media: ccs: Add support for analogue gain coefficient controls

Add four controls for reading CCS analogue gain coefficients. The
values are constants that are device specific.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l: uapi: ccs: Add controls for analogue gain constants

Add V4L2 controls for analogue gain constants required to control
analogue gain. The values are device specific and thus need to be obtained
from the driver.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation: ccs: Add user documentation for the CCS driver

Add user documentation for the CCS driver. This includes e.g. sub-devices
implemented by the driver.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l: uapi: Add user control base for CCS controls

Add a control base for CCS controls, and reserve 128 controls. Luckily
these numbers are cheap.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ccs: Only add analogue gain control if the device supports it

Some devices do not implement analogue gain this way. Only add the control
when a device does have the support.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ccs: Remove analogue gain field

The analogue gain control was stored to the device specific struct but was
never used. Remove it.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ccs: Add support for old-style SMIA digital gain

SMIA only has per-component digital gain. Add support for it.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ccs: Add digital gain support

CCS supports global (all-component) digital gain. Add support for it.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: mt9v111: Remove unneeded device-managed puts

Drivers don't need to explicitly call devm_{}_put on driver
removal, as it's automatically called by the device driver
resource management code.

Fixes: aab7ed1c3927 ("media: i2c: Add driver for Aptina MT9V111")
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: i2c: Add support for the OV8865 image sensor

The OV8865 is a 8 Mpx CMOS image sensor producing 3264x2448 at 30 fps.
Other modes (including some with sub-sampling) are available too.
It outputs 10-bit bayer CFA data through a MIPI CSI-2 interface with
up to 4 lanes supported.

Some register initialisation sequences are still needed for this driver,
as they cover registers for which no documentation is available.

This work is based on the first version of the driver submitted by
Kévin L'hôpital, which was adapted to mainline from the Allwinner BSP.
This version is a rewrite of the first version that matches the structure
of the OV5648 driver, with explicit PLL configuration, all the necessary
mode-specific fields, associatied registers and reduced static sequences.

It was tested with the Banana Pi Camera Board v3 and the Banana Pi M3.

Co-developed-by: Kévin L'hôpital <kevin.lhopital@bootlin.com>
Signed-off-by: Kévin L'hôpital <kevin.lhopital@bootlin.com>
Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: i2c: Add OV8865 bindings documentation

This introduces YAML bindings documentation for the OV8865
image sensor.

Co-developed-by: Kévin L'hôpital <kevin.lhopital@bootlin.com>
Signed-off-by: Kévin L'hôpital <kevin.lhopital@bootlin.com>
Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: i2c: Add support for the OV5648 image sensor

The OV5648 is a 5 Mpx CMOS image sensor, connected via MIPI CSI-2
in a one or two lane configuration.

Most of the features of the hardware are supported, including:
- Auto and manual exposition/gain
- Auto and manual white balance
- Horizontal and vertical flip
- Test patterns

But the following are still missing:
- Debanding, based on power source frequency;
- Exposition setting correlated to time units.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: i2c: Add OV5648 bindings documentation

This introduces YAML bindings documentation for the OV5648
image sensor.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ipu3-cio2: Fix mbus_code processing in cio2_subdev_set_fmt()

Loop was useless as it would always exit on the first iteration. Fix
it with right condition.

Signed-off-by: Pavel Machek (CIP) <pavel@denx.de>
Fixes: a86cf9b29e8b ("media: ipu3-cio2: Validate mbus format in setting subdev format")
Tested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: stable@vger.kernel.org # v4.16 and up
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: i2c: ov5670: Fix PIXEL_RATE minimum value

The driver currently reports a single supported value for
V4L2_CID_PIXEL_RATE and initializes the control's minimum value to 0,
which is very risky, as userspace might accidentally use it as divider
when calculating the time duration of a line.

Fix this by using as minimum the only supported value when registering
the control.

Fixes: 5de35c9b8dcd1 ("media: i2c: Add Omnivision OV5670 5M sensor support")
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation: media: Document clock handling in camera sensor drivers

Document pratices of handling clocks in camera sensor drivers on both DT
and ACPI.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ipu3-cio2: Build only for x86

According to the original code in the driver it was never assumed to work
with big page sizes: unsigned short type followed by PAGE_SHIFT and
PAGE_MASK which may be different on non-x86 architectures.

Recently LKP found an issue on non-x86 architectures due to above
mentioned limitations. Since Sakari acknowledges that it's not really
useful to be able to compile this elsewhere, mark it x86 only.

Fixes: a31d19f88932 ("media: ipu3: allow building it with COMPILE_TEST on non-x86 archs")
Reported-by: kernel test robot <lkp@intel.com>
Suggested-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l: ioctl: Use kmalloc to allocate a small chunk of memory

kvmalloc() was used to allocate the temporary memory buffer that was used
to contain both the IOCTL argument as well as a possible array argument
that could have been large. Now that the two are separated, the IOCTL
argument is known to be small in size. Use kmalloc to allocate it instead
of kvmalloc. Similarly for releasing it.

Suggested-by: Arnd Bergmann <arnd@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l: ioctl: Fix memory leak in video_usercopy

When an IOCTL with argument size larger than 128 that also used array
arguments were handled, two memory allocations were made but alas, only
the latter one of them was released. This happened because there was only
a single local variable to hold such a temporary allocation.

Fix this by adding separate variables to hold the pointers to the
temporary allocations.

Reported-by: Arnd Bergmann <arnd@kernel.org>
Reported-by: syzbot+1115e79c8df6472c612b@syzkaller.appspotmail.com
Fixes: d14e6d76ebf7 ("[media] v4l: Add multi-planar ioctl handling code")
Cc: stable@vger.kernel.org
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: uapi: Add an entity type for Image Signal Processors

Add and document a media entity type for Image Signal Processor devices.
Surprisingly we didn't have one, so add one now. More or less all ISP
drivers should use this type instead of what they currently are using (or
not using anything).

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l: fwnode: v4l2_async_notifier_parse_fwnode_endpoints is deprecated

Document that v4l2_async_notifier_parse_fwnode_endpoints() is deprecated.
Its functionality has been replaced by other, better functions. Also add a
reference to an example if someone ends up wandering here.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation: media: Update pixel rate formula for C-PHY

Update the formula to calculate the pixel rate on the link for C-PHY.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Remove 640x480 SBGGR8 mode

Capturing in 640x480 SBGGR8_1X8 hangs the system when capturing
with the unicam driver on RaspberryPi 4 platform.

Remove it and remove the support for multiple media bus codes in
the driver.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Support VIDIOC_SUBSCRIBE_EVENT

The driver reports the V4L2_SUBDEV_FL_HAS_EVENTS flag but does not
support subscribing and unsubscribing to events.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Constify oe_enable/disable reglist

Make the two register-value lists const.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Apply controls only when powered

Use pm_runtime_get_if_in_use() in s_ctrl to apply controls
only when the device is powered on.

Rework the control set function to balance the
pm_runtime_get_if_in_use() call with
pm_runtime_put() at the end of the function.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Rework s_stream() operation

Rework the s_stream() operation to turn the sensor on and
off at stream enable/disable time using the pm_runtime infrastructure.

Protect the stream on/off from being called multiple times in
sequence with a 'streaming' flag.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Use pm_runtime infrastructure

Use the pm_runtime framework to replace the legacy s_power() operation.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Advertise the correct exposure range

Exposure is clipped by the VTS of the mode, so it needs to be updated
when this is changed.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Support V4L2_CID_VBLANK control

Adds vblank control to allow for frame rate control.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Support V4L2_CID_HBLANK control

Add support for the V4L2_CID_HBLANK read-only control.

The implementation has been upported from RaspberryPi BSP commit:
commit d82f202156605 ("media: i2c: ov5647: Set V4L2_SUBDEV_FL_HAS_EVENTS flag")

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Support V4L2_CID_PIXEL_RATE

Clients need to know the pixel rate in order to compute exposure
and frame rate values. Advertise it.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Set V4L2_SUBDEV_FL_HAS_EVENTS flag

The ov5647 subdev can generate control events, therefore set
the V4L2_SUBDEV_FL_HAS_EVENTS flag.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Implement set_fmt pad operation

Now that the driver supports more than a single mode, implement the
.set_fmt pad operation and adjust the existing .get_fmt one to report
the currently applied format.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Use SBGGR10_1X10 640x480 as default

The SBGGR10_1X10 formats support more resolutions than SBGGR8_1X8.
Make it the default sensor format and set 2x2 binned 640x480 resolution
as default sensor size as it maximizes the FOV and framerate.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Add SGGBR10_1X10 modes

Add 4 additional sensor modes in SBGGR10_1X10 format.

Add the following resolutions:
- 2592x1944 full resolution
- 1920x1080 1080p cropped
- 1296x972 2x2 binned
- 640x480 2x2 binned, 2x2 subsampled

The register lists and modes definition have been upported from the
RaspberryPi BSP at revision:
commit 581dfda6d0a62 ("media: i2c: ov5647: Advertise the correct exposure range")

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Rename SBGGR8 VGA mode

Before adding new modes, rename the only existing one to report
the bit depth to distinguish it from future additions.

While at it, briefly describe the mode.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Add support for get_selection()

Support the get_selection() pad operation to report the device
full pixel array size, the currently applied analogue crop rectangle and
the active pixel array dimensions.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Break out format handling

Break format handling out from the main driver structure.

This commit prepares for the introduction of more sensor formats and
resolutions by instrumenting the existing operation to work on multiple
modes instead of assuming a single supported one.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Rationalize driver structure name

The driver structure name is referred to with different names ('ov5647',
'state', 'sensor') in different functions in the driver.

Polish this up by using 'struct ov5647 *sensor' everywhere.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Support gain, exposure and AWB controls

Add controls to support AWB, AEC and AGC. Also add control support to
set exposure (in lines) and analogue gain (as a register code).

Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Protect s_stream() with mutex

Use the driver mutex to protect s_stream() operations.
This will become more relevant once the sensor will support more formats
and set_format() could be issue concurrently to s_stream().

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Implement enum_frame_size()

Implement the .enum_frame_size subdev pad operation.

As the driver only supports one format and one resolution at the moment
the implementation is trivial.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Program mode at s_stream(1) time

Rename __sensor_init() function to ov5647_set_mode() as the function
is a regular one and the double underscores prefix shall be removed, and
then move it to program the mode at s_stream(1) time, not at sensor power
up.

Break out from __sensor_init() the stream_off() operation call at sensor
power up to coax the lanes in LP-11 state.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Fix return value from read/write

The ov5647_read()/ov5647_write() return in case of success the number
of bytes read or written respectively. This requires callers to check
if the return value is less than zero to detect an error. Unfortunately,
in several places, callers directly return the result of a read/write
call, causing issues when the returned valued is checked to be different
from zero to detect an error.

Fix this by returning zero if i2c_master_send() and i2c_master_read()
return a positive value (the number of bytes written or read).

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Replace license with SPDX identifier

Replace the boilerplate license text with the SPDX identifier.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Fix style issues

The driver has some obvious style issues which are worth fixing before
expanding the driver capabilities.

Fix:
- Variable declaration order
- Function parameters alignment
- Multi-line comments and spurious line breaks
- Use lowercase for hexadecimal values
- > 80 cols lines

Cosmetic change, no functional changes intended.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Fix format initialization

The driver currently support a single format. Fix its initialization to
use the only supported resolution.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Add set_fmt and get_fmt calls.

There's no way to query the subdevice for the supported
resolutions. Add set_fmt and get_fmt implementations. Since there's
only one format supported set_fmt does nothing and get returns single
format.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Add support for non-continuous clock mode

Add support for optional non-continuous clock mode to the ov5647
sensor driver.

Non-continuous clock saves a small amount of power and on some SoCs
is easier to interface with.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ov5647: Add support for PWDN GPIO.

Add support for an optional GPIO connected to PWDN on the sensor. This
allows the use of hardware standby mode where internal device clock
and circuit activities are halted.

Please note that power is off when PWDN is high.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: i2c: Rename ov5647.yaml

Rename 'ov5647.yaml' as 'ovti,ov5647.yaml' and update the
MAINTAINERS file entry accordingly.

Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: i2c: imx219: take lock in imx219_enum_mbus_code/frame_size

These two functions did not take the imx219->mutex lock, but
imx219_get_format_code checks that a lock is taken, so it issues
a warning:

[    8.738717] WARNING: CPU: 2 PID: 60 at drivers/media/i2c/imx219.c:653 imx219_get_format_code+0xac/0xc0
[    8.748113] Modules linked in:
[    8.751214] CPU: 2 PID: 60 Comm: kworker/2:1 Tainted: G        W         5.8.0-rc1-arm64 #148
[    8.759821] Hardware name: NVIDIA Jetson TX1 Developer Kit (DT)
[    8.765806] Workqueue: events deferred_probe_work_func
[    8.771003] pstate: 60000005 (nZCv daif -PAN -UAO BTYPE=--)
[    8.776635] pc : imx219_get_format_code+0xac/0xc0
[    8.781390] lr : imx219_get_format_code+0xa8/0xc0
[    8.786143] sp : ffff800012a538f0
[    8.789495] x29: ffff800012a538f0 x28: ffff800012838e90
[    8.794867] x27: ffff0000f28c5800 x26: ffff800011161c68
[    8.800237] x25: ffff0000f2a5a3f8 x24: 0000000000000018
[    8.805605] x23: ffff0000f284ef18 x22: ffff0000f2a5a080
[    8.810974] x21: ffff0000f284ff00 x20: ffff0000f2a5a080
[    8.816343] x19: 000000000000300f x18: 00000000ffffffff
[    8.821712] x17: ffff800011c77268 x16: 00000000000040d7
[    8.827081] x15: 00000000000040d8 x14: 0000000000000000
[    8.832451] x13: 00000000000040d4 x12: ffff800011d19300
[    8.837819] x11: 00000000000208c0 x10: 0000000000000004
[    8.843188] x9 : 000000003baa2ecd x8 : 000000008b3f9c73
[    8.848558] x7 : 0000000000000008 x6 : 0000000000000034
[    8.853929] x5 : 0000000000000000 x4 : 0000000000000001
[    8.859297] x3 : ffff800010a2a8a8 x2 : ffff0000f84a8000
[    8.864666] x1 : 0000000000000000 x0 : 0000000000000000
[    8.870034] Call trace:
[    8.872515]  imx219_get_format_code+0xac/0xc0
[    8.876921]  imx219_enum_mbus_code+0x38/0x60
[    8.881241]  call_enum_mbus_code+0x50/0x70
[    8.885387]  tegra_vi_graph_notify_complete+0x290/0x5e8
[    8.890670]  v4l2_async_notifier_try_complete.part.0+0x48/0x68
[    8.896563]  v4l2_async_register_subdev+0x100/0x1c0
[    8.901497]  v4l2_async_register_subdev_sensor_common+0x70/0xf0
[    8.907477]  imx219_probe+0x590/0x728
[    8.911184]  i2c_device_probe+0xe4/0x2b0
[    8.915151]  really_probe+0xd8/0x330
[    8.918768]  driver_probe_device+0x58/0xb8
[    8.922909]  __device_attach_driver+0x84/0xc8
[    8.927315]  bus_for_each_drv+0x78/0xc8
[    8.931193]  __device_attach+0xe4/0x140
[    8.935072]  device_initial_probe+0x14/0x20
[    8.939301]  bus_probe_device+0x9c/0xa8
[    8.943179]  deferred_probe_work_func+0x74/0xb0
[    8.947759]  process_one_work+0x2c4/0x740
[    8.951813]  worker_thread+0x4c/0x430
[    8.955518]  kthread+0x158/0x178
[    8.958786]  ret_from_fork+0x10/0x1c
[    8.962401] irq event stamp: 63536
[    8.965846] hardirqs last  enabled at (63535): [<ffff800010082398>] el1_irq+0xd8/0x180
[    8.973846] hardirqs last disabled at (63536): [<ffff8000100a6484>] do_debug_exception+0x16c/0x258
[    8.982895] softirqs last  enabled at (63534): [<ffff800010080d4c>] _stext+0x54c/0x594
[    8.990896] softirqs last disabled at (63525): [<ffff8000100c8350>] irq_exit+0x100/0x138
[    8.999066] ---[ end trace ebfbcd84b75ef921 ]---
[    9.004354] ------------[ cut here ]------------

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reviewed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation: v4l: Document that link_validate op is valid for sink only

The link_validate pad op will only be called on sink pads. Document this.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation: v4l: Remove reference to video ops

We no longer have format related callbacks in video ops. Remove the
reference to them.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: allegro: rename stream_id to dst_handle

The dst_handle field in the PUT_STREAM_BUFFER and ENCODE_FRAME_RESPONSE
is used to retrieve the V4L2 CAPTURE buffer. Calling it stream_id is
confusing. Call it dst_handle inspired by src_handle for the OUTPUT
buffer.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: allegro: remove custom drain state handling

The v4l2-m2m has various helpers for correctly handle the draining. Drop
the driver specific state machine and use the m2m helper functions.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: allegro,al5e: Convert to YAML

Convert the Allegro DVT video IP codec text binding to Yaml.

Add the converted binding to the MAINTAINERS file.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: allegro: move driver out of staging

The stateful encoder API was finalized. Nothing is blocking the driver
from being moved out of staging.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: allegro: Fix use after free on error

The "channel" is added to the "dev->channels" but then if
v4l2_m2m_ctx_init() fails then we free "channel" but it's still on the
list so it could lead to a use after free.  Let's not add it to the
list until after v4l2_m2m_ctx_init() succeeds.

Fixes: cc62c74749a3 ("media: allegro: add missed checks in allegro_open()")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: cec: add EPOLLPRI in poll() when dev is unregistered

If the CEC device was unregistered, then add EPOLLPRI to
the poll() mask. Otherwise a select() that only waits for
exceptions will not wake up. A select() that waits for
read and/or write events *will* wake up on an EPOLLERR, but
not (for some reason) if it just waits for exceptions.

Strangly the epoll functionality will wakeup on EPOLLERR if
you just wait for an exception, so in this respect select()
and epoll differ.

In the end it doesn't really matter, what matters is that
polling file handles are woken up on device unregistration.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-dev: add EPOLLPRI in v4l2_poll() when dev is unregistered

If the V4L2 device was unregistered, then add EPOLLPRI to
the poll mask. Otherwise a select() that only waits for
exceptions will not wake up. A select() that waits for
read and/or write events *will* wake up on an EPOLLERR, but
not (for some reason) if it just waits for exceptions.

Strangly the epoll functionality will wakeup on EPOLLERR if
you just wait for an exception, so in this respect select()
and epoll differ.

In the end it doesn't really matter, what matters is that
polling file handles are woken up on device unregistration.

It also improves the code a bit if vdev->fops->poll is NULL:
this didn't check for device unregistration.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vivid: call v4l2_event_wake_all() on disconnect

When the disconnect error injection control is set, then besides
faking unregistering the device nodes, also call v4l2_event_wake_all()
to ensure any userspace applications will wake up as per a 'normal'
unregister.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-dev/event: add v4l2_event_wake_all()

When unregistering a V4L2 device node, make sure any filehandles
that are waiting for an event are woken up.

Add v4l2_event_wake_all() to v4l2-event.c and call it from
video_unregister_device().

Otherwise userspace might never know that a device node was removed.

[hverkuil: checkpatch: replaced 'if (vdev == NULL)' by 'if (!vdev)']

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-mem2mem: always call poll_wait() on queues

do_poll()/do_select() seem to set the _qproc member of poll_table to
NULL the first time they are called on a given table, making subsequent
calls of poll_wait() on that table no-ops. This is a problem for mem2mem
which calls poll_wait() on the V4L2 queues' waitqueues only when a
queue-related event is requested, which may not necessarily be the case
during the first poll.

For instance, a stateful decoder is typically only interested in
EPOLLPRI events when it starts, and will switch to listening to both
EPOLLPRI and EPOLLIN after receiving the initial resolution change event
and configuring the CAPTURE queue. However by the time that switch
happens and v4l2_m2m_poll_for_data() is called for the first time,
poll_wait() has become a no-op and the V4L2 queues waitqueues thus
cannot be registered.

Fix this by moving the registration to v4l2_m2m_poll() and do it whether
or not one of the queue-related events are requested.

Signed-off-by: Alexandre Courbot <gnurou@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: videobuf2: always call poll_wait() on queues

do_poll()/do_select() seem to set the _qproc member of poll_table to
NULL the first time they are called on a given table, making subsequent
calls of poll_wait() on that table no-ops. This is a problem for vb2
which calls poll_wait() on the V4L2 queues' waitqueues only when a
queue-related event is requested, which may not necessarily be the case
during the first poll.

Fix this by making the call to poll_wait() happen first thing and
unconditionally in vb2_core_poll().

Signed-off-by: Alexandre Courbot <gnurou@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add custom V4L2 control V4L2_CID_TEGRA_SYNCPT_TIMEOUT_RETRY

This patch adds custom V4L2 control for syncpt timeout retry to continue
capture on error for specified retries count through this control.

This is useful for HDMI-to-CSI bridge debug purposes like for hotplug scenarios
or for ignoring captures till HDMI input is stabilized.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add support for x8 captures with gang ports

Tegra VI/CSI hardware don't have native 8 lane capture support.

Each CSI port has max 4 lanes only. So for x8 captures, consecutive
ports are ganged up for left half and right half captures on to each
x4 ports with buffer offsets based on source image split width to align
side-by-side.

All ports in gang are configured together during the corresponding
video device node streaming for x8 captures.

x8 capture with gang ports are supported with HDMI-to-CSI bridges
where they split 4K image into left half onto one x4 port and
right half onto second x4 port.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: tegra: Update csi data-lanes to maximum 8 lanes

Tegra VI/CSI hardware don't have native 8 lane CSI RX port.

But x8 capture can be supported by using consecutive x4 ports
simultaneously with HDMI-to-CSI bridges where source image is split
on to two x4 ports.

This patch updates dt-bindings for csi endpoint data-lane property
with maximum of 8 lanes.

Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-fwnode: Update V4L2_FWNODE_CSI2_MAX_DATA_LANES to 8

Some CSI2 receivers support 8 data lanes.

So, this patch updates CSI2 maximum data lanes to be 8.

Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Implement V4L2 device notify callback

Implement V4L2 device notify callback to handle
V4L2_EVENT_SOURCE_CHANGE event from subdevices.

HDMI-to-CSI bridge drivers trigger V4L2_EVENT_SOURCE_CHANGE when
source DV timing changes are detected or when HDMI hotplug happens.

Runtime source parameter changes during active streaming is not
allowed and Tegra video driver calls vb2_queue_error to signal a
fatal error if a notification of this event happens during an active
streaming.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add support for V4L2_EVENT_SOURCE_CHANGE

Current implementation uses v4l2_ctrl_subscribe_event() and this
does not handle V4L2_EVENT_SOURCE_CHANGE.

So, update driver to handle V4L2_EVENT_SOURCE_CHANGE.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add support for VIDIOC_LOG_STATUS ioctl

This patch adds support for log_status ioctl.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add support for EDID ioctl ops

This patch adds support for EDID get and set v4l2 ioctl ops to use
with HDMI to CSI bridges.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add DV timing support

This patch adds below v4l2 DV timing ioctls to support HDMI-to-CSI
bridges.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Add support for V4L2_PIX_FMT_NV16

NV16 are two-plane versions of YUV422 format.

VI/CSI surface0 registers corresponds to first Y plane and
surface1 registers corresponds to seconds UV plane.

This patch updates image size for NV16 format to include both planes
and programs VI/CSI surface1 registers for UV plane capture.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Fix V4L2 pixel format RGB and YUV

V4L2 pixel format is incorrect for RGB and YUV formats.

This patch fixes it.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Enable VI pixel transform for YUV and RGB formats

VI Pixel transforms converts source pixel data to selected
destination pixel formats in memory and aligns properly.

YUV and RGB formats need this pixel transform to be enabled.

RAW formats use T_R16_I destination pixel format in memory and
does not need pixel transform as they support direct write to
memory.

So, this patch enables pixel transform for YUV and RGB and keeps
it bypass for RAW formats.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: tegra-video: Use zero crop settings if subdev has no get_selection

Currently try format implementation doesn't check if subdevice has
get_selection ops implemented and returns -EINVAL on error from
direct v4l2_subdev_call of get_selection.

Selection API's are not mandatory for all V4L2 subdevices.

This patch fixes it by adding v4l2_subdev_has_ops check prior to
calling get_selection ops and continues with try or set format with
zero crop settings for subdevices that don't have get_selection and
returns -EINVAL only for subdevices that has get_selection ops.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Simplify the context API

Rework the context API exposed to cal-video.c to simplify it. The
configuration and enable steps are all grouped in a single
cal_ctx_start() function, and the DMA stop and IRQ disable are similarly
groupd in cal_ctx_stop(). The cal_ctx_wr_dma_addr() function is renamed
to cal_ctx_set_dma_addr() for consistency with the cal_ctx_ prefix of
the start and stop functions.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Store buffer DMA address in dma_addr_t

dma_addr_t is the correct type to store DMA addresses. Replace incorrect
usage of unsigned long and unsigned int.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Set cal_dmaqueue.pending to NULL when no pending buffer

When a pending buffer becomes active, the cal_dmaqueue.active field is
updated, but the pending field keeps the same value until a new buffer
becomes pending. This requires handling the special case of
pending == active in different places. Simplify the code by setting the
pending field to NULL when the pending buffer becomes active. Buffers
are now simply moved from queue to pending and from pending to active.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Group all DMA queue fields in struct cal_dmaqueue

The cal_dmaqueue structure only contains the list of queued buffers.
Move the other fields that are logically related to the DMA queue
(current and next buffer points, state, wait queue and lock) from
cal_ctx to cal_dmaqueue.

Take this as an opportunity to document the fields usage and to give
them more appropriate names. The 'active' field stored the list of all
queued buffers, not the active buffers, so rename it to 'queue'. The
'cur_frm' and 'next_frm' are respectively renamed to 'active' and
'pending' to better explain their purpose. The 'dma_state' and
'dma_wait' fields are stripped of their 'dma_' prefix as they're now
part of cal_dmaqueue. Finally, 'slock' is renamed to 'lock'.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Use list_first_entry()

Use the list_first_entry() macro where appropriate to replace manual
usage of list_entry(head.next).

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Remove DMA queue empty check at start streaming time

The vb2 queue ensures that the start streaming operation will only be
called with a minimal number of buffers queued to the driver. There's
thus no need to manually check if the DMA queue is empty. Remove the
check.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Drop unneeded check in cal_calc_format_size()

The cal_calc_format_size() function checks that the passed fmtinfo
argument is not NULL. All callers ensure that the pointer is valid. Drop
the check.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Drop V4L2_CAP_READWRITE

The V4L2 read/write API is inefficient and makes little sense for
devices that handle frame-based formats. Applications shouldn't use it,
drop its support from the driver.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Share buffer release code between start and stop

The cal_start_streaming() and cal_stop_streaming() functions duplicate
the same buffer release logic. split it to a separate function to share
the code.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Use spin_lock_irq() when starting or stopping stream

The cal_start_streaming() and cal_stop_streaming() functions are called
with interrupts enabled. spin_lock_irq() can thus be used instead of
spin_lock_irqsave().

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Stop write DMA without disabling PPI

When stopping the stream, the driver needs to ensure that ongoing DMA
completes and that no new DMA is started. It does so using a feature of
the PPI that can be stopped on a frame boundary. The downside of this
mechanism is that the DMA can't be stopped independently of the source,
which prevents usage of multiple contexts for the same source (to handle
CSI-2 virtual channels or data types).

Rework the stream stop mechanism to stop the write DMA without disabling
the PPI first. The new mechanism relies on the combination of a state
machine in the driver and shadowing of the CAL_WR_DMA_CTRL_j.MODE field
in the hardware. When a stop is requested, the DMA start interrupt
handler will request the hardware to stop at the end of the current
frame by disabling the write DMA context in the shadowed register, and
flag that a stop is in progress. The next DMA end interrupt will then
report that the stop is complete.

This makes it possible to stop the PPI after stopping the DMA, and fold
the cal_camerarx_ppi_disable() call into cal_camerarx_stop().

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Fold PPI enable in CAMERARX .s_stream()

To further decouple the context and CAMERARX components, move the call
to cal_camerarx_ppi_enable() from cal_start_streaming() to the CAMERARX
.s_stream() operation. The DMA destination address has to be set before
starting the CAMERARX, which is desirable anyway.

cal_camerarx_ppi_disable() will be addressed separately.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Refactor interrupt enable/disable

Interrupts are enabled and disabled by the cal_camerarx_enable_irqs()
and cal_camerarx_disable_irqs(). Despite their name, they deal with all
interrupts, not just the CAMERARX interrupts, and they hardcode the
assumption that the context index is identical to the CAMERARX index.

Split the context-related interrupt management to two new functions,
cal_ctx_enable_irqs() and cal_ctx_disable_irqs(), called from the
cal_start_streaming() and cal_stop_streaming() functions. The explicit
calls to cal_camerarx_enable_irqs() and cal_camerarx_disable_irqs() are
folded with the CAMERARX .s_stream() operation to simplify the CAMERARX
API.

Enabling the OCPO error interrupt is moved to the PM runtime resume
operation, as it's global to the device, not related to a CAMERARX or
context. The VC IRQ enable and disable are removed as they're not used,
the parent interrupt bit (CAL_HL_IRQ_VC_MASK) never being set.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Rename struct cal_fmt to cal_format_info

The cal_fmt structure stores information about a format. Its name is
ambiguous, as it could be understood as storing a format. Furthermore,
instances of the structure are called 'fmt' through the code, which
leads to confusion with the 'format' variables. Rename the structure to
cal_format_info, and the corresponding 'fmt' variables to 'fmtinfo'.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Don't pass format to cal_ctx_wr_dma_config()

The cal_ctx_wr_dma_config() function has access to the context, there's
no need to give it format-related values retrieved from the context.
Access the values internally, and reword internal comments. The comment
regarding the CAL_WR_DMA_OFST register not being well understood is
dropped, as the datasheet explicitly documents it as "Offset between two
consecutive line starts".

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Use CAMERARX subdev s_stream op in video device code

Replace calls to cal_camerarx_start() and cal_camerarx_stop() with usage
of the .s_stream() subdev operation. This requires calling the
.set_fmt() operation as the CAMERARX now relies on the format
information set through the subdev API instead of receiving it in the
cal_camerarx_start() function.

This change prepare for exposing the CAMERARX subdev operations to
userspace by using the same API within the kernel.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Implement subdev ops for CAMERARX

Implement subdev operations for the CAMERARX. They will be used to
replace calls to custom CAMERARX functions in the V4L2 video device
code, and will be exposed to userspace.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Iterate over correct number of CAMERARX instances

When performing operations on all CAMERARX instances, code usually
iterates over all cal->phy[] entries and skips the NULL pointers. The
number of available CAMERARX instances is however available through
cal->data->num_csi2_phy. Use it instead.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Replace hardcoded BIT() value with macro

Replace the hardcoded BIT(3) value with CAL_CSI2_PPI_CTRL_FRAME_MASK to
increase readability.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Rename MAX_(WIDTH|HEIGHT)_* macros with CAL_ prefix

The MAX_WIDTH_BYTES and MAX_HEIGHT_LINES macros have a generic name that
is prone to namespace clashes. Rename them with a CAL_ prefix. While at
it, expand their documentation and add CAL_MIN_(WIDTH|HEIGHT)_* macros
that will be used to implement CAMERARX subdev operations.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: Move format handling to cal.c and expose helpers

The cal_formats array contain the description of all formats supported
by the hardware. It's currently used by the V4L2 video device operations
only, but will be needed by the CAMERARX subdev code too. Move it from
cal-video.c to cal.c and add helper functions to access it.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>