Some PMBus chips support multiple phases, and report telemetry such
as input current, output current, or temperature for each phase.
Add support for such chips to the PMBus core.
Start with a maximum of 8 phases per page, and assume that supported
sensors per phase are similar for all pages. Only support per-phase
telemetry attributes, no limits or alarms.
As part of this patch, set the initial page variable to 0xff to ensure
that the page is updated when the first page command is issued. Also
only issue page commands if the chip supports more than one page.
Cc: Vadim Pasternak <vadimp@mellanox.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
In preparation for multi-phase support, add 'phase' parameter to read_word
and set_page functions. Actual multi-phase support will be added in
a subsequent patch.
Cc: Vadim Pasternak <vadimp@mellanox.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Extend "vrm_version" with the type for Intel IMVP9 and AMD 6.25mV VID
modes.
Add calculation for those types.
Signed-off-by: Vadim Pasternak <vadimp@mellanox.com>
Link: https://lore.kernel.org/r/20200113150841.17670-3-vadimp@mellanox.com
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Add support for VID protocol detection per page bases, instead of
detecting it based on "PMBU_VOUT" readout from page 0 for all the pages
supported by particular device.
The reason that some devices allows to configure different VID modes
per page within the same device.
Patch modifies the field "vrm_version" within the structure
"pmbus_driver_info" to be per page array.
Signed-off-by: Vadim Pasternak <vadimp@mellanox.com>
Link: https://lore.kernel.org/r/20200113150841.17670-2-vadimp@mellanox.com
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
If a chip is write protected, we can not change any limits, and we can
not clear status flags. This may be the reason why clearing status flags
is reported to not work for some chips. Detect the condition in the pmbus
core. If the chip is write protected, set limit attributes as read-only,
and set the flag indicating that the status flag should be ignored.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Some chips have attributes which exist on more than one page but the
attribute is not presently marked as paged. This causes the attributes
to be generated with the same label, which makes it impossible for
userspace to tell them apart.
Marking all such attributes as paged would result in the page suffix
being added regardless of whether they were present on more than one
page or not, which might break existing setups. Therefore, we add a
second check which treats the attribute as paged, even if not marked as
such, if it is present on multiple pages.
Fixes: b4ce237b7f ("hwmon: (pmbus) Introduce infrastructure to detect sensors and limit registers")
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
update_lock is a mutex intended to protect write operations. It was not
taken, however, when _pmbus_write_word_data is called from
pmbus_set_samples() function which may cause problems especially when
some PMBUS_VIRT_* operation is implemented as a read-modify-write cycle.
This patch makes sure the lock is held during the operation.
Fixes: 49c4455dcc ("hwmon: (pmbus) Introduce PMBUS_VIRT_*_SAMPLES registers")
Signed-off-by: Krzysztof Adamski <krzysztof.adamski@nokia.com>
Reviewed-by: Alexander Sverdlin <alexander.sverdlin@nokia.com>
[groeck: Declared and initialized missing 'data' variable]
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Based on 1 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details you
should have received a copy of the gnu general public license along
with this program if not write to the free software foundation inc
675 mass ave cambridge ma 02139 usa
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 441 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Reviewed-by: Richard Fontana <rfontana@redhat.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190520071858.739733335@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch makes it possible to pass custom struct attribute_group array
via the pmbus_driver_info struct so that those can be added to the
attribute groups passed to hwmon_device_register_with_groups().
This makes it possible to register custom sysfs attributes by PMBUS
drivers similar to how you can do this with most other busses/classes.
Signed-off-by: Krzysztof Adamski <krzysztof.adamski@nokia.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Those virtual registers can be used to export manufacturer specific
functionality for controlling the number of samples for average values
reported by the device.
Signed-off-by: Krzysztof Adamski <krzysztof.adamski@nokia.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Replace S_<PERMS> with octal values.
The conversion was done automatically with coccinelle. The semantic patches
and the scripts used to generate this commit log are available at
https://github.com/groeck/coccinelle-patches/.
This patch does not introduce functional changes. It was verified by
compiling the old and new files and comparing text and data sizes.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Devices with compatible="pmbus" field have zero initial page count,
and pmbus_clear_faults() being called before the page count auto-
detection does not actually clear faults because it depends on the
page count. Non-cleared faults in its turn may fail the subsequent
page count auto-detection.
This patch fixes this problem by calling pmbus_clear_fault_page()
for currently set page and calling pmbus_clear_faults() after the
page count was detected.
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Bazhenov <bazhenov.dn@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Pmbus client drivers, if they want to use debugfs, should use the same
root directory as the pmbus debugfs entries are using. Therefore, export
the device dentry for the pmbus client.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Some circumstances call for virtual pages, to expose multiple values
packed into an extended PMBus register in a manner non-compliant with
the PMBus standard. An example of this is the Maxim MAX31785 controller,
which extends the READ_FAN_SPEED_1 PMBus register from two to four bytes
to support tach readings for both rotors of a dual rotor fan. This extended
register contains two word-sized values, one reporting the rate of the
fastest rotor, the other the rate of the slowest. The concept of virtual
pages aids this situation by mapping the page number onto the value to be
selected from the vectored result.
We should not try to set virtual pages on the device as such a page
explicitly doesn't exist; add a flag so we can avoid doing so.
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Expose fanX_target, pwmX and pwmX_enable hwmon sysfs attributes.
Fans in a PMBus device are driven by the configuration of two registers,
FAN_CONFIG_x_y and FAN_COMMAND_x: FAN_CONFIG_x_y dictates how the fan
and the tacho operate (if installed), while FAN_COMMAND_x sets the
desired fan rate. The unit of FAN_COMMAND_x is dependent on the
operational fan mode, RPM or PWM percent duty, as determined by the
corresponding configuration in FAN_CONFIG_x_y.
The mapping of fanX_target, pwmX and pwmX_enable onto FAN_CONFIG_x_y and
FAN_COMMAND_x is implemented with the addition of virtual registers to
facilitate the necessary side-effects of each access:
1. PMBUS_VIRT_FAN_TARGET_x
2. PMBUS_VIRT_PWM_x
3. PMBUS_VIRT_PWM_ENABLE_x
Some complexity arises with the fanX_target and pwmX attributes both mapping
onto FAN_COMMAND_x: There is no general mapping between PWM percent duty and
RPM, so we can't display values in either attribute in terms of the other
(which in my mind is the intuitive, if impossible, behaviour). This problem
also affects the pwmX_enable attribute which allows userspace to switch between
full speed, manual PWM and a number of automatic control modes, possibly
including a switch to RPM behaviour (e.g. automatically adjusting PWM duty to
reach a RPM target, the behaviour of fanX_target).
The next most intuitive behaviour is for fanX_target and pwmX to simply be
independent, to retain their most recently set value even if that value is not
active on the hardware (due to switching to the alternative control mode). This
property of retaining the value independent of the hardware state has useful
results for both userspace and the kernel: Userspace always sees a sensible
value in the attribute (the last thing it was set to, as opposed to 0 or
receiving an error on read), and the kernel can use the attributes as a value
cache. This latter point eases the implementation of pwmX_enable, which can
look up the associated pmbus_sensor object, take its cached value and apply it
to hardware on changing control mode. This ensures we will not arbitrarily set
a PWM value as an RPM value or vice versa, and we can assume that the RPM or
PWM value set was sensible at least at some point in the past.
Finally, the DIRECT mode coefficients of some controllers is different between
RPM and PWM percent duty control modes, so PSC_PWM is introduced to capture the
necessary coefficients. As pmbus core had no PWM support previously PSC_FAN
continues to be used to capture the RPM DIRECT coefficients, but in order to
avoid falsely applying RPM scaling to PWM values I have introduced the
PMBUS_HAVE_PWM12 and PMB_BUS_HAVE_PWM34 feature bits. These feature bits allow
drivers to explicitly declare PWM support in order to have the attributes
exposed.
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Power values in the 100s of watt range can easily blow past
32bit math limits when processing everything in microwatts.
Use 64bit math instead to avoid these issues on common 32bit ARM
BMC platforms.
Fixes: 442aba7872 ("hwmon: PMBus device driver")
Signed-off-by: Robert Lippert <rlippert@google.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
The pmbus core may call read/write word data functions with a page value
of -1, intending to perform the operation without setting the page.
However, the read/write word data functions accept only unsigned 8-bit
page numbers, and therefore cannot check for negative page number to
avoid setting the page. This results in setting the page number to 0xFF.
This may result in errors or undefined behavior of some devices
(specifically the ir35221, which allows the page to be set to 0xFF,
but some subsequent operations to read registers may fail).
Switch the pmbus_set_page page parameter to an integer and perform the
check for negative page there. Make read/write functions consistent in
accepting an integer page number parameter.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Fixes: cbcdec6202 ("hwmon: (pmbus): Access word data for STATUS_WORD")
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
The below lists of VOUT_MODE command readout with their related VID
protocols, Digital to Analog Converter steps:
- VR13.0 mode, 10-mV DAC - 0x24
- VR13.0 mode, 5-mV DAC - 0x27
Signed-off-by: Vadim Pasternak <vadimp@mellanox.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Export all the available status registers through debugfs. This is
useful for hardware diagnostics, especially on multi-page pmbus devices,
as user-space access of the i2c space could corrupt the pmbus page
accounting.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Add PB_STATUS_INPUT as the generic alarm bit for iin and pin. We also
need to redo the status register checking before setting up the boolean
attribute, since it won't necessarily check STATUS_WORD if the device
doesn't support it, which we need for this bit.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Pmbus always reads byte data from the status register, even if
configured to use STATUS_WORD. Use a function pointer to read the
correct amount of data from the registers.
Also switch to try STATUS_WORD first before STATUS_BYTE on init.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Switch the storage of status registers to 16 bit values. This allows us
to store all the bits of STATUS_WORD.
Signed-off-by: Edward A. James <eajames@us.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
PMBus controllers optionally support PEC. Configure the driver
to use it if available to improve operational security.
Suggested-by: Michael Jones <mike@proclivis.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
pmbus_regulator_ops is not modified after initialized, so make it const.
Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Add support for simple on/off control of each channel.
To add regulator support, the pmbus part driver needs to add
regulator_desc information and number of regulators to its
pmbus_driver_info struct.
regulator_desc can be declared using default macro for a
regulator (PMBUS_REGULATOR) that is in pmbus.h
The regulator_init_data can be initialized from either
platform data or the device tree.
Signed-off-by: Alan Tull <atull@opensource.altera.com>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Some chips use different exponents for sensors on different pages
or rails. Detect and store exponent per page to support this situation.
This fixes a problem with wrong voltages seen on UCD90120.
Reported-by: Soren Brinkmann <soren.brinkmann@xilinx.com>
Tested-by: Soren Brinkmann <soren.brinkmann@xilinx.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
pmbus code currently crashes the kernel if it detects an internal
implementation error. While the detected condition suggests that there
is a bug in the code, it is hardly fatal. Therefore, it should not
trigger a crash. Replace BUG() with WARN().
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Use the wrapper function for retrieving the platform data instead of
accessing dev->platform_data directly.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
If krealloc() returns NULL, it *doesn't* free the original. So any code
of the form 'foo = krealloc(foo, …);' is almost certainly a bug.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Rearranged some data structures, and merged some common functions.
Overall code and data size reduction by more than 900 bytes.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Not all PMBus devices support the byte status register at 0x78.
Try to use the word status register at 0x79 instead if that is the case.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
For PMBus chips, modifying one limit register may affect other limits.
Since limits are all cached in the PMBus core driver, related changes
are not reflected in reported limits.
Introduce function to clear the attribute cache. After calling this function,
the core pmbus driver re-reads all cached values.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Some PMBus chips support monitoring an additional non-standard voltage. While
this voltage can in many cases be supported by simulating an additional sensor
page, this does not work in all cases. Specifically, it is problematic if the
data format is linear and the voltage is reported in LINEAR11 format. Since
output voltages use LINEAR16, and the exponent for LINEAR16 data is chip-wide
and fixed, this can result in overflows.
To solve this problem, add support for an additional virtual input voltage,
call it 'vmon', and treat this voltage as input voltage (which, when the chip
supports linear data format, uses LINEAR11).
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
So far, attribute memory was allocated by pre-calculating the maximum possible
amount of attributes. Not only does this waste memory, it is also risky because
the calculation might be wrong. It also requires a lot of defines to specify
the maximum number of attributes per class.
Allocate attribute memory using krealloc() instead. That means we have to use
kfree(), since devm_krealloc() does not exist, but that is still less costly
and less risky than trying to predict the number of attributes at the beginning.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Since memory is now allocated with dev_ functions, we no longer need to keep
track of allocated memory. Sensor memory allocation can therefore be
simplified significantly.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Boolean handling depends on storing the sensor data index in sensor_device_attr
as part of the index variable. This limits the number of sensor attributes to
256, and means the sensor sequence number actually has to be maintained to be
able to access sensor data from boolean functions.
Rework the code to store sensor pointers in the pmbus_boolean data structure
directly. With this approach, the number of supportable sensors is now
unlimited.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Since memory is now allocated with dev_ functions, we no longer need to keep
track of allocated memory. Memory allocation for booleans and labels can
therefore be simplified substantially by allocating it only as needed.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Fix:
ERROR: Macros with multiple statements should be enclosed in a do - while loop
by unwinding the problematic macros.
As a side effect, this patch reduces code size on x86_64 by 160 bytes and bss
size by 64 bytes.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
SENSORS_LIMIT and the generic clamp_val have the same functionality,
and clamp_val is more efficient.
This patch reduces text size by 9052 bytes and bss size by 11624 bytes
for x86_64 builds.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: George Joseph <george.joseph@fairview5.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
These drivers use no sleep or delay functions so they don't need to
include <linux/delay.h>.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Cc: Rudolf Marek <r.marek@assembler.cz>
Many hwmon drivers use jiffies but omit the inclusion of the header
file. Fix that, and also fix one driver which was including the header
file but didn't need it.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Luca Tettamanti <kronos.it@gmail.com>
Cc: Marc Hulsman <m.hulsman@tudelft.nl>
Cc: Rudolf Marek <r.marek@assembler.cz>
Guenter Roeck