From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
To: Crt Mori <cmo@melexis.com>
Cc: Jonathan Cameron <jic23@kernel.org>,
Linux Iio <linux-iio@vger.kernel.org>
Subject: Re: [PATCH 1/2] iio: temperature: Adding support for MLX90632
Date: Tue, 5 Dec 2017 11:23:37 +0000 [thread overview]
Message-ID: <20171205112337.00005dca@huawei.com> (raw)
In-Reply-To: <CAKv63usL=m73Ar5yduxHpWAMY9ztwrhYJXFdDo08mRNeuo+XyA@mail.gmail.com>
On Sat, 2 Dec 2017 21:30:54 +0100
Crt Mori <cmo@melexis.com> wrote:
> Hi Jonathan,
> Definitely did not think there will be so many comments inside, but I
> expected few of them indeed.
>
> See my replies inside.
>
> On 2 December 2017 at 15:33, Jonathan Cameron <jic23@kernel.org> wrote:
> > On Wed, 29 Nov 2017 23:07:49 +0100
> > Crt Mori <cmo@melexis.com> wrote:
> >
> >> Melexis has just released Infra Red temperature sensor MLX90632 used
> >> for contact-less temperature measurement. Driver provides basic
> >> functionality for reporting object (and ambient) temperature with
> >> support for object emissivity.
> >>
> >> Signed-off-by: Crt Mori <cmo@melexis.com>
> >
> > Various comments inline.
> >
> >> ---
> >> MAINTAINERS | 7 +
> >> drivers/iio/temperature/Kconfig | 12 +
> >> drivers/iio/temperature/Makefile | 1 +
> >> drivers/iio/temperature/mlx90632.c | 802 +++++++++++++++++++++++++++++++++++++
> >> 4 files changed, 822 insertions(+)
> >> create mode 100644 drivers/iio/temperature/mlx90632.c
> >>
> >> diff --git a/MAINTAINERS b/MAINTAINERS
> >> index 2d3d750b19c0..81aec02b08b8 100644
> >> --- a/MAINTAINERS
> >> +++ b/MAINTAINERS
> >> @@ -8690,6 +8690,13 @@ W: http://www.melexis.com
> >> S: Supported
> >> F: drivers/iio/temperature/mlx90614.c
> >>
> >> +MELEXIS MLX90632 DRIVER
> >> +M: Crt Mori <cmo@melexis.com>
> >> +L: linux-iio@vger.kernel.org
> >> +W: http://www.melexis.com
> >> +S: Supported
> >> +F: drivers/iio/temperature/mlx90632.c
> >> +
> >> MELFAS MIP4 TOUCHSCREEN DRIVER
> >> M: Sangwon Jee <jeesw@melfas.com>
> >> W: http://www.melfas.com
> >> diff --git a/drivers/iio/temperature/Kconfig b/drivers/iio/temperature/Kconfig
> >> index 5378976d6d27..82e4a62745e2 100644
> >> --- a/drivers/iio/temperature/Kconfig
> >> +++ b/drivers/iio/temperature/Kconfig
> >> @@ -43,6 +43,18 @@ config MLX90614
> >> This driver can also be built as a module. If so, the module will
> >> be called mlx90614.
> >>
> >> +config MLX90632
> >> + tristate "MLX90632 contact-less infrared sensor with medical accuracy"
> >> + depends on I2C
> >> + select REGMAP_I2C
> >> + help
> >> + If you say yes here you get support for the Melexis
> >> + MLX90632 contact-less infrared sensor with medical accuracy
> >> + connected with I2C.
> >> +
> >> + This driver can also be built as a module. If so, the module will
> >> + be called mlx90632.
> >> +
> >> config TMP006
> >> tristate "TMP006 infrared thermopile sensor"
> >> depends on I2C
> >> diff --git a/drivers/iio/temperature/Makefile b/drivers/iio/temperature/Makefile
> >> index ad1d668de546..44644fe01bc9 100644
> >> --- a/drivers/iio/temperature/Makefile
> >> +++ b/drivers/iio/temperature/Makefile
> >> @@ -5,6 +5,7 @@
> >> obj-$(CONFIG_HID_SENSOR_TEMP) += hid-sensor-temperature.o
> >> obj-$(CONFIG_MAXIM_THERMOCOUPLE) += maxim_thermocouple.o
> >> obj-$(CONFIG_MLX90614) += mlx90614.o
> >> +obj-$(CONFIG_MLX90632) += mlx90632.o
> >> obj-$(CONFIG_TMP006) += tmp006.o
> >> obj-$(CONFIG_TMP007) += tmp007.o
> >> obj-$(CONFIG_TSYS01) += tsys01.o
> >> diff --git a/drivers/iio/temperature/mlx90632.c b/drivers/iio/temperature/mlx90632.c
> >> new file mode 100644
> >> index 000000000000..05c7d943e504
> >> --- /dev/null
> >> +++ b/drivers/iio/temperature/mlx90632.c
> >> @@ -0,0 +1,802 @@
> >> +/*
> >> + * mlx90632.c - Melexis MLX90632 contactless IR temperature sensor
> >> + *
> >> + * Copyright (c) 2017 Melexis <cmo@melexis.com>
> >> + *
> >> + * This file is subject to the terms and conditions of version 2 of
> >> + * the GNU General Public License. See the file COPYING in the main
> >> + * directory of this archive for more details.
> >> + *
> >> + * Driver for the Melexis MLX90632 I2C 16-bit IR thermopile sensor
> >> + */
> >> +#include <asm/byteorder.h>
> > Don't think this should ever be included in a driver.
> > What do you need it for?
> >
> >> +#include <linux/delay.h>
> >> +#include <linux/err.h>
> >> +#include <linux/gpio/consumer.h>
> >> +#include <linux/i2c.h>
> >> +#include <linux/kernel.h>
> >> +#include <linux/module.h>
> >> +#include <linux/math64.h>
> >> +#include <linux/of.h>
> >> +#include <linux/pm_runtime.h>
> >> +#include <linux/regmap.h>
> >> +#include <linux/unaligned/be_byteshift.h>
> > why?
> >
> >> +
> >> +#include <linux/iio/iio.h>
> >> +#include <linux/iio/sysfs.h>
> >> +
> >> +/* Memory sections addresses */
> >> +#define MLX90632_ADDR_RAM 0x4000 /* Start address of ram */
> >> +#define MLX90632_ADDR_EEPROM 0x2480 /* Start address of user eeprom */
> >> +
> >> +/* EEPROM addresses - used at startup */
> >> +#define MLX90632_EE_CTRL 0x24d4 /* Control register initial value */
> >> +#define MLX90632_EE_I2C_ADDR 0x24d5 /* I2C address register initial value */
> >> +#define MLX90632_EE_VERSION 0x240b /* EEPROM version reg address */
> >> +#define MLX90632_EE_P_R 0x240c /* P_R calibration register 32bit */
> >> +#define MLX90632_EE_P_G 0x240e /* P_G calibration register 32bit */
> >> +#define MLX90632_EE_P_T 0x2410 /* P_T calibration register 32bit */
> >> +#define MLX90632_EE_P_O 0x2412 /* P_O calibration register 32bit */
> >> +#define MLX90632_EE_Aa 0x2414 /* Aa calibration register 32bit */
> >> +#define MLX90632_EE_Ab 0x2416 /* Ab calibration register 32bit */
> >> +#define MLX90632_EE_Ba 0x2418 /* Ba calibration register 32bit */
> >> +#define MLX90632_EE_Bb 0x241a /* Bb calibration register 32bit */
> >> +#define MLX90632_EE_Ca 0x241c /* Ca calibration register 32bit */
> >> +#define MLX90632_EE_Cb 0x241e /* Cb calibration register 32bit */
> >> +#define MLX90632_EE_Da 0x2420 /* Da calibration register 32bit */
> >> +#define MLX90632_EE_Db 0x2422 /* Db calibration register 32bit */
> >> +#define MLX90632_EE_Ea 0x2424 /* Ea calibration register 32bit */
> >> +#define MLX90632_EE_Eb 0x2426 /* Eb calibration register 32bit */
> >> +#define MLX90632_EE_Fa 0x2428 /* Fa calibration register 32bit */
> >> +#define MLX90632_EE_Fb 0x242a /* Fb calibration register 32bit */
> >> +#define MLX90632_EE_Ga 0x242c /* Ga calibration register 32bit */
> >> +
> >> +#define MLX90632_EE_Gb 0x242e /* Gb calibration register 16bit */
> >> +#define MLX90632_EE_Ka 0x242f /* Ka calibration register 16bit */
> >> +
> >> +#define MLX90632_EE_Ha 0x2481 /* Ha customer calib value reg 16bit */
> >> +#define MLX90632_EE_Hb 0x2482 /* Hb customer calib value reg 16bit */
> >> +
> >> +/* Register addresses - volatile */
> >> +#define MLX90632_REG_I2C_ADDR 0x3000 /* Chip I2C address register */
> >> +
> >> +/* Control register address - volatile */
> >> +#define MLX90632_REG_CONTROL 0x3001 /* Control Register address */
> >> +#define MLX90632_CFG_PWR_MASK GENMASK(2, 1) /* PowerMode Mask */
> >> +/* PowerModes statuses */
> >> +#define MLX90632_PWR_STATUS(ctrl_val) (ctrl_val << 1)
> >> +#define MLX90632_PWR_STATUS_HALT MLX90632_PWR_STATUS(0) /* hold */
> >> +#define MLX90632_PWR_STATUS_SLEEP_STEP MLX90632_PWR_STATUS(1) /* sleep step*/
> >> +#define MLX90632_PWR_STATUS_STEP MLX90632_PWR_STATUS(2) /* step */
> >> +#define MLX90632_PWR_STATUS_CONTINUOUS MLX90632_PWR_STATUS(3) /* continuous*/
> >> +
> >> +/* Device status register - volatile */
> >> +#define MLX90632_REG_STATUS 0x3fff /* Device status register */
> >> +#define MLX90632_STAT_BUSY BIT(10) /* Device busy indicator */
> >> +#define MLX90632_STAT_EE_BUSY BIT(9) /* EEPROM busy indicator */
> >> +#define MLX90632_STAT_BRST BIT(8) /* Brown out reset indicator */
> >> +#define MLX90632_STAT_CYCLE_POS GENMASK(6, 2) /* Data position */
> >> +#define MLX90632_STAT_DATA_RDY BIT(0) /* Data ready indicator */
> >> +
> >> +/* RAM_MEAS address-es for each channel */
> >> +#define MLX90632_RAM_1(meas_num) (MLX90632_ADDR_RAM + 3 * meas_num)
> >> +#define MLX90632_RAM_2(meas_num) (MLX90632_ADDR_RAM + 3 * meas_num + 1)
> >> +#define MLX90632_RAM_3(meas_num) (MLX90632_ADDR_RAM + 3 * meas_num + 2)
> >> +
> >> +/* Magic constants */
> >> +#define MLX90632_EEPROM_VERSION 0xff05 /* EEPROM DSP version for constants */
> >> +#define MLX90632_ID_MEDICAL 0x01ff /* EEPROM Medical device id */
> >> +#define MLX90632_ID_CONSUMER 0x02ff /* EEPROM Consumer device id */
> >> +#define MLX90632_EEPROM_WRITE_KEY 0x554C /* EEPROM write key 0x55 and 0x4c */
> >> +#define MLX90632_RESET_CMD 0x0006 /* Reset sensor (address or global) */
> >> +#define MLX90632_REF_12 12LL /**< ResCtrlRef value of Ch 1 or Ch 2 */
> >> +#define MLX90632_REF_3 12LL /**< ResCtrlRef value of Channel 3 */
> >> +
> >> +#define TENTO3 1000LL
> >> +#define TENTO4 10000LL
> >> +#define TENTO5 100000LL
> >> +#define TENTO6 1000000LL
> >> +#define TENTO7 10000000LL
> >> +#define TENTO10 10000000000LL
> >> +#define TENTO12 1000000000000LL
> > Umm. The numbers describe the constants rather better than the defines ;)
> > Just use the numbers if you need them!
> >
>
> This is way shorter (line length) and more readable in my opinion.
> Also writing 12 times 0 is error prone :D
Needs care but I really don't like these constants.
>
> >> +
> >> +struct mlx90632_data {
> >> + struct i2c_client *client;
> >> + struct mutex lock; /* Multiple reads for single measurement */
> >> + struct regmap *regmap;
> >> + u16 emissivity;
> >> +};
> >> +
> >> +static const struct regmap_range mlx90632_volatile_reg_range[] = {
> >> + regmap_reg_range(MLX90632_REG_CONTROL, MLX90632_REG_I2C_ADDR),
> >> + regmap_reg_range(MLX90632_REG_STATUS, MLX90632_REG_STATUS),
> >> + regmap_reg_range(MLX90632_RAM_1(0),
> >> + MLX90632_RAM_3(MLX90632_MAX_MEAS_NUM)),
> >> +};
> >> +
> >> +static const struct regmap_access_table mlx90632_volatile_regs_tbl = {
> >> + .yes_ranges = mlx90632_volatile_reg_range,
> >> + .n_yes_ranges = ARRAY_SIZE(mlx90632_volatile_reg_range),
> >> +};
> >> +
> >> +static const struct regmap_range mlx90632_read_reg_range[] = {
> >> + regmap_reg_range(MLX90632_EE_VERSION, MLX90632_EE_Ka),
> >> + regmap_reg_range(MLX90632_EE_CTRL, MLX90632_EE_I2C_ADDR),
> >> + regmap_reg_range(MLX90632_EE_Ha, MLX90632_EE_Hb),
> >> + regmap_reg_range(MLX90632_REG_CONTROL, MLX90632_REG_I2C_ADDR),
> >> + regmap_reg_range(MLX90632_REG_STATUS, MLX90632_REG_STATUS),
> >> + regmap_reg_range(MLX90632_RAM_1(0),
> >> + MLX90632_RAM_3(MLX90632_MAX_MEAS_NUM)),
> >> +};
> >> +
> >> +static const struct regmap_access_table mlx90632_readable_regs_tbl = {
> >> + .yes_ranges = mlx90632_read_reg_range,
> >> + .n_yes_ranges = ARRAY_SIZE(mlx90632_read_reg_range),
> >> +};
> >> +
> >> +static const struct regmap_range mlx90632_no_write_reg_range[] = {
> >> + regmap_reg_range(MLX90632_EE_VERSION, MLX90632_EE_Ka),
> >> + regmap_reg_range(MLX90632_RAM_1(0),
> >> + MLX90632_RAM_3(MLX90632_MAX_MEAS_NUM)),
> >> +};
> >> +
> >> +static const struct regmap_access_table mlx90632_writeable_regs_tbl = {
> >> + .no_ranges = mlx90632_no_write_reg_range,
> >> + .n_no_ranges = ARRAY_SIZE(mlx90632_no_write_reg_range),
> >> +};
> >> +
> >> +static const struct regmap_config mlx90632_regmap = {
> >> + .reg_bits = 16,
> >> + .val_bits = 16,
> >> +
> >> + .volatile_table = &mlx90632_volatile_regs_tbl,
> >> + .rd_table = &mlx90632_readable_regs_tbl,
> >> + .wr_table = &mlx90632_writeable_regs_tbl,
> >> +
> >> + .use_single_rw = true,
> >> + .reg_format_endian = REGMAP_ENDIAN_BIG,
> >> + .val_format_endian = REGMAP_ENDIAN_BIG,
> >> + .cache_type = REGCACHE_RBTREE,
> >> +};
> >> +
> >> +static u64 mlx90632_int_sqrt(u64 x)
> >> +{
> >> + u64 b, m, y = 0;
> >> +
> >> + if (BITS_PER_LONG != 32)
> >> + return int_sqrt(x);
> >
> > needs a comment on why...
> > Ideally propose a standard 64 bit routine if one is needed for
> > 32 bit machines.
> >
>
> Problem is that int_sqrt is not strongly typed. That means 64 bit
> number gets truncated to 32bit on 32bit machine as you mentioned
> and that caused quite many headaches in my testing (worked on
> mobile phone, but not on beaglebone black). This solution is just
> stronger typing of int_sqrt function.
>
> If you think I should put it onto math.h as a patch I can try it.
Sure, let's see if that goes anywhere. Seems the right approach to me.
>
> >> +
> >> + if (x <= 1)
> >> + return x;
> >> +
> >> + m = 1ULL << (64 - 2);
> >> + while (m != 0) {
> >> + b = y + m;
> >> + y >>= 1;
> >> +
> >> + if (x >= b) {
> >> + x -= b;
> >> + y += m;
> >> + }
> >> + m >>= 2;
> >> + }
> >> + return y;
> >> +}
> >> +
> >> +static s32 mlx90632_pwr_set_sleep_step(struct regmap *regmap)
> >> +{
> >> + return regmap_update_bits(regmap, MLX90632_REG_CONTROL,
> >> + MLX90632_CFG_PWR_MASK,
> >> + MLX90632_PWR_STATUS_SLEEP_STEP);
> >> +}
> >> +
> >> +static s32 mlx90632_pwr_set_step(struct regmap *regmap)
> >> +{
> >> + return regmap_update_bits(regmap, MLX90632_REG_CONTROL,
> >> + MLX90632_CFG_PWR_MASK,
> >> + MLX90632_PWR_STATUS_STEP);
> >> +}
> >> +
> >> +static s32 mlx90632_pwr_continuous(struct regmap *regmap)
> >> +{
> >> + return regmap_update_bits(regmap, MLX90632_REG_CONTROL,
> >> + MLX90632_CFG_PWR_MASK,
> >> + MLX90632_PWR_STATUS_CONTINUOUS);
> >> +}
> >> +
> >> +static int mlx90632_start_measurement(struct mlx90632_data *data)
> >
> > Looks superficially to me like this is actually waiting for a reading
> > to complete rather than just starting it? If so change the name
> > to reflect that.
> >
>
> ok
>
> >> +{
> >> + int ret, tries = 100;
> >> + unsigned int reg_status;
> >> +
> >> + ret = regmap_update_bits(data->regmap, MLX90632_REG_STATUS,
> >> + MLX90632_STAT_DATA_RDY, 0);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + while (tries-- > 0) {
> >> + ret = regmap_read(data->regmap, MLX90632_REG_STATUS,
> >> + ®_status);
> >> + if (ret < 0)
> >> + return ret;
> >> + if (reg_status & MLX90632_STAT_DATA_RDY)
> >> + break;
> >> + usleep_range(10000, 11000);
> >> + }
> >> +
> >> + if (tries < 0) {
> >> + dev_err(&data->client->dev, "data not ready");
> >> + return -ETIMEDOUT;
> >> + }
> >> +
> >> + return (reg_status & MLX90632_STAT_CYCLE_POS) >> 2;
> >
> > This is a non obvious return value so I would suggest adding
> > some documentation to say what is going on here..
> >
>
> Yes, it is indeed.
>
> >> +}
> >> +
> >> +static int mlx9032_channel_new_select(int ret, uint8_t *channel_new,
> >> + uint8_t *channel_old)
> >
> > don't use ret as a name for a variable being passed in. Rather
> > confusing!
> >
>
> OK. I think I should just copy setting of channel_old and new in the
> only spot it is used.
>
> >> +{
> >> + if (ret == 1) {
> >> + *channel_new = 1;
> >> + *channel_old = 2;
> >> + } else if (ret == 2) {
> >> + *channel_new = 2;
> >> + *channel_old = 1;
> >> + } else {
> >> + return ret;
> >> + }
> >> + return 0;
> > Can't get here...
> >
>
> Not true. You get here in every example other than else where channel_new
> and channel_old are successfully set. but I agree there is a bug in case ret=0
> when it will return 0, but it will not set channels.
Ah good point, I missed that. I would do it as a switch with direct returns.
>
> >> +}
> >> +
> >> +static int mlx90632_read_ambient_raw(struct regmap *regmap,
> >> + s16 *ambient_new_raw, s16 *ambient_old_raw)
> >> +{
> >> + int ret;
> >> + unsigned int read_tmp;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_3(1), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + *ambient_new_raw = (s16)read_tmp;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_3(2), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + *ambient_old_raw = (s16)read_tmp;
> >> +
> >> + return ret;
> >> +}
> >> +
> >> +static int mlx90632_read_object_raw(struct regmap *regmap,
> >> + int start_measurement_ret,
> >> + s16 *object_new_raw, s16 *object_old_raw)
> >> +{
> >> + int ret;
> >> + unsigned int read_tmp;
> >> + s16 read;
> >> + u8 channel = 0;
> >> + u8 channel_old = 0;
> >> +
> >> + ret = mlx90632_channel_new_select(start_measurement_ret, &channel,
> >> + &channel_old);
> >> + if (ret != 0)
> >> + return ret;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_2(channel), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + read = (s16)read_tmp;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_1(channel), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + *object_new_raw = (read + (s16)read_tmp) / 2;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_2(channel_old), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + read = (s16)read_tmp;
> >> +
> >> + ret = regmap_read(regmap, MLX90632_RAM_1(channel_old), &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + *object_old_raw = (read + (s16)read_tmp) / 2;
> >> +
> >> + return ret;
> >> +}
> >> +
> >> +static int mlx90632_read_all_channel(struct mlx90632_data *data,
> >> + s16 *ambient_new_raw, s16 *ambient_old_raw,
> >> + s16 *object_new_raw, s16 *object_old_raw)
> >> +{
> >> + s32 ret, tm_ret;
> >> +
> >> + mutex_lock(&data->lock);
> >> + tm_ret = mlx90632_start_measurement(data);
> >> + if (tm_ret >= 0) {
> >> + ret = mlx90632_read_ambient_raw(data->regmap, ambient_new_raw,
> >> + ambient_old_raw);
> >> + if (ret >= 0) {
> >> + ret = mlx90632_read_object_raw(data->regmap, tm_ret,
> >> + object_new_raw,
> >> + object_old_raw);
> >> + }
> >> + } else {
> >> + ret = tm_ret;
> >> + }
> > Use something cleaner like
> > mutex_lock(&data->lock);
> > ret = mlx90632_start_measurement(data);
> > if (ret < 0)
> > goto unlock;
> > ret = mlx90632_read_ambient_raw(data->regmap, ambient_new_raw,
> > ambient_old_raw);
> > if (ret < 0)
> > goto unlock;
> >
> > ret = mlx90632_read_object_raw(data->regmap, tm_ret,
> > object_new_raw,
> > object_old_raw);
> > unlock:
>
> Ok sorry, professional deformation (goto).
>
> >> + mutex_unlock(&data->lock);
> >> + return ret;
> >> +}
> >> +
> >> +static int mlx90632_read_ee_register(struct regmap *regmap, u16 reg_lsb,
> >> + s32 *reg_value)
> >> +{
> >> + s32 ret;
> >> + unsigned int read;
> >> + __le32 value;
> >> +
> >> + ret = regmap_read(regmap, reg_lsb, &read);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + value = cpu_to_le32(read);
> >
> > Why would you be converting to le32 to then do
> > calculations in here using it at cpu endianness?
> > (guessing you want le32_to_cpu..)
> >
> > hmm isn't any relevant endian conversion wrapped up in regmap
> > anyway?
> >
>
> Regmap reads 16 bit values, but some values are 32 bit. To avoid CPU
> endianess affecting this, I am putting it to known time and shift values.
> This way I ensured it is working on any CPU. le32_to_cpu is in return value.
>
I'm far from convinced this is a good idea. I would ensure all values
are in cpu endianness before combining them based on the read order
(which is endian independant)
> >> +
> >> + ret = regmap_read(regmap, reg_lsb + 1, &read);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + value = (cpu_to_le32(read) << 16) | (value & 0xffff);
> >> +
> >> + *reg_value = le32_to_cpu(value);
> >> + return 0;
> >> +}
> >> +
> >> +static s64 mlx90632_preprocess_temp_amb(s16 ambient_new_raw,
> >> + s16 ambient_old_raw, s16 Gb)
> >> +{
> >> + s64 VR_Ta, kGb, tmp;
> >> +
> >> + kGb = ((s64)Gb * TENTO3) >> 10ULL;
> >> + VR_Ta = (s64)ambient_old_raw * TENTO6 +
> >> + kGb * div64_s64(((s64)ambient_new_raw * TENTO3),
> >> + (MLX90632_REF_3));
> >> + tmp = div64_s64(
> >> + div64_s64(((s64)ambient_new_raw * TENTO12),
> >> + (MLX90632_REF_3)), VR_Ta);
> >> + return div64_s64(tmp << 19ULL, TENTO3);
> >> +}
> >> +
> >> +static s64 mlx90632_preprocess_temp_obj(s16 object_new_raw, s16 object_old_raw,
> >> + s16 ambient_new_raw,
> >> + s16 ambient_old_raw, s16 Ka)
> >> +{
> >> + s64 VR_IR, kKa, tmp;
> >> +
> >> + kKa = ((s64)Ka * TENTO3) >> 10ULL;
> >> + VR_IR = (s64)ambient_old_raw * TENTO6 +
> >> + kKa * div64_s64(((s64)ambient_new_raw * TENTO3),
> >> + (MLX90632_REF_3));
> >> + tmp = div64_s64(
> >> + div64_s64(((s64)((object_new_raw + object_old_raw) / 2)
> >> + * TENTO12), (MLX90632_REF_12)), VR_IR);
> >> + return div64_s64(tmp << 19ULL), TENTO3);
> >> +}
> >> +
> >> +static s32 mlx90632_calc_temp_ambient(s16 ambient_new_raw, s16 ambient_old_raw,
> >> + s32 P_T, s32 P_R, s32 P_G, s32 P_O,
> >> + s16 Gb)
> >> +{
> >> + s64 Asub, Bsub, Ablock, Bblock, Cblock, AMB, sum;
> >> +
> >> + AMB = mlx90632_preprocess_temp_amb(ambient_new_raw, ambient_old_raw,
> >> + Gb);
> >> + Asub = ((s64)P_T * TENTO10) >> 44ULL;
> >> + Bsub = AMB - (((s64)P_R * TENTO3) >> 8ULL);
> >> + Ablock = Asub * (Bsub * Bsub);
> >> + Bblock = (div64_s64(Bsub * TENTO7, P_G)) << 20ULL;
> >> + Cblock = ((s64)P_O * TENTO10) >> 8ULL;
> >> +
> >> + sum = div64_s64(Ablock, TENTO6) + Bblock + Cblock;
> >> +
> >> + return div64_s64(sum, TENTO7);
> >> +}
> >> +
> >> +static s32 mlx90632_calc_temp_object_iteration(s32 prev_object_temp, s64 object,
> >> + s64 TAdut, s32 Fa, s32 Fb,
> >> + s32 Ga, s16 Ha, s16 Hb,
> >> + u16 emissivity)
> >> +{
> >> + s64 calcedKsTO, calcedKsTA, ir_Alpha, TAdut4, Alpha_corr;
> >> + s64 Ha_customer, Hb_customer;
> >> +
> >> + Ha_customer = ((s64)Ha * TENTO6) >> 14ULL;
> >> + Hb_customer = ((s64)Hb * 100) >> 10ULL;
> >> +
> >> + calcedKsTO = ((s64)((s64)Ga * (prev_object_temp - 25 * TENTO3)
> >> + * TENTO3)) >> 36LL;
> >> + calcedKsTA = ((s64)(Fb * (TAdut - 25 * TENTO6))) >> 36LL;
> >> + Alpha_corr = div64_s64((((s64)(Fa * TENTO10) >> 46LL) * Ha_customer),
> >> + TENTO3);
> >> + Alpha_corr *= ((s64)(1 * TENTO6 + calcedKsTO + calcedKsTA));
> >> + Alpha_corr = emissivity * div64_s64(Alpha_corr, TENTO5);
> >> + Alpha_corr = div64_s64(Alpha_corr, TENTO3);
> >> + ir_Alpha = div64_s64((s64)object * TENTO7, Alpha_corr);
> >> + TAdut4 = (div64_s64(TAdut, TENTO4) + 27315) *
> >> + (div64_s64(TAdut, TENTO4) + 27315) *
> >> + (div64_s64(TAdut, TENTO4) + 27315) *
> >> + (div64_s64(TAdut, TENTO4) + 27315);
> >> +
> >> + return (mlx90632_int_sqrt(
> >> + mlx90632_int_sqrt(ir_Alpha * TENTO12 + TAdut4)
> >> + ) - 27315 - Hb_customer) * 10;
> >> +}
> >> +
> >> +static s32 mlx90632_calc_temp_object(s64 object, s64 ambient, s32 Ea, s32 Eb,
> >> + s32 Fa, s32 Fb, s32 Ga, s16 Ha, s16 Hb,
> >> + u16 tmp_emi)
> >> +{
> >> + s64 kTA, kTA0, TAdut;
> >> + s64 temp = 25000;
> >> + s8 i;
> >> +
> >> + kTA = (Ea * TENTO3) >> 16LL;
> >> + kTA0 = (Eb * TENTO3) >> 8LL;
> >> + TAdut = div64_s64(((ambient - kTA0) * TENTO6), kTA) + 25 * TENTO6;
> >> +
> >> + for (i = 0; i < 5; ++i) {
> > comment on why iterations are needed would be good here.
>
> This is described in datasheet (the whole calculation). Will add
> explanatory comment.
>
> >> + temp = mlx90632_calc_temp_object_iteration(temp, object, TAdut,
> >> + Fa, Fb, Ga, Ha, Hb,
> >> + tmp_emi);
> >> + }
> >> + return temp;
> >> +}
> >> +
> >> +static int mlx90632_calc_object_dsp105(struct mlx90632_data *data, int *val)
> >> +{
> >> + s32 ret;
> >> + s32 Ea, Eb, Fa, Fb, Ga;
> >> + unsigned int read_tmp;
> >> + s16 Ha, Hb, Gb, Ka;
> >> + s16 ambient_new_raw, ambient_old_raw, object_new_raw, object_old_raw;
> >> + s64 object, ambient;
> >> +
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_Ea, &Ea);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_Eb, &Eb);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_Fa, &Fa);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_Fb, &Fb);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_Ga, &Ga);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = regmap_read(data->regmap, MLX90632_EE_Ha, &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + Ha = (s16)read_tmp;
> >> + ret = regmap_read(data->regmap, MLX90632_EE_Hb, &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + Hb = (s16)read_tmp;
> >> + ret = regmap_read(data->regmap, MLX90632_EE_Gb, &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + Gb = (s16)read_tmp;
> >> + ret = regmap_read(data->regmap, MLX90632_EE_Ka, &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + Ka = (s16)read_tmp;
> >> +
> >> + ret = mlx90632_read_all_channel(data,
> >> + &ambient_new_raw, &ambient_old_raw,
> >> + &object_new_raw, &object_old_raw);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + ambient = mlx90632_preprocess_temp_amb(ambient_new_raw,
> >> + ambient_old_raw, Gb);
> >> + object = mlx90632_preprocess_temp_obj(object_new_raw,
> >> + object_old_raw,
> >> + ambient_new_raw,
> >> + ambient_old_raw, Ka);
> >> +
> >> + *val = mlx90632_calc_temp_object(object, ambient, Ea, Eb, Fa, Fb, Ga,
> >> + Ha, Hb, data->emissivity);
> >> + return 0;
> >> +}
> >> +
> >> +static int mlx90632_calc_ambient_dsp105(struct mlx90632_data *data, int *val)
> >> +{
> >> + s32 ret;
> >> + unsigned int read_tmp;
> >> + s32 PT, PR, PG, PO;
> >> + s16 Gb;
> >> + s16 ambient_new_raw, ambient_old_raw;
> >> +
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_P_R, &PR);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_P_G, &PG);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_P_T, &PT);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = mlx90632_read_ee_register(data->regmap, MLX90632_EE_P_O, &PO);
> >> + if (ret < 0)
> >> + return ret;
> >> + ret = regmap_read(data->regmap, MLX90632_EE_Gb, &read_tmp);
> >> + if (ret < 0)
> >> + return ret;
> >> + Gb = (s16)read_tmp;
> >> +
> >> + ret = mlx90632_read_ambient_raw(data->regmap, &ambient_new_raw,
> >> + &ambient_old_raw);
> >> + *val = mlx90632_calc_temp_ambient(ambient_new_raw, ambient_old_raw,
> >> + PT, PR, PG, PO, Gb);
> >> + return ret;
> >> +}
> >> +
> >> +static int mlx90632_read_raw(struct iio_dev *indio_dev,
> >> + struct iio_chan_spec const *channel, int *val,
> >> + int *val2, long mask)
> >> +{
> >> + struct mlx90632_data *data = iio_priv(indio_dev);
> >> + int ret;
> >> +
> >> + switch (mask) {
> >> + case IIO_CHAN_INFO_PROCESSED:
> >> + switch (channel->channel2) {
> >> + case IIO_MOD_TEMP_AMBIENT:
> >> + ret = mlx90632_calc_ambient_dsp105(data, val);
> >> + if (ret < 0)
> >> + return ret;
> >> + return IIO_VAL_INT;
> >> + case IIO_MOD_TEMP_OBJECT:
> >> + ret = mlx90632_calc_object_dsp105(data, val);
> >> + if (ret < 0)
> >> + return ret;
> >> + return IIO_VAL_INT;
> >> + default:
> >> + return -EINVAL;
> >> + }
> >> + case IIO_CHAN_INFO_CALIBEMISSIVITY:
> >> + if (data->emissivity == 1000) {
> >> + *val = 1;
> >> + *val2 = 0;
> >> + } else {
> >> + *val = 0;
> >> + *val2 = data->emissivity;
> > Odd given you are reporting as int + nano
> > this goes from 1.0 to 0.000000999 in one step?
> > Seems unlikely.. If it is true then a comment
> > is needed.
> >
> > The write is int + micro but this would still be wrong
> > without a factor of 1000.
> >
>
> In input (write) below I am adding factor of 1000. I do not know
> if resolution like that is needed, but you never know what they
> come up with in real life so I rather left some of the breating
> space.
The above still doesn't make much sense unless your range
is disjoint 0-0.000000999, 1.0 which would be surprising and
also doesn't correspond to the write.
>
> >> + }
> >> + return IIO_VAL_INT_PLUS_NANO;
> >> +
> >> + default:
> >> + return -EINVAL;
> >> + }
> >> +}
> >> +
> >> +static int mlx90632_write_raw(struct iio_dev *indio_dev,
> >> + struct iio_chan_spec const *channel, int val,
> >> + int val2, long mask)
> >> +{
> >> + struct mlx90632_data *data = iio_priv(indio_dev);
> >> +
> >> + switch (mask) {
> >> + case IIO_CHAN_INFO_CALIBEMISSIVITY:
> >> + if (val < 0 || val2 < 0 || val > 1 ||
> >> + (val == 1 && val2 != 0))
> >
> > I'd add a comment describing what this is doing. I think it
> > is checking for 0-1.0 inclusive?
> >
>
> Yes, exactly. In case you set emissivity outside of range it should
> not change it and return EINVAL.
>
> >> + return -EINVAL;
> >> + data->emissivity = val * 1000 + val2 / 1000;
> >> + return 0;
> >> + default:
> >> + return -EINVAL;
> >> + }
> >> +}
> >> +
> >> +static const struct iio_chan_spec mlx90632_channels[] = {
> >> + {
> >> + .type = IIO_TEMP,
> >> + .modified = 1,
> >> + .channel2 = IIO_MOD_TEMP_AMBIENT,
> >> + .info_mask_separate = BIT(IIO_CHAN_INFO_PROCESSED),
> >> + },
> >> + {
> >> + .type = IIO_TEMP,
> >> + .modified = 1,
> >> + .channel2 = IIO_MOD_TEMP_OBJECT,
> >> + .info_mask_separate = BIT(IIO_CHAN_INFO_PROCESSED) |
> >> + BIT(IIO_CHAN_INFO_CALIBEMISSIVITY),
> >> + },
> >> +};
> >> +
> >> +static const struct iio_info mlx90632_info = {
> >> + .read_raw = mlx90632_read_raw,
> >> + .write_raw = mlx90632_write_raw,
> >> +};
> >> +
> >> +#ifdef CONFIG_PM
> >> +static int mlx90632_sleep(struct mlx90632_data *data)
> >> +{
> >> + dev_dbg(&data->client->dev, "Requesting sleep");
> >> + return mlx90632_pwr_set_sleep_step(data->regmap);
> >> +}
> >> +
> >> +static int mlx90632_wakeup(struct mlx90632_data *data)
> >> +{
> >> + dev_dbg(&data->client->dev, "Requesting wake-up");
> >> + return mlx90632_pwr_continuous(data->regmap);
> >> +}
> >> +#endif
> >> +
> >> +static int mlx90632_probe(struct i2c_client *client,
> >> + const struct i2c_device_id *id)
> >> +{
> >> + struct iio_dev *indio_dev;
> >> + struct mlx90632_data *mlx90632;
> >> + struct regmap *regmap;
> >> + int ret;
> >> + unsigned int read;
> >> +
> >> + indio_dev = devm_iio_device_alloc(&client->dev, sizeof(*mlx90632));
> >> + if (!indio_dev) {
> >> + dev_err(&client->dev, "Failed to allocate device");
> >> + return -ENOMEM;
> >> + }
> >> +
> >> + regmap = devm_regmap_init_i2c(client, &mlx90632_regmap);
> >> + if (IS_ERR(regmap)) {
> >> + ret = PTR_ERR(regmap);
> >> + dev_err(&client->dev, "Failed to allocate regmap: %d\n", ret);
> >> + return ret;
> >> + }
> >> +
> >> + mlx90632 = iio_priv(indio_dev);
> >> + i2c_set_clientdata(client, indio_dev);
> >> + mlx90632->client = client;
> >> + mlx90632->regmap = regmap;
> >> +
> >> + mutex_init(&mlx90632->lock);
> >> + mlx90632_wakeup(mlx90632);
> >> +
> >> + indio_dev->dev.parent = &client->dev;
> >> + indio_dev->name = id->name;
> >> + indio_dev->modes = INDIO_DIRECT_MODE;
> >> + indio_dev->info = &mlx90632_info;
> >> + indio_dev->channels = mlx90632_channels;
> >> + indio_dev->num_channels = ARRAY_SIZE(mlx90632_channels);
> >> +
> >> + ret = regmap_read(mlx90632->regmap, MLX90632_EE_VERSION, &read);
> >> + if (ret < 0) {
> >> + dev_err(&client->dev, "read of version failed: %d\n", ret);
> >> + return ret;
> >> + }
> >> + if (read == (MLX90632_EEPROM_VERSION & MLX90632_ID_MEDICAL)) {
> >
> > This is odd. Why the bitwise and of what look to be two different things entirely?
> >
>
> If you look at how they are set (high vs low bits) you will see they
> are complimentary and
> prepared exactly for bitwise check. ID_MEDICAL might not be best name, but
> EEPROM_ID_MEDICAL was just too long. You read 16 bits and that is why
> lower 8 bits
> are in fact EEPROM_VERSION while higher 8 are ID of calibration. Since
> calculations
> are currently dependent on lower 8 bits, I left some of the wiggle
> space for future.
> In case we add another one in future, but I would rather have a strict
> check for this to
> avoid any kind of confusion what is supported.
Surely should be | then?
read == (0xff05 & 0x01ff)
read == (0x105)
So only 3 bits of overlap?
Definitely needs a comment to explain this int the code.
>
> >> + dev_dbg(&client->dev,
> >> + "Detected Medical EEPROM calibration %x", read);
> >> + } else if (read == (MLX90632_EEPROM_VERSION & MLX90632_ID_CONSUMER)) {
> >> + dev_dbg(&client->dev,
> >> + "Detected Consumer EEPROM calibration %x", read);
> >> + } else {
> >> + dev_err(&client->dev,
> >> + "Chip EEPROM version mismatch %x (expected %x)",
> >> + read, MLX90632_EEPROM_VERSION);
> >> + return -EPROTONOSUPPORT;
> >> + }
> >> +
> >> + mlx90632->emissivity = 1000;
> >> +
> >> + return devm_iio_device_register(&client->dev, indio_dev);
> >
> > Don't use devm version. You are (correctly) manually unwinding
> > this in the remove (as you have pm to deal with after removing the
> > interfaces). This will result in a double free I think...
> > return iio_device_register is the way to go.
> >
> > I'm a bit confused that you don't seem to set up runtime pm anywhere...
> > I would assume we would be looking at autosuspend for a device
> > like this but it isn't enabled..
> >
>
> OK, will check how to use iio_device_register instead.
>
> I did not do much of PM on device, so I can't really say. That is
> why autosuspend is not enabled.
If you are dropping the PM for now, then devm_iio_device_register is
fine, just drop the remove function entirely as it won't need to do
anything.
>
> >> +}
> >> +
> >> +static int mlx90632_remove(struct i2c_client *client)
> >> +{
> >> + struct iio_dev *indio_dev = i2c_get_clientdata(client);
> >> + struct mlx90632_data *data = iio_priv(indio_dev);
> >> +
> >> + iio_device_unregister(indio_dev);
> >> +
> >> + pm_runtime_disable(&client->dev);
> >> + if (!pm_runtime_status_suspended(&client->dev))
> >> + mlx90632_sleep(data);
> >> + pm_runtime_set_suspended(&client->dev);
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +static const struct i2c_device_id mlx90632_id[] = {
> >> + { "mlx90632", 0 },
> >> + { }
> >> +};
> >> +MODULE_DEVICE_TABLE(i2c, mlx90632_id);
> >> +
> >> +static const struct of_device_id mlx90632_of_match[] = {
> >> + { .compatible = "melexis,mlx90632" },
> >> + { }
> >> +};
> >> +MODULE_DEVICE_TABLE(of, mlx90632_of_match);
> >> +
> >> +#ifdef CONFIG_PM_SLEEP
> >> +static int mlx90632_pm_suspend(struct device *dev)
> >> +{
> >> + struct iio_dev *indio_dev = i2c_get_clientdata(to_i2c_client(dev));
> >> + struct mlx90632_data *data = iio_priv(indio_dev);
> >> +
> >> + if (pm_runtime_active(dev))
> >> + return mlx90632_sleep(data);
> >
> > I'm a little confused as to why, if the device is powered
> > up fully and a suspend comes in we have to do less than we do
> > in runtime pm case.
> >
> > Am I missing something?
> >
>
> I think I have mostly tested runtime pm case, so this might have
> been missed. I did get wrong results if I did not mark regcache as dirty.
>
> I will just remove the whole PM thing to start with and add it up later, OK?
Makes sense.
>
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +static int mlx90632_pm_resume(struct device *dev)
> >> +{
> >> + struct iio_dev *indio_dev = i2c_get_clientdata(to_i2c_client(dev));
> >> + struct mlx90632_data *data = iio_priv(indio_dev);
> >> + int err;
> >> +
> >> + err = mlx90632_wakeup(data);
> >> + if (err < 0)
> >> + return err;
> >> +
> >> + pm_runtime_disable(dev);
> >> + pm_runtime_set_active(dev);
> >> + pm_runtime_enable(dev);
> >> +
> >> + return 0;
> >> +}
> >> +#endif
> >> +
> >> +#ifdef CONFIG_PM
> >> +static int mlx90632_pm_runtime_suspend(struct device *dev)
> >> +{
> >> + struct iio_dev *indio_dev = i2c_get_clientdata(to_i2c_client(dev));
> >> + struct mlx90632_data *mlx90632 = iio_priv(indio_dev);
> >> +
> >> + regcache_sync(mlx90632->regmap);
> >> +
> >> + return mlx90632_sleep(mlx90632);
> >> +}
> >> +
> >> +static int mlx90632_pm_runtime_resume(struct device *dev)
> >> +{
> >> + s32 ret;
> >> + struct iio_dev *indio_dev = i2c_get_clientdata(to_i2c_client(dev));
> >> + struct mlx90632_data *mlx90632 = iio_priv(indio_dev);
> >> +
> >> + regcache_mark_dirty(mlx90632->regmap);
> >> + regcache_cache_only(mlx90632->regmap, false);
> >> + ret = regcache_sync(mlx90632->regmap);
> >> + if (ret < 0) {
> >> + dev_err(dev, "Failed to sync regmap registers: %d\n", ret);
> >> + return ret;
> >> + }
> >> +
> >> + return mlx90632_wakeup(mlx90632);
> >> +}
> >> +#endif
> >> +
> >> +static const struct dev_pm_ops mlx90632_pm_ops = {
> >> + SET_SYSTEM_SLEEP_PM_OPS(mlx90632_pm_suspend, mlx90632_pm_resume)
> >> + SET_RUNTIME_PM_OPS(mlx90632_pm_runtime_suspend,
> >> + mlx90632_pm_runtime_resume, NULL)
> >> +};
> >> +
> >> +static struct i2c_driver mlx90632_driver = {
> >> + .driver = {
> >> + .name = "mlx90632",
> >> + .of_match_table = mlx90632_of_match,
> >> + .pm = &mlx90632_pm_ops,
> >> + },
> >> + .probe = mlx90632_probe,
> >> + .remove = mlx90632_remove,
> >> + .id_table = mlx90632_id,
> >> +};
> >> +module_i2c_driver(mlx90632_driver);
> >> +
> >> +MODULE_AUTHOR("Crt Mori <cmo@melexis.com>");
> >> +MODULE_DESCRIPTION("Melexis MLX90632 contactless Infra Red temperature sensor driver");
> >> +MODULE_LICENSE("GPL v2");
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-iio" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-iio" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2017-12-05 11:23 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-29 22:07 [PATCH 1/2] iio: temperature: Adding support for MLX90632 Crt Mori
2017-12-02 14:33 ` Jonathan Cameron
2017-12-02 20:30 ` Crt Mori
2017-12-05 11:23 ` Jonathan Cameron [this message]
2017-12-05 15:59 ` Crt Mori
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20171205112337.00005dca@huawei.com \
--to=jonathan.cameron@huawei.com \
--cc=cmo@melexis.com \
--cc=jic23@kernel.org \
--cc=linux-iio@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox