media: uvcvideo: Pass allocation size directly to uvc_alloc_urb_buffer

The uvc_alloc_urb_buffer() function implicitly depended on the
stream->urb_size field, which was set by its caller,
uvc_alloc_urb_buffers(). This implicit data flow makes the code harder
to follow.

More importantly, stream->urb_size was updated within the allocation
loop before the allocation was confirmed to be successful. If the
allocation failed, the stream object would be left with a urb_size that
doesn't correspond to valid, allocated URB buffers.

Refactor uvc_alloc_urb_buffer() to accept the buffer size as an explicit
argument. This makes the function's dependencies clear and improves the
robustness of the error handling path. The stream->urb_size is now set only
after a complete and successful allocation.

This is a pure refactoring and introduces no functional changes.

Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Itay Chamiel <itay.chamiel@q.ai>
Link: https://patch.msgid.link/20260114-uvc-alloc-urb-v1-2-cedf3fb66711@chromium.org
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: uvcvideo: Fix allocation for small frame sizes

If a frame has size of less or equal than one packet size
uvc_alloc_urb_buffers() is unable to allocate memory for it due to a
off-by-one error.

Fix the off-by-one-error and now that we are at it, make sure that
stream->urb_size has always a valid value when we return from the
function, even when an error happens.

Fixes: efdc8a9585ce ("V4L/DVB (10295): uvcvideo: Retry URB buffers allocation when the system is low on memory.")
Reported-by: Itay Chamiel <itay.chamiel@q.ai>
Closes: https://lore.kernel.org/linux-media/CANiDSCsSoZf2LsCCoWAUbCg6tJT-ypXR1B85aa6rAdMVYr2iBQ@mail.gmail.com/T/#t
Co-developed-by: Itay Chamiel <itay.chamiel@q.ai>
Signed-off-by: Itay Chamiel <itay.chamiel@q.ai>
Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Itay Chamiel <itay.chamiel@q.ai>
Link: https://patch.msgid.link/20260114-uvc-alloc-urb-v1-1-cedf3fb66711@chromium.org
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: uvcvideo: Return queued buffers on start_streaming() failure

Return buffers if streaming fails to start due to uvc_pm_get() error.

This bug may be responsible for a warning I got running

    while :; do yavta -c3 /dev/video0; done

on an xHCI controller which failed under this workload.
I had no luck reproducing this warning again to confirm.

xhci_hcd 0000:09:00.0: HC died; cleaning up
usb 13-2: USB disconnect, device number 2
WARNING: CPU: 2 PID: 29386 at drivers/media/common/videobuf2/videobuf2-core.c:1803 vb2_start_streaming+0xac/0x120

Fixes: 7dd56c47784a ("media: uvcvideo: Remove stream->is_streaming field")
Cc: stable@vger.kernel.org
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Reviewed-by: Ricardo Ribalda <ribalda@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20251015133642.3dede646.michal.pecio@gmail.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: uvcvideo: Create an ID namespace for streaming output terminals

Some devices, such as the Grandstream GUV3100 and the LSK Meeting Eye
for Business & Home, exhibit entity ID collisions between units and
streaming output terminals.

The UVC specification requires unit and terminal IDs to be unique, and
uses the ID to reference entities:

- In control requests, to identify the target entity
- In the UVC units and terminals descriptors' bSourceID field, to
  identify source entities
- In the UVC input header descriptor's bTerminalLink, to identify the
  terminal associated with a streaming interface

Entity ID collisions break accessing controls and make the graph
description in the UVC descriptors ambiguous. However, collisions where
one of the entities is a streaming output terminal and the other entity
is not a streaming terminal are less severe. Streaming output terminals
have no controls, and, as they are the final entity in pipelines, they
are never referenced in descriptors as source entities. They are
referenced by ID only from innput header descriptors, which by
definition only reference streaming terminals.

For these reasons, we can work around the collision by giving streaming
output terminals their own ID namespace. Do so by setting bit
UVC_TERM_OUTPUT (15) in the uvc_entity.id field, which is normally never
set as the ID is a 8-bit value.

This ID change doesn't affect the entity name in the media controller
graph as the name isn't constructed from the ID, so there should not be
any impact on the uAPI.

Although this change handles some ID collisions automagically, keep
printing an error in uvc_alloc_new_entity() when a camera has invalid
descriptors. Hopefully this message will help vendors fix their invalid
descriptors.

This new method of handling ID collisions includes a revert of commit
758dbc756aad ("media: uvcvideo: Use heuristic to find stream entity")
that attempted to fix the problem urgently due to regression reports.

Suggested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Lili Orosz <lily@floofy.city>
Co-developed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20251113210400.28618-1-laurent.pinchart@ideasonboard.com
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add HEVC support for the VDPU383 variant

The VDPU383 decoder is used on the RK3576 SoC and has support for HEVC.

This patch also moves some functions to a common rkvdec-hevc-common.c
file and adds a specific scaling matrix flatten function.

The fluster score for JCT-VC-HEVC_V1 is 146/147.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add HEVC support for the VDPU381 variant

The VDPU381 supports HEVC decoding up to 7680x4320@30fps.
It could double that when using both decoder cores.

It support YUV420 (8 and 10 bits) as well as AFBC (not implemented
here)

The fluster score is 146/147 for JCT-VC-HEVC_V1, tested on ROCK 5B.
None of the other test suites works.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add H264 support for the VDPU383 variant

This variant is used on the RK3576 SoC.

The moving vectors size requirements are slightly different so support
for a colmv_size function per variant is added.

Also, the link registers are used to start the decoder and read IRQ status.

The fluster score is 128/135 for JVT-AVC_V1, with MPS_MW_A failing in
addition to the usual ones.
The other test suites are not supported yet.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add H264 support for the VDPU381 variant

This decoder variant is found in Rockchip RK3588 SoC family.

Like for rkvdec on rk3399, it supports the NV12, NV15, NV16 and NV20
output formats and level up to 6.0.

The maximum width and height have been significantly increased
supporting up to 65520 pixels for both.

Support for named register sections is added for this variant and future
ones.

Fluster score for JVT-AVC_V1 is 129/135.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Disable multicore support

Similarly to what is done in Hantro, avoid exposing equal video codecs to
userspace. Equal video codecs allow scheduling work between the cores.
For that kernel support is required, which does not yet exist.
Until that is implemented, avoid exposing each core separately to
userspace so that multicore can be added in the future without breaking
userspace ABI.

This currently applies only to RK3588 which has 2 equal VDPU381 decoders,
but will be applied for all SoC supported by rkvdec that has multiple DTS
nodes with the same compatible.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Enable all clocks without naming them

For other variants, the clock names and number will differ.

There is no need to keep track of the clock names in the driver so drop
them to avoid having a list for each variant.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Support per-variant interrupt handler

Prepare for supporting different variants with different interrupt
managers.

To support other variants specific function type later, introduce the
rkvdec_variant_ops struct.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add RCB and SRAM support

The RCB (Rows and Cols Buffers) are a set of buffers used by other
variations of the decoder to store temporary data.

Those variation come with a dedicated SRAM area used to store those
buffers for better performances.

The buffer sizes are either the width or height of the frame being
decoded multiplied by a documented factor and can be stored either
in SRAM or RAM.
A fallback to RAM is provided if the SRAM is full (e.g.: multiple
streams are being decoded at the same time).

To manage the different kind of allocation, an enum is added to the
rkvdec_aux_buf struct to specify how the buffer was allocated, and
so, how to free it.

This commit is in preparation of other variants support.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Add variant specific coded formats list

Prepare for adding new variants of the decoder and support specific
formats and format ops per variant.

This removes the need of capability flags for variants, so remove them.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Move hevc functions to common file

This is a preparation commit to add support for new variants of the
decoder.

The functions will later be shared with vdpu381 (rk3588) and vdpu383
(rk3576).

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Move h264 functions to common file

This is a preparation commit to add support for new variants of the
decoder.

The functions will later be shared with vdpu381 (rk3588) and vdpu383
(rk3576).

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Use structs to represent the HW RPS

This is in preparation to add support for other variants of the decoder.

Moving to struct representation is mainly to prepare for multicore
support that is present in e.g. rk3588.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Move cabac tables to their own source file

This is in preparation to add support for new variants that will use the
same tables.

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkvdec: Switch to using structs instead of writel

In an effort to merge the rkvdec2 driver [1] with this one, switch from
writel() calls to using structs to represent the register mappings.

This is done in order to have all supported decoders use the same format
in the future and ease reading of the code.

Using structs also improves stability as the hardware is tested and
validated downstream using a similar method.
It was noticed, on decoders, that:
 - Some registers require to be writen in increasing order [2]
 - Some registers, even if unrelated, need to be written to their reset
   values (it was the case here for axi_ddr_[rw]data).

Using structs can also help improving performance later when, e.g.
multicore support is added on RK3588.

Performance seems to be slightly improved, but at least, not made worse.
Running fluster's JVT-AVC_V1 test suite with GStreamer on the Radxa ROCK
PI 4 SE gives the following times:

Before this patch:

- --jobs 1: Ran 129/135 tests successfully               in 77.167 secs
- --jobs 6: Ran 129/135 tests successfully               in 23.046 secs

With this patch:
- --jobs 1: Ran 129/135 tests successfully               in 70.698 secs
- --jobs 6: Ran 129/135 tests successfully               in 22.917 secs

This also shows that the fluster score hasn't changed.

[1]: https://lore.kernel.org/all/20250325213303.826925-1-detlev.casanova@collabora.com/
[2]: https://lore.kernel.org/all/20200127143009.15677-5-andrzej.p@collabora.com/

Tested-by: Diederik de Haas <didi.debian@cknow.org> # Rock 5B
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: visl: Add HEVC short and long term RPS sets

Log the recently added v4l2 controls to set HEVC short and long term RPS
sets with 2 new ftrace entries.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: v4l2-ctrls: Add hevc_ext_sps_[ls]t_rps controls

The vdpu381 decoder found on newer Rockchip SoC need the information
from the long term and short term ref pic sets from the SPS.

So far, it wasn't included in the v4l2 API, so add it with new dynamic
sized controls.

Each element of the hevc_ext_sps_lt_rps array contains the long term ref
pic set at that index.
Each element of the hevc_ext_sps_st_rps contains the short term ref pic
set at that index, as the raw data.
It is the role of the drivers to calculate the reference sets values.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: uapi: HEVC: Add v4l2_ctrl_hevc_ext_sps_[ls]t_rps controls

Some hardware (e.g.: Rockchip's rk3588 hevc decoder) need to fully parse
the slice header, which cannot be passed with the current controls.
There is also no skip method similar as to what can be found in
verisilicon HW.

The SPS is therefore extended through these 2 new controls, providing the
long and short term reference information from the slice header for HEVC
decoding.

These now controls are similar as what is passed in the Vulkan Video API
with the StdVideoH265ShortTermRefPicSet and StdVideoH265LongTermRefPicsSps
structures embedded in the SPS struct.

Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: verisilicon: AV1: Fix tile info buffer size

Each tile info is composed of: row_sb, col_sb, start_pos
and end_pos (4 bytes each). So the total required memory
is AV1_MAX_TILES * 16 bytes.
Use the correct #define to allocate the buffer and avoid
writing tile info in non-allocated memory.

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Fixes: 727a400686a2c ("media: verisilicon: Add Rockchip AV1 decoder")
Cc: stable@vger.kernel.org
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: synopsys: add driver for the designware mipi csi-2 receiver

The Synopsys DesignWare MIPI CSI-2 Receiver is a CSI-2 bridge with
one input port and one output port. It receives the data with the
help of an external MIPI PHY (C-PHY or D-PHY) and passes it to e.g.,
the Rockchip Video Capture (VICAP) block on recent Rockchip SoCs.

Add a V4L2 subdevice driver for this unit.

Signed-off-by: Michael Riesch <michael.riesch@wolfvision.net>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Mehdi Djait <mehdi.djait@linux.intel.com>
Signed-off-by: Michael Riesch <michael.riesch@collabora.com>
[Sakari Ailus: Make sparse and smatch happy.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: dt-bindings: add rockchip mipi csi-2 receiver

Add documentation for the Rockchip MIPI CSI-2 Receiver.

Signed-off-by: Michael Riesch <michael.riesch@wolfvision.net>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Michael Riesch <michael.riesch@collabora.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: staging: media: imx6-mipi-csi2: use devm_platform_ioremap_resource() simplify code

Use devm_platform_ioremap_resource() simplify code. No functional change.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260116-stage-csi2-cleanup-v2-5-a56e9cb25196@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: staging: media: imx6-mipi-csi2: use devm_mutex_init() to simplify code

Use devm_mutex_init() to simplify the code. No functional change.

Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20260116-stage-csi2-cleanup-v2-2-a56e9cb25196@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: staging: media: imx6-mipi-csi2: replace spaces with tabs for alignment

Replace spaces with tabs to align register value definitions, making it
easier to add new entries and maintain consistent formatting.

Also use a space between the type and field in struct csi2_dev.

No functional change.

Signed-off-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20260116-stage-csi2-cleanup-v2-1-a56e9cb25196@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: nxp: use cleanup __free(fwnode_handle) simplify code

Use cleanup __free(fwnode_handle) simplify code. No functional change.

Signed-off-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20260116-cam_cleanup-v4-3-29ce01640443@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: nxp: imx8-isi: use devm_pm_runtime_enable() to simplify code

Use devm_pm_runtime_enable() to simplify code. Change to use
dev_err_probe() because previous goto change to return.

No functional change.

Signed-off-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20260116-cam_cleanup-v4-2-29ce01640443@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: nxp: use dev_err_probe() to simplify code

Use dev_err_probe() to simplify the code. Drop the explicit error message
after returning from imx8mq_mipi_csi_parse_dt(), as the error is already
reported by this helper.

No functional change.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Alexander Stein <alexander.stein@ew.tq-group.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260116-cam_cleanup-v4-1-29ce01640443@nxp.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkisp1: Fix filter mode register configuration

The rkisp1_flt_config() function performs an initial direct write to
RKISP1_CIF_ISP_FILT_MODE without including the RKISP1_CIF_ISP_FLT_ENA
bit, which clears the filter enable bit in the hardware.

The subsequent read/modify/write sequence then reads back the register
with the enable bit already cleared and cannot restore it, resulting in
the filter being inadvertently disabled.

Remove the redundant direct write. The read/modify/write sequence alone
correctly preserves the existing enable bit state while updating the
DNR mode and filter configuration bits.

Signed-off-by: Rui Wang <rui.wang@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/20260105171142.147792-2-rui.wang@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: rkisp1: Discard pm_runtime_put() return value

Printing error messages on pm_runtime_put() returning negative values
is not particularly useful.

Returning an error code from pm_runtime_put() merely means that it has
not queued up a work item to check whether or not the device can be
suspended and there are many perfectly valid situations in which that
can happen, like after writing "on" to the devices' runtime PM "control"
attribute in sysfs for one example.

Accordingly, update rkisp1_vb2_stop_streaming() to simply discard the
return value of pm_runtime_put().

This will facilitate a planned change of the pm_runtime_put() return
type to void in the future.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patch.msgid.link/2356323.iZASKD2KPV@rafael.j.wysocki
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: imx8-isi: Drop unneeded module alias

The imx8-isi driver has a module alias named "ISI". This is not
required, as there is no reason to load this module through an alias:
the device is probed through OF, and the module has never been named
"ISI". Drop the alias.

Suggested-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20251102234438.8022-1-laurent.pinchart@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: dt-bindings: media: renesas,fcp: Allow three clocks for RZ/V2N SoC

Update the FCP DT schema to permit three clock inputs for the RZ/V2N SoC.
The FCP block on this SoC requires three separate clocks, unlike other
variants which use only one.

Fixes: f42eddf44fbf ("media: dt-bindings: media: renesas,fcp: Document RZ/V2N SoC")
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://patch.msgid.link/20251103194554.54313-1-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Introduce vpu ops for vpu4 with necessary hooks

Add power sequence for vpu4 by reusing from previous generation wherever
possible. Hook up vpu4 op with vpu4 specific implemtation or resue from
earlier generation wherever feasible, like clock calculation in this
case.

Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Move vpu35 specific api to common to use for vpu4

Some of the sequence and calculations for vpu4 is identical to vpu35,
namely power sequence for vpu controller and the clock frequency
calculation. Move those to common file that can be shared for both vpu35
and vpu4. This patch prepares for power sequence for vpu4 which is added
in subsequent patch.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Move vpu register defines to common header file

Some of vpu4 register defines are common with vpu3x. Move those into the
common register defines header. This is done to reuse the defines for
vpu4 in subsequent patch which enables the power sequence for vpu4.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Co-developed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Introduce buffer size calculations for vpu4

Introduces vp4 buffer size calculation for both encoder and decoder.
Reuse the buffer size calculation which are common, while adding the
vpu4 ones separately.

Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Add support for multiple TZ content protection(CP) configs

vpu4 needs an additional configuration with respect to CP regions. Make
the CP configuration as array such that the multiple configuration can be
managed per platform.

Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Add support for multiple clock sources

vpu4 depends on more than one clock source. Thus far hardware versions
up to vpu3x have been clocked by a single source.
This adds support for multiple clocks by,
- Adding a lookup table
- Configuring OPP table for video device with different video clocks
- Setting OPP for multiple clocks during dev_pm_opp_set_opp()

This patch extends the support for multiple clocks in driver, which
would be used in subsequent patch for kaanapali, when the platform data
is prepared.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Co-developed-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Add intra refresh support for encoder

Add support for V4L2_CID_MPEG_VIDEO_INTRA_REFRESH_PERIOD and
V4L2_CID_MPEG_VIDEO_INTRA_REFRESH_PERIOD_TYPE controls.

Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Add flip support for encoder

Add support for V4L2_CID_HFLIP and V4L2_CID_VFLIP controls in encoder.

Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Add rotation support for encoder

Add rotation control for encoder, enabling V4L2_CID_ROTATE and handling
 90/180/270 degree rotation.

Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Co-developed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Add scale support for encoder

Add members enc_scale_width, enc_scale_height to the struct iris_inst to
support scale requirements.

Add output width and height settings in iris_venc_s_fmt_output to
enable scaling functionality.

Add VPSS buffer to platform data, which the scale function requires.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Improve crop_offset handling for encoder

The setting of HFI_PROP_CROP_OFFSETS for the OUTPUT port is correct,
but on the CAPTURE port it is used to inform the firmware about the
region of interest, so crop_offset needs to be handled accordingly.

Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: qcom: iris: Improve format alignment for encoder

Add members enc_raw_width, enc_raw_height to the struct iris_inst to
support codec alignment requirements.

HFI_PROP_RAW_RESOLUTION needs to be set to the actual YUV resolution.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK
Signed-off-by: Wangao Wang <wangao.wang@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Change psc properties message to debug level

The message for attempting to set already configured input or output
parameter subscribe change properties is not an error condition.
Move it from error to debug level and make it more descriptive by
indicating which specific input or output psc was already set.

Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Document difference in size during allocation

As we get ready for kzalloc checking for invalid sizes, let's add
documentation for the cases where the size is different but valid.

This patch fixes this cocci warning:
./platform/qcom/iris/iris_hfi_gen2_command.c:1215:9-25: WARNING: casting value returned by memory allocation function to (struct iris_inst *) is useless.

Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Prevent output buffer queuing before stream-on completes

During normal playback, stream-on for input is followed by output, and
only after input stream-on does actual streaming begin. However, when
gst-play performs a seek, both input and output streams are stopped,
and on restart, output stream-on occurs first. At this point, firmware
has not yet started streaming. Queuing output buffers before the firmware
begins streaming causes it to process buffers in an invalid state, leading
to an error response. These buffers are returned to the driver as errors,
forcing the driver into an error state and stopping playback.

Fix this by deferring output buffer queuing until stream-on completes.
Input buffers can still be queued before stream-on as required.

Fixes: 92e007ca5ab6 ("media: iris: Add V4L2 streaming support for encoder video device")
Signed-off-by: Vishnu Reddy <busanna.reddy@oss.qualcomm.com>
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: gen2: Add sanity check for session stop

In iris_kill_session, inst->state is set to IRIS_INST_ERROR and
session_close is executed, which will kfree(inst_hfi_gen2->packet).
If stop_streaming is called afterward, it will cause a crash.

Add a NULL check for inst_hfi_gen2->packet before sendling STOP packet
to firmware to fix that.

Fixes: 11712ce70f8e ("media: iris: implement vb2 streaming ops")
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: gen1: Destroy internal buffers after FW releases

After the firmware releases internal buffers, the driver was not
destroying them. This left stale allocations that were no longer used,
especially across resolution changes where new buffers are allocated per
the updated requirements. As a result, memory was wasted until session
close.

Destroy internal buffers once the release response is received from the
firmware.

Fixes: 73702f45db81 ("media: iris: allocate, initialize and queue internal buffers")
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Skip resolution set on first IPSC

The resolution property is not supposed to be set during reconfig.
Existing iris_drc_pending(inst) check is insufficient, as it doesn't
cover the first port setting change.

Extend the conditional check to also skip resolution setting when
the instance is in IRIS_INST_SUB_FIRST_IPSC.

Fixes: caf205548769 ("media: iris: Avoid updating frame size to firmware during reconfig")
Reviewed-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Add buffer to list only after successful allocation

Move `list_add_tail()` to after `dma_alloc_attrs()` succeeds when creating
internal buffers. Previously, the buffer was enqueued in `buffers->list`
before the DMA allocation. If the allocation failed, the function returned
`-ENOMEM` while leaving a partially initialized buffer in the list, which
could lead to inconsistent state and potential leaks.

By adding the buffer to the list only after `dma_alloc_attrs()` succeeds,
we ensure the list contains only valid, fully initialized buffers.

Fixes: 73702f45db81 ("media: iris: allocate, initialize and queue internal buffers")
Reviewed-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Add missing platform data entries for SM8750

Two platform-data fields for SM8750 were missed:

  - get_vpu_buffer_size = iris_vpu33_buf_size
    Without this, the driver fails to allocate the required internal
    buffers, leading to basic decode/encode failures during session
    bring-up.

  - max_core_mbps = ((7680 * 4320) / 256) * 60
    Without this capability exposed, capability checks are incomplete and
    v4l2-compliance for encoder fails.

Fixes: a5925a2ce077 ("media: iris: add VPU33 specific encoding buffer calculation")
Fixes: a6882431a138 ("media: iris: Add support for ENUM_FRAMESIZES/FRAMEINTERVALS for encoder")
Cc: stable@vger.kernel.org
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: remove v4l2_m2m_ioctl_{de,en}coder_cmd API usage during STOP handling

Currently v4l2_m2m_ioctl_{de,enc}coder_cmd is being invoked during STOP
command handling. However, this is not required as the iris driver has
its own drain and stop handling mechanism in place.

Using the m2m command API in this context leads to incorrect behavior,
where the LAST flag is prematurely attached to a capture buffer,
when there are no buffers in m2m source queue. But, in this scenario
even though the source buffers are returned to client, hardware might
still need to process the pending capture buffers.

Attaching LAST flag prematurely can result in the capture buffer being
removed from the destination queue before the hardware has finished
processing it, causing issues when the buffer is eventually returned by
the hardware.

To prevent this, remove the m2m API usage in stop handling.

Fixes: d09100763bed ("media: iris: add support for drain sequence")
Fixes: 75db90ae067d ("media: iris: Add support for drain sequence in encoder video device")
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: use fallback size when S_FMT is called without width/height

According to 4.5.1.5 of the M2M stateful decoder UAPI documentation,
providing the width and the height to S_FMT is "required only if it
cannot be parsed from the stream", otherwise they can be left as 0
and the S_FMT implementation is expected to return a valid placeholder
resolution that would let REQBUFS succeed.

iris was missing the fallback, so clients like rpi-ffmpeg wouldn't work.
Fix by adding an explicit fallback to defaults.

Fixes: b530b95de22c ("media: iris: implement s_fmt, g_fmt and try_fmt ioctls")
Link: https://github.com/jc-kynesim/rpi-ffmpeg/issues/103
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Signed-off-by: Val Packett <val@packett.cool>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Fix fps calculation

iris_venc_s_param() uses do_div to divide two 64 bits operators, this is
wrong. Luckily for us, both of the operators fit in 32 bits, so we can use
a normal division.

Now that we are at it, mark the fps smaller than 1 as invalid, the code
does not seem to handle them properly.

The following cocci warning is fixed with this patch:
./platform/qcom/iris/iris_venc.c:378:1-7: WARNING: do_div() does a 64-by-32 division, please consider using div64_u64 instead

Fixes: 4ff586ff28e3 ("media: iris: Add support for G/S_PARM for encoder video device")
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: iris: Fix ffmpeg corrupted frame error

When the ffmpeg decoder is running, the driver receives the
V4L2_BUF_FLAG_KEYFRAME flag in the input buffer. The driver then forwards
this flag information to the firmware. The firmware, in turn, copies the
input buffer flags directly into the output buffer flags. Upon receiving
the output buffer from the firmware, the driver observes that the buffer
contains the HFI_BUFFERFLAG_DATACORRUPT flag. The root cause is that both
V4L2_BUF_FLAG_KEYFRAME and HFI_BUFFERFLAG_DATACORRUPT are the same value.
As a result, the driver incorrectly interprets the output frame as
corrupted, even though the frame is actually valid. This misinterpretation
causes the driver to report an error and skip good frames, leading to
missing frames in the final video output and triggering ffmpeg's "corrupt
decoded frame" error.

To resolve this issue, the input buffer flags should not be sent to the
firmware during decoding, since the firmware does not require this
information.

Fixes: 17f2a485ca67 ("media: iris: implement vb2 ops for buf_queue and firmware response")
Cc: stable@vger.kernel.org
Signed-off-by: Vishnu Reddy <quic_bvisredd@quicinc.com>
Reviewed-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

Revert "media: iris: Add sanity check for stop streaming"

This reverts commit ad699fa78b59241c9d71a8cafb51525f3dab04d4.

Revert the check that skipped stop_streaming when the instance was in
IRIS_INST_ERROR, as it caused multiple regressions:

1. Buffers were not returned to vb2 when the instance was already in
   error state, triggering warnings in the vb2 core because buffer
   completion was skipped.

2. If a session failed early (e.g. unsupported configuration), the
   instance transitioned to IRIS_INST_ERROR. When userspace attempted
   to stop streaming for cleanup, stop_streaming was skipped due to the
   added check, preventing proper teardown and leaving the firmware
   in an inconsistent state.

Fixes: ad699fa78b59 ("media: iris: Add sanity check for stop streaming")
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: venus: assign unique bus_info strings for encoder and decoder

The Venus encoder and decoder video devices currently report the same
bus_info string ("platform:qcom-venus").

Assign unique bus_info identifiers by appending ":dec" and ":enc" to the
parent device name. With this change v4l2-ctl will display two separate
logical devices

Signed-off-by: Jorge Ramirez-Ortiz <jorge.ramirez@oss.qualcomm.com>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: venus: vdec: restrict EOS addr quirk to IRIS2 only

On SM8250 (IRIS2) with firmware older than 1.0.087, the firmware could
not handle a dummy device address for EOS buffers, so a NULL device
address is sent instead. The existing check used IS_V6() alongside a
firmware version gate:

    if (IS_V6(core) && is_fw_rev_or_older(core, 1, 0, 87))
        fdata.device_addr = 0;
    else
fdata.device_addr = 0xdeadb000;

However, SC7280 which is also V6, uses a firmware string of the form
"1.0.<commit-hash>", which the version parser translates to 1.0.0. This
unintentionally satisfies the `is_fw_rev_or_older(..., 1, 0, 87)`
condition on SC7280. Combined with IS_V6() matching there as well, the
quirk is incorrectly applied to SC7280, causing VP9 decode failures.

Constrain the check to IRIS2 (SM8250) only, which is the only platform
that needed this quirk, by replacing IS_V6() with IS_IRIS2(). This
restores correct behavior on SC7280 (no forced NULL EOS buffer address).

Fixes: 47f867cb1b63 ("media: venus: fix EOS handling in decoder stop command")
Cc: stable@vger.kernel.org
Reported-by: Mecid <mecid@mecomediagroup.de>
Closes: https://github.com/qualcomm-linux/kernel-topics/issues/222
Co-developed-by: Renjiang Han <renjiang.han@oss.qualcomm.com>
Signed-off-by: Renjiang Han <renjiang.han@oss.qualcomm.com>
Signed-off-by: Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>
Tested-by: Renjiang Han <renjiang.han@oss.qualcomm.com>
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: venus: vdec: fix error state assignment for zero bytesused

When hfi_session_flush is issued, all queued buffers are returned to
the V4L2 driver. Some of these buffers are not processed and have
bytesused = 0. Currently, the driver marks such buffers as error even
during drain operations, which can incorrectly flag EOS buffers.

Only capture buffers with zero payload (and not EOS) should be marked
with VB2_BUF_STATE_ERROR. The check is performed inside the non-EOS
branch to ensure correct handling.

Fixes: 51df3c81ba10b ("media: venus: vdec: Mark flushed buffers with error state")
Signed-off-by: Renjiang Han <renjiang.han@oss.qualcomm.com>
Reviewed-by: Vikash Garodia <vikash.garodia@oss.qualcomm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bod@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

Documentation: admin-guide: media: mgb4: Add GMSL1 & GMSL3-coax modules info

Add the mgb4 GMSL1 and GMSL3-coax modules info.

Signed-off-by: Martin Tůma <martin.tuma@digiteqautomotive.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: pci: mgb4: Add support for GMSL3 coaxial modules

Add support for GMSL3 modules with a coaxial interface, a variant of the
already existing module with the id #4. The FW is the same as for all other
GMSL3 modules.

Signed-off-by: Martin Tůma <martin.tuma@digiteqautomotive.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: pci: mgb4: Add support for GMSL1 modules

Add support for GMSL1 modules. GMSL1 modules have 2 inputs, 2 outputs and
require a separate FW (id #3).

Signed-off-by: Martin Tůma <martin.tuma@digiteqautomotive.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: os05b10: Few minor improvements

Return 0 from disable_streams, use div_u64() instead of do_div(), use
dev_dbg() for a debug message and move loop variable definition into the
loop itself.

Signed-off-by: Himanshu Bhavani <himanshu.bhavani@siliconsignals.io>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov9282: dynamic flash_duration maximum

This patch sets the current exposure time as maximum for the
flash_duration control. As Flash/Strobes which are longer than the
exposure time have no effect.

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
[Sakari Ailus: Some lines rewrapped.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov9282: implement try_ctrl for strobe_duration

As the granularity of the hardware supported values is lower than the
control value, implement a try_ctrl() function for
V4L2_CID_FLASH_DURATION. This function calculates the nearest possible
µs strobe duration for the given value and returns it back to the
caller.

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov9282: add strobe_duration v4l2 control

Add V4L2_CID_FLASH_DURATION support using the "strobe_frame_span"
feature of the sensor. This is implemented by transforming the given µs
value by an interpolated formula to a "span step width" value and
writing it to register PWM_CTRL_25, PWM_CTRL_26, PWM_CTRL_27,
PWM_CTRL_28 (0x3925, 0x3926, 0x3927, 0x3928).

The maximum control value is set to the period of the current default
framerate.

All register values are based on the OV9281 datasheet v1.53 (jan 2019)
and tested using an ov9281 VisionComponents module.

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
[Sakari Ailus: Remove redundant assignment of ret, declare ret as last.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov9282: add strobe output enable v4l2 control

Add V4L2_CID_FLASH_STROBE_OE enable/disable support using the
"strobe output enable" feature of the sensor.

All values are based on the OV9281 datasheet v1.53 (january 2019) and
tested using an ov9281 VisionComponents module.

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov9282: add output enable register definitions

Add #define's for the output enable registers (0x3004, 0x3005, 0x3006),
also known as SC_CTRL_04, SC_CTRL_05, SC_CTRL_04. Use those register
definitions instead of the raw values in the `common_regs` struct.

All values are based on the OV9281 datasheet v1.53 (january 2019).

Reviewed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

Documentation: uAPI: media: add V4L2_CID_FLASH_{DURATION,STROBE_OE}

Add the new strobe duration and hardware strobe output enable to v4l
uAPI documentation. Additionally add labels for cross-referencing v4l
controls.

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: v4l: ctrls: add a control for enabling strobe output

Add a control V4L2_CID_FLASH_STROBE_OE to en- or disable the
strobe output of v4l2 devices (most likely sensors).

Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: v4l: ctrls: add a control for flash/strobe duration

Add a V4L2_CID_FLASH_DURATION control to set the duration of a
flash/strobe pulse. This controls the length of the flash/strobe pulse
output by device (typically a camera sensor) and connected to the flash
controller. This is different to the V4L2_CID_FLASH_TIMEOUT control,
which is implemented by the flash controller and defines a limit after
which the flash is "forcefully" turned off again.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Richard Leitner <richard.leitner@linux.dev>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: imx219: Inline imx219_update_pad_format() in its caller

The imx219_update_pad_format() is short and called from a single place,
in imx219_set_pad_format(). Inline the code in the caller to keep all
format adjustments grouped in a single place and improve readability.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Make supported mbus code bitmask a u64

Prepare for adding more media bus codes.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Remove which parameter from ccs_propagate

ccs_propagate() no longer stores information in the driver's context
struct. The which parameter can thus be removed.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Compute scaling configuration from sub-device state

Compute scaling configuration from sub-device state instead of storing it
to the driver's device context struct.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Compute binning configuration from sub-device state

Calculate binning configuration from sub-device state so the state related
configuration can be removed from the driver's device context struct.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Switch to guard and scoped_guard

Replace the use of mutex_{,un}lock() by guard() and scoped_guard() where
it makes sense (i.e. everywhere).

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Rely on sub-device state locking

Rely on sub-device state locking to serialise access to driver's data
structures. The driver-provided mutex is used as the state lock for all
driver sub-devices.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Remove ccs_get_crop_compose helper

As it's now easier to obtain the necessary information on crop and compose
rectangles after moving to sub-device state, remove the
ccs_get_crop_compose helper.

Also remove the comp arguments of the compose goodness calculators and
make related local variables and function arguments const where
applicable.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov01a10: Fix digital gain range

Digital gain wraps-around at the maximum of 16838 / 0x3fff.
Fix the maximum digital gain by setting it to 0x3fff.

Signed-off-by: Mehdi Djait <mehdi.djait@linux.intel.com>
Reviewed-by: Hans de Goede <johannes.goede@oss.qualcomm.com>
Fixes: 0827b58dabff ("media: i2c: add ov01a10 image sensor driver")
Cc: stable@vger.kernel.org
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ov01a10: Fix indentation

Fix indentation in ov01a10_get_pm_resources().

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ipu6: Update e-mail addresses

Remove Hongju's e-mail address that no longer works.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: bytecap: remove useless cmier variable

The struct dcmipp_bytecap_device embeds a cmier variable which is
actually only set once and then used in several places for interrupt
mask or interrupt enable/disable. Use the DCMIPP_CMIER_P0ALL
bitfield macro directly instead of using it through the cmier
variable.

There is also no need to mask again vcap->cmsr2 against enabled
interrupts bit since this has already been done when storing the
interrupt status into the cmsr2 variable.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: byteproc: only allow compose with 8/16bit fmts

Since the byteproc is working at byte level, it is not possible to
perform decimation with formats other than 8 or 16 bit per pixel.
Ensure any other format won't be able to set a compose.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: add Y10-Y12-Y14 in all subdevs

DCMIPP can handle Y8 to Y14 input formats however until now only
Y8 was handled. Add support for Y10-Y12-Y14 in all relevant
subdevs.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: Add RGB888 format support overall

Add entries within several subdevs in order to handle RGB888 format.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: byteproc: disable compose for all bayers

Avoid possibility to perform compose on all frames which mbus code is
within the bayer range or jpeg format.

Fixes: 822c72eb1519 ("media: stm32: dcmipp: add bayer 10~14 bits formats")
Cc: stable@vger.kernel.org
Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: bytecap: clear all interrupts upon stream stop

Ensure that there are no pending interrupts after we have stopped the
pipeline. Indeed, it could happen that new interrupt has been generated
during the stop_streaming processing hence clear them in order to avoid
getting a new interrupt right from the start of a next start_streaming.

Fixes: 28e0f3772296 ("media: stm32-dcmipp: STM32 DCMIPP camera interface driver")
Cc: stable@vger.kernel.org
Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: Add note to prevent buggy code re-use

adv7604 and et8ek8 sensor drivers have mixed up logical and line level
for reset/powerdown signal.  They call it a reset signal (it indeed
behaves like that), but drivers assert the reset to operate which is
clearly incorrect and relies on wrong ACTIVE_HIGH flag in the DTS.

People in discussions copy existing poor code and claim they can repeat
same mistake, so add a note to prevent that.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: stm32: dcmipp: avoid naming clock if only one is needed

When DCMIPP requires only a single clock (kclk), avoid relying on its
name to obtain it. The introduction of MP25 support added the mclk,
which necessitated naming the first clock kclk. However, this breaks
backward compatibility with existing MP13 device trees that do not
specify clock names.

Fixes: 686f27f7ea37 ("media: stm32: dcmipp: add core support for the stm32mp25")
Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Cc: Stable@vger.kernel.org # 6.14.x: 7f487562af49 media: stm32: dcmipp: correct ret type in dcmipp_graph_notify_bound
Cc: Stable@vger.kernel.org # 6.14.x: c715dd62da30 media: stm32: dcmipp: add has_csi2 & needs_mclk in match data
Cc: Stable@vger.kernel.org # 6.14.x:
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

MAINTAINERS: Promote Frank Li as i.MX7/8 media maintainer

Move Rui Miguel Silva to reviewer since he does not have sufficient time to
maintain i.MX7/8 media drivers. Frank Li has the hardware, complete
documentation, and has participated actively in patch review. Promote him
as the i.MX7/8 media drivers maintainer.

Reviewed-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Accommodate C-PHY into the calculation

We need to set correct mode for PLL to calculate correct frequency.
Signalling mode is known at this point, so use it for that.

Fixes: 47b6eaf36eba ("media: ccs-pll: Differentiate between CSI-2 D-PHY and C-PHY")
Reviewed-by: Mehdi Djait <mehdi.djait@linux.intel.com>
Signed-off-by: David Heidelberg <david@ixit.cz>
[Sakari Ailus: Drop extra newline.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Use devm_v4l2_sensor_clk_get()

Rely on devm_v4l2_sensor_clk_get() to obtain the sensor's external clock.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: ccs: Clean up ccs_get_hwcfg() a little

Fix coding style issues and do other small cleanups in ccs_get_hwcfg().

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov5647: switch to {enable,disable}_streams

Switch from s_stream to enable_streams and disable_streams callbacks.

Signed-off-by: Xiaolei Wang <xiaolei.wang@windriver.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tarang Raval <tarang.raval@siliconsignals.io>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov5647: Switch to using the sub-device state lock

Switch to using the sub-device state lock and properly call
v4l2_subdev_init_finalize() / v4l2_subdev_cleanup() on probe() /
remove().

Signed-off-by: Xiaolei Wang <xiaolei.wang@windriver.com>
Reviewed-by: Tarang Raval <tarang.raval@siliconsignals.io>
[Sakari Ailus: Remove now-redundant assignment of control handler lock.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>

media: i2c: ov5647: Convert to CCI register access helpers

Use the new common CCI register access helpers to replace the private
register access helpers in the ov5647 driver. This simplifies the driver
by reducing the amount of code.

Signed-off-by: Xiaolei Wang <xiaolei.wang@windriver.com>
Reviewed-by: Tarang Raval <tarang.raval@siliconsignals.io>
Reviewed-by: Hans de Goede <johannes.goede@oss.qualcomm.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>