somewhat odd requirements of:
1. 8 byte alignment of the provided buffer.
2. space for an 8 byte naturally aligned timestamp to be inserted at the end.

Unfortuantely there were rather a lot of these left, but time to bite the bullet

Unfortunately

Quoted 14 lines

and clean them up.

As discussed previous in
https://lore.kernel.org/linux-iio/[email protected]/
it is not easy to fix the alignment issue without requiring a bounce buffer (see part 4 of the alignment fixes for a proposal for that where it is absolutely necessary).

Part 2 is the simple cases where the timestamp position is always the same (as not more than 8 bytes of other scan elements) and where it is either in a structure that is kzalloc'd or there was already an explicit memset that would clear any holes in the structure (and hence avoid a potential kernel data leak).

Quoted 3 lines
Cc Peter Meerwald <[email protected]>
Cc Andy Shevchenko <[email protected]>

1

Quoted 4 lines
Cc Srinivas Pandruvada <[email protected]>
Cc Daniel Baluta <[email protected]>
Cc Andy Shevchenko <[email protected]>

2

One time is enough :-)

Anyway, I looked at them all and haven't seen any issues on the first glance, so

Quoted 71 lines
Cc Stefan-Gabriel Mirea <[email protected]>
Cc Sanchayan Maity <[email protected]>
Cc Stephan Gerhold <[email protected]>
Cc Andreas Klinger <[email protected]>
Cc Matt Ranostay <[email protected]>
Cc Linus Walleij <[email protected]>

Jonathan Cameron (19):
iio: accel: bma180: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: bma220: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: hid: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: kxcjk-1013: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: mxc4005: Fix overread of data and alignment issue.
iio: accel: stk8312: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: stk8ba50: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: ti-ads1015: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: vf610: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: gyro: bmg160: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: humidity: am2315: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: srf08: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: pulsed-light: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: as3935: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: hmc5843: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: bmc150: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: isl29125: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3414: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3472: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()

 drivers/iio/accel/bma180.c                    | 10 +++++---
 drivers/iio/accel/bma220_spi.c                | 10 +++++---
 drivers/iio/accel/hid-sensor-accel-3d.c       | 13 ++++++----
 drivers/iio/accel/kxcjk-1013.c                | 24 +++++++++++--------
 drivers/iio/accel/mxc4005.c                   | 10 +++++---
 drivers/iio/accel/stk8312.c                   | 12 ++++++----
 drivers/iio/accel/stk8ba50.c                  | 17 +++++++------
 drivers/iio/adc/ti-ads1015.c                  | 12 ++++++----
 drivers/iio/adc/vf610_adc.c                   | 10 +++++---
 drivers/iio/gyro/bmg160_core.c                | 10 +++++---
 drivers/iio/humidity/am2315.c                 | 16 ++++++++-----
 drivers/iio/light/isl29125.c                  | 10 +++++---
 drivers/iio/light/tcs3414.c                   | 10 +++++---
 drivers/iio/light/tcs3472.c                   | 10 +++++---
 drivers/iio/magnetometer/bmc150_magn.c        | 11 +++++----
 drivers/iio/magnetometer/hmc5843.h            |  8 +++++--
 drivers/iio/magnetometer/hmc5843_core.c       |  4 ++--
 drivers/iio/proximity/as3935.c                | 10 +++++---
 .../iio/proximity/pulsedlight-lidar-lite-v2.c | 10 +++++---
 drivers/iio/proximity/srf08.c                 | 14 +++++------
 20 files changed, 148 insertions(+), 83 deletions(-)

--
2.31.1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author maintainer
#34 lore raw
Quoted 38 lines · On Sat, 1 May 2021 22:12:40 +0300 Andy Shevchenko <[email protected]>…
Quoted 11 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

I finally got around to do a manual audit of all the calls to iio_push_to_buffers_with_timestamp() which has the somewhat odd requirements of:
1. 8 byte alignment of the provided buffer.
2. space for an 8 byte naturally aligned timestamp to be inserted at the end.

Unfortuantely there were rather a lot of these left, but time to bite the bullet

Unfortunately

Quoted 14 lines

and clean them up.

As discussed previous in
https://lore.kernel.org/linux-iio/[email protected]/
it is not easy to fix the alignment issue without requiring a bounce buffer (see part 4 of the alignment fixes for a proposal for that where it is absolutely necessary).

Part 2 is the simple cases where the timestamp position is always the same (as not more than 8 bytes of other scan elements) and where it is either in a structure that is kzalloc'd or there was already an explicit memset that would clear any holes in the structure (and hence avoid a potential kernel data leak).

Quoted 3 lines
Cc Peter Meerwald <[email protected]>
Cc Andy Shevchenko <[email protected]>

1

Quoted 2 lines
Cc Srinivas Pandruvada <[email protected]>
Cc Daniel Baluta <[email protected]>
Quoted 8 lines
Quoted 1 line
Cc Andy Shevchenko <[email protected]>

2

One time is enough :-)

Anyway, I looked at them all and haven't seen any issues on the first glance, so

Thanks. Whole series applied though seems I was being more than a bit careless in some of the early ones so I took a very close look at all them whilst applying.

Do too many similar patches, and mistakes slip in :(

Thanks,

Jonathan

Quoted 74 lines
Quoted 72 lines
Cc Stefan-Gabriel Mirea <[email protected]>
Cc Sanchayan Maity <[email protected]>
Cc Stephan Gerhold <[email protected]>
Cc Andreas Klinger <[email protected]>
Cc Matt Ranostay <[email protected]>
Cc Linus Walleij <[email protected]>

Jonathan Cameron (19):
iio: accel: bma180: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: bma220: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: hid: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: kxcjk-1013: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: mxc4005: Fix overread of data and alignment issue.
iio: accel: stk8312: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: accel: stk8ba50: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: ti-ads1015: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: adc: vf610: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: gyro: bmg160: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: humidity: am2315: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: srf08: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: pulsed-light: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: prox: as3935: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: hmc5843: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: magn: bmc150: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: isl29125: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3414: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()
iio: light: tcs3472: Fix buffer alignment in
iio_push_to_buffers_with_timestamp()

 drivers/iio/accel/bma180.c                    | 10 +++++---
 drivers/iio/accel/bma220_spi.c                | 10 +++++---
 drivers/iio/accel/hid-sensor-accel-3d.c       | 13 ++++++----
 drivers/iio/accel/kxcjk-1013.c                | 24 +++++++++++--------
 drivers/iio/accel/mxc4005.c                   | 10 +++++---
 drivers/iio/accel/stk8312.c                   | 12 ++++++----
 drivers/iio/accel/stk8ba50.c                  | 17 +++++++------
 drivers/iio/adc/ti-ads1015.c                  | 12 ++++++----
 drivers/iio/adc/vf610_adc.c                   | 10 +++++---
 drivers/iio/gyro/bmg160_core.c                | 10 +++++---
 drivers/iio/humidity/am2315.c                 | 16 ++++++++-----
 drivers/iio/light/isl29125.c                  | 10 +++++---
 drivers/iio/light/tcs3414.c                   | 10 +++++---
 drivers/iio/light/tcs3472.c                   | 10 +++++---
 drivers/iio/magnetometer/bmc150_magn.c        | 11 +++++----
 drivers/iio/magnetometer/hmc5843.h            |  8 +++++--
 drivers/iio/magnetometer/hmc5843_core.c       |  4 ++--
 drivers/iio/proximity/as3935.c                | 10 +++++---
 .../iio/proximity/pulsedlight-lidar-lite-v2.c | 10 +++++---
 drivers/iio/proximity/srf08.c                 | 14 +++++------
 20 files changed, 148 insertions(+), 83 deletions(-)

--
2.31.1