| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Thu, 26 Oct 2023 22:10:29 -0700
From: Guenter Roeck <linux@...ck-us.net>
To: Daniel Matyas <daniel.matyas@...log.com>
Cc: Jean Delvare <jdelvare@...e.com>, Jonathan Corbet <corbet@....net>,
linux-hwmon@...r.kernel.org, linux-doc@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 4/4] hwmon: max31827: Add custom attribute for
resolution
On 10/26/23 07:44, Daniel Matyas wrote:
> Added custom channel-specific (temp1) attribute for resolution. The wait
> time for a conversion in one-shot mode (enable = 0) depends on the
> resolution.
>
> When resolution is 12-bit, the conversion time is 140ms, but the minimum
> update_interval is 125ms. Handled this problem by waiting an additional
> 15ms (125ms + 15ms = 140ms).
>
> Added 'mask' parameter to the shutdown_write() function. Now it can
> either write or update bits, depending on the value of mask. This is
> needed, because for alarms a write is necessary, but for resolution only
> the resolution bits should be updated.
>
> Signed-off-by: Daniel Matyas <daniel.matyas@...log.com>
> ---
>
> v4 -> v5: No change.
>
> v3 -> v4: No change.
>
> v2 -> v3: Fixed indentation problems in .rst.
>
> v1 -> v2: Changed subject. Separated patch. Removed timeout sysfs
> attribute and kept only resolution. Added temp1_ prefix to resolution.
> Changed value of resolution from bits to milli-degrees Celsius. Added
> appropriate documentation.
>
>
> Documentation/hwmon/max31827.rst | 29 ++++++--
> drivers/hwmon/max31827.c | 121 ++++++++++++++++++++++++++++---
> 2 files changed, 134 insertions(+), 16 deletions(-)
>
> diff --git a/Documentation/hwmon/max31827.rst b/Documentation/hwmon/max31827.rst
> index a8bbfb85dd02..44ab9dc064cb 100644
> --- a/Documentation/hwmon/max31827.rst
> +++ b/Documentation/hwmon/max31827.rst
> @@ -90,11 +90,28 @@ the data sheet are:
>
> Enabling the device when it is already enabled has the side effect of setting
> the conversion frequency to 1 conv/s. The conversion time varies depending on
> -the resolution. The conversion time doubles with every bit of increased
> -resolution. For 10 bit resolution 35ms are needed, while for 12 bit resolution
> -(default) 140ms. When chip is in shutdown mode and a read operation is
> -requested, one-shot is triggered, the device waits for 140 (conversion time) ms,
> -and only after that is the temperature value register read.
> +the resolution.
> +
> +The conversion time doubles with every bit of increased resolution. The
> +available resolutions are:
> +
> +- 8 bit -> 8.75 ms conversion time
> +- 9 bit -> 17.5 ms conversion time
> +- 10 bit -> 35 ms conversion time
> +- 12 bit (default) -> 140 ms conversion time
> +
> +There is a temp1_resolution attribute which indicates the unit change in the
> +input temperature in milli-degrees C.
> +
> +- 1000 mC -> 8 bit
> +- 500 mC -> 9 bit
> +- 250 mC -> 10 bit
> +- 62 mC -> 12 bit (default) - actually this is 62.5, but the fil returns 62
fil ? Should that be "attribute" ?
> +
> +When chip is in shutdown mode and a read operation is requested, one-shot is
> +triggered, the device waits for <conversion time> ms, and only after that is
> +the temperature value register read. Note that the conversion times are rounded
> +up to the nearest possible integer.
>
> The LSB of the temperature values is 0.0625 degrees Celsius, but the values of
> the temperatures are displayed in milli-degrees. This means, that some data is
> @@ -117,4 +134,4 @@ corresponding status bits.
> Notes
> -----
>
> -PEC and resolution are not implemented.
> +PEC is not implemented.
> diff --git a/drivers/hwmon/max31827.c b/drivers/hwmon/max31827.c
> index 8ba387b51157..c9fa6d4d6194 100644
> --- a/drivers/hwmon/max31827.c
> +++ b/drivers/hwmon/max31827.c
> @@ -37,6 +37,9 @@
> #define MAX31827_CONFIGURATION_U_TEMP_STAT_MASK BIT(14)
> #define MAX31827_CONFIGURATION_O_TEMP_STAT_MASK BIT(15)
>
> +#define MAX31827_8_BIT_CNV_TIME 9
> +#define MAX31827_9_BIT_CNV_TIME 18
> +#define MAX31827_10_BIT_CNV_TIME 35
> #define MAX31827_12_BIT_CNV_TIME 140
>
> #define MAX31827_ALRM_POL_HIGH 0x1
> @@ -68,6 +71,27 @@ static const u16 max31827_conversions[] = {
> [MAX31827_CNV_8_HZ] = 125,
> };
>
> +enum max31827_resolution {
> + MAX31827_RES_8_BIT = 0,
> + MAX31827_RES_9_BIT,
> + MAX31827_RES_10_BIT,
> + MAX31827_RES_12_BIT,
> +};
> +
> +static const u16 max31827_resolutions[] = {
> + [MAX31827_RES_8_BIT] = 1000,
> + [MAX31827_RES_9_BIT] = 500,
> + [MAX31827_RES_10_BIT] = 250,
> + [MAX31827_RES_12_BIT] = 62,
> +};
> +
> +static const u16 max31827_conv_times[] = {
> + [MAX31827_RES_8_BIT] = MAX31827_8_BIT_CNV_TIME,
> + [MAX31827_RES_9_BIT] = MAX31827_9_BIT_CNV_TIME,
> + [MAX31827_RES_10_BIT] = MAX31827_10_BIT_CNV_TIME,
> + [MAX31827_RES_12_BIT] = MAX31827_12_BIT_CNV_TIME,
> +};
> +
> struct max31827_state {
> /*
> * Prevent simultaneous access to the i2c client.
> @@ -75,6 +99,8 @@ struct max31827_state {
> struct mutex lock;
> struct regmap *regmap;
> bool enable;
> + unsigned int resolution;
> + unsigned int update_interval;
> };
>
> static const struct regmap_config max31827_regmap = {
> @@ -91,9 +117,9 @@ static int shutdown_write(struct max31827_state *st, unsigned int reg,
> int ret;
>
> /*
> - * Before the Temperature Threshold Alarm and Alarm Hysteresis Threshold
> - * register values are changed over I2C, the part must be in shutdown
> - * mode.
> + * Before the Temperature Threshold Alarm, Alarm Hysteresis Threshold
> + * and Resolution bits from Configuration register are changed over I2C,
> + * the part must be in shutdown mode.
> *
> * Mutex is used to ensure, that some other process doesn't change the
> * configuration register.
> @@ -211,9 +237,18 @@ static int max31827_read(struct device *dev, enum hwmon_sensor_types type,
> mutex_unlock(&st->lock);
> return ret;
> }
> -
> - msleep(MAX31827_12_BIT_CNV_TIME);
> + msleep(max31827_conv_times[st->resolution]);
> }
> +
> + /*
> + * For 12-bit resolution the conversion time is 140 ms,
> + * thus an additional 15 ms is needed to complete the
> + * conversion: 125 ms + 15 ms = 140 ms
> + */
> + if (max31827_resolutions[st->resolution] == 12 &&
> + st->update_interval == 125)
> + usleep_range(15000, 20000);
> +
> ret = regmap_read(st->regmap, MAX31827_T_REG, &uval);
>
> mutex_unlock(&st->lock);
> @@ -371,10 +406,14 @@ static int max31827_write(struct device *dev, enum hwmon_sensor_types type,
> res = FIELD_PREP(MAX31827_CONFIGURATION_CNV_RATE_MASK,
> res);
>
> - return regmap_update_bits(st->regmap,
> - MAX31827_CONFIGURATION_REG,
> - MAX31827_CONFIGURATION_CNV_RATE_MASK,
> - res);
> + ret = regmap_update_bits(st->regmap,
> + MAX31827_CONFIGURATION_REG,
> + MAX31827_CONFIGURATION_CNV_RATE_MASK,
> + res);
> + if (ret)
> + return ret;
> +
> + st->update_interval = val;
> }
> break;
>
> @@ -385,6 +424,68 @@ static int max31827_write(struct device *dev, enum hwmon_sensor_types type,
> return -EOPNOTSUPP;
> }
>
> +static ssize_t temp1_resolution_show(struct device *dev,
> + struct device_attribute *devattr,
> + char *buf)
> +{
> + struct max31827_state *st = dev_get_drvdata(dev);
> + unsigned int val;
> + int ret;
> +
> + ret = regmap_read(st->regmap, MAX31827_CONFIGURATION_REG, &val);
> + if (ret)
> + return ret;
> +
> + val = FIELD_GET(MAX31827_CONFIGURATION_RESOLUTION_MASK, val);
> +
> + return scnprintf(buf, PAGE_SIZE, "%u\n", max31827_resolutions[val]);
> +}
> +
> +static ssize_t temp1_resolution_store(struct device *dev,
> + struct device_attribute *devattr,
> + const char *buf, size_t count)
> +{
> + struct max31827_state *st = dev_get_drvdata(dev);
> + unsigned int idx = 0;
> + unsigned int val;
> + int ret;
> +
> + ret = kstrtouint(buf, 10, &val);
> + if (ret)
> + return ret;
> +
> + /*
> + * Convert the desired resolution into register
> + * bits. idx is already initialized with 0.
> + *
> + * This was inspired by lm73 driver.
> + */
> + while (idx < ARRAY_SIZE(max31827_resolutions) &&
> + val < max31827_resolutions[idx])
> + idx++;
> +
> + if (idx == ARRAY_SIZE(max31827_resolutions) ||
> + val != max31827_resolutions[idx])
> + return -EOPNOTSUPP;
-EINVAL
Question here is if this should return the closest value/index
or return an error. Traditionally, for ranges like this, drivers
would accept all values and select the closest supported value.
[ just as the lm73 driver, which is listed as example here, does ]
There are even helper functions for that - find_closest()
or find_closest_descending(). I would very much prefer the
more flexible approach and not burden the user with having to find
supported values.
> +
> + st->resolution = idx;
> +
> + ret = shutdown_write(st, MAX31827_CONFIGURATION_REG,
> + MAX31827_CONFIGURATION_RESOLUTION_MASK,
> + FIELD_PREP(MAX31827_CONFIGURATION_RESOLUTION_MASK,
> + idx));
> +
> + return (ret) ? ret : count;
drop () around ret
> +}
> +
> +static DEVICE_ATTR_RW(temp1_resolution);
> +
> +static struct attribute *max31827_attrs[] = {
> + &dev_attr_temp1_resolution.attr,
> + NULL
> +};
> +ATTRIBUTE_GROUPS(max31827);
> +
> static const struct i2c_device_id max31827_i2c_ids[] = {
> { "max31827", max31827 },
> { "max31828", max31828 },
> @@ -501,7 +602,7 @@ static int max31827_probe(struct i2c_client *client)
>
> hwmon_dev = devm_hwmon_device_register_with_info(dev, client->name, st,
> &max31827_chip_info,
> - NULL);
> + max31827_groups);
>
> return PTR_ERR_OR_ZERO(hwmon_dev);
> }
Powered by blists - more mailing lists