Message-ID: <CAFXKEHZtG-y0xq=ejHg66N-gztukzVQRaGDjzaN5sdAJOL_G5g@mail.gmail.com>
Date: Tue, 26 Nov 2024 22:53:12 +0100
From: Lothar Rubusch <l.rubusch@...il.com>
To: Jonathan Cameron <jic23@...nel.org>
Cc: lars@...afoo.de, Michael.Hennerich@...log.com, linux-iio@...r.kernel.org, 
	linux-kernel@...r.kernel.org, eraretuya@...il.com
Subject: Re: [PATCH v2 15/22] iio: accel: adxl345: reset the FIFO on error

On Sun, Nov 24, 2024 at 7:54 PM Jonathan Cameron <jic23@...nel.org> wrote:
>
> On Sun, 17 Nov 2024 18:26:44 +0000
> Lothar Rubusch <l.rubusch@...il.com> wrote:
>
> > Add a function to empty the FIFO and reset the INT_SOURCE register.
> > Reading out is used to reset the fifo again. For cleanup also a read
> > on the INT_SOURCE register is needed to allow the adxl345 to issue
> > interrupts again. Without clearing the fields no further interrupts
> > will happen.
> >
> > Signed-off-by: Lothar Rubusch <l.rubusch@...il.com>
> > ---
> >  drivers/iio/accel/adxl345_core.c | 75 ++++++++++++++++++++++++++++----
> >  1 file changed, 67 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/iio/accel/adxl345_core.c b/drivers/iio/accel/adxl345_core.c
> > index 40e78dbdb0..82bd5c2b78 100644
> > --- a/drivers/iio/accel/adxl345_core.c
> > +++ b/drivers/iio/accel/adxl345_core.c
> > @@ -356,6 +356,61 @@ static int adxl345_get_fifo_entries(struct adxl34x_state *st, int *fifo_entries)
> >       return 0;
> >  }
> >
> > +/**
> > + * adxl345_read_fifo_elements() - Read fifo_entries number of elements.
> > + * @st: The instance of the state object of this sensor.
> > + * @fifo_entries: The number of lines in the FIFO referred to as fifo_entry,
> > + * a fifo_entry has 3 elements for X, Y and Z direction of 2 bytes each.
> > + *
> > + * The FIFO of the sensor is read linewise. The read measurement values are
> > + * queued in the corresponding data structure in *st.
> > + *
> > + * It is recommended that a multiple-byte read of all registers be performed to
> > + * prevent a change in data between reads of sequential registers. That is to
> > + * read out the data registers X0, X1, Y0, Y1, Z0, Z1 at once.
>
> To ensure this, set avail_scan_modes.
> Then if the user requests a subset, the IIO core code will extract what is necessary
> from the read of everythign.

Very intersting. Unfortunately, I could not find "avail_scan_modes".
Did you mean "avail_scan_masks"? I saw some drivers operating with
such. Could you give me some more keywords, that I could have a look
how this is done, pls?

This must be (one of) the reason(s) why with noinc read I only managed
to read 3 directions from the FIFO at a time. Actually, I guess, it
should be possible to dump the entire FIFO.

Lothar

> > + *
> > + * Return: 0 or error value.
> > + */
> > +static int adxl345_read_fifo_elements(struct adxl34x_state *st, int fifo_entries)
> > +{
> > +     size_t count, ndirs = 3;
> > +     int i, ret;
> > +
> > +     count = 2 * ndirs; /* 2 byte per direction */
> sizeof(st->fifo_buf[0] * ndirs);
>
> > +     for (i = 0; i < fifo_entries; i++) {
> > +             ret = regmap_noinc_read(st->regmap, ADXL345_REG_XYZ_BASE,
> > +                             st->fifo_buf + (i * count / 2), count);
> > +             if (ret) {
> > +                     pr_warn("%s(): regmap_noinc_read() failed\n", __func__);
> > +                     return -EFAULT;
> > +             }
> > +     }
> > +
> > +     return 0;
> > +}
> > +
> > +/**
> > + * adxl345_empty_fifo() - Empty the FIFO.
> > + * @st: The instance to the state object of the sensor.
> > + *
> > + * Reading all elements of the FIFO linewise empties the FIFO. Reading th
> > + * interrupt source register resets the sensor. This is needed also in case of
> > + * overflow or error handling to reenable the sensor to issue interrupts.
> > + */
> > +static void adxl345_empty_fifo(struct adxl34x_state *st)
> > +{
> > +     int regval;
> > +     int fifo_entries;
> > +
> > +     /* In case the HW is not "clean" just read out remaining elements */
> > +     adxl345_get_fifo_entries(st, &fifo_entries);
> > +     if (fifo_entries > 0)
> > +             adxl345_read_fifo_elements(st, fifo_entries);
> > +
> > +     /* Reset the INT_SOURCE register by reading the register */
> > +     regmap_read(st->regmap, ADXL345_REG_INT_SOURCE, &regval);
> > +}
> > +
> >  static const struct iio_buffer_setup_ops adxl345_buffer_ops = {
> >  };
> >
> > @@ -401,30 +456,34 @@ static irqreturn_t adxl345_trigger_handler(int irq, void *p)
> >
> >       ret = adxl345_get_status(st, &int_stat);
> >       if (ret < 0)
> > -             goto done;
> > +             goto err;
> All this churn just makes things less readable.  Better to have the bulk
> of the addition of fifo handling in one patch. It won't be too large
> for review.
> >
> >       /* Ignore already read event by reissued too fast */
> >       if (int_stat == 0x0)
> > -             goto done;
> > +             goto err;
> >
> >       /* evaluation */
> >
> >       if (int_stat & ADXL345_INT_OVERRUN) {
> >               pr_debug("%s(): OVERRUN event detected\n", __func__);
> > -             goto done;
> > +             goto err;
> >       }
> >
> >       if (int_stat & (ADXL345_INT_DATA_READY | ADXL345_INT_WATERMARK)) {
> >               pr_debug("%s(): WATERMARK or DATA_READY event detected\n", __func__);
> >               if (adxl345_get_fifo_entries(st, &fifo_entries) < 0)
> > -                     goto done;
> > -     }
> > -     goto done;
> > -done:
> > +                     goto err;
> >
> > -     if (indio_dev)
> >               iio_trigger_notify_done(indio_dev->trig);
> > +     }
> >
> > +     goto done;
> > +err:
> > +     iio_trigger_notify_done(indio_dev->trig);
> > +     adxl345_empty_fifo(st);
> > +     return IRQ_NONE;
> NONE is probably a bad idea. Something went wrong but that doesn't
> mean it wasn't our interrupt.  In most cases it is better to just
> return that we handled it.  IRQ_NONE might be valid if the status
> said it wasn't ours.
>
> > +
> > +done:
> >       return IRQ_HANDLED;
> return where you have goto done and get rid of this.
>
> >  }
> >
>

Powered by blists - more mailing lists