Re: [PATCH v2 3/6] hwmon: (pmbus/pim4328) Add support for reading direct format coefficients
From: Erik Rosen <hidden>
Date: 2021-05-23 17:53:01
Also in:
linux-hwmon, lkml
On Sat, May 22, 2021 at 3:41 PM Guenter Roeck [off-list ref] wrote:
On 5/22/21 3:55 AM, Erik Rosen wrote:quoted
Add support for reading and decoding direct format coefficients to the PMBus core driver. If the new flag PMBUS_USE_COEFFICIENTS_CMD is set, the driver will use the COEFFICIENTS register together with the information in the pmbus_sensor_attr structs to initialize relevant coefficients for the direct mode format. Signed-off-by: Erik Rosen <redacted> --- drivers/hwmon/pmbus/pmbus_core.c | 93 ++++++++++++++++++++++++++++++++ include/linux/pmbus.h | 8 +++ 2 files changed, 101 insertions(+)diff --git a/drivers/hwmon/pmbus/pmbus_core.c b/drivers/hwmon/pmbus/pmbus_core.c index 460cbfd716e4..03c169bf5633 100644 --- a/drivers/hwmon/pmbus/pmbus_core.c +++ b/drivers/hwmon/pmbus/pmbus_core.c@@ -2177,6 +2177,38 @@ static int pmbus_find_attributes(struct i2c_client *client, return ret; } +static int pmbus_init_coefficients(struct i2c_client *client, + struct pmbus_data *data, int page,This seems wrong. Coefficients are not maintained per page but per class, and (re-)reading them for each supported page doesn't really add value or even make sense.quoted
+ enum pmbus_sensor_classes sensor_class, + const struct pmbus_sensor_attr *attrs, + int nattrs) +{ + int i, status; + + for (i = 0; i < nattrs; i++) { + if (attrs->class == sensor_class && + (attrs->func & data->info->func[page])) { + status = pmbus_read_coefficients(client, + (struct pmbus_driver_info *)data->info, + sensor_class, + attrs->reg); + if (status < 0) { + dev_err(&client->dev, + "Failed to read coefficients for register: %x\n", + attrs->reg); + return status; + } + return 0; + } + attrs++; + } + + dev_err(&client->dev, "No coefficients found for register: %x\n", + attrs->reg); +attrs points beyond the array size here, so attrs->reg does not point to a valid array element. The problem would also not be the register this happens to point to, but the class (ie the chip does not support a sensor of the requested class). Not sure if this should trigger a message or error in the first place. It won't matter since the chip will never need those coefficients. If anything, this would be a misconfiguration (the driver should not set direct format for this sensor class), and the return value should be -EINVAL. Either case, I wonder if this can be handled with less complex code, ie without having to check data->info->func[] for all pages. How about just walking through attrs and try all class matches until one is found that works (ie not return on error but keep trying) ?
Ok, I'll send a new version based on your comments. I'm not entirely comfortable with just silently ignoring any failure to retrieve the coefficients for a sensor class. I mean it could be due to any reason; a bus error for instance. I'll return a -EINVAL for now if you don't disagree. /Erik
quoted
+ return -ENODEV; +} + /* * Identify chip parameters. * This function is called for all chips.@@ -2185,6 +2217,7 @@ static int pmbus_identify_common(struct i2c_client *client, struct pmbus_data *data, int page) { int vout_mode = -1; + int ret; if (pmbus_check_byte_register(client, page, PMBUS_VOUT_MODE)) vout_mode = _pmbus_read_byte_data(client, page,@@ -2214,6 +2247,66 @@ static int pmbus_identify_common(struct i2c_client *client, } } + if (data->flags & PMBUS_USE_COEFFICIENTS_CMD) {I think there should be a separate function to handle that, to be called only once, not once per page.quoted
+ if (!i2c_check_functionality(client->adapter, + I2C_FUNC_SMBUS_BLOCK_PROC_CALL)) + return -ENODEV; + + if (data->info->format[PSC_VOLTAGE_IN] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_VOLTAGE_IN, + voltage_attributes, + ARRAY_SIZE(voltage_attributes)); + if (ret) + return ret; + }It might be useful to have a little structure with {class, attribute list pointer, attribute list size} and walk through that in a loop instead of repeating essentially the same code multiple times.quoted
+ + if (data->info->format[PSC_VOLTAGE_OUT] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_VOLTAGE_OUT, + voltage_attributes, + ARRAY_SIZE(voltage_attributes)); + if (ret) + return ret; + } + + if (data->info->format[PSC_CURRENT_IN] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_CURRENT_IN, + current_attributes, + ARRAY_SIZE(current_attributes)); + if (ret) + return ret; + } + + if (data->info->format[PSC_CURRENT_OUT] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_CURRENT_OUT, + current_attributes, + ARRAY_SIZE(current_attributes)); + if (ret) + return ret; + } + + if (data->info->format[PSC_POWER] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_POWER, + power_attributes, + ARRAY_SIZE(power_attributes)); + if (ret) + return ret; + } + + if (data->info->format[PSC_TEMPERATURE] == direct) { + ret = pmbus_init_coefficients(client, data, page, + PSC_TEMPERATURE, + temp_attributes, + ARRAY_SIZE(temp_attributes)); + if (ret) + return ret; + } + } + pmbus_clear_fault_page(client, page); return 0; }diff --git a/include/linux/pmbus.h b/include/linux/pmbus.h index f720470b1bab..7fdc282dab5a 100644 --- a/include/linux/pmbus.h +++ b/include/linux/pmbus.h@@ -52,6 +52,14 @@ */ #define PMBUS_NO_WRITE_PROTECT BIT(4) +/* + * PMBUS_USE_COEFFICIENTS_CMD + * + * When this flag is set the PMBus core driver will use the COEFFICIENTS + * register to initialize the coefficients for the direct mode format. + */ +#define PMBUS_USE_COEFFICIENTS_CMD BIT(5) + struct pmbus_platform_data { u32 flags; /* Device specific flags */