Prepare input updates for 5.4 merge window.

Looks like the Bios of the Lenovo Legion Y7000 is using ELAN061B
when the actual device is supposed to be used with hid-multitouch.

Remove it from the list of the supported device, hoping that
no one will complain about the loss in functionality.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=203467
Fixes: 738c06d0 ("Input: elan_i2c - add hardware ID for multiple Lenovo laptops")
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Don't populate the array seq on the stack but instead make it
static const. Makes the object code smaller by 30 bytes.

Before:
   text	   data	    bss	    dec	    hex	filename
  22284	   3184	      0	  25468	   637c	drivers/input/joystick/sidewinder.o

After:
   text	   data	    bss	    dec	    hex	filename
  22158	   3280	      0	  25438	   635e	drivers/input/joystick/sidewinder.o

(gcc version 9.2.1, amd64)
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

We need to reset input device's timestamp on input_sync(), otherwise
drivers not using input_set_timestamp() will end up with a stale
timestamp after their clients consume first input event.

Fixes: 3b51c44b ("Input: allow drivers specify timestamp for input events")
Reported-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

This switches the driver over to the standard touchscreen properties for
coordinate transformation, while keeping old bindings working as well.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

MT-B protocol is more efficient and everyone expects it. We use in-kernel
tracking to identify contacts.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

If the touchscreen is configured as wakeup source we should not be cutting
off power to it.

Also, now that the driver relies on I2C client to supply IRQ, we do not
need to explicitly enable and disable IRQ for wakeup: if device is created
as wakeup source, I2C core will mark interrupt as wakeup one.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Instead of trying to map INT GPIO to interrupt, let's use one supplied by
I2C client. If there is none - bail. This will also allow us to treat INT
GPIO as optional, as per the binding.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

There are no current users of the platform data in the tree, and
any new users should either use device tree, or static device
properties to describe the device.

This change drop the platform data definition and handling and moves the
driver over to generic device properties API. We also drop support for the
external clock. If it is needed we will have to extend the bindings to
supply the clock reference and handle it properly in the driver.

Also, wakeup setting should be coming from I2C client.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

This allows trimming error unwinding and device removal handling.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

The comments for individual functions in the driver do not provide any
additional information beyond what function names indicate.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Instead if #ifdef-ing out suspend and resume methods, let's mark
them as __maybe_unused to get better compile time coverage.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

"bu21013_data" and "struct bu21013_ts_data" are a tad long, let's call them
"ts" and "struct bu21013_ts".

Also rename retval to error in bu21013_init_chip() and adjust formatting;
i2c_smbus_write_byte_data() returns negative on error and 0 on success, so
we simply test if whether erro is 0 or not.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

This driver can use GPIO descriptors rather than GPIO numbers
without any problems, convert it. Name the field variables after
the actual pins on the chip rather than the "reset" and "touch"
names from the devicetree bindings that are vaguely inaccurate.

No in-tree users pass GPIO numbers in platform data so drop
this. Descriptor tables can be used to get these GPIOs from a board
file if need be.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

In preparation to update to bu21013_tp driver properly annotate GPIOs
property (the INT GPIOs are active low, not open drain), and also define
interrupt lines so we do not have to have special conversion in the driver.
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

We don't know when the device will be added with device_add() in
serio_add_port() because serio_add_port() is called from a workqueue
that this driver schedules by calling serio_register_port(). The best we
can know is that the device will definitely not have been added yet when
the start callback is called on the serio device.

While it hasn't been shown to be a problem, proactively move the wakeup
enabling calls to the start hook so that we don't race with the
workqueue calling device_add(). This will avoid racy situations where
code tries to add wakeup sysfs attributes for this device from
dpm_sysfs_add() but the path in device_set_wakeup_capable() has already
done so.
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

The registration of gpio-keys device can be written much shorter
by using the platform_device_register_resndata() helper.
Signed-off-by: Enrico Weigelt <info@metux.net>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Recently we had a building error if we enable the MOUSE_PS2_ALPS while
disable the MOUSE_PS2_TRACKPOINT, and was fixed by 49e6979e
("Input: psmouse - fix build error of multiple definition").

We could improve that fix by dropping all unneeded functions and
CONFIG_MOUSE_ guards from the header, it is safe to do that since
those functions are not directly called by psmouse-base.c anymore.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Separating "normal" and "polled" input devices was a mistake, as often we
want to allow the very same device work on both interrupt-driven and
polled mode, depending on the board on which the device is used.

This introduces new APIs:

- input_setup_polling
- input_set_poll_interval
- input_set_min_poll_interval
- input_set_max_poll_interval

These new APIs allow switching an input device into polled mode with sysfs
attributes matching drivers using input_polled_dev APIs that will be
eventually removed.
Tested-by: Michal Vokáč <michal.vokac@ysoft.com>
Acked-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

GCC warns that the output of our call to 'snprintf' in 'w8001_connect'
may be truncated since both 'serio->phys' and 'w8001->phys' are 32 bytes
in length. Increase the amount of space allocated for the latter to
compensate.
Signed-off-by: Jason Gerecke <jason.gerecke@wacom.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

In the previous patch we didn't mask out event_type in case statement,
so switches are always picked instead of buttons, which results in
ChromeOS devices misbehaving when power button is pressed.
This patch adds back the missing mask.

Fixes: d096aa3e ("Input: cros_ec_keyb: mask out extra flags in event_type")
Signed-off-by: Fei Shao <fshao@chromium.org>
Reviewed-by: Nicolas Boichat <drinkcat@chromium.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

We don't need dev_err() messages when platform_get_irq() fails now that
platform_get_irq() prints an error message itself when something goes
wrong. Let's remove these prints with a simple semantic patch.

// <smpl>
@@
expression ret;
struct platform_device *E;
@@

ret =
(
platform_get_irq(E, ...)
|
platform_get_irq_byname(E, ...)
);

if ( \( ret < 0 \| ret <= 0 \) )
{
(
-if (ret != -EPROBE_DEFER)
-{ ...
-dev_err(...);
-... }
|
...
-dev_err(...);
)
...
}
// </smpl>

While we're here, remove braces on if statements that only have one
statement (manually).
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

When calling debugfs functions, there is no need to ever check the
return value.  The function can work or not, but the code logic should
never do something different based on this.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

When calling debugfs functions, there is no need to ever check the
return value.  The function can work or not, but the code logic should
never do something different based on this.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Use use device_property_count_u32() directly, that makes code neater.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Use use device_property_count_u32() directly, that makes code neater.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Use use device_property_count_u32() directly, that makes code neater.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Use use device_property_count_u32() directly, that makes code neater.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Drivers now have the option to have the driver core create and remove any
needed sysfs attribute files.  So take advantage of that and do not
register "by hand" a sysfs group of attributes.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Drivers now have the option to have the driver core create and remove any
needed sysfs attribute files.  So take advantage of that and do not
register "by hand" a bunch of sysfs files.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Merge tag 'dev_groups_all_drivers' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core into next

dev_groups added to struct driver

Persistent tag for others to pull this branch from

This is the first patch in a longer series that adds the ability for the
driver core to create and remove a list of attribute groups
automatically when the device is bound/unbound from a specific driver.

See:
	https://lore.kernel.org/r/20190731124349.4474-2-gregkh@linuxfoundation.org
for details on this patch, and examples of how to use it in other
drivers.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Sync up with mainline to bring in device_property_count_u32 andother
newer APIs.

Pull dax fixes from Dan Williams:
 "A filesystem-dax and device-dax fix for v5.3.

  The filesystem-dax fix is tagged for stable as the implementation has
  been mistakenly throwing away all cow pages on any truncate or hole
  punch operation as part of the solution to coordinate device-dma vs
  truncate to dax pages.

  The device-dax change fixes up a regression this cycle from the
  introduction of a common 'internal per-cpu-ref' implementation.

  Summary:

   - Fix dax_layout_busy_page() to not discard private cow pages of
     fs/dax private mappings.

   - Update the memremap_pages core to properly cleanup on behalf of
     internal reference-count users like device-dax"

* tag 'dax-fixes-5.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  mm/memremap: Fix reuse of pgmap instances with internal references
  dax: dax_layout_busy_page() should not unmap cow pages

Pull NTB fix from Jon Mason:
 "Bug fix for NTB MSI kernel compile warning"

* tag 'ntb-5.3-bugfixes' of git://github.com/jonmason/ntb:
  NTB/msi: remove incorrect MODULE defines

Pull RISC-V updates from Paul Walmsley:
 "A few minor RISC-V updates for v5.3-rc4:

   - Remove __udivdi3() from the 32-bit Linux port, converting the only
     upstream user to use do_div(), per Linux policy

   - Convert the RISC-V standard clocksource away from per-cpu data
     structures, since only one is used by Linux, even on a multi-CPU
     system

   - A set of DT binding updates that remove an obsolete text binding in
     favor of a YAML binding, fix a bogus compatible string in the
     schema (thus fixing a "make dtbs_check" warning), and clarifies the
     future values expected in one of the RISC-V CPU properties"

* tag 'riscv/for-v5.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  dt-bindings: riscv: fix the schema compatible string for the HiFive Unleashed board
  dt-bindings: riscv: remove obsolete cpus.txt
  RISC-V: Remove udivdi3
  riscv: delay: use do_div() instead of __udivdi3()
  dt-bindings: Update the riscv,isa string description
  RISC-V: Remove per cpu clocksource

Pull x86 fixes from Thomas Gleixner:
 "A few fixes for x86:

   - Don't reset the carefully adjusted build flags for the purgatory
     and remove the unwanted flags instead. The 'reset all' approach led
     to build fails under certain circumstances.

   - Unbreak CLANG build of the purgatory by avoiding the builtin
     memcpy/memset implementations.

   - Address missing prototype warnings by including the proper header

   - Fix yet more fall-through issues"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/lib/cpu: Address missing prototypes warning
  x86/purgatory: Use CFLAGS_REMOVE rather than reset KBUILD_CFLAGS
  x86/purgatory: Do not use __builtin_memcpy and __builtin_memset
  x86: mtrr: cyrix: Mark expected switch fall-through
  x86/ptrace: Mark expected switch fall-through

Pull perf tooling fixes from Thomas Gleixner:
 "Perf tooling fixes all over the place:

   - Fix the selection of the main thread COMM in db-export

   - Fix the disassemmbly display for BPF in annotate

   - Fix cpumap mask setup in perf ftrace when only one CPU is present

   - Add the missing 'cpu_clk_unhalted.core' event

   - Fix CPU 0 bindings in NUMA benchmarks

   - Fix the module size calculations for s390

   - Handle the gap between kernel end and module start on s390
     correctly

   - Build and typo fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf pmu-events: Fix missing "cpu_clk_unhalted.core" event
  perf annotate: Fix s390 gap between kernel end and module start
  perf record: Fix module size on s390
  perf tools: Fix include paths in ui directory
  perf tools: Fix a typo in a variable name in the Documentation Makefile
  perf cpumap: Fix writing to illegal memory in handling cpumap mask
  perf ftrace: Fix failure to set cpumask when only one cpu is present
  perf db-export: Fix thread__exec_comm()
  perf annotate: Fix printing of unaugmented disassembled instructions from BPF
  perf bench numa: Fix cpu0 binding

Pull scheduler fixes from Thomas Gleixner:
 "Three fixlets for the scheduler:

   - Avoid double bandwidth accounting in the push & pull code

   - Use a sane FIFO priority for the Pressure Stall Information (PSI)
     thread.

   - Avoid permission checks when setting the scheduler params for the
     PSI thread"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/psi: Do not require setsched permission from the trigger creator
  sched/psi: Reduce psimon FIFO priority
  sched/deadline: Fix double accounting of rq/running bw in push & pull

Pull irq fix from Thomas Gleixner:
 "A small fix for the affinity spreading code.

  It failed to handle situations where a single vector was requested
  either due to only one CPU being available or vector exhaustion
  causing only a single interrupt to be granted.

  The fix is to simply remove the requirement in the affinity spreading
  code for more than one interrupt being available"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq/affinity: Create affinity mask for single vector