The only thing which atomisp_ospm_dphy_down() does is disable the CSI
pins, but if we failed to probe the ISP then these will never have
been enabled (because the ISP never started streaming).
So the atomisp_ospm_dphy_down() call in the probe error path is
unnecessary, remove it.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_css_init() is always called after calling atomisp_power_on()
either directly or through getting a runtime-pm reference.
Likewise atomisp_css_uninit() is always called after calling
atomisp_power_off().
Move the call site of these 2 functions to inside atomisp_power_on() /
atomisp_power_off() to make this more explicit.
Note this makes atomisp_reset() also set isp_fatal_error on
atomisp_power_on() errors, where as before it only did this on
atomisp_css_init() errors. This behavior change is for the better,
since power-on failing is pretty fatal too.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_suspend() contains a 1:1 copy of atomisp_runtime_suspend() and
the same goes for the resume() functions.
Rename the runtime functions to atomisp_power_on()/_off() and use these
as runtime-pm handlers as well as helper in other places to remove
the code duplication.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Remove the unnecessary sw_contex.power_state checks:
1. atomisp_freq_scaling() and atomisp_stop_streaming() only run when the
ISP is powered up through runtime-pm, so the checks are not necessary
2. atomisp_mrfld_pre_power_down() and atomisp_runtime_resume() are only
called through the driver-core pm handling code which already
guarantees they are not called when already powered down / up.
3. atomisp_isr() also checks isp->css_initialized which only gets set
by atomisp_css_init() which runs *after* powering up the ISP and which
gets cleared by atomisp_css_uninit() *before* powering down the ISP.
So all the checks are unnecessary, remove them as well as the
sw_contex.power_state field itself.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_css_suspend() is a 1:1 copy of atomisp_css_uninit() and
atomisp_css_resume() is a 1:1 copy of atomisp_css_init().
Remove the 2 copies and have their one caller just call
atomisp_css_uninit() / atomisp_css_init() instead.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_css_uninit() only runs when all streams are stopped and
atomisp_css_stop() already clears the config, so the clearing
of the config can be dropped from atomisp_css_uninit().
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_mrfld_power_down()/_up() are unnecessary wrappers around
atomisp_mrfld_power() remove them and just call atomisp_mrfld_power()
directly.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_reset() calls atomisp_mrfld_power_down() after calling
atomisp_runtime_suspend(), which already calls
atomisp_mrfld_power_down() itself.
And the some goes for atomisp_runtime_resume() / atomisp_mrfld_power_up().
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
After the conversion to videobuf2 userptr support is no longer needed,
drop it.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
After the conversion to videobuf2 a bunch of ia_css_frame_*()
functions are unused, remove them.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Remove atomisp_css_yuvpp_configure_viewfinder(), it is not used anywhere.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Remove ia_css_pipe_get_acc_stage_desc() and sh_css_flush(),
after removing the accelerator /dev/video# node and related ioctls
these are no longer used.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_release() was taking pipe->vb_queue_mutex + isp->mutex at the
same time. But if the /dev/video# node is closed while still streaming
then vb2_queue_release() will call atomisp_stop_streaming() which takes
isp->mutex itself, leading to a deadlock.
To fix this only take isp->mutex after cleaning up the v4l2_fh /
the vb2_queue. While at it switch to vb2_fop_release() which will take
pipe->vb_queue_mutex for us, which also resolves a FIXME comment.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
I managed to trigger an atomisp_css_start() error by pushing my test
system towards an OOM situation, this resulted in the following errors:
atomisp-isp2 0000:00:03.0: alloc pages err...
atomisp-isp2 0000:00:03.0: hmm_bo_alloc_pages failed.
atomisp-isp2 0000:00:03.0: stream[0] start error.
But it is not entirely clear what the root cause of
the "alloc pages err..." error is. I suspect the root cause is
alloc_pages_bulk_array() failing. Add a log message to make
the root cause more clear if this is hit again.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
I managed to trigger an atomisp_css_start() error by pushing my test system
towards an OOM situation, this triggered the following WARN_ON() in
__vb2_queue_cancel() in videobuf2-core.c:
/*
* If you see this warning, then the driver isn't cleaning up properly
* after a failed start_streaming(). See the start_streaming()
* documentation in videobuf2-core.h for more information how buffers
* should be returned to vb2 in start_streaming().
*/
if (WARN_ON(atomic_read(&q->owned_by_drv_count))) {
Fix this by calling atomisp_flush_video_pipe() to return any queued buffers
back to the videobuf2-core on an atomisp_css_start() error.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
With the accel code gone this is unused, remove it.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The ATOMISP_ACC_* custom ioctls and the ACC device node have been removed
in commit a5c17adbadcb ("media: atomisp: Remove the ACC device node").
This means that pipe_configs[pipe_id].acc_extension now never gets set
which causes atomisp_compat_css20.c: __create_pipe() to always skip
creation of pipes with a pipe_id of IA_CSS_PIPE_ID_ACC / a mode of
IA_CSS_PIPE_MODE_ACC.
This allows removing of the acc_pipe creation / handling code
from mainly sh_css.c and a bunch of other places.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Standard v4l2 userspace apps do not consume the s3a statistics block
data. Until we have a userspace consumer for this (libcamera), which
might also involve changing the API for this, lower the log level
of these messages to dev_dbg() to avoid them filling up the logs.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Depending on which order userspace makes various v4l2 calls, the sensor
might still be powered down when set_fmt is called.
What should really happen here is delay the writing of the mode-related
registers till streaming is started, but for now use the same quick fix
as the atomisp_ov2680 code and call power_up() from set_fmt() in
combination with keeping track of the power-state to avoid doing the
power-up sequence twice.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Remove the complicated __atomisp_get_pipe() helper, atomisp_buf_done()
only needs the pipe pointer in cases where it has a frame, so we can
simply get the pipe from the frame using the vb_to_pipe() helper.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Make atomisp_g_fmt_cap() default to YUV420 so that it matches with what
atomisp_try_fmt_cap() and atomisp_queue_setup() do when they need to
pick a default pixelformat.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The atomisp_try_fmt() call in atomisp_try_fmt_cap() replaces
the pixelformat passed by userspace with the sensors native pixelformat.
Which always gets replaced by V4L2_PIX_FMT_YUV420 by atomisp_adjust_fmt()
because raw sensor formats are not supported.
This needs to be fixed so that userspace which does a try_fmt call before
s_fmt does not end up always getting YUV420 even if it passed something
else into the try_fmt call.
To fix this restore the userspace requested pixelformat before
the atomisp_adjust_fmt() call. atomisp_adjust_fmt() will replace this
with V4L2_PIX_FMT_YUV420 in case an unsupported format is requested.
Note this relies on the "media: atomisp: Refactor atomisp_adjust_fmt()"
change, before that atomisp_adjust_fmt() would return -EINVAL for
unsupported pixelformats.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Refactor atomisp_adjust_fmt():
1. The block starting at "format_bridge = atomisp_get_format_bridge(...)"
and ending with "if (field == V4L2_FIELD_ANY) field = V4L2_FIELD_NONE;"
is duplicated. With only the second block:
a) Properly checking that format_bridge is not NULL; amd
b) Having the special handling for IA_CSS_FRAME_FORMAT_RAW
Remove the first block.
2. On a NULL return from atomisp_get_format_bridge(f->fmt.pix.pixelformat)
fall back to V4L2_PIX_FMT_YUV420 just like in the IA_CSS_FRAME_FORMAT_RAW
case. atomisp_adjust_fmt() is used in VIDIOC_TRY_FMT handling and that
should jusy pick a fmt rather then returning -EINVAL.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
camorama calls VIDIOC_REQBUFS(V4L2_MEMORY_MMAP) to test if MMAP support
works (this was added specifically because of the previously broken
MMAP support in atomisp).
Currently this fails because atomisp_get_css_frame_info() fails in this
case. Although it is weird to call VIDIOC_REQBUFS before VIDIOC_S_FMT,
it is allowed to do this.
Fix this not working by doing a S_FMT to V4L2_PIX_FMT_YUV420 + the highest
supported resolution.
Note this will cause camorama to use mmap mode, which means it will also
use libv4l2 to do format conversion. libv4l2 will pick V4L2_PIX_FMT_RGB565
as input format and this will lead to a garbled video display. This is
a libv4lconvert bug, the RGB565 -> RGB24 path in libv4lconvert assumes
that stride == width which is not true on the atomisp.
I've already send out a libv4lconvert fix for this. Also this can be worked
around by passing --dont-use-libv4l2 as argument to camorama.
Link: https://git.linuxtv.org/v4l-utils.git/commit/?id=aecfcfccfc2f78d7531456ffa5465666c6bc641e
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
To fix atomisp_queue_setup() sometimes failing it needs to be able to call
atomisp_set_fmt(), but atomisp_queue_setup() (VIDIOC_REQBUFS) does not get
passed a file handle by the videobuf2 core.
Partly revert commit b3be98f984d4 ("media: atomisp: Remove a couple of not
useful function wrappers") so that atomisp_set_fmt() can be used
without a file handle.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Convert atomisp to use videobuf2.
This fixes mmap not working and in general moving over to
the more modern videobuf2 is a good plan.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Several places rely on the [frame_]info member being the first member of
struct ia_css_frame, so that &frame->info will yield NULL when frame is
NULL (some places already explicitly check for a NULL frame pointer but
not nearly all).
For videobuf2 support the vb2_v4l2_buffer struct needs to be embedded
in the frame struct and it needs to be the first member. Breaking the
assumption that &frame->info will yield NULL when frame is NULL.
Add a ia_css_frame_get_info() helper to return either the ia_css_frame_info
struct embedded in the frame, or NULL when the frame pointer is NULL and
use this in places where a ia_css_frame_info ptr or NULL is expected.
To make sure that we catch all uses of the info field this patch also
renames the info field to frame_info.
This is a preparation patch for converting the driver to videobuf2.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Use the new atomisp_flush_video_pipe() helper instead of open-coding it.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
With the videobuf2 conversion accessing the index of a frame is going
to become more involved and writing this out each time is undesired.
Add an 'i' helper variable for the index and assing this once as
preparation for the videobuf2 conversion.
This also makes use of the new rules wrt max line-lengths in the kernel
to avoid breaking up a bunch of lines. Not breaking these lines results
in better readable code (IMHO).
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Instead of using an integer to keep count of how many buffers have
been handed over to the ISP (buffers_in_css) move buffers handed
over to the ISP to a new buffers_in_css list_head so that we can
easily loop over them.
This removes the need for atomisp_flush_video_pipe() to loop over
all buffers and then (ab)use the state to figure out if they
were handed over to the ISP.
Since the buffers are now always on a list when owned by the driver
this also allows the buffer_done path on flush vs normal completion
to be unified (both now need a list_del()) and this common code can
now be factored out into a new atomisp_buffer_done() helper.
This is a preparation patch for moving the driver over to
the videobuf2 framework.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The atomisp_buffers_queued[_pipe] functions are not used anywhere,
remove them.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
There is no guarantee that when we stop the pipeline all buffers owned
by the CSS are cleanly returned to the videobuf queue.
This is a problem with videobuf2 which will complain loudly when not
all buffers have been returned after the streamoff() queue op has
returned.
And this also allows removing a WARN() in the continuous mode path.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
atomisp_q_video_buffers_to_css() is only used insidd atomisp_fops.c, make
it static.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Add a function to initialize (rather then alloc/create) a
ia_css_frame struct based on an ia_css_frame_info struct.
This is a preparation patch for adding videobuf2 support.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Add a new hmm creating function to create a vmm object from a vmalloc-ed
kernel buffer. This is a preparation patch for adding videobuf2 (and
working MMAP mode) support.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The Atmel ISC driver is not compliant with media controller specification.
In order to evolve this driver, it has to move to media controller, to
support enhanced features and future products which embed it.
The move to media controller involves several changes which are
not backwards compatible with the current usability of the driver.
The best example is the way the format is propagated from the top video
driver /dev/videoX down to the sensor.
In a simple configuration sensor ==> isc , the isc just calls subdev s_fmt
and controls the sensor directly. This is achieved by having a lot of code
inside the driver that will query the subdev at probe time and make a list
of formats which are usable.
Basically the user has nothing to configure, as the isc will handle
everything at the top level. This is an easy way to capture, but also comes
with the drawback of lack of flexibility.
In a more complicated pipeline
sensor ==> controller 1 ==> controller 2 ==> isc
this will not be achievable, as controller 1 and controller 2 might be
media-controller configurable, and will not propagate the formats down to
the sensor.
After discussions with the media maintainers, the decision is to move
Atmel ISC to staging as-is, to keep the Kconfig symbols and the users
to the driver in staging. Thus, all the existing users of the non
media-controller paradigm will continue to be happy and use the old config
way.
The new driver was added in the media subsystem with a different
symbol, with the conversion to media controller done, and new users
of the driver will be able to use all the new features.
This patch is merely a file move to staging, not affecting any of the
users.
The exported symbols had to be renamed to atmel_* to avoid duplication with
the new Microchip ISC driver.
Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
This cleans up a checkstyle warning where the vb2_ops struct definitions
should be const.
Signed-off-by: Ian Cowan <ian@linux.cowan.aero>
Reviewed-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Testing reference video TSUNEQBD_A_MAIN10_Technicolor_2 has show that
Cedrus is capable of decoding frames with different chroma and luma bit
depths.
Relax restrictions so only highest depth is checked.
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
10-bit frames needs extra buffer space when 8-bit capture format is
used. Use previously prepared infrastructure to adjust buffer size.
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
In some cases decoding engine needs extra space in capture buffers. This
is the case for decoding 10-bit HEVC frames into 8-bit capture format.
This commit only adds infrastructure for such cases.
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Use of standard min() helper macro is preferred over using ternary
operator for logical evaluation and value assignment. This issue is
identified by coccicheck using the minmax.cocci file.
Signed-off-by: Deepak R Varma <drv@mailo.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Some Allwinner platforms come with an Image Signal Processor, which
supports various features in order to enhance and transform data
received by image sensors into good-looking pictures. In most cases,
the data is raw bayer, which gets internally converted to RGB and
finally YUV, which is what the hardware produces.
This driver supports ISPs that are similar to the A31 ISP, which was
the first standalone ISP found in Allwinner platforms. Simpler ISP
blocks were found in the A10 and A20, where they are tied to a CSI
controller. Newer generations of Allwinner SoCs (starting with the
H6, H616, etc) come with a new camera subsystem and revised ISP.
Even though these previous and next-generation ISPs are somewhat
similar to the A31 ISP, they have enough significant differences to
be out of the scope of this driver.
While the ISP supports many features, including 3A and many
enhancement blocks, this implementation is limited to the following:
- V3s (V3/S3) platform support;
- Bayer media bus formats as input;
- Semi-planar YUV (NV12/NV21) as output;
- Debayering with per-component gain and offset configuration;
- 2D noise filtering with configurable coefficients.
Since many features are missing from the associated uAPI, the driver
is aimed to integrate staging until all features are properly
described.
On the technical side, it uses the v4l2 and media controller APIs,
with a video node for capture, a processor subdev and a video node
for parameters submission. A specific uAPI structure and associated
v4l2 meta format are used to configure parameters of the supported
modules.
One particular thing about the hardware is that configuration for
module registers needs to be stored in a DMA buffer and gets copied
to actual registers by the hardware at the next vsync, when instructed
by a flag. This is handled by the "state" mechanism in the driver.
Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Replace custom implementation of the device_match_fwnode().
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQRSrcquik9wuZrNjXJrQl33mxwedgUCY2U2ZQAKCRBrQl33mxwe
dhaLAP4t7vpFHF6UwPKhUsnBjykDE9cqX5tFY5hV/P4B3dhm3wEAmJFEi38QFQlb
gkISSnYIx5DZGP2iBXT96ftGlPI7qgg=
=wwvs
-----END PGP SIGNATURE-----
Merge tag 'br-v6.2e' of git://linuxtv.org/hverkuil/media_tree into media_stage
Tag branch
* tag 'br-v6.2e' of git://linuxtv.org/hverkuil/media_tree: (29 commits)
media: davinci/vpbe: Fix a typo ("defualt_mode")
media: sun6i-csi: Remove unnecessary print function dev_err()
media: Documentation: Drop deprecated bytesused == 0
media: platform: exynos4-is: fix return value check in fimc_md_probe()
media: dvb-core: remove variable n, turn for-loop to while-loop
media: vivid: fix compose size exceed boundary
media: rkisp1: make const arrays ae_wnd_num and hist_wnd_num static
media: dvb-core: Fix UAF due to refcount races at releasing
media: rkvdec: Add required padding
media: aspeed: Extend debug message
media: aspeed: Support aspeed mode to reduce compressed data
media: Documentation: aspeed-video: Add user documentation for the aspeed-video driver
media: v4l2-ctrls: Reserve controls for ASPEED
media: v4l: Add definition for the Aspeed JPEG format
staging: media: tegra-video: fix device_node use after free
staging: media: tegra-video: fix chan->mipi value on error
media: cedrus: initialize controls a bit later
media: cedrus: prefer untiled capture format
media: cedrus: Remove cedrus_codec enum
media: cedrus: set codec ops immediately
...
The addresses of two elements of the segmap[][] member are passed to the
hardware which expects 128-bit aligned addresses. However, without this
patch offsetof(struct rkvdec_vp9_priv_tbl, segmap[0]) is an odd number
(2421) but the hardware just ignores the 5 least significant bits of the
address. As a result, the hardware writes the segmentation map to incorrect
locations.
Inserting 11 bytes of padding corrects this situation by making the said
addresses divisible by 16 (i.e. aligned on a 128-bit boundary).
Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Fixes: f25709c4ff ("media: rkvdec: Add the VP9 backend")
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
At probe time this code path is followed:
* tegra_csi_init
* tegra_csi_channels_alloc
* for_each_child_of_node(node, channel) -- iterates over channels
* automatically gets 'channel'
* tegra_csi_channel_alloc()
* saves into chan->of_node a pointer to the channel OF node
* automatically gets and puts 'channel'
* now the node saved in chan->of_node has refcount 0, can disappear
* tegra_csi_channels_init
* iterates over channels
* tegra_csi_channel_init -- uses chan->of_node
After that, chan->of_node keeps storing the node until the device is
removed.
of_node_get() the node and of_node_put() it during teardown to avoid any
risk.
Fixes: 1ebaeb0983 ("media: tegra-video: Add support for external sensor capture")
Cc: stable@vger.kernel.org
Cc: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
chan->mipi takes the return value of tegra_mipi_request() which can be a
valid pointer or an error. However chan->mipi is checked in several places,
including error-cleanup code in tegra_csi_channels_cleanup(), as 'if
(chan->mipi)', which suggests the initial intent was that chan->mipi should
be either NULL or a valid pointer, never an error. As a consequence,
cleanup code in case of tegra_mipi_request() errors would dereference an
invalid pointer.
Fix by ensuring chan->mipi always contains either NULL or a void pointer.
Also add that to the documentation.
Fixes: 523c857e34 ("media: tegra-video: Add CSI MIPI pads calibration")
Cc: stable@vger.kernel.org
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
While it doesn't matter if controls are initialized before or after
queues and formats from open handler standpoint, initializing them last
helps keeping s_ctrl handler simpler, since everything has already valid
values.
This is just preparation for follow up changes. No functional change is
intended.
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
While all generations of display engine on Allwinner SoCs support
untiled format, only first generation supports tiled format. Let's
move untiled format up, so it can be picked before tiled one. If
Cedrus variant doesn't support untiled format, tiled will still be
picked as default format.
Acked-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
