Re: [PATCH] iio: imu: inv_mpu6050: read the full fifo when processing data

[Jonathan Cameron <Jonathan.Cameron@Huawei.com>]

On Mon, 3 Jul 2023 15:19:52 +0000
Jean-Baptiste Maneyrol <Jean-Baptiste.Maneyrol@tdk.com> wrote:

> Hello Jonathan,
> 
> any news about this patch?
> 
> Thanks for your feedback,

LGTM.  I'm traveling however, so I've just been marking stuff that
is ready for queuing up in patchwork. Will send replies when I actually
queue them.

https://patchwork.kernel.org/project/linux-iio/list/

See patches labeled 'queued' which I'm using for this purpose.
Normally I'd just use applied and push the tree out the same
day.

Jonathan

> JB
> 
> 
> From: INV Git Commit <INV.git-commit@tdk.com>
> Sent: Friday, June 23, 2023 10:29
> To: jic23@kernel.org <jic23@kernel.org>
> Cc: lars@metafoo.de <lars@metafoo.de>; linux-iio@vger.kernel.org <linux-iio@vger.kernel.org>; Jean-Baptiste Maneyrol <Jean-Baptiste.Maneyrol@tdk.com>
> Subject: [PATCH] iio: imu: inv_mpu6050: read the full fifo when processing data 
>  
> From: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com>
> 
> When processing data read the full fifo data in 1 time. If there
> are several samples in the FIFO, it means we are experiencing
> system delay. In this case, it is better to read all data with 1
> bus access than to add additional latency by doing several ones.
> 
> This requires to use a bigger buffer depending on chip FIFO size
> and do an additional local data copy before sending. But the cost
> is minimal and behavior is still better like this under system
> heavy load.
> 
> Signed-off-by: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com>
> ---
>  drivers/iio/imu/inv_mpu6050/inv_mpu_core.c |  3 +++
>  drivers/iio/imu/inv_mpu6050/inv_mpu_iio.h  |  4 ++--
>  drivers/iio/imu/inv_mpu6050/inv_mpu_ring.c | 19 +++++++++++++------
>  3 files changed, 18 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/iio/imu/inv_mpu6050/inv_mpu_core.c b/drivers/iio/imu/inv_mpu6050/inv_mpu_core.c
> index 13086b569b90..29f906c884bd 100644
> --- a/drivers/iio/imu/inv_mpu6050/inv_mpu_core.c
> +++ b/drivers/iio/imu/inv_mpu6050/inv_mpu_core.c
> @@ -1345,6 +1345,9 @@ static int inv_check_and_setup_chip(struct inv_mpu6050_state *st)
>          st->reg = hw_info[st->chip_type].reg;
>          memcpy(&st->chip_config, hw_info[st->chip_type].config,
>                 sizeof(st->chip_config));
> +       st->data = devm_kzalloc(regmap_get_device(st->map), st->hw->fifo_size, GFP_KERNEL);
> +       if (st->data == NULL)
> +               return -ENOMEM;
>  
>          /* check chip self-identification */
>          result = regmap_read(st->map, INV_MPU6050_REG_WHOAMI, &regval);
> diff --git a/drivers/iio/imu/inv_mpu6050/inv_mpu_iio.h b/drivers/iio/imu/inv_mpu6050/inv_mpu_iio.h
> index a51d103a57ca..ed5a96e78df0 100644
> --- a/drivers/iio/imu/inv_mpu6050/inv_mpu_iio.h
> +++ b/drivers/iio/imu/inv_mpu6050/inv_mpu_iio.h
> @@ -179,7 +179,7 @@ struct inv_mpu6050_hw {
>   *  @magn_raw_to_gauss:        coefficient to convert mag raw value to Gauss.
>   *  @magn_orient:       magnetometer sensor chip orientation if available.
>   *  @suspended_sensors:        sensors mask of sensors turned off for suspend
> - *  @data:             dma safe buffer used for bulk reads.
> + *  @data:             read buffer used for bulk reads.
>   */
>  struct inv_mpu6050_state {
>          struct mutex lock;
> @@ -203,7 +203,7 @@ struct inv_mpu6050_state {
>          s32 magn_raw_to_gauss[3];
>          struct iio_mount_matrix magn_orient;
>          unsigned int suspended_sensors;
> -       u8 data[INV_MPU6050_OUTPUT_DATA_SIZE] __aligned(IIO_DMA_MINALIGN);
> +       u8 *data;
>  };
>  
>  /*register and associated bit definition*/
> diff --git a/drivers/iio/imu/inv_mpu6050/inv_mpu_ring.c b/drivers/iio/imu/inv_mpu6050/inv_mpu_ring.c
> index d83f61a99504..66d4ba088e70 100644
> --- a/drivers/iio/imu/inv_mpu6050/inv_mpu_ring.c
> +++ b/drivers/iio/imu/inv_mpu6050/inv_mpu_ring.c
> @@ -52,6 +52,7 @@ irqreturn_t inv_mpu6050_read_fifo(int irq, void *p)
>          u16 fifo_count;
>          u32 fifo_period;
>          s64 timestamp;
> +       u8 data[INV_MPU6050_OUTPUT_DATA_SIZE];
>          int int_status;
>          size_t i, nb;
>  
> @@ -105,24 +106,30 @@ irqreturn_t inv_mpu6050_read_fifo(int irq, void *p)
>                  goto flush_fifo;
>          }
>  
> -       /* compute and process all complete datum */
> +       /* compute and process only all complete datum */
>          nb = fifo_count / bytes_per_datum;
> +       fifo_count = nb * bytes_per_datum;
>          /* Each FIFO data contains all sensors, so same number for FIFO and sensor data */
>          fifo_period = NSEC_PER_SEC / INV_MPU6050_DIVIDER_TO_FIFO_RATE(st->chip_config.divider);
>          inv_sensors_timestamp_interrupt(&st->timestamp, fifo_period, nb, nb, pf->timestamp);
>          inv_sensors_timestamp_apply_odr(&st->timestamp, fifo_period, nb, 0);
> +
> +       /* clear internal data buffer for avoiding kernel data leak */
> +       memset(data, 0, sizeof(data));
> +
> +       /* read all data once and process every samples */
> +       result = regmap_noinc_read(st->map, st->reg->fifo_r_w, st->data, fifo_count);
> +       if (result)
> +               goto flush_fifo;
>          for (i = 0; i < nb; ++i) {
> -               result = regmap_noinc_read(st->map, st->reg->fifo_r_w,
> -                                          st->data, bytes_per_datum);
> -               if (result)
> -                       goto flush_fifo;
>                  /* skip first samples if needed */
>                  if (st->skip_samples) {
>                          st->skip_samples--;
>                          continue;
>                  }
> +               memcpy(data, &st->data[i * bytes_per_datum], bytes_per_datum);
>                  timestamp = inv_sensors_timestamp_pop(&st->timestamp);
> -               iio_push_to_buffers_with_timestamp(indio_dev, st->data, timestamp);
> +               iio_push_to_buffers_with_timestamp(indio_dev, data, timestamp);
>          }
>  
>  end_session: