In former code (before the inclusion of the driver in mainline kernel),
hif_join() could run asynchronously. When a join request was in
progress, it was forbidden to launch hif_scan().

Now, hif_join() is always run synchronously. There is no more reasons to
keep a protection against this case.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-13-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Keys sent to the firmware are never reset. So, it is not necessary to
re-upload them after hif_reset(). Thus, it is no more necessary to keep
a copy of the keys in struct wfx_dev.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-12-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Currently code handle WEP keys manually. It is far easier to use
the set_default_unicast_key() callback provided by mac80211.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-11-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Basic Rates are already set by hif_join(). hif_join() is also able to
manage possible changes after association.

Firmware also allows to change the Basic Rates with
hif_set_association_mode() but it does not bring anything in our case.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-10-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The structure hif_req_set_bss_params come from hardware API. It is not
intended to be manipulated in upper layers of the driver.

In add, current code for hif_req_set_bss_params() is too dumb. It should
pack data with hardware representation instead of leaving all work to
the caller.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-9-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The field bss_params.operational_rate_set is ignore by the firmware. It
is not necessary to send it. This change greatly simplify
wfx_join_finalize(). It is no more necessary to get sta and it allow to
save a RCU.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-8-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Dual CTS is only necessary when sending/receiving STBC data. However,
the chip does not support STBC, so it is never necessary to enable
double CTS.

We can simplify the code.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-7-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Current code skip some configuration during joining an IBSS network.
Indeed, it seems that this configuration is not used in IBSS. However,
it would be harmless to set them. In add, we would prefer to keep
association processes for ad-hoc and managed networks the closest as
possible. It also ensures the values of internal parameters of the
firmware.

Therefore, apply them unconditionally.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-6-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Currently, events from firmware are handled in a work queue with a
complex event queue mechanism. It is probably overkill since there is
only two events to handle: bss_loss and CQM events. Handling these
events synchronously is sufficient.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-5-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Since wfx_bss_params_work() does not exist anymore, there is no more
reason to keep a copy of bss_params in struct wfx_dev. A local instance
in wfx_join_finalize() is sufficient.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-4-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The firmware is able to filter beacons and send a notification if one or
multiple beacons are not received. Note that it send this notification
only once. Only if it receive beacons gain, it send a new notification.

Currently, the driver handle the connection loss itself (see
wfx_cqm_bssloss_sm()). It send null frames and watch the answers.

This patch fixes all this mess:
  - settle firmware to send a notification on the first beacon loss
  - call ieee80211_beacon_loss() and let mac80211 handle all the process
  - since we do have notification for each beacon loss, add a period
    task that call ieee80211_beacon_loss() until we receive "REGAIN"
    notification.

Thus, we can drop the ugly wfx_cqm_bssloss_sm() and
wfx_bss_params_work().
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-3-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Firmware dislikes the driver enables PS when it is not yet associated.
The current check for that condition is more complex than necessary.
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Link: https://lore.kernel.org/r/20200420160311.57323-2-Jerome.Pouiller@silabs.comSigned-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'iio-for-5.8a' of git://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next

Jonathan writes:

First set of new IIO device support, features and cleanup for the 5.8 cycle

Usual mix of new drivers, new support in old drivers and lots of minor
cleanup. Yaml conversions also continue to trickle in (plenty more to go!)

New device support
* ad7476
  - Add ad7091 support (ID only)
* ad9467
  - New driver for this 200/250 MSPS adi-axi-adc and SPI attached ADC.
* adi-axi-adc
  - New driver to support ADIs generic AXI ADC IP core, used to provide
    high speed interfaces to ADCs (JESD204B/C or parallel interfaces). Usually
    paired with a device using a slow configuration interface (spi etc)
    Includes DT bindings and some fixes for fpga headers.
* bmg160
  - Add support for BMI088 (ID only)
* max1241
  - New driver for this ADC.
* st_sensors
  - Add support for LIS2HH12 accelerometer
* sx9310
  - New driver supporting sx9310 and sx9311 proximity sensors.

Yaml DT binding conversions
* rockchip-saradc (including adding some missing parts)
* stm32-dac
* tsl2563
* vcnl4000

Features
* st_lsm6dsx
  - Add LIS3MDL as a possible sensor hub device.
* vcnl4000
  - Add new concept of near level (from DT) to provide to userspace which
    often needs to have some calibrated concept of 'near'.

Cleanups, minor fixes etc.
* core
  - Use snprintf for functions where strings are built and there is
    potential for overflow.
  - Correct docs to indicate mlock should not be used directly by drivers.
  - Fix up accidental dropping of a patch to use bitmap_zalloc.
  - Stop allowing enabling of buffers with no channels enabled.
  - Drop unused 'stufftoread' from iio_buffer.
  - Drop scan_el_attrs form iio_buffer as unused.
  - Reorder sanity checks in __iio_device_register to fail earlier.
  - Drop all the devm_ runregister / free functions from IIO as they
    were never used and encourage poor design.
* dma-buffer
  - Tidy up includes.
* dma-engine-buffer
  - Provide dev-managed allocator.
  - Fix an issue with printing a size_t
* cross subsystem (kxsd9, bmg160, mpu3050, bmi160, mpu6050, bmc150)
  - Replace some unnecessary casts of error ptrs and whilst there.
    use the %pe printf parameter to print them in a more useful fashion.
* cross subsystem
  - Drop casts in calls to regmap_bulk_read as they make no sense.
  - Use devm_platform_ioremap_resource to reduce boilerplate.
  - Fix typos in Analog Devices.
* counters/104-quad
  - Add Syed Nayyar Waris as an additional maintainer.
* ad7476
  - Generate CONVST signal internally rather than requiring external
    trigger.  Add sysfs read back as can now do so.
  - use devm_add_action_or_reset to tidy up error and remove handling.
* ad7793
  - Switch to read_avail from explicit attribute. Mostly done to avoid
    confusing people with a - sign (without surounding spaces) that
    was correct but checkpatch didn't like.
* adis library
  - Add missing newlines at end of error messages.
* adis16400
  - Use DEFINE_DEBUGS_ATTRIBUTE rather than DEFINE_SIMPLE_ATTR.
* adis16460
  - Use DEFINE_DEBUGS_ATTRIBUTE rather than DEFINE_SIMPLE_ATTR.
* ad_sigma_delta
  - Move some channel definitions into individual drivers to avoid
    having to deal with complex options.
* ak8974
  - Silence an error on deffered probe.
* bmp280
  - Harden against IRQ before registration.
  - Convert to read_avail instead of opencoding the equivalent.
  - Explicitly mark GPIO as optional.
  - Cleanup casts.
  - Remove line breaks from strings.
* htts221
  - Replace direct access to platform_data with dev_get_platdata().
  - Use device properties rather than device tree ones to allow ACPI
    probing.
  - Casting cleanups.
* intel_mrfld_adc
  - Don't use unaligned accessor for aligned data.
* isl29125
  - Reorder buer pre and post hooks to allow for coming core rework.
* ltc2983
  - Remove comp to bool.
* max1363
  - Stop using mlock from the iio_dev directly in favour of a local lock
    with clearly defined scope.
* max30100
  - Use generic device properties to allow ACPI probe.
* mpu6050
  - Convert to i2c_new_client_device.
  - Add debugfs register access.
* st_lsm6dsx
  - Provide means of configuring full scale on slave device if supported.
  - Drop include of st_sensors header to get one value. Its not otherwise
    used by this driver.
* st-sensors
  - Replace direct access to platform_data with dev_get_platdata().
  - Casting cleanups.
  - Avoid splitting strings.
* st_uvis25
  - Casting cleanups.
* tsl2563
  - Typo fix.
* tsl2772
  - scnprintf in a non obvious string building usecase. Note also 'fixes'
    a wrong calculation of remaining space that couldn't actually cause
    any trouble as there was lots of room.
* xilinx-xadc
  - Fix Lars-Peter spelling his own name wrong :) + additional typos.

* tag 'iio-for-5.8a' of git://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio: (101 commits)
  iio: magnetometer: ak8974: Silence deferred-probe error
  Documentation: ABI: document IIO in_proximity_nearlevel file
  iio: vcnl4000: Export near level property for proximity sensor
  dt-bindings: iio: light: vcnl4000: Add proximity-near-level
  dt-bindings: iio: Introduce common properties for iio sensors
  dt-bindings: iio: vcnl4000: convert bindings to YAML format
  iio: Fix misspellings of "Analog Devices"
  iio: light: isl29125: fix iio_triggered_buffer_{predisable,postenable} positions
  iio: adc: fsl-imx25-gcq: Use devm_platform_ioremap_resource
  iio: adc: at91-adc: Use devm_platform_ioremap_resource
  iio: adc: sun4i-gpadc-iio: Use devm_platform_ioremap_resource
  iio:light:ltr501: Drop unnecessary cast of parameter in regmap_bulk_read
  iio:magn:mmc35240: Drop unnecessary casts of val parameter in regmap_bulk*
  iio:imu:mpu6050: Tidy up parameters to regmap_bulk functions.
  iio:chemical:bme680: Tidy up parameters to regmap_bulk_read
  iio:chemical:atlas-sensor: Drop unnecessary explicit casts in regmap_bulk_read calls
  iio:accel:mxc4005: Drop unnecessary explicit casts in regmap_bulk_read calls
  iio: imu: st_lsm6dsx: drop huge include in sensor-hub driver
  iio: buffer: drop devm_iio_kfifo_free() API call
  iio: buffer: drop devm_iio_hw_consumer_free() API call
  ...

It's not uncommon that voltage regulator becomes available later during
kernel's boot process. This patch adds info message about unavailable
regulators in a case of the deferred-probe error and also amends the
error message with a error code.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

The vcnl4000 IIO driver introduced a new attribute
"in_proximity_nearlevel".  This adds it to the list of documented ABI
for sysfs-bus-iio.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

When an object can be considered close to the sensor is hardware
dependent. Allowing to configure the property via device tree
allows to configure this device specific value.

This is useful for e.g. iio-sensor-proxy to indicate to userspace
if an object is close to the sensor.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

This value indicates when userspace should consider an object
near to the sensor/device.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Introduce a file for common properties of iio sensors. So far this
contains the new proximity-near-level property for proximity sensors
that indicates when an object should be considered near.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Convert the vcnl4000 device tree bindings to the new YAML format.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

According to https://www.analog.com/, the company name is spelled
"Analog Devices".
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

The iio_triggered_buffer_{predisable,postenable} functions attach/detach
the poll functions.

For the predisable hook, the disable code should occur before detaching
the poll func, and for the postenable hook, the poll func should be
attached before the enable code.

This change reworks the predisable/postenable hooks so that the pollfunc is
attached/detached in the correct position.
It also balances the calls a bit, by grouping the preenable and the
iio_triggered_buffer_postenable() into a single
isl29125_buffer_postenable() function.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Use the helper function that wraps the calls to
platform_get_resource() and devm_ioremap_resource()
together. It reduces boilerplate and suggested by coccinelle.
Signed-off-by: Aishwarya Ramakrishnan <aishwaryarj100@gmail.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Use the helper function that wraps the calls to
platform_get_resource() and devm_ioremap_resource()
together. It reduces boilerplate and suggested by coccinelle.
Signed-off-by: Aishwarya Ramakrishnan <aishwaryarj100@gmail.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Use the helper function that wraps the calls to platform_get_resource()
and devm_ioremap_resource() together.  Reduces boilerplate and suggested
by coccinelle.
Signed-off-by: Aishwarya Ramakrishnan <aishwaryarj100@gmail.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

This only occurs once in the driver and isn't needed in this case either,
so drop it.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>

Given the parameter is void *, casting to u8 * is unnecessary.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Daniel Baluta <daniel.baluta@gmail.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>

A mixture of:
* Casts to u8 * for the val parameter which is void * and so needs no cast.
* Explicit sizes when we can use sizeof on the destination / source variable.

Not broken in anyway, I'm just trying to tidy up the is in the subsystem to
avoid it getting cut and paste into new drivers.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>

A mixture of:
* Unnecessary casts of val parameter to u8 * which makes little sense as
  the function take a void *.
* Explicit sizes where we can use the destination type to define that.
  Note that's not true in all cases as we do read 3 bytes into 4 byte
  destinations.

Note that noting was broken here, I'm just trying to ensure this doesn't
get cut and paste into more drivers so cleaning these out subsystem wide.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Himanshu Jha <himanshujha199640@gmail.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>

regmap_bulk_read takes a void * for its val parameter. It certainly
makes no sense to cast to a (u8 *) + no need to explicitly cast
at all when converting another pointer type to void *.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Matt Ranostay <matt.ranostay@konsulko.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>

regmap_bulk_read takes a void * for its val parameter. It certainly
makes no sense to cast to a (u8 *) + no need to explicitly cast
at all when converting another pointer type to void *.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>

st_lsm6dsx is a standalone driver for STM IMU sensors and does not rely
on st_sensor common framework, so it does not include st_sensor common
definitions.
In st_lsm6dsx_shub driver st_sensors.h is used just to introduce the
default wai address for LIS3MDL sensor.
Drop this largely unconnected include file and introduce the default wai
address for LIS3MDL in st_lsm6dsx_ext_dev_settings register map
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.

This is the last user of 'devm_iio_device_match()', so it can be removed as
well in this patch.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

It's unused so far, so it can be removed. Also makes sense to remove it
to discourage weird uses of this call during review.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Moves this to be the first check, as it's very simple and fails the
registration earlier, instead of potentially initializing the
'indio_dev->label' and checking for duplicate indexes, and then failing
with this simple-check.

This is a minor optimization, since '__iio_device_register()' will waste
fewer validation cycles in case 'indio_dev->info' is NULL.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>