Re: [PATCH v3] iio: adc: add buffering support to the TI LMP92064 ADC driver

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Leonard,

LGTM, only one comment from me. See below...
On Mon, 2023-07-03 at 12:39 +0200, Leonard Göhrs wrote:
> Enable buffered reading of samples from the LMP92064 ADC.
> The main benefit of this change is being able to read out current and
> voltage measurements in a single transfer, allowing instantaneous power
> measurements.
> 
> Reads into the buffer can be triggered by any software triggers, e.g.
> the iio-trig-hrtimer:
> 
>     $ mkdir /sys/kernel/config/iio/triggers/hrtimer/my-trigger
>     $ cat /sys/bus/iio/devices/iio\:device3/name
>     lmp92064
>     $ iio_readdev -t my-trigger -b 16 iio:device3 | hexdump
>     WARNING: High-speed mode not enabled
>     0000000 0000 0176 0101 0001 5507 abd5 7645 1768
>     0000010 0000 016d 0101 0001 ee1e ac6b 7645 1768
>     ...
> 
> Signed-off-by: Leonard Göhrs <l.goehrs@xxxxxxxxxxxxxx>
> ---
> Changes v1->v2:
> 
>   - Remove superfluous .shift = 0 initialization in scan_type.
>   - Replace kmalloc buffer allocation for every read with a stack
>     allocated structure.
>     A struct is used to ensure correct alignment of the timestamp field.
>     See e.g. adc/rockchip_saradc.c for other users of the same pattern.
>   - Use available_scan_masks and always push both voltage and current
>     measurements to the buffer.
> 
> Changes v2->v3:
> 
>   - Handle errors returned by lmp92064_read_meas() out-of-line via
>     a goto err;
> 
> ---
>  drivers/iio/adc/ti-lmp92064.c | 51 +++++++++++++++++++++++++++++++++++
>  1 file changed, 51 insertions(+)
> 
> diff --git a/drivers/iio/adc/ti-lmp92064.c b/drivers/iio/adc/ti-lmp92064.c
> index c30ed824924f3..73504a27a8e22 100644
> --- a/drivers/iio/adc/ti-lmp92064.c
> +++ b/drivers/iio/adc/ti-lmp92064.c
> @@ -16,7 +16,10 @@
>  #include <linux/spi/spi.h>
>  
>  #include <linux/iio/iio.h>
> +#include <linux/iio/buffer.h>
>  #include <linux/iio/driver.h>
> +#include <linux/iio/triggered_buffer.h>
> +#include <linux/iio/trigger_consumer.h>
>  
>  #define TI_LMP92064_REG_CONFIG_A 0x0000
>  #define TI_LMP92064_REG_CONFIG_B 0x0001
> @@ -91,6 +94,12 @@ static const struct iio_chan_spec lmp92064_adc_channels[] = {
>                 .address = TI_LMP92064_CHAN_INC,
>                 .info_mask_separate =
>                         BIT(IIO_CHAN_INFO_RAW) | BIT(IIO_CHAN_INFO_SCALE),
> +               .scan_index = TI_LMP92064_CHAN_INC,
> +               .scan_type = {
> +                       .sign = 'u',
> +                       .realbits = 12,
> +                       .storagebits = 16,
> +               },
>                 .datasheet_name = "INC",
>         },
>         {
> @@ -98,8 +107,20 @@ static const struct iio_chan_spec lmp92064_adc_channels[] = {
>                 .address = TI_LMP92064_CHAN_INV,
>                 .info_mask_separate =
>                         BIT(IIO_CHAN_INFO_RAW) | BIT(IIO_CHAN_INFO_SCALE),
> +               .scan_index = TI_LMP92064_CHAN_INV,
> +               .scan_type = {
> +                       .sign = 'u',
> +                       .realbits = 12,
> +                       .storagebits = 16,
> +               },
>                 .datasheet_name = "INV",
>         },
> +       IIO_CHAN_SOFT_TIMESTAMP(2),
> +};
> +
> +static const unsigned long lmp92064_scan_masks[] = {
> +       BIT(TI_LMP92064_CHAN_INC) | BIT(TI_LMP92064_CHAN_INV),
> +       0
>  };
>  
>  static int lmp92064_read_meas(struct lmp92064_adc_priv *priv, u16 *res)
> @@ -171,6 +192,30 @@ static int lmp92064_read_raw(struct iio_dev *indio_dev,
>         }
>  }
>  
> +static irqreturn_t lmp92064_trigger_handler(int irq, void *p)
> +{
> +       struct iio_poll_func *pf = p;
> +       struct iio_dev *indio_dev = pf->indio_dev;
> +       struct lmp92064_adc_priv *priv = iio_priv(indio_dev);
> +       struct {
> +               u16 values[2];
> +               int64_t timestamp;
> +       } data;
> +       int ret;
> +

I think you need to explicitly __align(8) your timestamp data. The reason is that
(IIRC) for x86 32 bits, a long long is not naturally aligned to 64bits but 32.

You also need to explicitly memset(0) your data so we don't leak data to userspace.

- Nuno Sá 




[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Input]     [Linux Kernel]     [Linux SCSI]     [X.org]

  Powered by Linux