media: meson: vdec: align stride on 32 bytes

The HEVC/VP9 aligns the plane stride on 32, so align the planes stride
for all codecs to 32 to satisfy HEVC/VP9 decoding using the "HEVC" HW.

This fixes VP9 decoding of streams with following (not limited) widths:
- 264
 -288
- 350
- 352
- 472
- 480
- 528
- 600
- 720
- 800
- 848
- 1440

Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Tested-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: meson: vdec: add H.264 decoding support

Add support for the H264 compressed format (V4L2_PIX_FMT_H264).

Signed-off-by: Maxime Jourdan <mjourdan@baylibre.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: meson: vdec: bring up to compliance

Add all the necessary bits to pass v4l2-compliance in stateful decoding
mode.

Mostly includes tracking the state of the decoder, allowing the OUTPUT
queue to stream while the CAPTURE queue is inactive, handling resolution
change events, draining with V4L2_DEC_CMD_STOP, copying more metadata
from the src buffers to the dst buffers, etc.

Signed-off-by: Maxime Jourdan <mjourdan@baylibre.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vicodec: use v4l2-mem2mem draining, stopped and next-buf-is-last states handling

Use the previously introduced v4l2-mem2mem core APIs to handle the drainig,
stopped and next-buf-is-last states.

With these changes, the v4l2-compliance still passes with the following
commands :
>>>><><><><><><><><><><><><><><><><>< 15.53 fps
 15.53 fps
><><><><><><><><><><><><>< 13.99 fps
 13.99 fps
><><><><><><><><><><><>< 13.52 fps
 13.52 fps
><><><><><><><><><><><><>< 13.41 fps
 13.41 fps
><><><><><><><><><><><><>< 13.21 fps
 13.21 fps
><><><><><><><><><><><>< 13.09 fps
 13.09 fps
><><><><><><><
STOP ENCODER
<<<
EOS EVENT

v4l2-compliance SHA: 7ead0e1856b89f2e19369af452bb03fd0cd16793, 64 bits
[...]
Total for vicodec device /dev/video0: 50, Succeeded: 50, Failed: 0, Warnings: 0

The full output is available at [1]

v4l2-compliance SHA: 7ead0e1856b89f2e19369af452bb03fd0cd16793, 64 bits
[...]
Total for vicodec device /dev/video1: 50, Succeeded: 50, Failed: 0, Warnings: 0

The full output is available at [2]

No functional changes should be noticed.

[1] https://termbin.com/25nn
[2] https://termbin.com/dza4

Suggested-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Suggested-by: Maxime Jourdan <mjourdan@baylibre.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-mem2mem: handle draining, stopped and next-buf-is-last states

Since the draining and stop phase of the HW decoder mem2mem bahaviour is
now clearly defined, we can move handling of the following states to the
common v4l2-mem2mem core code:
- draining
- stopped
- next-buf-is-last

By introducing the following v4l2-mem2mem APIs:
- v4l2_m2m_encoder_cmd/v4l2_m2m_ioctl_encoder_cmd to handle start/stop command
- v4l2_m2m_decoder_cmd/v4l2_m2m_ioctl_decoder_cmd to handle start/stop command
- v4l2_m2m_update_start_streaming_state to update state on start of streaming
of the de/encoder queue
- v4l2_m2m_update_stop_streaming_state to update state on stop of streaming
of the de/encoder queue
- v4l2_m2m_last_buffer_done to make the current dest buffer as the last one

And inline helpers:
- v4l2_m2m_mark_stopped to mark the de/encoding process as stopped
- v4l2_m2m_clear_state to clear the de/encoding state
- v4l2_m2m_dst_buf_is_last to detect the current dequeued dst_buf is the last
- v4l2_m2m_has_stopped to detect the de/encoding stopped state
- v4l2_m2m_is_last_draining_src_buf to detect the current source buffer should
 be the last processing before stopping the de/encoding process

The special next-buf-is-last when min_buffers != 1 case is also handled
in v4l2_m2m_qbuf() by reusing the other introduced APIs.

This state management has been stolen from the vicodec implementation,
and is no-op for drivers not calling the v4l2_m2m_encoder_cmd or
v4l2_m2m_decoder_cmd and v4l2_m2m_update_start/stop_streaming_state.

The vicodec will be the first one to be converted as an example.

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

media: v4l2-ctrls: v4l2_ctrl_request_complete() should always set ref->req

When the request is completed, all controls are copied to the request object.
However, when VIDIOC_G_EXT_CTRLS attempts to read control values from the
request it will read the current value instead for any control reference that
has a NULL ref->req pointer. But that's wrong: after completing the request
*all* controls should have a non-NULL ref->req pointer since they are after
all copied to the request.

So set ref->req if it wasn't set already.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: media: usbvision: Add missing MEDIA_USB_SUPPORT dependency

VIDEO_USBVISION driver depends on MEDIA_USB_SUPPORT

Fixes: 8fb12ce2ec9d ("media: usbvision: deprecate driver")

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: tegra-vde: Use devm_platform_ioremap_resource_byname()

There is a new devm_platform_ioremap_resource_byname() helper in the
kernel now, which helps to make code cleaner a tad by replacing few
"boilerplate" lines of code with a single line. Let's utilize that
new helper in the VDE driver.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: ti-vpe: cal: fix disable_irqs to only the intended target

disable_irqs() was mistakenly disabling all interrupts when called.
This cause all port stream to stop even if only stopping one of them.

Cc: stable <stable@vger.kernel.org>
Signed-off-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: imx: Propagate quantization and encoding in CSI

Unlike the PRPENC and PRPVF subdevices, the CSI's cannot convert
quantization from sink to source, or do any kind of Y'CbCr <-> RGB
encoding. So the CSI's cannot allow quantization and ycbcr_enc to be
selectable by the user and must be propagated from sink to source.

Fixes: 4791bd7d6adc4 ("media: imx: Try colorimetry at both sink and source pads")

Reported-by: Tim Harvey <tharvey@gateworks.com>
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Tim Harvey <tharvey@gateworks.com>
Reviewed-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2: Switch to new kerneldoc syntax for named variable macro arguments

The new syntax is available since commit 43756e347f21
("scripts/kernel-doc: Add support for named variable macro arguments").

The same HTML output is produced with and without this patch.

Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: spi: gs1662: Use new structure for SPI transfer delays

In a recent change to the SPI subsystem [1], a new `delay` struct was added
to replace the `delay_usecs`. This change replaces the current
`delay_usecs` with `delay` for this driver.

The `spi_transfer_delay_exec()` function [in the SPI framework] makes sure
that both `delay_usecs` & `delay` are used (in this order to preserve
backwards compatibility).

[1] commit bebcfd272df6 ("spi: introduce `delay` field for
`spi_transfer` + spi_transfer_delay_exec()")

Signed-off-by: Sergiu Cuciurean <sergiu.cuciurean@analog.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-core: fix entity initialization in device_register_subdev

The entity variable was being initialized in the wrong place, before the
parameters have been checked.
To solve this, completely removed the entity variable and replaced it
with the initialization value : &sd->entity.
This will avoid dereferencing 'sd' pointer before it's being checked if
it's NULL.

Fixes: 61f5db549dde ("[media] v4l: Make v4l2_subdev inherit from media_entity")

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: v4l2-core: fix a use-after-free bug of sd->devnode

sd->devnode is released after calling
v4l2_subdev_release. Therefore it should be set
to NULL so that the subdev won't hold a pointer
to a released object. This fixes a reference
after free bug in function
v4l2_device_unregister_subdev

Fixes: 0e43734d4c46e ("media: v4l2-subdev: add release() internal op")

Cc: stable@vger.kernel.org
Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vimc: use-after-free fix - release vimc in the v4l_device release

A use-after-free bug occures when unbinding the device while it streams.
The 'struct vimc_ent_device' allocated for the 'Sensor A' is freed
when calling the sensor's 'rm' callback but the freed pointer is
later accessed in the function 'vimc_streamer_pipeline_terminate'.
To fix this bug, move the release callback of the vimc entities
and vimc_device to the release callback of v4l2_device.
The .rm callback of vimc_ent_config is replaced by two callbacks:

.unregister - this is called upon removing the device and
it unregisters the entity. This is an optional callback since
subdevices don't need to implement it because they are already
unregistered in v4l2_device_unregister.

.release - this is called from the release callback of v4l2_device
and it frees the entity.

This ensures that the entities will be released when the last fh
of any of the devices is closed.

The commands that cause the crash and the KASAN report:

media-ctl -d platform:vimc -V '"Sensor A":0[fmt:SBGGR8_1X8/640x480]'
media-ctl -d platform:vimc -V '"Debayer A":0[fmt:SBGGR8_1X8/640x480]'
v4l2-ctl -z platform:vimc -d "RGB/YUV Capture" -v width=1920,height=1440
v4l2-ctl -z platform:vimc -d "Raw Capture 0" -v pixelformat=BA81
v4l2-ctl --stream-mmap --stream-count=1000 -d /dev/video2 &
sleep 1
echo -n vimc.0 >/sys/bus/platform/drivers/vimc/unbind

[  188.417934] BUG: KASAN: use-after-free in vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.420182] Read of size 8 at addr ffff8881e9c26008 by task bash/185
[  188.421800]
[  188.422223] CPU: 0 PID: 185 Comm: bash Not tainted 5.5.0-rc1+ #1
[  188.423681] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
[  188.425938] Call Trace:
[  188.426610]  dump_stack+0x75/0xa0
[  188.427519]  ? vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.429057]  print_address_description.constprop.6+0x16/0x220
[  188.430462]  ? vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.431979]  ? vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.433455]  __kasan_report.cold.9+0x1a/0x40
[  188.434518]  ? vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.436010]  kasan_report+0xe/0x20
[  188.436859]  vimc_streamer_pipeline_terminate+0x75/0x140 [vimc]
[  188.438339]  vimc_streamer_s_stream+0x8b/0x3c0 [vimc]
[  188.439576]  vimc_cap_stop_streaming+0x22/0x40 [vimc]
[  188.440863]  __vb2_queue_cancel+0x65/0x560 [videobuf2_common]
[  188.442391]  vb2_core_queue_release+0x19/0x50 [videobuf2_common]
[  188.443974]  vimc_cap_rm+0x10/0x20 [vimc]
[  188.444986]  vimc_rm_subdevs+0x9e/0xe0 [vimc]
[  188.446179]  vimc_remove+0x19/0x70 [vimc]
[  188.447301]  platform_drv_remove+0x2f/0x50
[  188.448468]  device_release_driver_internal+0x133/0x260
[  188.449814]  unbind_store+0x121/0x150
[  188.450726]  kernfs_fop_write+0x142/0x230
[  188.451724]  ? sysfs_kf_bin_read+0x100/0x100
[  188.452826]  vfs_write+0xdc/0x230
[  188.453760]  ksys_write+0xaf/0x140
[  188.454702]  ? __ia32_sys_read+0x40/0x40
[  188.455773]  ? __do_page_fault+0x473/0x620
[  188.456780]  do_syscall_64+0x5e/0x1a0
[  188.457711]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  188.459079] RIP: 0033:0x7f80f1f13504
[  188.459969] Code: 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b3 0f 1f 80 00 00 00 00 48 8d 05 f9 61 0d 00 8b 00 85 c0 75 13 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 54 c3 0f 1f 00 41 54 49 89 d4 55 48 89 f5 53
[  188.464445] RSP: 002b:00007ffd7e843b58 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
[  188.466276] RAX: ffffffffffffffda RBX: 0000000000000006 RCX: 00007f80f1f13504
[  188.467999] RDX: 0000000000000006 RSI: 000055ef2eb21b10 RDI: 0000000000000001
[  188.469708] RBP: 000055ef2eb21b10 R08: 00007f80f1fe68c0 R09: 00007f80f1e26740
[  188.471407] R10: 000055ef2eade010 R11: 0000000000000246 R12: 00007f80f1fe5760
[  188.473381] R13: 0000000000000006 R14: 00007f80f1fe0760 R15: 0000000000000006
[  188.475107]
[  188.475500] Allocated by task 473:
[  188.476351]  save_stack+0x19/0x80
[  188.477201]  __kasan_kmalloc.constprop.6+0xc1/0xd0
[  188.478507]  vimc_sen_add+0x36/0x309 [vimc]
[  188.479649]  vimc_probe+0x1e2/0x530 [vimc]
[  188.480776]  platform_drv_probe+0x46/0xa0
[  188.481829]  really_probe+0x16c/0x520
[  188.482732]  driver_probe_device+0x114/0x170
[  188.483783]  device_driver_attach+0x85/0x90
[  188.484800]  __driver_attach+0xa8/0x190
[  188.485734]  bus_for_each_dev+0xe4/0x140
[  188.486702]  bus_add_driver+0x223/0x2d0
[  188.487715]  driver_register+0xca/0x140
[  188.488767]  0xffffffffc037003d
[  188.489635]  do_one_initcall+0x86/0x28f
[  188.490702]  do_init_module+0xf8/0x340
[  188.491773]  load_module+0x3766/0x3a10
[  188.492811]  __do_sys_finit_module+0x11a/0x1b0
[  188.494059]  do_syscall_64+0x5e/0x1a0
[  188.495079]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  188.496481]
[  188.496893] Freed by task 185:
[  188.497670]  save_stack+0x19/0x80
[  188.498493]  __kasan_slab_free+0x125/0x170
[  188.499486]  kfree+0x8c/0x230
[  188.500254]  v4l2_subdev_release+0x64/0x70 [videodev]
[  188.501498]  v4l2_device_release_subdev_node+0x1c/0x30 [videodev]
[  188.502976]  device_release+0x3c/0xd0
[  188.503867]  kobject_put+0xf4/0x240
[  188.507802]  vimc_rm_subdevs+0x9e/0xe0 [vimc]
[  188.508846]  vimc_remove+0x19/0x70 [vimc]
[  188.509792]  platform_drv_remove+0x2f/0x50
[  188.510752]  device_release_driver_internal+0x133/0x260
[  188.512006]  unbind_store+0x121/0x150
[  188.512899]  kernfs_fop_write+0x142/0x230
[  188.513874]  vfs_write+0xdc/0x230
[  188.514698]  ksys_write+0xaf/0x140
[  188.515523]  do_syscall_64+0x5e/0x1a0
[  188.516543]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  188.517710]
[  188.518034] The buggy address belongs to the object at ffff8881e9c26000
[  188.518034]  which belongs to the cache kmalloc-4k of size 4096
[  188.520528] The buggy address is located 8 bytes inside of
[  188.520528]  4096-byte region [ffff8881e9c26000, ffff8881e9c27000)
[  188.523015] The buggy address belongs to the page:
[  188.524357] page:ffffea0007a70800 refcount:1 mapcount:0 mapping:ffff8881f6402140 index:0x0 compound_mapcount: 0
[  188.527058] raw: 0200000000010200 dead000000000100 dead000000000122 ffff8881f6402140
[  188.528983] raw: 0000000000000000 0000000000040004 00000001ffffffff 0000000000000000
[  188.530883] page dumped because: kasan: bad access detected
[  188.532336]
[  188.532720] Memory state around the buggy address:
[  188.533871]  ffff8881e9c25f00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  188.535631]  ffff8881e9c25f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  188.537370] >ffff8881e9c26000: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  188.538996]                       ^
[  188.539812]  ffff8881e9c26080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  188.541549]  ffff8881e9c26100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vimc: allocate vimc_device dynamically

In future patch, the release of the device will move
to the release callback of v4l2_device. Therefore the
device will be released only when the last fh will be
closed. Dynamic allocation will then be needed since
when the device is unbounded and then bounded again,
it might be that the probe callback will run before
the release of the last device is finished. In that
case both operations will run on the same memory
concurrently and cause memory corruption.
This patch also removes the pdev field of
vimc_device since it is not needed anymore.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vimc: replace vimc->pdev.dev with vimc->mdev.dev

replace 'vimc->pdev.dev' with 'vimc->mdev.dev'
in debug prints and in assignment to
vimc_ent_device.dev. This helps to unify the debug
statements. This will also eliminate the need to use
the pdev field in vimc_device in future patch.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation:media:v4l-drivers: Update vivid documentation.

Add metadata capture/output and v4l-touch support in vivid
documentation.

Signed-off-by: Vandana BN <bnvandana@gmail.com>
[hverkuil-cisco@xs4all.nl: fixed some typos, made some small tweaks of the text]
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: rcar_drif: Do not print error in case of EPROBE_DEFER for dma channel

If the dma channel request error code is EPROBE_DEFER there is no need to
print error message.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: aspeed: add AST2600 support

Video engine in AST2600 has the exactly same register set with
AST2500 except VR084 register which provides more precise JPEG
size read back. This commit adds support for the difference and
adds 'aspeed,ast2600-video-engine' compatible OF string.

Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com>
Reviewed-by: Joel Stanley <joel@jms.id.au>
Reviewed-by: Eddie James <eajames@linux.ibm.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: capture: remove support to userptr memory

VB2_USERPTR support doesn't make sense for drivers based on
vb2_dma_contig, so just drop it.

Otherwise, queues using vb2_dma_contig_memops causes v4l2-compliance
to fail for VB2_USERPTR mode, due to buffer misalignment.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
[hverkuil-cisco@xs4all.nl: replaced commit log with Helen's suggestion]
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: videobuf2-core.h: improve buf_struct_size documentation

Explicitly document that the driver-specific buffer structure
must start with the subsystem-specific struct (vb2_v4l2_buffer
in the case of V4L2).

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

media: staging: rkisp1: remove serialization item in the TODO file

The item 'Fix serialization on subdev ops.' was solved,
so remove it from the TODO file.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: videobuf2-dma-contig.c: remove spurious 'b' in message

Remove a spurious 'b' in the "contiguous chunk is too small  %lu/%lu b"
message.

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

media: mtk-vcodec: reset segment data then trig decoder

VP9 bitstream specification indicate segment data should reset to
default when meet key frames, intra only frames or enable error
resilience mode. So memset segmentation map buffer before every
decode process is not appropriate.

Reset segment data only when needed, then start decoder hardware

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

media: dt-bindings: media: Add Allwinner A64 deinterlace compatible

Allwinner A64 SoC also contains deinterlace core, compatible to H3.

Add compatible string for it.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: go7007: Fix URB type for interrupt handling

Josef reported that his old-and-good Plextor ConvertX M402U video
converter spews lots of WARNINGs on the recent kernels, and it turned
out that the device uses a bulk endpoint for interrupt handling just
like 2250 board.

For fixing it, generalize the check with the proper verification of
the endpoint instead of hard-coded board type check.

Fixes: 7e5219d18e93 ("[media] go7007: Fix 2250 urb type")
Reported-and-tested-by: Josef Möllers <josef.moellers@suse.com>
BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1162583
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=206427
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: sun8i: Add Allwinner A83T Rotate driver

Allwinner A83T contains rotation core which can rotate and flip images.

Add a driver for it.

Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
[hverkuil-cisco@xs4all.nl: MAINTAINERS paths were out of date, fix that]
[hverkuil-cisco@xs4all.nl: VFL_TYPE_GRABBER -> _VIDEO]
[hverkuil-cisco@xs4all.nl: Fix module build]
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: Add Allwinner A83T Rotate driver

Some Allwinner SoCs like A83T and A64 contain rotate core which can
rotate and flip images.

Add a binding for it.

Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
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: Split v4l2_pipeline_pm_use into v4l2_pipeline_pm_{get, put}

Currently, v4l2_pipeline_pm_use() prototype is:

  int v4l2_pipeline_pm_use(struct media_entity *entity, int use)

Where the 'use' argument shall only be set to '1' for enable/power-on,
or to '0' for disable/power-off. The integer return is specified
as only meaningful when 'use' is set to '1'.

Let's enforce this semantic by splitting the function in two:
v4l2_pipeline_pm_get and v4l2_pipeline_pm_put. This is done
for several reasons.

It makes the API easier to use (or harder to misuse).
It removes the constraint on the values the 'use' argument
shall take. Also, it removes the need to constraint
the return value, by making v4l2_pipeline_pm_put void return.

And last, it's more consistent with other kernel APIs, such
as the runtime pm APIs, which makes the code more symmetric.

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

media: usbvision: deprecate driver

The driver is deprecated and scheduled for removal by the end
of 2020. The reason is that this driver is for old and obsolete
hardware, and it produces a continuous stream of syzbot errors due
to poor code.

In order to prevent removal the following actions would have to
be taken:

- clean up the code
- convert to the vb2 framework
- fix the disconnect and free-on-last-user handling (i.e., add
  a release callback for struct v4l2_device and rework the code
  to use that correctly).

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

media: uapi: h264: Add DPB entry field reference flags

Using the field information attached to v4l2 buffers is not enough to
determine the type of field referenced by a DPB entry: the decoded
frame might contain the full picture (both top and bottom fields)
but the reference only point to one of them.
Let's add new V4L2_H264_DPB_ENTRY_FLAG_ flags to express that.

[Keep only 2 flags and add some details about they mean]

Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: rtl28xxu: add missing sleep before probing slave demod

Slave demod needs some time to wake up otherwise it may not respond to the
following probe commands. This problem manifested randomly on my Astrometa
DVB-T2 dongle.

Signed-off-by: Mario Hros <git@reversity.org>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: arm64: dts: amlogic: add rc-videostrong-kii-pro keymap

videostrong kii pro comes with a nec rc, add the keymap to the dts

Signed-off-by: Mohammad Rasim <mohammad.rasim96@gmail.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: add new kii pro key map

Add new entry for rc-videostrong-kii-pro in linux,rc-map-name

Signed-off-by: Mohammad Rasim <mohammad.rasim96@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: rc: iguanair: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: rc: fix spelling mistake "to" -> "too"

There is a spelling mistake in a dev_dbg message. Fix it.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: vsp1: tidyup VI6_HGT_LBn_H() macro

The address of VSP2_VI6_HGT_LBx_H are
VSP2_VI6_HGT_LB0_H : 0x3428
VSP2_VI6_HGT_LB1_H : 0x3430
VSP2_VI6_HGT_LB2_H : 0x3438
VSP2_VI6_HGT_LB3_H : 0x3440

Thus, VI6_HGT_LBn_H() macro should start from 0x3420 instead of 0x3430.
This patch fixes it.

Fixes: 26e0ca22c3b8 ("[media] v4l: Renesas R-Car VSP1 driver")

Reported-by: Koji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: smiapp: Move definitions under driver directory

include/media/i2c/smiapp.h was meant to serve systems where the sensor is
enumerated through platform data. That's no longer necessary, hopefully
not even in out-of-tree use cases. Move the definitions to the appropriate
headers.

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

media: smiapp: Refactor reading SMIA limits

Combine the two trivial functions reading limits into one. Also rename
smiapp_get_all_limits() as smiapp_read_all_smia_limits().

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

media: smiapp: Move SMIA limit reading up

Move SMIA limit reading up, where other limit and capability handling
takes place.

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

media: smiapp: Turn limit lookup into a function

Instead of direct array access, turn accessing limit information into a
function. Going forward, more elaborate CCS limits will replace most SMIA
limits, and conversion will be less complicated this way.

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

media: smiapp: Use unaligned get and put functions

Use get_unaligned_be* and put_unaligned_be* functions to convert register
values to CPU endianness. Consequently, two instances of BUG() are
removed.

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

media: smiapp: Simplify condition for choosing 8-bit access

Use the only8 boolean to determine whether 8-bit access is required for
reading.

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

media: omap3isp: Prevent enabling CCDC when stopping streaming

Commit ... prevented restarting CCDC through its interrupt handler when
it's about to be disabled. It missed to address the case when CCDC might
be enabled due to queueing a buffer. Do that now.

Fixes: dd12ed17ce9e ("omap3isp: Don't restart CCDC if we're about to stop")

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

media: MAINTAINERS: Sort entries in database for TI VPE/CAL

Run parse-maintainers.pl and choose TI VPE/CAL record. Fix it accordingly.

Note, this is urgent fix, without which parse-maintainers.pl throws
an exception:

Odd non-pattern line '  Documentation/devicetree/bindings/media/ti,cal.yaml
' for 'TI VPE/CAL DRIVERS' at scripts/parse-maintainers.pl line 147, <$file> line 16770.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: staging: media: ipu3: Change 'unsigned long int' to 'unsigned long'

This patch fixes the checkpatch.pl warning:

WARNING: Prefer 'unsigned long' over 'unsigned long int' as the int is
unnecessary

Signed-off-by: Jean-Baptiste Jouband <jj.jouband@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: i2c: Add driver for Sony IMX219 sensor

Adds a driver for the 8MPix Sony IMX219 CSI2 sensor.
Whilst the sensor supports 2 or 4 CSI2 data lanes, this driver
currently only supports 2 lanes.
8MPix @ 15fps, 1080P @ 30fps (cropped FOV), and 1640x1232 (2x2 binned)
@ 30fps are currently supported.

[Sakari Ailus: make imx219_check_hwcfg static]

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: dt-bindings: media: i2c: Add IMX219 CMOS sensor binding

Add YAML device tree binding for IMX219 CMOS image sensor, and
the relevant MAINTAINERS entries.

Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: ipu3.rst: add imgu pipe config tool link

A specific tool which can be used to generate imgu intermedia
resolutions now is public on github, this patch adds this information
into ipu3.rst.

Signed-off-by: Bingbu Cao <bingbu.cao@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: ov5675: add vflip/hflip control support

- Add V4L2 controls: horizontal/vertical flip,
  keep SGRBG10 Bayer order output (via change v/hflip)
- Fix Bayer order output in 1296x972 binning registers

Signed-off-by: Shawn Tu <shawnx.tu@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: dt-bindings: media: venus: Add sc7180 DT schema

Add new qcom,sc7180-venus DT binding schema.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Dikshita Agarwal <dikshita@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: core: add sc7180 DT compatible and resource struct

This add DT compatible string and resource structure for sc7180.

Signed-off-by: Dikshita Agarwal <dikshita@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: venus: delete old binding document

After transitioning to YAML DT schema we don't need this old-style
document.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: core: add sdm845-v2 DT compatible and resource struct

In order to use dynamic video codec assignment add a new sdm845-v2
DT compatible and new venus_resource structure to cover the binding
where all pmdomains and clocks are under the venus_core control.

Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: venus: Add sdm845v2 DT schema

Add new qcom,sdm845-venus-v2 DT binding schema.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: venus: Convert sdm845 to DT schema

Convert qcom,sdm845-venus Venus binding to DT schema.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: venus: Convert msm8996 to DT schema

Convert qcom,msm8996-venus Venus binding to DT schema.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: media: venus: Convert msm8916 to DT schema

Convert qcom,msm8916-venus Venus binding to DT schema

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: vdec: handle 10bit bitstreams

Handle 10bit video streams in the decoder by using dithering, i.e
the decoder output buffers will be in 8bit format.

The runtime handling is implemented by sending v4l2 event to
userspace application, then the application should stop the
streaming on capture queue and initiate format negotiation, and
start streaming again.

Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Co-developed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: introduce core selection

Presently the core (vcodec pipelines) assignment is static. Here we
introduce dynamic load balancing across the cores depending on the
current session load. The load on earch core is calculated and core
with minimum load is assigned to given instance. This will be
applicable on Venus v4 with more than one vcodec cores.

Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Co-developed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: redesign clocks and pm domains control

Redesign core (vcodec) clock control to give the venus core more
freedom to control them in order to make possible to use core
selection feature on Venus IP v4.

Move all clock and pmdomain functions in separate file and abstract
power control with common operations per Venus IP version.

Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: cache vb payload to be used by clock scaling

Instead of iterate over previously queued buffers in clock
scaling code do cache the payload in instance context structure
for later use when calculating new clock rate.

This will avoid to use spin locks during buffer list iteration
in clock_scaling.

This fixes following kernel Oops:

 Unable to handle kernel paging request at virtual address deacfffffffffd6c
 Mem abort info:
   ESR = 0x96000004
   EC = 0x25: DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
 Data abort info:
   ISV = 0, ISS = 0x00000004
   CM = 0, WnR = 0
 [deacfffffffffd6c] address between user and kernel address ranges
 Internal error: Oops: 96000004 [#1] PREEMPT SMP
 CPU: 7 PID: 5763 Comm: V4L2DecoderThre Tainted: G S      W         5.4.11 #8
 pstate: 20400009 (nzCv daif +PAN -UAO)
 pc : load_scale_v4+0x4c/0x2bc [venus_core]
 lr : session_process_buf+0x18c/0x1c0 [venus_core]
 sp : ffffffc01376b8d0
 x29: ffffffc01376b8d0 x28: ffffff80cf1b0220
 x27: ffffffc01376bba0 x26: ffffffd8f562b2d8
 x25: ffffff80cf1b0220 x24: 0000000000000005
 x23: ffffffd8f5620d98 x22: ffffff80ca01c800
 x21: ffffff80cf1b0000 x20: ffffff8149490080
 x19: ffffff8174b2c010 x18: 0000000000000000
 x17: 0000000000000000 x16: ffffffd96ee3a0dc
 x15: 0000000000000026 x14: 0000000000000026
 x13: 00000000000055ac x12: 0000000000000001
 x11: deacfffffffffd6c x10: dead000000000100
 x9 : ffffff80ca01cf28 x8 : 0000000000000026
 x7 : 0000000000000000 x6 : ffffff80cdd899c0
 x5 : ffffff80cdd899c0 x4 : 0000000000000008
 x3 : ffffff80ca01cf28 x2 : ffffff80ca01cf28
 x1 : ffffff80d47ffc00 x0 : ffffff80cf1b0000
 Call trace:
  load_scale_v4+0x4c/0x2bc [venus_core]
  session_process_buf+0x18c/0x1c0 [venus_core]
  venus_helper_vb2_buf_queue+0x7c/0xf0 [venus_core]
  __enqueue_in_driver+0xe4/0xfc [videobuf2_common]
  vb2_core_qbuf+0x15c/0x338 [videobuf2_common]
  vb2_qbuf+0x78/0xb8 [videobuf2_v4l2]
  v4l2_m2m_qbuf+0x80/0xf8 [v4l2_mem2mem]
  v4l2_m2m_ioctl_qbuf+0x2c/0x38 [v4l2_mem2mem]
  v4l_qbuf+0x48/0x58
  __video_do_ioctl+0x2b0/0x39c
  video_usercopy+0x394/0x710
  video_ioctl2+0x38/0x48
  v4l2_ioctl+0x6c/0x80
  do_video_ioctl+0xb00/0x2874
  v4l2_compat_ioctl32+0x5c/0xcc
  __se_compat_sys_ioctl+0x100/0x2074
  __arm64_compat_sys_ioctl+0x20/0x2c
  el0_svc_common+0xa4/0x154
  el0_svc_compat_handler+0x2c/0x38
  el0_svc_compat+0x8/0x10
 Code: eb0a013f 54000200 aa1f03e8 d10e514b (b940016c)
 ---[ end trace e11304b46552e0b9 ]---

Fixes: c0e284ccfeda ("media: venus: Update clock scaling")

Cc: stable@vger.kernel.org # v5.5+
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: venc: blacklist two encoder properties

Those two properties are not implemented for Venus v4 (sdm845),
thus don't set them to firmware.

Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: firmware: Use %pR to print IO resource

Replace explicit use of members of struct resource by %pR to print
the resource.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: venus: hfi_parser: Ignore HEVC encoding for V1

Some older MSM8916 Venus firmware versions also seem to indicate
support for encoding HEVC, even though they really can't.
This will lead to errors later because hfi_session_init() fails
in this case.

HEVC is already ignored for "dec_codecs", so add the same for
"enc_codecs" to make these old firmware versions work correctly.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: MAINTAINERS: Add dt-binding files for Venus

Adds a path to devicetree dt-binding files.

Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging/imx: Missing assignment in imx_media_capture_device_register()

There was supposed to be a "ret = " assignment here, otherwise the
error handling on the next line won't work.

Fixes: 64b5a49df486 ("[media] media: imx: Add Capture Device Interface")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: Documentation/media/uapi: more readable unions

Avoid adding an extra columns when describing unions in the documentation.
That makes it much harder to read. See e.g. VIDIOC_QUERY_EXT_CTRLS.

Instead start off a union with 'union {' and end it with an extra row
containing '}'.

This leaves a lot more space for the description of the fields.

This formatting technique was used in a few places already, but this
patches fixes all remaining occurrences of 'union' in the media uAPI.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Nicolas Dufresne <nicolas@ndufresne.ca>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: add serialization to the resizer subdev ops

For subdevices drivers, the drivers themself are responsible
for serializing their operations.
This patch adds serialization to the resizer subdevice.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: add serialization to the isp subdev ops

For subdevices drivers, the drivers themself are responsible
for serializing their operations.
This patch adds serialization to the isp subdevice.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: isp: check for dphy bus before initializations in s_stream

In rkisp1_isp_s_stream it is better to return error in case the
bus type is not dphy before initializing the registers.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: change function to return void instead of int

There are functions that return int but actually return always 0.
Change them to return void and then there is no need to check
for error return value.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: isp: do not set invalid mbus code for pad

When setting source pad, check if the given mbus code is indeed valid
for source pad, if not, then set the default code.
Same for sink pad.

Fixes: d65dd85281fb ("media: staging: rkisp1: add Rockchip ISP1 base driver")
Reported-by: Wojciech Zabolotny <wzab01@gmail.com>
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: rsz: return to userspace the crop value in bayer mainpath

If bayer format is set in mainpath, then crop is not supported.
The core was just resetting crop but wasn't returning the values back to
user space.

Return values taken in consideration by the driver to userpace.

Signed-off-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: rsz: don't ignore set format in bayer mainpath

Fix issue when, in case of raw bayer format in mainpath,
VIDIOC_SUBDEV_S_FMT ioctl didn't have any effect, is wasn't updating the
format of the pad.

There is no crop for mainpath bayer raw data. Remove leftover check in
set_format, since check is already done in crop rkisp1_rsz_set_sink_crop()
function.

Fixes: d65dd85281fb ("media: staging: rkisp1: add Rockchip ISP1 base driver")
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: stats: use consistent bus_info string

Stats entity is using a slightly different bus_info string
"platform: rkisp1" (with a space) instead of "platform:rkisp1" used by
the rest of rkisp1 code.
This causes errors when using v4l2-util tools that uses the bus_info
string to identify the device.

Fixes: 9a28dbd65a88 ("media: staging: rkisp1: add capture device for statistics")
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: use consistent bus_info string for media_dev

Media device is using a slightly different bus_info string
"platform: rkisp1" (with a space) instead of "platform:rkisp1" used by
the rest of rkisp1 code.
This causes errors when using v4l2-util tools that uses the bus_info
string to identify the device.

Fixes: d65dd85281fb ("media: staging: rkisp1: add Rockchip ISP1 base driver")
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: improve inner documentation in rkisp1-isp.c

Improve the documentation in the beginning of the file
rkisp1-isp1.c

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: fix test of return value of media_entity_get_fwnode_pad

media_entity_get_fwnode_pad returns negative value on error
and the pad numeber on success. Therefore change the error test
from 'if (ret)' to 'if (ret < 0)' .

Fixes: d65dd85281fb ("media: staging: rkisp1: add Rockchip ISP1 base driver")
Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: rkisp1: don't lock media's graph_mutex when calling rkisp1_create_links

The function rkisp1_create_links calls media_entity_get_fwnode_pad
and media_create_pad_link in the mc api. Those calls don't require
locking the graph_mutex so remove the locking.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: mc-entity.c: use WARN_ON, validate link pads

Use WARN_ON instead of BUG_ON.

Add two new WARN_ONs to verify that the source pad is really a source
and that the sink pad is really a sink.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil-cisco@xs4all.nl: use ! instead of == NULL for source and sink]
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dt-bindings: Add binding for rk3228 rga

Indicate that the rk3228 rga is compatible with that of the rk3288.

But if any rk3228-specific quirks are identified in future that
require handling logic that differs from what is provided for the
rk3288, then allow for the compatibility string "rockchip,rk3228-rga"
to be matched instead of "rockchip,rk3288-rga".

Signed-off-by: Justin Swartz <justin.swartz@risingedge.co.za>
Acked-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: hantro: Prevent encoders from using post-processing

The post-processing feature is meant to be used by decoding
only. Prevent encoding jobs from enabling it.

Fixes: 8c2d66b036c77 ("media: hantro: Support color conversion via post-processing")
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: hantro: fix extra MV/MC sync space calculation

Add space for MVs and MC sync data to the capture buffers depending on
whether the post processor will be enabled for the new capture format
passed to TRY_FMT, not the currently set capture format.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: hantro: Write quantization table registers in increasing addresses order

Luma and chroma qtables need to be written into two 16-register blocks,
each table consisting of 64 bytes total. The blocks are contiguous and
start at offset 0 for luma and at offset 0x40 for chroma.

The seemingly innocent optimization of writing the two blocks using one
loop causes side effects which result in improper values of quantization
tables being used by the hardware during encoding. Visually this results
in macroblocking artifacts around contrasting edges in encoded images. The
artifacts look like horizontally flipped shadows of the said edges.
Changing the write operations to non-relaxed variant doesn't help.

This patch removes this premature optimization and after this change the
macroblocking artifacts around contrasting edges are gone.

Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: hantro: Write the quantization tables in proper order

The quantization tables as defined in the file (luma_q_table,
chroma_q_table) are in fact in linear order. The JPEG file header, which is
not generated by the hardware, but must be programatically created with the
CPU, expects the table in zigzag order. On the other hand, the hardware
doesn't expect neither linear, nor zigzag order. Instead it expects the
quantization tables in vertical groups of four quantization parameters,
and the groups are organized in blocks of two vertically adjacent groups.
On top of that the blocks must be provided to the hardware in this order:
leftmost top block, leftmost bottom block, second leftmost top block,
second leftmost bottom block and so on. So, if this is the quantization
table in linear order:

0x10, 0x0b, 0x0a, 0x10, 0x18, 0x28, 0x33, 0x3d,
0x0c, 0x0c, 0x0e, 0x13, 0x1a, 0x3a, 0x3c, 0x37,
0x0e, 0x0d, 0x10, 0x18, 0x28, 0x39, 0x45, 0x38,
0x0e, 0x11, 0x16, 0x1d, 0x33, 0x57, 0x50, 0x3e,
0x12, 0x16, 0x25, 0x38, 0x44, 0x6d, 0x67, 0x4d,
0x18, 0x23, 0x37, 0x40, 0x51, 0x68, 0x71, 0x5c,
0x31, 0x40, 0x4e, 0x57, 0x67, 0x79, 0x78, 0x65,
0x48, 0x5c, 0x5f, 0x62, 0x70, 0x64, 0x67, 0x63

then the hardware expects this in its consecutive registers:

0x100c0e0e,
0x0b0c0d11,
0x12183148,
0x1623405c,
0x0a0e1016,
0x1013181d,
0x25374e5f,
0x38405762,

and so on.

Consequently, the same area of memory cannot be used both for dumping it
into the JPEG file header and writing its contents to the hardware
registers. Instead, a separate pair of arrays is added for properly
reordered quantization tables, to be read with get_unaligned_be32()
and linearly written to the registers.

The "ctx" parameter is not needed any more for hantro_jpeg_get_qtable().

Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: hantro: Use standard luma quantization table

The table is actually different in the document than in this file, so align
this file with the document.

Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: hantro: Read be32 words starting at every fourth byte

Since (luma/chroma)_qtable is an array of unsigned char, indexing it
returns consecutive byte locations, but we are supposed to read the arrays
in four-byte words. Consequently, we should be pointing
get_unaligned_be32() at consecutive word locations instead.

Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Cc: stable@vger.kernel.org
Fixes: 00c30f42c7595f "media: rockchip vpu: remove some unused vars"
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: allegro: fix spelling mistake "to" -> "too"

There is a spelling mistake in a v4l2_err message. Fix it.

Signed-off-by: Colin Ian King <colin.king@canonical.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: vimc: streamer: if kthread_stop fails, ignore the error

Ignore errors returned from kthread_stop since the
vimc subdevices should still be notified that
streaming stopped so they can release the memory for
the streaming, and also kthread should be set to NULL.
kthread_stop can return -EINTR in case the thread
did not yet run. This can happen if userspace calls
streamon and streamoff right after.

Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Acked-by: Helen Koike <helen.koike@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: staging: media: rkisp1: make links immutable by default

The only places which make sese to allow users to enable or disable
links are:

* between sensors and isp:
So users can select which sensor should be used while streaming

* between isp and the resizers:
              |
              v here
rkisp1_isp:2 -> rkisp1_resizer_mainpath -> rkisp1_mainpath (capture)
            \-> rkisp1_resizer_selfpath -> rkisp1_selfpath (capture)
              ^ here
              |

So users can disable one of the capture paths when unused, to avoid
worring about matching formats.

Make the following links immutable to simplify userspace:

rkisp1_resizer_mainpath -> rkisp1_mainpath
rkisp1_resizer_selfpath -> rkisp1_selfpath
rkisp1_params           -> rkisp1_isp
rkisp1_isp              -> rkisp1_stats

Signed-off-by: Helen Koike <helen.koike@collabora.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: davinci: fix incorrect pix_fmt assignment

There's a mistakenly written self assignment in
`static int vpfe_enum_fmt_vid_cap(..)`.

Fixed it according to Prabhakar Lad's feedback.

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

media: sun8i: Remove redundant platform_get_irq error message

platform_get_irq() will call dev_err() itself on failure,
so there is no need for the driver to also do this.
This is detected by coccinelle.

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

media: v4l2-dev.h: remove VFL_TYPE_GRABBER

The last driver was converted to use VFL_TYPE_VIDEO, so the old
_GRABBER enum can now be removed.

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

media: media/platform: rename VFL_TYPE_GRABBER to _VIDEO

'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.

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

media: media/pci: rename VFL_TYPE_GRABBER to _VIDEO

'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.

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

media: media/i2c/video-i2c: rename VFL_TYPE_GRABBER to _VIDEO

'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Cc: Matt Ranostay <matt.ranostay@konsulko.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: media/common/saa7146: rename VFL_TYPE_GRABBER to _VIDEO

'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.

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

media: media/usb: rename VFL_TYPE_GRABBER to _VIDEO

'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.

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