| 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
| ||
|
Message-ID: <ZDFg8n4CEgJySGbv@equiv.tech>
Date: Sat, 08 Apr 2023 12:41:23 +0000 (UTC)
From: James Seo <james@...iv.tech>
To: Guenter Roeck <linux@...ck-us.net>
Cc: Jean Delvare <jdelvare@...e.com>,
platform-driver-x86@...r.kernel.org, linux-hwmon@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] hwmon: add HP WMI Sensors driver
Hello,
Apologies for not replying to this [PATCH] noting that I had sent a
superseding [PATCH v2]. I'll address your comments here anyway, as
they are still applicable and don't touch upon my changes in v2.
On Fri, Apr 07, 2023 at 02:52:36PM -0700, Guenter Roeck wrote:
> On 4/3/23 01:48, James Seo wrote:
>> Hewlett-Packard business-class computers report hardware monitoring
>> information via WMI. This driver pulls that information into hwmon.
>> Initial support is implemented for temperature, voltage, current,
>> and fan speed sensor types.
>>
>> HP's WMI implementation permits many other types of numeric sensors.
>> For that reason, a debugfs interface is also provided to enumerate
>> all numeric sensors visible on the WMI side. This should facilitate
>> adding support for other sensor types in the future.
>>
>> Tested on a HP Z420 and a HP EliteOne 800 G1.
>>
>> Note that these models have only temperature and fan speed sensors,
>> so other sensor types have not been confirmed working yet. However,
>> the driver should still work as expected. A 2005 HP whitepaper
>> describes the relevant MOF definition and sensor value calculation,
>> and reverse engineering confirms HP's official Performance Advisor
>> utility continues to comply with both of them.
>>
>> Link: https://h20331.www2.hp.com/hpsub/downloads/cmi_whitepaper.pdf
>> Signed-off-by: James Seo <james@...iv.tech>
>> ---
>> Documentation/hwmon/hp-wmi-sensors.rst | 97 +++
>> Documentation/hwmon/index.rst | 1 +
>> MAINTAINERS | 7 +
>> drivers/hwmon/Kconfig | 11 +
>> drivers/hwmon/Makefile | 1 +
>> drivers/hwmon/hp-wmi-sensors.c | 1043 ++++++++++++++++++++++++
>> 6 files changed, 1160 insertions(+)
>> create mode 100644 Documentation/hwmon/hp-wmi-sensors.rst
>> create mode 100644 drivers/hwmon/hp-wmi-sensors.c
>>
>> diff --git a/Documentation/hwmon/hp-wmi-sensors.rst b/Documentation/hwmon/hp-wmi-sensors.rst
>> new file mode 100644
>> index 000000000000..b2e166c8e771
>> --- /dev/null
>> +++ b/Documentation/hwmon/hp-wmi-sensors.rst
>> @@ -0,0 +1,97 @@
>> +.. SPDX-License-Identifier: GPL-2.0-or-later
>> +
>> +.. include:: <isonum.txt>
>> +
>> +Linux HP WMI Sensors Driver
>> +===========================
>> +
>> +:Copyright: |copy| 2023 James Seo <james@...iv.tech>
>> +
>> +Description
>> +-----------
>> +
>> +Hewlett-Packard business-class computers report hardware monitoring information
>> +via Windows Management Instrumentation (WMI). This driver exposes that
>> +information to the Linux ``hwmon`` subsystem, allowing familiar userspace
>> +utilities like ``sensors`` to gather numeric sensor readings.
>> +
>> +``sysfs`` interface
>> +-------------------
>> +
>> +When the driver is loaded, it discovers the sensors available on the current
>> +system and creates the following read-only ``sysfs`` attributes as appropriate
>> +within ``/sys/class/hwmon/hwmonX``:
>> +
>> +================ ===================================
>> +Name Description
>> +================ ===================================
>> +curr[X]_input Current in milliamperes (mA).
>> +curr[X]_label Current sensor label.
>> +fan[X]_input Fan speed in RPM.
>> +fan[X]_label Fan sensor label.
>> +fan[X]_fault Fan sensor fault indicator.
>> +in[X]_input Voltage in millivolts (mV).
>> +in[X]_label Voltage sensor label.
>> +temp[X]_input Temperature in millivolts (mV).
>> +temp[X]_label Temperature sensor label.
>> +temp[X]_fault Temperature sensor fault indicator.
>> +================ ===================================
>> +
>> +Here, ``X`` is some number that depends on other available sensors and on other
>> +system hardware components.
>> +
>> +``fault`` attributes
>> + Reading ``1`` instead of ``0`` as the ``fault`` attribute for a sensor
>> + indicates that the sensor has encountered some issue during operation, and
>> + that measurements from it should no longer be trusted.
>> +
>> +``debugfs`` interface
>> +---------------------
>> +
>> +The standard ``hwmon`` interface in ``sysfs`` exposes sensors of several common
>> +types that are connected and operating normally as of driver initialization.
>> +However, there are usually other sensors on the WMI side that do not meet these
>> +criteria. This driver therefore provides a ``debugfs`` interface in
>> +``/sys/kernel/debug/hp-wmi-sensors-X`` that allows read-only access to *all* HP
>> +WMI sensors on the current system.
>> +
>
> That doesn't explain why not all standard attributes are utilized. "Caution"
> would presumably match _warning attributes, and "Critical" would match "_crit"
> attributes. There are standard attributes for intrusion detection and for humidity.
>
I presume you meant "_alarm" attributes instead of "_warning".
First, sorry for any confusion caused by my imprecision in an earlier
reply to Armin when I said that I would filter out sensors that were
"Not Present". In the code, this will take the form of filtering out
sensors with an OperationalStatus of "No Contact".
In the HP_BIOSNumericSensor object, CurrentState and PossibleStates[]
are freeform strings that can and do vary by system. Using them for
threshold-based attributes like _crit, _emergency, and _max isn't
feasible, as they aren't guaranteed to have any particular values.
The only certainty according to that whitepaper I referenced as [1]
is that CurrentState is one of the PossibleStates[].
It's true that for the particular system on which I gathered the
example values below, CurrentState will probably change from "Normal"
to "Caution" at some threshold, etc. Perhaps this sort of sensor is
even present on all HP systems. But finding the thresholds can be
done only indirectly by polling WMI in the background to watch
CurrentState for changes that wouldn't be expected to happen in
normal use. And foreknowledge of PossibleStates[] would be required.
Since CurrentState and PossibleStates[] shouldn't be considered, the
only remaining HP_BIOSNumericSensor property that might be useful for
other hwmon attributes is OperationalStatus. I'll discuss my use of
it and _alarm attributes further below in my response re: _fault.
The PWM sensor type is out because HP_BIOSNumericSensor objects are
read-only.
The humidity type is out for now due to a combination of lack of
hardware and my not wanting to guess how HP would implement a
humidity sensor (unlike voltage and current sensors, for which the
implementation is quite obvious). But assuming hardware exists in the
wild, adding support would be straightforward.
The intrusion type is also out for the same two reasons, in addition
to intrusion sensors being implemented as HP_BIOSStateSensor objects,
which this driver does not deal with. But it could be added given
demand, hardware, and testers.
Does that sufficiently explain my choices?
>> +.. warning:: The ``debugfs`` interface is only available when the kernel is
>> + compiled with option ``CONFIG_DEBUG_FS``, and its implementation
>> + is subject to change without notice at any time.
>> +
>> +One numbered entry is created per sensor with the following attributes:
>> +
>> +=============================== ==========================================
>> +Name Example
>> +=============================== ==========================================
>> +name ``CPU0 Fan``
>> +description ``Reports CPU0 fan speed``
>> +sensor_type ``12``
>> +other_sensor_type ``(null)``
>> +operational_status ``2``
>> +current_state ``Normal``
>> +possible_states ``Normal\nCaution\nCritical\nNot Present``
>> +base_units ``19``
>> +unit_modifier ``0``
>> +current_reading ``1008``
>> +=============================== ==========================================
>> +
>> +These represent the properties of the underlying ``HP_BIOSNumericSensor`` WMI
>> +object, some of which may vary in contents and formatting (but not presence or
>> +semantics) between systems. See [#]_ for more details.
>> +
>> +Known issues and limitations
>> +----------------------------
>> +
>> +- Non-numeric HP sensor types such as intrusion sensors that belong to the
>> + ``HP_BIOSStateSensor`` WMI object type are not supported.
>> +- It is intended that the ``debugfs`` interface will facilitate supporting more
>> + types in the future. Whether systems that implement more than the types
>> + already supported exist in the wild is unknown.
>> +
>> +References
>> +----------
>> +
>> +.. [#] Hewlett-Packard Development Company, L.P.,
>> + "HP Client Management Interface Technical White Paper", 2005. [Online].
>> + Available: https://h20331.www2.hp.com/hpsub/downloads/cmi_whitepaper.pdf
>> diff --git a/Documentation/hwmon/index.rst b/Documentation/hwmon/index.rst
>> index f1fe75f596a5..f8f3c0bef6ed 100644
>> --- a/Documentation/hwmon/index.rst
>> +++ b/Documentation/hwmon/index.rst
>> @@ -77,6 +77,7 @@ Hardware Monitoring Kernel Drivers
>> gl518sm
>> gxp-fan-ctrl
>> hih6130
>> + hp-wmi-sensors
>> ibmaem
>> ibm-cffps
>> ibmpowernv
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 1dc8bd26b6cf..5038dc59d3f7 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -9372,6 +9372,13 @@ L: platform-driver-x86@...r.kernel.org
>> S: Orphan
>> F: drivers/platform/x86/hp/tc1100-wmi.c
>> +HP WMI HARDWARE MONITOR DRIVER
>> +M: James Seo <james@...iv.tech>
>> +L: linux-hwmon@...r.kernel.org
>> +S: Maintained
>> +F: Documentation/hwmon/hp-wmi-sensors.rst
>> +F: drivers/hwmon/hp-wmi-sensors.c
>> +
>> HPET: High Precision Event Timers driver
>> M: Clemens Ladisch <clemens@...isch.de>
>> S: Maintained
>> diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
>> index 5b3b76477b0e..78798e0ed5e6 100644
>> --- a/drivers/hwmon/Kconfig
>> +++ b/drivers/hwmon/Kconfig
>> @@ -2399,6 +2399,17 @@ config SENSORS_ASUS_EC
>> This driver can also be built as a module. If so, the module
>> will be called asus_ec_sensors.
>> +config SENSORS_HP_WMI
>> + tristate "HP WMI Sensors"
>> + depends on ACPI_WMI
>> + help
>> + If you say yes here you get support for the ACPI hardware monitoring
>> + interface found in HP business-class computers. Temperature, voltage,
>> + current, and fan speed sensor types are supported if present.
>> +
>> + This driver can also be built as a module. If so, the module
>> + will be called hp_wmi_sensors.
>> +
>> endif # ACPI
>> endif # HWMON
>> diff --git a/drivers/hwmon/Makefile b/drivers/hwmon/Makefile
>> index 88712b5031c8..05cce16f37f6 100644
>> --- a/drivers/hwmon/Makefile
>> +++ b/drivers/hwmon/Makefile
>> @@ -11,6 +11,7 @@ obj-$(CONFIG_SENSORS_ACPI_POWER) += acpi_power_meter.o
>> obj-$(CONFIG_SENSORS_ATK0110) += asus_atk0110.o
>> obj-$(CONFIG_SENSORS_ASUS_EC) += asus-ec-sensors.o
>> obj-$(CONFIG_SENSORS_ASUS_WMI) += asus_wmi_sensors.o
>> +obj-$(CONFIG_SENSORS_HP_WMI) += hp-wmi-sensors.o
>> # Native drivers
>> # asb100, then w83781d go first, as they can override other drivers' addresses.
>> diff --git a/drivers/hwmon/hp-wmi-sensors.c b/drivers/hwmon/hp-wmi-sensors.c
>> new file mode 100644
>> index 000000000000..76af70b5a0e0
>> --- /dev/null
>> +++ b/drivers/hwmon/hp-wmi-sensors.c
>> @@ -0,0 +1,1043 @@
>> +// SPDX-License-Identifier: GPL-2.0-or-later
>> +/*
>> + * hwmon driver for HP business-class computers that report numeric
>> + * sensor data via Windows Management Instrumentation (WMI).
>> + *
>> + * Copyright (C) 2023 James Seo <james@...iv.tech>
>> + *
>> + * References:
>> + * [1] Hewlett-Packard Development Company, L.P.,
>> + * "HP Client Management Interface Technical White Paper", 2005. [Online].
>> + * Available: https://h20331.www2.hp.com/hpsub/downloads/cmi_whitepaper.pdf
>> + */
>> +
>> +#include <linux/acpi.h>
>> +#include <linux/debugfs.h>
>> +#include <linux/hwmon.h>
>> +#include <linux/jiffies.h>
>> +#include <linux/mutex.h>
>> +#include <linux/units.h>
>> +#include <linux/wmi.h>
>> +#include <linux/workqueue.h>
>> +
>> +/*
>> + * MOF definition of the HP_BIOSNumericSensor WMI object [1]:
>> + *
>> + * #pragma namespace("\\\\.\\root\\HP\\InstrumentedBIOS");
>> + *
>> + * [abstract]
>> + * class HP_BIOSSensor
>> + * {
>> + * [read] string Name;
>> + * [read] string Description;
>> + * [read, ValueMap {"0","1","2","3","4","5","6","7","8","9",
>> + * "10","11","12"}, Values {"Unknown","Other","Temperature",
>> + * "Voltage","Current","Tachometer","Counter","Switch","Lock",
>> + * "Humidity","Smoke Detection","Presence","Air Flow"}]
>> + * uint32 SensorType;
>> + * [read] string OtherSensorType;
>> + * [read, ValueMap {"0","1","2","3","4","5","6","7","8","9",
>> + * "10","11","12","13","14","15","16","17","18","..",
>> + * "0x8000.."}, Values {"Unknown","Other","OK","Degraded",
>> + * "Stressed","Predictive Failure","Error",
>> + * "Non-Recoverable Error","Starting","Stopping","Stopped",
>> + * "In Service","No Contact","Lost Communication","Aborted",
>> + * "Dormant","Supporting Entity in Error","Completed",
>> + * "Power Mode","DMTF Reserved","Vendor Reserved"}]
>> + * uint32 OperationalStatus;
>> + * [read] string CurrentState;
>> + * [read] string PossibleStates[];
>> + * };
>> + *
>> + * class HP_BIOSNumericSensor : HP_BIOSSensor
>> + * {
>> + * [read, ValueMap {"0","1","2","3","4","5","6","7","8","9",
>> + * "10","11","12","13","14","15","16","17","18","19","20",
>> + * "21","22","23","24","25","26","27","28","29","30","31",
>> + * "32","33","34","35","36","37","38","39","40","41","42",
>> + * "43","44","45","46","47","48","49","50","51","52","53",
>> + * "54","55","56","57","58","59","60","61","62","63","64",
>> + * "65"}, Values {"Unknown","Other","Degrees C","Degrees F",
>> + * "Degrees K","Volts","Amps","Watts","Joules","Coulombs",
>> + * "VA","Nits","Lumens","Lux","Candelas","kPa","PSI",
>> + * "Newtons","CFM","RPM","Hertz","Seconds","Minutes",
>> + * "Hours","Days","Weeks","Mils","Inches","Feet",
>> + * "Cubic Inches","Cubic Feet","Meters","Cubic Centimeters",
>> + * "Cubic Meters","Liters","Fluid Ounces","Radians",
>> + * "Steradians","Revolutions","Cycles","Gravities","Ounces",
>> + * "Pounds","Foot-Pounds","Ounce-Inches","Gauss","Gilberts",
>> + * "Henries","Farads","Ohms","Siemens","Moles","Becquerels",
>> + * "PPM (parts/million)","Decibels","DbA","DbC","Grays",
>> + * "Sieverts","Color Temperature Degrees K","Bits","Bytes",
>> + * "Words (data)","DoubleWords","QuadWords","Percentage"}]
>> + * uint32 BaseUnits;
>> + * [read] sint32 UnitModifier;
>> + * [read] uint32 CurrentReading;
>> + * };
>> + *
>> + */
>> +
>> +#define HP_WMI_BIOS_GUID "5FB7F034-2C63-45E9-BE91-3D44E2C707E4"
>> +#define HP_WMI_NUMERIC_SENSOR_GUID "8F1F6435-9F42-42C8-BADC-0E9424F20C9A"
>> +
>> +/* These limits are arbitrary. The WMI implementation may vary by model. */
>> +
>> +#define HP_WMI_MAX_STR_SIZE 128U
>> +#define HP_WMI_MAX_PROPERTIES 32U
>> +#define HP_WMI_MAX_INSTANCES 32U
>> +
>> +enum hp_wmi_type {
>> + HP_WMI_TYPE_OTHER = 1,
>> + HP_WMI_TYPE_TEMPERATURE = 2,
>> + HP_WMI_TYPE_VOLTAGE = 3,
>> + HP_WMI_TYPE_CURRENT = 4,
>> + HP_WMI_TYPE_AIR_FLOW = 12,
>
> This seems odd. Fan speed is measured in RPM, which seems to
> match "Revolutions" or "RPM" above, not air flow (which is
> typically a volume, such as CFM). In the text above there is
> "Tachometer" which would presumably the fan speed, and "Air Flow"
> which isn't really defined. Even if an "Air Flow" sensor would return
> RPM, that doesn't mean it is a fan. What about "Tachometer" (numeric
> value 5) ?
>
I agree it seems odd, but that is how HP does it. I'll add a comment
to that effect later on in the code.
>> +};
>> +
>> +enum hp_wmi_status {
>> + HP_WMI_STATUS_OK = 2,
>> +};
>> +
>> +enum hp_wmi_units {
>> + HP_WMI_UNITS_DEGREES_C = 2,
>> + HP_WMI_UNITS_DEGREES_F = 3,
>> + HP_WMI_UNITS_DEGREES_K = 4,
>> + HP_WMI_UNITS_VOLTS = 5,
>> + HP_WMI_UNITS_AMPS = 6,
>> + HP_WMI_UNITS_RPM = 19,
>> +};
>> +
>> +enum hp_wmi_property {
>> + HP_WMI_PROPERTY_NAME = 0,
>> + HP_WMI_PROPERTY_DESCRIPTION = 1,
>> + HP_WMI_PROPERTY_SENSOR_TYPE = 2,
>> + HP_WMI_PROPERTY_OTHER_SENSOR_TYPE = 3,
>> + HP_WMI_PROPERTY_OPERATIONAL_STATUS = 4,
>> + HP_WMI_PROPERTY_CURRENT_STATE = 5,
>> + HP_WMI_PROPERTY_POSSIBLE_STATES = 6,
>> + HP_WMI_PROPERTY_BASE_UNITS = 7,
>> + HP_WMI_PROPERTY_UNIT_MODIFIER = 8,
>> + HP_WMI_PROPERTY_CURRENT_READING = 9,
>> +};
>> +
>> +static const acpi_object_type hp_wmi_property_map[] = {
>> + [HP_WMI_PROPERTY_NAME] = ACPI_TYPE_STRING,
>> + [HP_WMI_PROPERTY_DESCRIPTION] = ACPI_TYPE_STRING,
>> + [HP_WMI_PROPERTY_SENSOR_TYPE] = ACPI_TYPE_INTEGER,
>> + [HP_WMI_PROPERTY_OTHER_SENSOR_TYPE] = ACPI_TYPE_STRING,
>> + [HP_WMI_PROPERTY_OPERATIONAL_STATUS] = ACPI_TYPE_INTEGER,
>> + [HP_WMI_PROPERTY_CURRENT_STATE] = ACPI_TYPE_STRING,
>> + [HP_WMI_PROPERTY_POSSIBLE_STATES] = ACPI_TYPE_STRING,
>> + [HP_WMI_PROPERTY_BASE_UNITS] = ACPI_TYPE_INTEGER,
>> + [HP_WMI_PROPERTY_UNIT_MODIFIER] = ACPI_TYPE_INTEGER,
>> + [HP_WMI_PROPERTY_CURRENT_READING] = ACPI_TYPE_INTEGER,
>> +};
>> +
>> +static const enum hwmon_sensor_types hp_wmi_hwmon_type_map[] = {
>> + [HP_WMI_TYPE_TEMPERATURE] = hwmon_temp,
>> + [HP_WMI_TYPE_VOLTAGE] = hwmon_in,
>> + [HP_WMI_TYPE_CURRENT] = hwmon_curr,
>> + [HP_WMI_TYPE_AIR_FLOW] = hwmon_fan,
>> +};
>> +
>> +static const u32 hp_wmi_hwmon_attributes[hwmon_max] = {
>> + [hwmon_chip] = HWMON_C_REGISTER_TZ,
>> + [hwmon_temp] = HWMON_T_INPUT | HWMON_T_LABEL | HWMON_T_FAULT,
>> + [hwmon_in] = HWMON_I_INPUT | HWMON_I_LABEL,
>> + [hwmon_curr] = HWMON_C_INPUT | HWMON_C_LABEL,
>> + [hwmon_fan] = HWMON_F_INPUT | HWMON_F_LABEL | HWMON_F_FAULT,
>> +};
>> +
>> +/*
>> + * struct hp_wmi_numeric_sensor - a HP_BIOSNumericSensor instance
>> + *
>> + * Contains WMI object instance properties. See MOF definition [1].
>> + */
>> +struct hp_wmi_numeric_sensor {
>> + const char *name;
>> + const char *description;
>> + u32 sensor_type;
>> + const char *other_sensor_type; /* Explains "Other" SensorType. */
>> + u32 operational_status;
>> + const char *current_state;
>> + const char **possible_states; /* Count may vary. */
>> + u32 base_units;
>> + s32 unit_modifier;
>> + u32 current_reading;
>> +
>> + u8 possible_states_count;
>> +};
>> +
>> +/*
>> + * struct hp_wmi_info - sensor info
>> + * @nsensor: numeric sensor properties
>> + * @instance: its WMI instance number
>> + * @is_active: whether the following fields are valid
>> + * @type: its hwmon sensor type
>> + * @cached_val: current sensor reading value, scaled for hwmon
>> + * @last_updated: when these readings were last updated
>> + */
>> +struct hp_wmi_info {
>> + struct hp_wmi_numeric_sensor nsensor;
>> + u8 instance;
>> +
>> + bool is_active;
>> + enum hwmon_sensor_types type;
>> + long cached_val;
>> + unsigned long last_updated; /* in jiffies */
>> +};
>> +
>> +/*
>> + * struct hp_wmi_sensors - driver state
>> + * @wdev: pointer to the parent WMI device
>> + * @debugfs: root directory in debugfs
>> + * @info: sensor info structs for all sensors visible in WMI
>> + * @info_map: access info structs by hwmon type and channel number
>> + * @count: count of all sensors visible in WMI
>> + * @channel_count: count of hwmon channels by hwmon type
>> + * @lock: mutex to lock polling WMI and changes to driver state
>> + */
>> +struct hp_wmi_sensors {
>> + struct wmi_device *wdev;
>> + struct dentry *debugfs;
>> + struct hp_wmi_info info[HP_WMI_MAX_INSTANCES];
>> + struct hp_wmi_info **info_map[hwmon_max];
>> + u8 count;
>> + u8 channel_count[hwmon_max];
>> +
>> + struct mutex lock; /* lock polling WMI, driver state changes */
>> +};
>> +
>> +/* hp_wmi_strdup - devm_kstrdup, but length-limited */
>> +static char *hp_wmi_strdup(struct device *dev, const char *src, u32 len)
>> +{
>> + char *dst;
>> +
>> + len = min(len, HP_WMI_MAX_STR_SIZE - 1);
>> +
>> + dst = devm_kmalloc(dev, (len + 1) * sizeof(*dst), GFP_KERNEL);
>> + if (!dst)
>> + return NULL;
>> +
>> + strscpy(dst, src, len + 1);
>> +
>> + return dst;
>> +}
>> +
>> +/*
>> + * hp_wmi_get_wobj - poll WMI for a HP_BIOSNumericSensor object instance
>> + * @state: pointer to driver state
>> + * @instance: WMI object instance number
>> + *
>> + * Returns a new WMI object instance on success, or NULL on error.
>> + * Caller must kfree the result.
>> + */
>> +static union acpi_object *hp_wmi_get_wobj(struct hp_wmi_sensors *state,
>> + u8 instance)
>> +{
>> + struct wmi_device *wdev = state->wdev;
>> +
>> + return wmidev_block_query(wdev, instance);
>> +}
>> +
>> +/*
>> + * check_wobj - validate a HP_BIOSNumericSensor WMI object instance
>> + * @wobj: pointer to WMI object instance to check
>> + * @possible_states_count: out pointer to count of possible states
>> + *
>> + * Returns 0 on success, or a negative error code on error.
>> + */
>> +static int check_wobj(const union acpi_object *wobj, u8 *possible_states_count)
>> +{
>> + acpi_object_type type = wobj->type;
>> + int prop = HP_WMI_PROPERTY_NAME;
>> + acpi_object_type valid_type;
>> + union acpi_object *elements;
>> + u32 elem_count;
>> + u8 count = 0;
>> + u32 i;
>> +
>> + if (type != ACPI_TYPE_PACKAGE)
>> + return -EINVAL;
>> +
>> + elem_count = wobj->package.count;
>> + if (elem_count > HP_WMI_MAX_PROPERTIES)
>> + return -EINVAL;
>> +
>> + elements = wobj->package.elements;
>> + for (i = 0; i < elem_count; i++, prop++) {
>> + if (prop > HP_WMI_PROPERTY_CURRENT_READING)
>> + return -EINVAL;
>> +
>> + type = elements[i].type;
>> + valid_type = hp_wmi_property_map[prop];
>> + if (type != valid_type)
>> + return -EINVAL;
>> +
>> + /*
>> + * elements is a variable-length array of ACPI objects, one for
>> + * each property of the WMI object instance, except that the
>> + * strs in PossibleStates[] are flattened into this array, and
>> + * their count is found in the WMI BMOF. We don't decode the
>> + * BMOF, so find the count by finding the next int.
>> + */
>> +
>> + if (prop == HP_WMI_PROPERTY_CURRENT_STATE) {
>> + prop = HP_WMI_PROPERTY_POSSIBLE_STATES;
>> + valid_type = hp_wmi_property_map[prop];
>> + for (; i + 1 < elem_count; i++, count++) {
>> + type = elements[i + 1].type;
>> + if (type != valid_type)
>> + break;
>> + }
>> + }
>> + }
>> +
>> + if (!count || prop <= HP_WMI_PROPERTY_CURRENT_READING)
>> + return -EINVAL;
>> +
>> + *possible_states_count = count;
>> +
>> + return 0;
>> +}
>> +
>> +static int numeric_sensor_has_fault(const struct hp_wmi_numeric_sensor *nsensor)
>> +{
>> + u32 operational_status = nsensor->operational_status;
>> + u32 current_reading = nsensor->current_reading;
>> +
>> + return operational_status != HP_WMI_STATUS_OK || !current_reading;
>
> Any status but "OK" is a fault, including "Starting", "Stopping", "Completed",
> "In Service", and others ? Really ?
>
Indeed. The documentation says that a fault condition means that the
measurement for a channel shouldn't be trusted. Reading through the
OperationalStatus definitions in [1], I was able to come up with a
plausible reason why literally every single one except for "OK"
carried with it some implication of uncertainty.
Let's go through just the statuses you mentioned. Why should I trust
a "Starting" sensor that doesn't consider itself ready yet? Or a
"Stopping" sensor that isn't stopping at my request? An "In Service"
sensor is being "configured, maintained, cleaned, or otherwise
administered" - doesn't sound too trustworthy to me. As for
"Completed" (17), it's supposed to be "combined" (by OR-ing, I
assume) with "OK" (2), "Error" (6), or "Degraded" (3), and it's
my clear favorite of the bunch. Try to make those values work and
you'll see why.
At any rate, it sounds like you'd prefer to see a more nuanced
handling of at least fault conditions. So the next version will
signal fault only for what I feel are the more obviously faulty
statuses.
I didn't implement alarm conditions because the documentation just
says that they are chip-determined. Since there is no chip here, the
determination would fall to my poor/blind judgment, with no extra
information coming from WMI about e.g. a "Non-Recoverable Error" to
tell me what kind of error and why it was non-recoverable, etc. And
I'm not altogether comfortable with that. But I'll do it if you like.
Please let me know your thoughts as to whether I should, and more
generally re: what an alarm condition is supposed to mean.
>> +}
>> +
>> +/* scale_numeric_sensor - scale sensor reading for hwmon */
>> +static long scale_numeric_sensor(const struct hp_wmi_numeric_sensor *nsensor)
>> +{
>> + u32 current_reading = nsensor->current_reading;
>> + s32 unit_modifier = nsensor->unit_modifier;
>> + u32 sensor_type = nsensor->sensor_type;
>> + u32 base_units = nsensor->base_units;
>> + s32 target_modifier;
>> + long val;
>> +
>> + /* Fan readings are in RPM units; others are in milliunits. */
>> + target_modifier = sensor_type == HP_WMI_TYPE_AIR_FLOW ? 0 : -3;
>> +
>> + val = current_reading;
>> +
>> + for (; unit_modifier < target_modifier; unit_modifier++)
>> + val = DIV_ROUND_CLOSEST(val, 10);
>> +
>> + for (; unit_modifier > target_modifier; unit_modifier--) {
>> + if (val > LONG_MAX / 10) {
>> + val = LONG_MAX;
>> + break;
>> + }
>> + val *= 10;
>> + }
>> +
>> + if (sensor_type == HP_WMI_TYPE_TEMPERATURE) {
>> + switch (base_units) {
>> + case HP_WMI_UNITS_DEGREES_F:
>> + val -= 32 * MILLI;
>> + val = val <= LONG_MAX / 5 ? (val * 5) / 9 :
>> + (val / 9) * 5;
>> + break;
>> +
>> + case HP_WMI_UNITS_DEGREES_K:
>> + val = milli_kelvin_to_millicelsius(val);
>> + break;
>> + }
>> + }
>> +
>> + return val;
>> +}
>> +
>> +/*
>> + * classify_numeric_sensor - classify a numeric sensor
>> + * @nsensor: pointer to numeric sensor struct
>> + *
>> + * Returns an enum hp_wmi_type value on success,
>> + * or a negative value if the sensor type is unsupported.
>> + */
>> +static int classify_numeric_sensor(const struct hp_wmi_numeric_sensor *nsensor)
>> +{
>> + u32 sensor_type = nsensor->sensor_type;
>> + u32 base_units = nsensor->base_units;
>> +
>> + switch (sensor_type) {
>> + case HP_WMI_TYPE_TEMPERATURE:
>> + if (base_units == HP_WMI_UNITS_DEGREES_C ||
>> + base_units == HP_WMI_UNITS_DEGREES_F ||
>> + base_units == HP_WMI_UNITS_DEGREES_K)
>> + return HP_WMI_TYPE_TEMPERATURE;
>> + break;
>> +
>> + case HP_WMI_TYPE_VOLTAGE:
>> + if (base_units == HP_WMI_UNITS_VOLTS)
>> + return HP_WMI_TYPE_VOLTAGE;
>> + break;
>> +
>> + case HP_WMI_TYPE_CURRENT:
>> + if (base_units == HP_WMI_UNITS_AMPS)
>> + return HP_WMI_TYPE_CURRENT;
>> + break;
>> +
>> + case HP_WMI_TYPE_AIR_FLOW:
>> + if (base_units == HP_WMI_UNITS_RPM)
>> + return HP_WMI_TYPE_AIR_FLOW;
>
> I really don't know what those sensors are, but this just seems wrong.
> Why would the BIOS report fan speeds as "Air Flow" and not as
> "Tachometer" like every other system ? If this ever reports anything but
> RPM (like CFM, for which there is a define), this is obviously not a fan
> but an air flow sensor, and it should not be reported as fan speed
> even if it reports the air flow in RPM (whatever that would mean).
>
I'll add a comment here that this is how HP does fan RPM sensors.
I can't rule out that HP also does Tachometer fan RPM sensors, but
I'm inclined to wait for sightings in the wild before accommodating
them.
In the same vein, I can't rule out the existence of Air Flow CFM
sensors, but hwmon wouldn't support such sensors anyway, and the code
would remain as is to disambiguate based on units.
>> + break;
>> + }
>> +
>> + return -EINVAL;
>> +}
>> +
>> +static int
>> +populate_numeric_sensor_from_wobj(struct device *dev,
>> + struct hp_wmi_numeric_sensor *nsensor,
>> + const union acpi_object *wobj)
>> +{
>> + const union acpi_object *element;
>> + const char **possible_states;
>> + u8 possible_states_count;
>> + acpi_object_type type;
>> + const char *string;
>> + u32 value;
>> + int prop;
>> + int err;
>> +
>> + err = check_wobj(wobj, &possible_states_count);
>> + if (err)
>> + return err;
>> +
>> + possible_states = devm_kcalloc(dev, possible_states_count,
>> + sizeof(*possible_states),
>> + GFP_KERNEL);
>> + if (!possible_states)
>> + return -ENOMEM;
>> +
>> + element = wobj->package.elements;
>> + nsensor->possible_states = possible_states;
>> + nsensor->possible_states_count = possible_states_count;
>> +
>> + for (prop = 0; prop <= HP_WMI_PROPERTY_CURRENT_READING; prop++) {
>> + type = hp_wmi_property_map[prop];
>> +
>> + switch (type) {
>> + case ACPI_TYPE_INTEGER:
>> + value = element->integer.value;
>> + break;
>> +
>> + case ACPI_TYPE_STRING:
>> + string = hp_wmi_strdup(dev, element->string.pointer,
>> + element->string.length);
>> + if (!string)
>> + return -ENOMEM;
>> + break;
>> +
>> + default:
>> + return -EINVAL;
>> + }
>> +
>> + element++;
>> +
>> + switch (prop) {
>> + case HP_WMI_PROPERTY_NAME:
>> + nsensor->name = string;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_DESCRIPTION:
>> + nsensor->description = string;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_SENSOR_TYPE:
>> + if (value > HP_WMI_TYPE_AIR_FLOW)
>> + return -EINVAL;
>> +
>> + nsensor->sensor_type = value;
>> +
>> + /* Skip OtherSensorType if it will be meaningless. */
>> + if (value != HP_WMI_TYPE_OTHER) {
>> + element++;
>> + prop++;
>> + }
>> +
>> + break;
>> +
>> + case HP_WMI_PROPERTY_OTHER_SENSOR_TYPE:
>> + nsensor->other_sensor_type = string;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_OPERATIONAL_STATUS:
>> + nsensor->operational_status = value;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_STATE:
>> + nsensor->current_state = string;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_POSSIBLE_STATES:
>> + *possible_states++ = string;
>> + if (--possible_states_count)
>> + prop--;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_BASE_UNITS:
>> + nsensor->base_units = value;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_UNIT_MODIFIER:
>> + /* UnitModifier is signed. */
>> + nsensor->unit_modifier = (s32)value;
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_READING:
>> + nsensor->current_reading = value;
>> + break;
>> +
>> + default:
>> + return -EINVAL;
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +/* update_numeric_sensor_from_wobj - update fungible sensor properties */
>> +static void
>> +update_numeric_sensor_from_wobj(struct device *dev,
>> + struct hp_wmi_numeric_sensor *nsensor,
>> + const union acpi_object *wobj)
>> +{
>> + const union acpi_object *elements;
>> + const union acpi_object *element;
>> + const char *string;
>> + u32 length;
>> + u8 offset;
>> +
>> + elements = wobj->package.elements;
>> +
>> + element = &elements[HP_WMI_PROPERTY_OPERATIONAL_STATUS];
>> + nsensor->operational_status = element->integer.value;
>> +
>> + element = &elements[HP_WMI_PROPERTY_CURRENT_STATE];
>> + string = element->string.pointer;
>> +
>> + if (strcmp(string, nsensor->current_state)) {
>> + length = element->string.length;
>> + devm_kfree(dev, nsensor->current_state);
>> + nsensor->current_state = hp_wmi_strdup(dev, string, length);
>> + }
>> +
>> + /* Offset reads into the elements array after PossibleStates[0]. */
>> + offset = nsensor->possible_states_count - 1;
>> +
>> + element = &elements[HP_WMI_PROPERTY_UNIT_MODIFIER + offset];
>> + nsensor->unit_modifier = (s32)element->integer.value;
>> +
>> + element = &elements[HP_WMI_PROPERTY_CURRENT_READING + offset];
>> + nsensor->current_reading = element->integer.value;
>> +}
>> +
>> +/*
>> + * interpret_info - interpret sensor for hwmon
>> + * @info: pointer to sensor info struct
>> + *
>> + * Should be called after the numeric sensor member has been updated.
>> + */
>> +static void interpret_info(struct hp_wmi_info *info)
>> +{
>> + const struct hp_wmi_numeric_sensor *nsensor = &info->nsensor;
>> +
>> + info->cached_val = scale_numeric_sensor(nsensor);
>> + info->last_updated = jiffies;
>> +}
>> +
>> +/*
>> + * hp_wmi_update_info - poll WMI to update sensor info
>> + * @state: pointer to driver state
>> + * @info: pointer to sensor info struct
>> + *
>> + * Returns 0 on success, or a negative error code on error.
>> + */
>> +static int hp_wmi_update_info(struct hp_wmi_sensors *state,
>> + struct hp_wmi_info *info)
>> +{
>> + struct hp_wmi_numeric_sensor *nsensor = &info->nsensor;
>> + struct device *dev = &state->wdev->dev;
>> + const union acpi_object *wobj;
>> + u8 instance = info->instance;
>> + int ret = 0;
>> +
>> + if (time_after(jiffies, info->last_updated + HZ)) {
>> + wobj = hp_wmi_get_wobj(state, instance);
>> + if (!wobj) {
>> + ret = -EIO;
>> + goto out_free_wobj;
>> + }
>> +
>> + update_numeric_sensor_from_wobj(dev, nsensor, wobj);
>> +
>> + interpret_info(info);
>> +
>> +out_free_wobj:
>> + kfree(wobj);
>> + }
>> +
>> + return ret;
>> +}
>> +
>> +#if CONFIG_DEBUG_FS
>> +
>> +static int basic_string_show(struct seq_file *seqf, void *ignored)
>> +{
>> + const char *str = seqf->private;
>> +
>> + seq_printf(seqf, "%s\n", str);
>> +
>> + return 0;
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(basic_string);
>> +
>> +static int fungible_show(struct seq_file *seqf, enum hp_wmi_property prop)
>> +{
>> + struct hp_wmi_numeric_sensor *nsensor;
>> + struct hp_wmi_sensors *state;
>> + struct hp_wmi_info *info;
>> + int err;
>> +
>> + switch (prop) {
>> + case HP_WMI_PROPERTY_OPERATIONAL_STATUS:
>> + nsensor = container_of(seqf->private,
>> + struct hp_wmi_numeric_sensor,
>> + operational_status);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_STATE:
>> + nsensor = container_of(seqf->private,
>> + struct hp_wmi_numeric_sensor,
>> + current_state);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_UNIT_MODIFIER:
>> + nsensor = container_of(seqf->private,
>> + struct hp_wmi_numeric_sensor,
>> + unit_modifier);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_READING:
>> + nsensor = container_of(seqf->private,
>> + struct hp_wmi_numeric_sensor,
>> + current_reading);
>> + break;
>> +
>> + default:
>> + return -EOPNOTSUPP;
>> + }
>> +
>> + info = container_of(nsensor, struct hp_wmi_info, nsensor);
>> + state = container_of(info, struct hp_wmi_sensors, info[info->instance]);
>> +
>> + mutex_lock(&state->lock);
>> +
>> + err = hp_wmi_update_info(state, info);
>> +
>> + mutex_unlock(&state->lock);
>> +
>> + if (err)
>> + return err;
>> +
>> + switch (prop) {
>> + case HP_WMI_PROPERTY_OPERATIONAL_STATUS:
>> + seq_printf(seqf, "%u\n", nsensor->operational_status);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_STATE:
>> + seq_printf(seqf, "%s\n", nsensor->current_state);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_UNIT_MODIFIER:
>> + seq_printf(seqf, "%d\n", nsensor->unit_modifier);
>> + break;
>> +
>> + case HP_WMI_PROPERTY_CURRENT_READING:
>> + seq_printf(seqf, "%u\n", nsensor->current_reading);
>> + break;
>> +
>> + default:
>> + break;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int operational_status_show(struct seq_file *seqf, void *ignored)
>> +{
>> + return fungible_show(seqf, HP_WMI_PROPERTY_OPERATIONAL_STATUS);
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(operational_status);
>> +
>> +static int current_state_show(struct seq_file *seqf, void *ignored)
>> +{
>> + return fungible_show(seqf, HP_WMI_PROPERTY_CURRENT_STATE);
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(current_state);
>> +
>> +static int possible_states_show(struct seq_file *seqf, void *ignored)
>> +{
>> + struct hp_wmi_numeric_sensor *nsensor = seqf->private;
>> + u8 i;
>> +
>> + for (i = 0; i < nsensor->possible_states_count; i++)
>> + seq_printf(seqf, "%s\n", nsensor->possible_states[i]);
>> +
>> + return 0;
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(possible_states);
>> +
>> +static int unit_modifier_show(struct seq_file *seqf, void *ignored)
>> +{
>> + return fungible_show(seqf, HP_WMI_PROPERTY_UNIT_MODIFIER);
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(unit_modifier);
>> +
>> +static int current_reading_show(struct seq_file *seqf, void *ignored)
>> +{
>> + return fungible_show(seqf, HP_WMI_PROPERTY_CURRENT_READING);
>> +}
>> +DEFINE_SHOW_ATTRIBUTE(current_reading);
>> +
>> +/* hp_wmi_devm_debugfs_remove - devm callback for debugfs cleanup */
>> +static void hp_wmi_devm_debugfs_remove(void *res)
>> +{
>> + debugfs_remove_recursive(res);
>> +}
>> +
>> +/* hp_wmi_debugfs_init - create and populate debugfs directory tree */
>> +static void hp_wmi_debugfs_init(struct hp_wmi_sensors *state)
>> +{
>> + struct device *dev = &state->wdev->dev;
>> + struct hp_wmi_info *info = state->info;
>> + struct hp_wmi_numeric_sensor *nsensor;
>> + char buf[HP_WMI_MAX_STR_SIZE];
>> + struct dentry *dir;
>> + int err;
>> + u8 i;
>> +
>> + /* dev_name() gives a not-very-friendly GUID for WMI devices. */
>> + scnprintf(buf, sizeof(buf), "%s-%u", "hp-wmi-sensors", dev->id);
>> +
>
> Why the first "%s" instead of "hp-wmi-sensors-%u" ?
I'll change this.
>
>> + state->debugfs = debugfs_create_dir(buf, NULL);
>
> This isn't used outside this function, and keeping a pointer to it in
> 'state' is therefore pointless.
>
Good point. I'll remove it.
>> + if (IS_ERR(state->debugfs))
>> + return;
>> +
>> + err = devm_add_action(dev, hp_wmi_devm_debugfs_remove, state->debugfs);
>> + if (err) {
>> + debugfs_remove(state->debugfs);
>> + return;
>> + }
>> +
>> + for (i = 0; i < state->count; i++, info++) {
>> + nsensor = &info->nsensor;
>> +
>> + scnprintf(buf, sizeof(buf), "%u", i);
>> + dir = debugfs_create_dir(buf, state->debugfs);
>> +
>> + debugfs_create_file("name", 0444, dir,
>> + (void *)nsensor->name,
>> + &basic_string_fops);
>> +
>> + debugfs_create_file("description", 0444, dir,
>> + (void *)nsensor->description,
>> + &basic_string_fops);
>> +
>> + debugfs_create_u32("sensor_type", 0444, dir,
>> + &nsensor->sensor_type);
>> +
>> + debugfs_create_file("other_sensor_type", 0444, dir,
>> + (void *)nsensor->other_sensor_type,
>> + &basic_string_fops);
>> +
>> + debugfs_create_file("operational_status", 0444, dir,
>> + &nsensor->operational_status,
>> + &operational_status_fops);
>> +
>> + debugfs_create_file("current_state", 0444, dir,
>> + (void *)&nsensor->current_state,
>> + ¤t_state_fops);
>> +
>> + debugfs_create_file("possible_states", 0444, dir,
>> + nsensor, &possible_states_fops);
>> +
>> + debugfs_create_u32("base_units", 0444, dir,
>> + &nsensor->base_units);
>> +
>> + debugfs_create_file("unit_modifier", 0444, dir,
>> + &nsensor->unit_modifier,
>> + &unit_modifier_fops);
>> +
>> + debugfs_create_file("current_reading", 0444, dir,
>> + &nsensor->current_reading,
>> + ¤t_reading_fops);
>> + }
>> +}
>> +
>> +#else
>> +
>> +static void hp_wmi_debugfs_init(struct hp_wmi_sensors *state)
>> +{
>> +}
>> +
>> +#endif
>> +
>> +static umode_t hp_wmi_hwmon_is_visible(const void *drvdata,
>> + enum hwmon_sensor_types type,
>> + u32 attr, int channel)
>> +{
>> + const struct hp_wmi_sensors *state = drvdata;
>> +
>> + if (!state->info_map[type] || !state->info_map[type][channel])
>> + return 0;
>> +
>> + return 0444;
>> +}
>> +
>> +static int hp_wmi_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
>> + u32 attr, int channel, long *val)
>> +{
>> + struct hp_wmi_sensors *state = dev_get_drvdata(dev);
>> + const struct hp_wmi_numeric_sensor *nsensor;
>> + struct hp_wmi_info *info;
>> + int err;
>> +
>> + info = state->info_map[type][channel];
>> + nsensor = &info->nsensor;
>> +
>> + mutex_lock(&state->lock);
>> +
>> + err = hp_wmi_update_info(state, info);
>> +
>> + mutex_unlock(&state->lock);
>> +
>> + if (err)
>> + return err;
>> +
>> + else if ((type == hwmon_temp && attr == hwmon_temp_fault) ||
>> + (type == hwmon_fan && attr == hwmon_fan_fault))
>> + *val = numeric_sensor_has_fault(nsensor);
>> +
>
> else after return is unnecessary, confusing, and static analyzers
> love complaining about it.
>
I'll rewrite this to stop doing else after return.
>> + else
>> + *val = info->cached_val;
>> +
>> + return 0;
>> +}
>> +
>> +static int hp_wmi_hwmon_read_string(struct device *dev,
>> + enum hwmon_sensor_types type, u32 attr,
>> + int channel, const char **str)
>> +{
>> + const struct hp_wmi_sensors *state = dev_get_drvdata(dev);
>> + const struct hp_wmi_info *info;
>> +
>> + info = state->info_map[type][channel];
>> + *str = info->nsensor.name;
>> +
>> + return 0;
>> +}
>> +
>> +static int add_channel_info(struct device *dev,
>> + struct hwmon_channel_info *channel_info,
>> + u8 count, enum hwmon_sensor_types type)
>> +{
>> + u32 attr = hp_wmi_hwmon_attributes[type];
>> + u32 *config;
>> +
>> + config = devm_kcalloc(dev, count + 1, sizeof(*config), GFP_KERNEL);
>> + if (!config)
>> + return -ENOMEM;
>> +
>> + channel_info->type = type;
>> + channel_info->config = config;
>> + memset32(config, attr, count);
>> +
>> + return 0;
>> +}
>> +
>> +static const struct hwmon_ops hp_wmi_hwmon_ops = {
>> + .is_visible = hp_wmi_hwmon_is_visible,
>> + .read = hp_wmi_hwmon_read,
>> + .read_string = hp_wmi_hwmon_read_string,
>> +};
>> +
>> +static struct hwmon_chip_info hp_wmi_chip_info = {
>> + .ops = &hp_wmi_hwmon_ops,
>> + .info = NULL,
>> +};
>> +
>> +static int hp_wmi_sensors_init(struct hp_wmi_sensors *state)
>> +{
>> + struct hp_wmi_info *active_info[HP_WMI_MAX_INSTANCES];
>> + const struct hwmon_channel_info **ptr_channel_info;
>> + struct hwmon_channel_info *channel_info;
>> + struct device *dev = &state->wdev->dev;
>> + struct hp_wmi_info *info = state->info;
>> + struct hp_wmi_numeric_sensor *nsensor;
>> + u8 channel_count[hwmon_max] = {};
>> + enum hwmon_sensor_types type;
>> + union acpi_object *wobj;
>> + struct device *hwdev;
>> + u8 type_count = 0;
>> + u8 channel;
>> + int wtype;
>> + int err;
>> + u8 i;
>> +
>> + for (i = 0, channel = 0; i < HP_WMI_MAX_INSTANCES; i++, info++) {
>> + wobj = hp_wmi_get_wobj(state, i);
>> + if (!wobj)
>> + break;
>> +
>> + info->instance = i;
>> + nsensor = &info->nsensor;
>> +
>> + err = populate_numeric_sensor_from_wobj(dev, nsensor, wobj);
>> + if (err)
>> + goto out_free_wobj;
>> +
>> + if (numeric_sensor_has_fault(nsensor))
>> + goto out_free_wobj;
>> +
>> + wtype = classify_numeric_sensor(nsensor);
>> + if (wtype < 0)
>> + goto out_free_wobj;
>> +
>> + type = hp_wmi_hwmon_type_map[wtype];
>> + if (!channel_count[type])
>> + type_count++;
>> + channel_count[type]++;
>> +
>> + info->is_active = true;
>> + info->type = type;
>> +
>> + interpret_info(info);
>> +
>> + active_info[channel++] = info;
>> +
>> +out_free_wobj:
>> + kfree(wobj);
>> +
>> + if (err)
>> + return err;
>> + }
>> +
>> + dev_dbg(dev, "Found %u sensors (%u active, %u types)\n",
>> + i, channel, type_count);
>> +
>> + state->count = i;
>> + if (!state->count)
>> + return -ENODATA;
>> +
>> + hp_wmi_debugfs_init(state);
>> +
>> + if (!channel)
>> + return 0; /* Not an error, but debugfs only. */
>> +
>> + if (channel_count[hwmon_temp]) {
>> + channel_count[hwmon_chip] = 1;
>> + type_count++;
>> + }
>> +
>> + memcpy(state->channel_count, channel_count, sizeof(channel_count));
>> +
>> + channel_info = devm_kcalloc(dev, type_count,
>> + sizeof(*channel_info),
>> + GFP_KERNEL);
>> + if (!channel_info)
>> + return -ENOMEM;
>> +
>> + ptr_channel_info = devm_kcalloc(dev, type_count + 1,
>> + sizeof(*ptr_channel_info),
>> + GFP_KERNEL);
>> + if (!ptr_channel_info)
>> + return -ENOMEM;
>> +
>> + hp_wmi_chip_info.info = ptr_channel_info;
>> +
>> + channel_info += type_count - 1;
>> + ptr_channel_info += type_count - 1;
>> +
>> + for (type = hwmon_max; type > hwmon_chip;) {
>> + if (!channel_count[--type])
>> + continue;
>> +
>> + err = add_channel_info(dev, channel_info,
>> + channel_count[type], type);
>> + if (err)
>> + return err;
>> +
>> + *ptr_channel_info-- = channel_info--;
>> +
>> + state->info_map[type] = devm_kcalloc(dev, channel_count[type],
>> + sizeof(*state->info_map),
>> + GFP_KERNEL);
>> + if (!state->info_map[type])
>> + return -ENOMEM;
>> + }
>> +
>> + while (channel > 0) {
>> + type = active_info[--channel]->type;
>> + i = --channel_count[type];
>> + state->info_map[type][i] = active_info[channel];
>> + }
>> +
>> + hwdev = devm_hwmon_device_register_with_info(dev, "hp_wmi_sensors",
>> + state, &hp_wmi_chip_info,
>> + NULL);
>> + return PTR_ERR_OR_ZERO(hwdev);
>> +}
>> +
>> +static int hp_wmi_sensors_probe(struct wmi_device *wdev, const void *context)
>> +{
>> + struct device *dev = &wdev->dev;
>> + struct hp_wmi_sensors *state;
>> + int err;
>> +
>> + /* Sanity check. */
>> + if (!wmi_has_guid(HP_WMI_NUMERIC_SENSOR_GUID) ||
>> + !wmi_has_guid(HP_WMI_BIOS_GUID)) {
>
> As mentioned by others, HP_WMI_BIOS_GUID is not in the ID table,
> and checking for it is pointless. Since only HP_WMI_NUMERIC_SENSOR_GUID
> is in the ID table, the entire check is pointless. Yes, it appears that
> other drivers do similar checks, but the argument that it is ok to do
> something unnecessary or wrong because others do it is just like saying
> that it would be ok to <pick your crime> because others do it.
>
There's enough of that in the world already. It's out.
>> + err = -ENODEV;
>> + goto out_err;
>
> Please please please no goto to a return statement. It suggests that
> some cleanup is needed when that is not the case. That is just confusing.
> Yes, I know, some people like doing that, but that doesn't make it better.
>
I can see how it would. I'll remove such gotos.
>> + }
>> +
>> + state = devm_kzalloc(dev, sizeof(*state), GFP_KERNEL);
>> + if (!state) {
>> + err = -ENOMEM;
>> + goto out_err;
>> + }
>> +
>> + state->wdev = wdev;
>> +
>> + mutex_init(&state->lock);
>> +
>> + dev_set_drvdata(dev, state);
>> +
>> + err = hp_wmi_sensors_init(state);
>> +
>> +out_err:
>> + return err;
>> +}
>> +
>> +static const struct wmi_device_id hp_wmi_sensors_id_table[] = {
>> + { HP_WMI_NUMERIC_SENSOR_GUID, NULL },
>> + {},
>> +};
>> +
>> +static struct wmi_driver hp_wmi_sensors_driver = {
>> + .driver = { .name = "hp-wmi-sensors" },
>> + .id_table = hp_wmi_sensors_id_table,
>> + .probe = hp_wmi_sensors_probe,
>> +};
>> +module_wmi_driver(hp_wmi_sensors_driver);
>> +
>> +MODULE_AUTHOR("James Seo <james@...iv.tech>");
>> +MODULE_DESCRIPTION("HP WMI Sensors driver");
>> +MODULE_LICENSE("GPL");
>>
>> base-commit: 197b6b60ae7bc51dd0814953c562833143b292aa
>
Again, thanks for reviewing. I really do appreciate your time.
James
Powered by blists - more mailing lists