lkml.org 
[lkml]   [2023]   [Jan]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH v2 3/5] iio: pressure: bmp280: Add support for new sensor BMP580
From
Date
On Fri, 2022-12-30 at 18:45 +0000, Jonathan Cameron wrote:
> On Mon, 26 Dec 2022 15:29:22 +0100
> Angel Iglesias <ang.iglesiasg@gmail.com> wrote:
>
> > Adds compatibility with the new sensor generation, the BMP580.
> >
> > The measurement and initialization codepaths are adapted from
> > the device datasheet and the repository from manufacturer at
> > https://github.com/boschsensortec/BMP5-Sensor-API.
> >
> > Signed-off-by: Angel Iglesias <ang.iglesiasg@gmail.com>
>
> Hi Angel,
>
> Some comments inline,
>
> Thanks,
>
> Jonathan
>
> >
> > diff --git a/drivers/iio/pressure/Kconfig b/drivers/iio/pressure/Kconfig
> > index c9453389e4f7..1c18e3b2c501 100644
> > --- a/drivers/iio/pressure/Kconfig
> > +++ b/drivers/iio/pressure/Kconfig
> > @@ -17,14 +17,14 @@ config ABP060MG
> >           will be called abp060mg.
> >  
> >  config BMP280
> > -       tristate "Bosch Sensortec BMP180/BMP280/BMP380 pressure sensor I2C
> > driver"
> > +       tristate "Bosch Sensortec BMP180/BMP280/BMP380/BMP580 pressure
> > sensor I2C driver"
> Future reference:  If this gets much longer we'll have to shorten to
> "BMP280 and similar" like we have already had to do for a bunch of other
> drivers
> that support too many parts to fit in the short description.  We carry on
> listing
> them all in the help though.
>
> I think this is the last time we can get aways with it.
>
> I2C driver?  Looks to be handling SPI as well.

Welp, you're right, last time I was here forgot to update the wording.

> >         depends on (I2C || SPI_MASTER)
> >         select REGMAP
> >         select BMP280_I2C if (I2C)
> >         select BMP280_SPI if (SPI_MASTER)
> >         help
> > -         Say yes here to build support for Bosch Sensortec BMP180, BMP280
> > and
> > -         BMP380 pressure and temperature sensors. Also supports the BME280
> > with
> > +         Say yes here to build support for Bosch Sensortec BMP180, BMP280,
> > BMP380
> > +         and BMP580 pressure and temperature sensors. Also supports the
> > BME280 with
> >           an additional humidity sensor channel.
> >  
> >           To compile this driver as a module, choose M here: the core module
>
> ...
>
>
> > +
> > +/*
> > + * Helper function to send a command to BMP5XX sensors.
> > + *
> > + * BMP5xx sensors have a series of commands actionable
> > + * writing specific sequences on the CMD register:
> > + * SOFT_RESET: performs a reset of the system.
> > + * NVM_READ: read the contents of a user position of the nvm memory.
> > + * NVM_WRITE: write new data to a user position of the nvm memory.
> > + * EXT_MODE: enable extended mode with additional debug pages.
> > + */
> > +static int bmp580_cmd(struct bmp280_data *data, enum bmp580_commands cmd)
>
> Is there an advantage in rolling these up in one function?  There seems to be
> no real
> shared code.  Also most of this isn't used currently I think, so perhaps
> should be
> introduced alongside code that uses it.

Probably, would be best to break up this code in two helpers functions for reset
and NVM operations and drop the code to setup debugging ext_mode. I'm not using
the ext_mode and there's no documentation about the new registers exposed in
that mode.

> > +{
> > +       unsigned long deadline;
> > +       unsigned int reg;
> > +       int ret;
> > +
> > +       switch (cmd) {
> > +       case BMP580_SOFT_RESET_CMD:
> > +               /* Send reset word */
> > +               ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > BMP580_CMD_SOFT_RESET);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to send reset command to
> > device\n");
> > +                       return ret;
> > +               }
> > +               /* Wait 2ms for reset completion */
> > +               usleep_range(2000, 2500);
>
> blank line.
>
> > +               /* Dummy read of chip_id */
> > +               ret = regmap_read(data->regmap, BMP580_REG_CHIP_ID, &reg);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to reestablish comms
> > after reset\n");
> > +                       return ret;
> > +               }
>
> blank line etc.  See below for reasoning.
>
> > +               /* Check if POR bit is set on interrupt reg */
> > +               ret = regmap_read(data->regmap, BMP580_REG_INT_STATUS,
> > &reg);
> > +               if (ret) {
> > +                       dev_err(data->dev, "error reading interrupt status
> > register\n");
> > +                       return ret;
> > +               }
> > +               if (!(reg & BMP580_INT_STATUS_POR_MASK)) {
> > +                       dev_err(data->dev, "error resetting sensor\n");
> > +                       return -EINVAL;
> > +               }
> > +               break;
> > +       case BMP580_NVM_WRITE_CMD:
> > +       case BMP580_NVM_READ_CMD:
> > +               /* Check nvm ready flag */
> > +               ret = regmap_read(data->regmap, BMP580_REG_STATUS, &reg);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to check nvm status\n");
> > +                       return ret;
> > +               }
> > +               if (!(reg & BMP580_STATUS_NVM_RDY_MASK)) {
> > +                       dev_err(data->dev, "sensor's nvm is not ready\n");
> > +                       return -EIO;
> > +               }
> > +               /* Send NVM operation sequence */
> > +               ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > BMP580_CMD_NVM_OP_SEQ_0);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to send nvm operation's
> > first sequence\n");
> > +                       return ret;
> > +               }
> > +               if (cmd == BMP580_NVM_WRITE_CMD) {
> > +                       /* Send write sequence */
> > +                       ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > +                                          BMP580_CMD_NVM_WRITE_SEQ_1);
> > +                       if (ret) {
> > +                               dev_err(data->dev, "failed to send nvm write
> > sequence\n");
> > +                               return ret;
> > +                       }
> > +                       /* Datasheet says on 4.8.1.2 it takes approximately
> > 10ms */
> > +                       usleep_range(10000, 10500);
> > +                       deadline = jiffies + msecs_to_jiffies(10);
> > +               } else {
> > +                       /* Send read sequence */
> > +                       ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > +                                          BMP580_CMD_NVM_READ_SEQ_1);
> > +                       if (ret) {
> > +                               dev_err(data->dev, "failed to send nvm read
> > sequence\n");
> > +                               return ret;
> > +                       }
> > +                       /* Datasheet says on 4.8.1.1 it takes approximately
> > 200us */
> > +                       usleep_range(200, 250);
> > +                       deadline = jiffies + usecs_to_jiffies(200);
> > +               }
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to write command
> > sequence\n");
> > +                       return -EIO;
> > +               }
> > +               /* Wait until NVM is ready again */
> > +               do {
> > +                       ret = regmap_read(data->regmap, BMP580_REG_STATUS,
> > &reg);
> > +                       if (ret) {
> > +                               dev_err(data->dev, "failed to check nvm
> > status\n");
> > +                               reg &= ~BMP580_STATUS_NVM_RDY_MASK;
> > +                       }
> > +               } while (time_before(jiffies, deadline) && !(reg &
> > BMP580_STATUS_NVM_RDY_MASK));
> > +
> > +               if (!(reg & BMP580_STATUS_NVM_RDY_MASK)) {
> > +                       dev_err(data->dev,
> > +                               "reached timeout waiting for nvm operation
> > completion\n");
> > +                       return -ETIMEDOUT;
> > +               }
> > +               /* Checks nvm error flags */
> > +               if ((reg & BMP580_STATUS_NVM_ERR_MASK) || (reg &
> > BMP580_STATUS_NVM_CMD_ERR_MASK)) {
> > +                       dev_err(data->dev, "error processing nvm
> > operation\n");
> > +                       return -EIO;
> > +               }
> > +               break;
> > +       case BMP580_EXT_MODE_CMD:
> > +               ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > BMP580_CMD_EXTMODE_SEQ_0);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to send ext_mode first
> > sequence\n");
> > +                       return ret;
> > +               }
> > +               ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > BMP580_CMD_EXTMODE_SEQ_1);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to send ext_mode second
> > sequence\n");
> > +                       return ret;
> > +               }
> > +               ret = regmap_write(data->regmap, BMP580_REG_CMD,
> > BMP580_CMD_EXTMODE_SEQ_2);
> > +               if (ret) {
> > +                       dev_err(data->dev, "failed to send ext_mode second
> > sequence\n");
> > +                       return ret;
> > +               }
> > +               break;
>
> Might as well return from each of the instead of break.  slightly reduces the
> code
> anyone interested in a particular flow needs to look at.
>
>
> > +       }
> > +
> > +       return 0;
> > +}
>
> ...
>
> > +
> > +static int bmp580_read_press(struct bmp280_data *data, int *val, int *val2)
> > +{
> > +       u32 raw_press;
> > +       int ret;
> > +
> > +       ret = regmap_bulk_read(data->regmap, BMP580_REG_PRESS_XLSB, data-
> > >buf,
> > +                              sizeof(data->buf));
> > +       if (ret) {
> > +               dev_err(data->dev, "failed to read pressure\n");
> > +               return ret;
> > +       }
> > +
> > +       raw_press = get_unaligned_le24(data->buf);
> > +       if (raw_press == BMP580_PRESS_SKIPPED) {
> > +               dev_err(data->dev, "reading pressure skipped\n");
> > +               return -EIO;
> > +       }
> > +       /*
> > +        * Pressure is returned in Pascals in fractional form down 2^16.
> > +        * We reescale /1000 to convert to kilopascal to respect IIO ABI.
> > +        */
> > +       *val = raw_press;
> > +       *val2 = 64000; // 2^6 * 1000
>
> /* */ syntax for comments in IIO.
>
> > +       return IIO_VAL_FRACTIONAL;
> > +}
>
> ...
>
> > +static int bmp580_preinit(struct bmp280_data *data)
> > +{
> > +       unsigned int reg;
> > +       int ret;
> > +
> > +       /* Issue soft-reset command */
> > +       ret = bmp580_cmd(data, BMP580_SOFT_RESET_CMD);
> > +       if (ret)
> > +               return ret;
> Blank line here and similar places.  Nice to keep unrelated blocks of
> code separate.
>
> > +       /* Post powerup sequence */
> > +       ret = regmap_read(data->regmap, BMP580_REG_CHIP_ID, &reg);
> > +       if (ret)
> > +               return ret;
> > +       if (reg != BMP580_CHIP_ID) {
> > +               dev_err(data->dev, "preinit: unexpected chip_id\n");
>
> I'd rather this was just an info print and carry on.  If Bosch brings
> another device out that is sufficiently compatible and we use a fallback
> dt compatible for it - so that it works with older kernels then we might
> want to indicate we aren't sure we can handle the chip correctly but then
> assume

This sounds a great idea, but this would need a few extra changes to work as
intended here:
* In the common probe, before calling to preinit, the device ID is read and
compared to expected ID stored in the chip_info. If the IDs are different, we
stop initialization with an error
* I should backport this, at least, to the BMP380 preinit call, as I have
pending testing if this works well with the BMP390, and sounds like a good
candidate. Which, after looking to datasheet, seems to have 0x60 as its chip ID!

> the DT description was correct and carry on anyway.
>
> > +               return -EINVAL;
> > +       }
> > +       ret = regmap_read(data->regmap, BMP580_REG_STATUS, &reg);
> > +       if (ret)
> > +               return ret;
> > +       /* Check nvm status */
> > +       if (!(reg & BMP580_STATUS_NVM_RDY_MASK) || (reg &
> > BMP580_STATUS_NVM_ERR_MASK)) {
> > +               dev_err(data->dev, "preinit: nvm error on powerup
> > sequence\n");
> > +               return -EIO;
> > +       }
> > +
> > +       return 0;
> > +}
> > +
>
> ...
>
> > +static const int bmp580_oversampling_avail[] = { 1, 2, 4, 8, 16, 32, 64,
> > 128 };
> > +static const int bmp580_iir_filter_coeffs_avail[] = { 1, 2, 4, 8, 16, 32,
> > 64, 128 };
>
> Up to you, but you could take advantage of the fact this array matches the
> bmp380 one.
> It is arguable that the code is clearer with it not being reused though so
> your choice.

Hum yes, I could reuse the array for the BMP380, maybe a should use a more
generic name for that array to avoid confusion? Something like
bmp280_iir_filter_coeffs_avail, refering to the driver name instead of the
concrete sensor?

>
> > +
> > +static const struct bmp280_chip_info bmp580_chip_info = {
> > +       .id_reg = BMP580_REG_CHIP_ID,
> > +       .chip_id = BMP580_CHIP_ID,
> > +       .start_up_time = 2000,
> > +       .channels = bmp380_channels,
> > +       .num_channels = 2,
> > +
> > +       .oversampling_temp_avail = bmp580_oversampling_avail,
> > +       .num_oversampling_temp_avail =
> > ARRAY_SIZE(bmp580_oversampling_avail),
> > +       .oversampling_temp_default = ilog2(1),
> > +
> > +       .oversampling_press_avail = bmp580_oversampling_avail,
> > +       .num_oversampling_press_avail =
> > ARRAY_SIZE(bmp580_oversampling_avail),
> > +       .oversampling_press_default = ilog2(4),
> > +
> > +       .sampling_freq_avail = bmp580_odr_table,
> > +       .num_sampling_freq_avail = ARRAY_SIZE(bmp580_odr_table) * 2,
> > +       .sampling_freq_default = BMP580_ODR_50HZ,
> > +
> > +       .iir_filter_coeffs_avail = bmp580_iir_filter_coeffs_avail,
> > +       .num_iir_filter_coeffs_avail =
> > ARRAY_SIZE(bmp580_iir_filter_coeffs_avail),
> > +       .iir_filter_coeff_default = 2,
> > +
> > +       .chip_config = bmp580_chip_config,
> > +       .read_temp = bmp580_read_temp,
> > +       .read_press = bmp580_read_press,
> > +       .read_calib = NULL,
>
> As in previous patch, I'd prefer not to see explicit NULL being set for
> optional elements. Their absence should be enough.
>
> > +       .preinit = bmp580_preinit,
> > +};
> > +
> >  static int bmp180_measure(struct bmp280_data *data, u8 ctrl_meas)
> >  {
> >         const int conversion_time_max[] = { 4500, 7500, 13500, 25500 };
> > @@ -1713,6 +2142,9 @@ int bmp280_common_probe(struct device *dev,
> >         case BMP380:
> >                 chip_info = &bmp380_chip_info;
> >                 break;
> > +       case BMP580:
> > +               chip_info = &bmp580_chip_info;
> > +               break;
> >         default:
> >                 return -EINVAL;
> >         }
> > @@ -1779,9 +2211,10 @@ int bmp280_common_probe(struct device *dev,
> >          */
> >         if (data->chip_info->preinit) {
> >                 ret = data->chip_info->preinit(data);
> > -               dev_err(dev, "error running preinit tasks");
> > -               if (ret < 0)
> > +               if (ret) {
> > +                       dev_err(dev, "error running preinit tasks\n");
> >                         return ret;
> Wrong patch.  + use dev_error_probe() like the case below already does.
>
> > +               }
> >         }
> >  
> >         ret = data->chip_info->chip_config(data);
> > @@ -1795,11 +2228,12 @@ int bmp280_common_probe(struct device *dev,
> >          * non-volatile memory during production". Let's read them out at
> > probe
> >          * time once. They will not change.
> >          */
> > -
> > -       ret = data->chip_info->read_calib(data);
> > -       if (ret < 0)
> > -               return dev_err_probe(data->dev, ret,
> > -                                    "failed to read calibration
> > coefficients\n");
> > +       if (data->chip_info->read_calib) {
>
> Small thing, but ideally this no op refactoring should have been a precursor
> patch
> rather than buried in here. It's small enough though that I don't care enough
> to insist
> you break it out.

Sure

>
> > +               ret = data->chip_info->read_calib(data);
> > +               if (ret < 0)
> > +                       return dev_err_probe(data->dev, ret,
> > +                                            "failed to read calibration
> > coefficients\n");
> > +       }
> >  
> >         /*
> >          * Attempt to grab an optional EOC IRQ - only the BMP085 has this
> > diff --git a/drivers/iio/pressure/bmp280-i2c.c
> > b/drivers/iio/pressure/bmp280-i2c.c
> > index 59921e8cd592..c52d2b477bb7 100644
> > --- a/drivers/iio/pressure/bmp280-i2c.c
> > +++ b/drivers/iio/pressure/bmp280-i2c.c
> > @@ -22,6 +22,9 @@ static int bmp280_i2c_probe(struct i2c_client *client)
> >         case BMP380:
> >                 regmap_config = &bmp380_regmap_config;
> >                 break;
> > +       case BMP580:
> > +               regmap_config = &bmp580_regmap_config;
>
> whilst here, similar to below, it would be nice to switch to generic
> firmware methods as default way to get the match data then fallback to the
> i2c_device_id table if that fails (as we did old school i2c probing from
> a board file or from userspace).
>
> > +               break;
> >         default:
> >                 return -EINVAL;
> >         }
> > @@ -45,6 +48,7 @@ static const struct of_device_id bmp280_of_i2c_match[] = {
> >         { .compatible = "bosch,bmp280", .data = (void *)BMP280 },
> >         { .compatible = "bosch,bme280", .data = (void *)BME280 },
> >         { .compatible = "bosch,bmp380", .data = (void *)BMP380 },
> > +       { .compatible = "bosch,bmp580", .data = (void *)BMP580 },
> >         { },
> >  };
> >  MODULE_DEVICE_TABLE(of, bmp280_of_i2c_match);
> > @@ -55,6 +59,7 @@ static const struct i2c_device_id bmp280_i2c_id[] = {
> >         {"bmp280", BMP280 },
> >         {"bme280", BME280 },
> >         {"bmp380", BMP380 },
> > +       {"bmp580", BMP580 },
> >         { },
> >  };
> >  MODULE_DEVICE_TABLE(i2c, bmp280_i2c_id);
>
> > diff --git a/drivers/iio/pressure/bmp280-spi.c
> > b/drivers/iio/pressure/bmp280-spi.c
> > index 4a2df5b5d838..5653c3c33081 100644
> > --- a/drivers/iio/pressure/bmp280-spi.c
> > +++ b/drivers/iio/pressure/bmp280-spi.c
> > @@ -69,6 +69,9 @@ static int bmp280_spi_probe(struct spi_device *spi)
> >         case BMP380:
> >                 regmap_config = &bmp380_regmap_config;
> >                 break;
> > +       case BMP580:
> > +               regmap_config = &bmp580_regmap_config;
>
> If using a pointer as suggested you'd get this for free :)
>
> I'd also like to see this driver move to using the of_device_id table
> via device_get_match_data() first then fallback to the spi_device_id table
> only if that fails.   Otherwise for dt supprot, we are relying on a match
> based on stripping the bosch prefix of the of compatible and that is nasty.

Ok! I'll have a look at it and add this as a precursor patch.

>
> > +               break;
> >         default:
> >                 return -EINVAL;
> >         }
> > @@ -96,6 +99,7 @@ static const struct of_device_id bmp280_of_spi_match[] = {
> >         { .compatible = "bosch,bmp280", },
> >         { .compatible = "bosch,bme280", },
> >         { .compatible = "bosch,bmp380", },
> > +       { .compatible = "bosch,bmp580", },
> >         { },
> >  };
> >  MODULE_DEVICE_TABLE(of, bmp280_of_spi_match);
> > @@ -106,6 +110,7 @@ static const struct spi_device_id bmp280_spi_id[] = {
> >         { "bmp280", BMP280 },
> >         { "bme280", BME280 },
> >         { "bmp380", BMP380 },
> > +       { "bmp580", BMP580 },
> >         { }
> >  };
> >  MODULE_DEVICE_TABLE(spi, bmp280_spi_id);
> > diff --git a/drivers/iio/pressure/bmp280.h b/drivers/iio/pressure/bmp280.h
> > index efc31bc84708..27d2abc17d01 100644
> > --- a/drivers/iio/pressure/bmp280.h
> > +++ b/drivers/iio/pressure/bmp280.h
> > @@ -3,6 +3,107 @@
>
> ...
>
> > +#define BMP580_CHIP_ID                 0x50
> >  #define BMP380_CHIP_ID                 0x50
>
> Well given we can't order them by ID value, probably best
> to have the set with a given ID in alphabetical order.  So
> swap the two above.

Got it

>
> >  #define BMP180_CHIP_ID                 0x55
> >  #define BMP280_CHIP_ID                 0x58
> > @@ -197,12 +299,14 @@ enum bmp280_variant {
> >         BMP280,
> >         BME280,
> >         BMP380,
> > +       BMP580,
> >  };
> >  
> >  /* Regmap configurations */
> >  extern const struct regmap_config bmp180_regmap_config;
> >  extern const struct regmap_config bmp280_regmap_config;
> >  extern const struct regmap_config bmp380_regmap_config;
> > +extern const struct regmap_config bmp580_regmap_config;
> >  
> >  /* Probe called from different transports */
> >  int bmp280_common_probe(struct device *dev,
>
Thanks for your time!
Angel

\
 
 \ /
  Last update: 2023-03-26 23:24    [W:0.132 / U:0.272 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site