[PATCH v5 0/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[PATCH v5 0/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[Eddie James]

Corruption of the MEAS_CFG register has been observed soon after
system boot. In order to recover this scenario, check MEAS_CFG if
measurement isn't ready, and if it's incorrect, reset the DPS310
and execute the startup procedure. Include a patch to move the
startup procedure into a function.

Changes since v4:
 - Just check for rc rather than rc < 0 in some cases
 - Split declaration and init of rc

Changes since v3:
 - Don't check regmap* return codes for < 0
 - Fix comment spelling

Changes since v2:
 - Add some comments
 - Fix the clunky control flow

Changes since v1:
 - Separate into two patches
 - Rename 'dps310_verify_meas_cfg' to 'dps310_check_reset_meas_cfg'

Eddie James (2):
  iio: pressure: dps310: Refactor startup procedure
  iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

 drivers/iio/pressure/dps310.c | 277 +++++++++++++++++++++-------------
 1 file changed, 173 insertions(+), 104 deletions(-)

-- 
2.31.1

[PATCH v5 1/2] iio: pressure: dps310: Refactor startup procedure

[Eddie James]

Move the startup procedure into a function, and correct a missing
check on the return code for writing the PRS_CFG register.

Signed-off-by: Eddie James <eajames@linux.ibm.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Joel Stanley <joel@jms.id.au>
---
 drivers/iio/pressure/dps310.c | 188 ++++++++++++++++++----------------
 1 file changed, 99 insertions(+), 89 deletions(-)

diff --git a/drivers/iio/pressure/dps310.c b/drivers/iio/pressure/dps310.c
index 36fb7ae0d0a9..c706a8b423b5 100644
--- a/drivers/iio/pressure/dps310.c
+++ b/drivers/iio/pressure/dps310.c
@@ -159,6 +159,102 @@ static int dps310_get_coefs(struct dps310_data *data)
 	return 0;
 }
 
+/*
+ * Some versions of the chip will read temperatures in the ~60C range when
+ * it's actually ~20C. This is the manufacturer recommended workaround
+ * to correct the issue. The registers used below are undocumented.
+ */
+static int dps310_temp_workaround(struct dps310_data *data)
+{
+	int rc;
+	int reg;
+
+	rc = regmap_read(data->regmap, 0x32, &reg);
+	if (rc)
+		return rc;
+
+	/*
+	 * If bit 1 is set then the device is okay, and the workaround does not
+	 * need to be applied
+	 */
+	if (reg & BIT(1))
+		return 0;
+
+	rc = regmap_write(data->regmap, 0x0e, 0xA5);
+	if (rc)
+		return rc;
+
+	rc = regmap_write(data->regmap, 0x0f, 0x96);
+	if (rc)
+		return rc;
+
+	rc = regmap_write(data->regmap, 0x62, 0x02);
+	if (rc)
+		return rc;
+
+	rc = regmap_write(data->regmap, 0x0e, 0x00);
+	if (rc)
+		return rc;
+
+	return regmap_write(data->regmap, 0x0f, 0x00);
+}
+
+static int dps310_startup(struct dps310_data *data)
+{
+	int rc;
+	int ready;
+
+	/*
+	 * Set up pressure sensor in single sample, one measurement per second
+	 * mode
+	 */
+	rc = regmap_write(data->regmap, DPS310_PRS_CFG, 0);
+	if (rc)
+		return rc;
+
+	/*
+	 * Set up external (MEMS) temperature sensor in single sample, one
+	 * measurement per second mode
+	 */
+	rc = regmap_write(data->regmap, DPS310_TMP_CFG, DPS310_TMP_EXT);
+	if (rc)
+		return rc;
+
+	/* Temp and pressure shifts are disabled when PRC <= 8 */
+	rc = regmap_write_bits(data->regmap, DPS310_CFG_REG,
+			       DPS310_PRS_SHIFT_EN | DPS310_TMP_SHIFT_EN, 0);
+	if (rc)
+		return rc;
+
+	/* MEAS_CFG doesn't update correctly unless first written with 0 */
+	rc = regmap_write_bits(data->regmap, DPS310_MEAS_CFG,
+			       DPS310_MEAS_CTRL_BITS, 0);
+	if (rc)
+		return rc;
+
+	/* Turn on temperature and pressure measurement in the background */
+	rc = regmap_write_bits(data->regmap, DPS310_MEAS_CFG,
+			       DPS310_MEAS_CTRL_BITS, DPS310_PRS_EN |
+			       DPS310_TEMP_EN | DPS310_BACKGROUND);
+	if (rc)
+		return rc;
+
+	/*
+	 * Calibration coefficients required for reporting temperature.
+	 * They are available 40ms after the device has started
+	 */
+	rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG, ready,
+				      ready & DPS310_COEF_RDY, 10000, 40000);
+	if (rc)
+		return rc;
+
+	rc = dps310_get_coefs(data);
+	if (rc)
+		return rc;
+
+	return dps310_temp_workaround(data);
+}
+
 static int dps310_get_pres_precision(struct dps310_data *data)
 {
 	int rc;
@@ -677,52 +773,12 @@ static const struct iio_info dps310_info = {
 	.write_raw = dps310_write_raw,
 };
 
-/*
- * Some verions of chip will read temperatures in the ~60C range when
- * its actually ~20C. This is the manufacturer recommended workaround
- * to correct the issue. The registers used below are undocumented.
- */
-static int dps310_temp_workaround(struct dps310_data *data)
-{
-	int rc;
-	int reg;
-
-	rc = regmap_read(data->regmap, 0x32, &reg);
-	if (rc < 0)
-		return rc;
-
-	/*
-	 * If bit 1 is set then the device is okay, and the workaround does not
-	 * need to be applied
-	 */
-	if (reg & BIT(1))
-		return 0;
-
-	rc = regmap_write(data->regmap, 0x0e, 0xA5);
-	if (rc < 0)
-		return rc;
-
-	rc = regmap_write(data->regmap, 0x0f, 0x96);
-	if (rc < 0)
-		return rc;
-
-	rc = regmap_write(data->regmap, 0x62, 0x02);
-	if (rc < 0)
-		return rc;
-
-	rc = regmap_write(data->regmap, 0x0e, 0x00);
-	if (rc < 0)
-		return rc;
-
-	return regmap_write(data->regmap, 0x0f, 0x00);
-}
-
 static int dps310_probe(struct i2c_client *client,
 			const struct i2c_device_id *id)
 {
 	struct dps310_data *data;
 	struct iio_dev *iio;
-	int rc, ready;
+	int rc;
 
 	iio = devm_iio_device_alloc(&client->dev,  sizeof(*data));
 	if (!iio)
@@ -747,54 +803,8 @@ static int dps310_probe(struct i2c_client *client,
 	if (rc)
 		return rc;
 
-	/*
-	 * Set up pressure sensor in single sample, one measurement per second
-	 * mode
-	 */
-	rc = regmap_write(data->regmap, DPS310_PRS_CFG, 0);
-
-	/*
-	 * Set up external (MEMS) temperature sensor in single sample, one
-	 * measurement per second mode
-	 */
-	rc = regmap_write(data->regmap, DPS310_TMP_CFG, DPS310_TMP_EXT);
-	if (rc < 0)
-		return rc;
-
-	/* Temp and pressure shifts are disabled when PRC <= 8 */
-	rc = regmap_write_bits(data->regmap, DPS310_CFG_REG,
-			       DPS310_PRS_SHIFT_EN | DPS310_TMP_SHIFT_EN, 0);
-	if (rc < 0)
-		return rc;
-
-	/* MEAS_CFG doesn't update correctly unless first written with 0 */
-	rc = regmap_write_bits(data->regmap, DPS310_MEAS_CFG,
-			       DPS310_MEAS_CTRL_BITS, 0);
-	if (rc < 0)
-		return rc;
-
-	/* Turn on temperature and pressure measurement in the background */
-	rc = regmap_write_bits(data->regmap, DPS310_MEAS_CFG,
-			       DPS310_MEAS_CTRL_BITS, DPS310_PRS_EN |
-			       DPS310_TEMP_EN | DPS310_BACKGROUND);
-	if (rc < 0)
-		return rc;
-
-	/*
-	 * Calibration coefficients required for reporting temperature.
-	 * They are available 40ms after the device has started
-	 */
-	rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG, ready,
-				      ready & DPS310_COEF_RDY, 10000, 40000);
-	if (rc < 0)
-		return rc;
-
-	rc = dps310_get_coefs(data);
-	if (rc < 0)
-		return rc;
-
-	rc = dps310_temp_workaround(data);
-	if (rc < 0)
+	rc = dps310_startup(data);
+	if (rc)
 		return rc;
 
 	rc = devm_iio_device_register(&client->dev, iio);
-- 
2.31.1

[PATCH v5 2/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[Eddie James]

Corruption of the MEAS_CFG register has been observed soon after
system boot. In order to recover this scenario, check MEAS_CFG if
measurement isn't ready, and if it's incorrect, reset the DPS310
and execute the startup procedure.

Signed-off-by: Eddie James <eajames@linux.ibm.com>
Reviewed-by: Joel Stanley <joel@jms.id.au>
---
 drivers/iio/pressure/dps310.c | 89 +++++++++++++++++++++++++++++------
 1 file changed, 74 insertions(+), 15 deletions(-)

diff --git a/drivers/iio/pressure/dps310.c b/drivers/iio/pressure/dps310.c
index c706a8b423b5..bbeb2f3bcc8a 100644
--- a/drivers/iio/pressure/dps310.c
+++ b/drivers/iio/pressure/dps310.c
@@ -393,6 +393,45 @@ static int dps310_get_temp_k(struct dps310_data *data)
 	return scale_factors[ilog2(rc)];
 }
 
+/*
+ * Called with lock held. Returns a negative value on error, a positive value
+ * when the device is not ready, and zero when the device is ready.
+ */
+static int dps310_check_reset_meas_cfg(struct dps310_data *data, int ready_bit)
+{
+	int rc;
+	int meas_cfg;
+
+	rc = regmap_read(data->regmap, DPS310_MEAS_CFG, &meas_cfg);
+	if (rc)
+		return rc;
+
+	/* Device is ready, proceed to measurement */
+	if (meas_cfg & ready_bit)
+		return 0;
+
+	/* Device is OK, just not ready */
+	if (meas_cfg & (DPS310_PRS_EN | DPS310_TEMP_EN | DPS310_BACKGROUND))
+		return 1;
+
+	/* DPS310 register state corrupt, better start from scratch */
+	rc = regmap_write(data->regmap, DPS310_RESET, DPS310_RESET_MAGIC);
+	if (rc)
+		return rc;
+
+	/* Wait for device chip access: 2.5ms in specification */
+	usleep_range(2500, 12000);
+
+	/* Reinitialize the chip */
+	rc = dps310_startup(data);
+	if (rc)
+		return rc;
+
+	dev_info(&data->client->dev,
+		 "recovered from corrupted MEAS_CFG=%02x\n", meas_cfg);
+	return 1;
+}
+
 static int dps310_read_pres_raw(struct dps310_data *data)
 {
 	int rc;
@@ -405,16 +444,26 @@ static int dps310_read_pres_raw(struct dps310_data *data)
 	if (mutex_lock_interruptible(&data->lock))
 		return -EINTR;
 
-	rate = dps310_get_pres_samp_freq(data);
-	timeout = DPS310_POLL_TIMEOUT_US(rate);
-
-	/* Poll for sensor readiness; base the timeout upon the sample rate. */
-	rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG, ready,
-				      ready & DPS310_PRS_RDY,
-				      DPS310_POLL_SLEEP_US(timeout), timeout);
-	if (rc)
+	rc = dps310_check_reset_meas_cfg(data, DPS310_PRS_RDY);
+	if (rc < 0)
 		goto done;
 
+	if (rc > 0) {
+		rate = dps310_get_pres_samp_freq(data);
+		timeout = DPS310_POLL_TIMEOUT_US(rate);
+
+		/*
+		 * Poll for sensor readiness; base the timeout upon the sample
+		 * rate.
+		 */
+		rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG,
+					      ready, ready & DPS310_PRS_RDY,
+					      DPS310_POLL_SLEEP_US(timeout),
+					      timeout);
+		if (rc)
+			goto done;
+	}
+
 	rc = regmap_bulk_read(data->regmap, DPS310_PRS_BASE, val, sizeof(val));
 	if (rc < 0)
 		goto done;
@@ -454,16 +503,26 @@ static int dps310_read_temp_raw(struct dps310_data *data)
 	if (mutex_lock_interruptible(&data->lock))
 		return -EINTR;
 
-	rate = dps310_get_temp_samp_freq(data);
-	timeout = DPS310_POLL_TIMEOUT_US(rate);
-
-	/* Poll for sensor readiness; base the timeout upon the sample rate. */
-	rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG, ready,
-				      ready & DPS310_TMP_RDY,
-				      DPS310_POLL_SLEEP_US(timeout), timeout);
+	rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY);
 	if (rc < 0)
 		goto done;
 
+	if (rc > 0) {
+		rate = dps310_get_temp_samp_freq(data);
+		timeout = DPS310_POLL_TIMEOUT_US(rate);
+
+		/*
+		 * Poll for sensor readiness; base the timeout upon the sample
+		 * rate.
+		 */
+		rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG,
+					      ready, ready & DPS310_TMP_RDY,
+					      DPS310_POLL_SLEEP_US(timeout),
+					      timeout);
+		if (rc)
+			goto done;
+	}
+
 	rc = dps310_read_temp_ready(data);
 
 done:
-- 
2.31.1

Re: [PATCH v5 2/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[Andy Shevchenko]

On Mon, Aug 15, 2022 at 5:57 PM Eddie James <eajames@linux.ibm.com> wrote:
>
> Corruption of the MEAS_CFG register has been observed soon after
> system boot. In order to recover this scenario, check MEAS_CFG if
> measurement isn't ready, and if it's incorrect, reset the DPS310
> and execute the startup procedure.

...

> + * Called with lock held. Returns a negative value on error, a positive value
> + * when the device is not ready, and zero when the device is ready.

Can we have

#define DPS310_DEVICE_NOT_READY  1

(or anonymous enum) and return it instead of abstract 1 or any other
positive number?

...

> +       rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY);
>         if (rc < 0)
>                 goto done;
>
> +       if (rc > 0) {
> +               rate = dps310_get_temp_samp_freq(data);
> +               timeout = DPS310_POLL_TIMEOUT_US(rate);
> +
> +               /*
> +                * Poll for sensor readiness; base the timeout upon the sample
> +                * rate.
> +                */
> +               rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG,
> +                                             ready, ready & DPS310_TMP_RDY,
> +                                             DPS310_POLL_SLEEP_US(timeout),
> +                                             timeout);
> +               if (rc)
> +                       goto done;
> +       }

But have you tried to make a helper that takes a pointer to the
respective function?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v5 2/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[Jonathan Cameron]

On Thu, 18 Aug 2022 23:16:55 +0300
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Mon, Aug 15, 2022 at 5:57 PM Eddie James <eajames@linux.ibm.com> wrote:
> >
> > Corruption of the MEAS_CFG register has been observed soon after
> > system boot. In order to recover this scenario, check MEAS_CFG if
> > measurement isn't ready, and if it's incorrect, reset the DPS310
> > and execute the startup procedure.  
> 
> ...
> 
> > + * Called with lock held. Returns a negative value on error, a positive value
> > + * when the device is not ready, and zero when the device is ready.  
> 
> Can we have
> 
> #define DPS310_DEVICE_NOT_READY  1
> 
> (or anonymous enum) and return it instead of abstract 1 or any other
> positive number?

Perhaps make it even clearer by returning the need to wait via another parameter
rather than this being (probably) the only place in driver with a postive rc.

bool reset_done;

rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY, &reset_done);

if (reset_done) {
}

> 
> ...
> 
> > +       rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY);
> >         if (rc < 0)
> >                 goto done;
> >
> > +       if (rc > 0) {
> > +               rate = dps310_get_temp_samp_freq(data);
> > +               timeout = DPS310_POLL_TIMEOUT_US(rate);
> > +
> > +               /*
> > +                * Poll for sensor readiness; base the timeout upon the sample
> > +                * rate.
> > +                */
> > +               rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG,
> > +                                             ready, ready & DPS310_TMP_RDY,
> > +                                             DPS310_POLL_SLEEP_US(timeout),
> > +                                             timeout);
> > +               if (rc)
> > +                       goto done;
> > +       }  
> 
> But have you tried to make a helper that takes a pointer to the
> respective function?
> 

Re: [PATCH v5 2/2] iio: pressure: dps310: Reset chip if MEAS_CFG is corrupt

[Eddie James]

On 8/20/22 07:02, Jonathan Cameron wrote:
> On Thu, 18 Aug 2022 23:16:55 +0300
> Andy Shevchenko <andy.shevchenko@gmail.com> wrote:
>
>> On Mon, Aug 15, 2022 at 5:57 PM Eddie James <eajames@linux.ibm.com> wrote:
>>> Corruption of the MEAS_CFG register has been observed soon after
>>> system boot. In order to recover this scenario, check MEAS_CFG if
>>> measurement isn't ready, and if it's incorrect, reset the DPS310
>>> and execute the startup procedure.
>> ...
>>
>>> + * Called with lock held. Returns a negative value on error, a positive value
>>> + * when the device is not ready, and zero when the device is ready.
>> Can we have
>>
>> #define DPS310_DEVICE_NOT_READY  1
>>
>> (or anonymous enum) and return it instead of abstract 1 or any other
>> positive number?
> Perhaps make it even clearer by returning the need to wait via another parameter
> rather than this being (probably) the only place in driver with a postive rc.
>
> bool reset_done;
>
> rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY, &reset_done);
>
> if (reset_done) {
> }


Thanks for the feedback.

Unfortunately our issue is still presenting itself even with a correct 
MEAS_CFG register. The sensor ready bit never gets set. So I'm going to 
rework this patch to reset the device after a timeout.


Thanks,

Eddie


>
>> ...
>>
>>> +       rc = dps310_check_reset_meas_cfg(data, DPS310_TMP_RDY);
>>>          if (rc < 0)
>>>                  goto done;
>>>
>>> +       if (rc > 0) {
>>> +               rate = dps310_get_temp_samp_freq(data);
>>> +               timeout = DPS310_POLL_TIMEOUT_US(rate);
>>> +
>>> +               /*
>>> +                * Poll for sensor readiness; base the timeout upon the sample
>>> +                * rate.
>>> +                */
>>> +               rc = regmap_read_poll_timeout(data->regmap, DPS310_MEAS_CFG,
>>> +                                             ready, ready & DPS310_TMP_RDY,
>>> +                                             DPS310_POLL_SLEEP_US(timeout),
>>> +                                             timeout);
>>> +               if (rc)
>>> +                       goto done;
>>> +       }
>> But have you tried to make a helper that takes a pointer to the
>> respective function?
>>