| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Variable initialization code in notify_src_chg() is incorrectly
indented. Fix it.
Signed-off-by: Laurent Pinchart <[email protected]>
Reviewed-by: Sakari Ailus <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
This device only supports video capture, so bail out if invalid
selection type is passed.
Signed-off-by: Alexander Stein <[email protected]>
Reviewed-by: Laurent Pinchart <[email protected]>
Reviewed-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
'field' is zero-initialized to V4L2_FIELD_ANY, which is an invalid value
to return to userspace. Instead default to non-interleaving.
Signed-off-by: Alexander Stein <[email protected]>
Reviewed-by: Laurent Pinchart <[email protected]>
Reviewed-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The CSI requires a connected source subdev to operate. If
fwnode_graph_get_endpoint_by_id() fails and returns NULL, there is no
point in going on. Print an error message and abort instead.
Also we don't need to check for an existing asd. Any failure of
v4l2_async_nf_add_fwnode_remote() should abort the probe.
Suggested-by: Laurent Pinchart <[email protected]>
Signed-off-by: Frieder Schrempf <[email protected]>
Reviewed-by: Laurent Pinchart <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <[email protected]>
Reviewed-by: Xavier Roumegue <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
Convert platform_get_resource(), devm_ioremap_resource() to a single
call to devm_platform_get_and_ioremap_resource(), as this is exactly
what this function does.
Signed-off-by: Ye Xingchen <[email protected]>
Reviewed-by: Xavier Roumegue <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The call of mxc_jpeg_get_plane_size() from mxc_jpeg_dec_irq() sets
plane_no argument to 1. The compiler sees that it's possible to end up
with an access beyond the bounds of sizeimage, if mem_planes was too
large:
if (plane_no >= fmt->mem_planes) // mem_planes = 2+
return 0;
if (fmt->mem_planes == fmt->comp_planes) // comp_planes != mem_planes
return q_data->sizeimage[plane_no];
if (plane_no < fmt->mem_planes - 1) // mem_planes = 2
return q_data->sizeimage[plane_no];
comp_planes == 0 or 1 is safe. comp_planes > 2 would be out of bounds.
(This isn't currently possible given the contents of mxc_formats, though.)
Silence the warning by bounds checking comp_planes for future
robustness. Seen with GCC 13:
In function 'mxc_jpeg_get_plane_size',
inlined from 'mxc_jpeg_dec_irq' at ../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.c:729:14:
../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.c:641:42: warning: array subscript 2 is above array bounds of 'u32[2]' {aka 'unsigned int[2]'} [-Warray-bounds=]
641 | size += q_data->sizeimage[i];
| ~~~~~~~~~~~~~~~~~^~~
In file included from ../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg-hw.h:112,
from ../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.c:63:
../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.h: In function 'mxc_jpeg_dec_irq':
../drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.h:84:41: note: while referencing 'sizeimage'
84 | u32 sizeimage[MXC_JPEG_MAX_PLANES];
| ^~~~~~~~~
Cc: Mirela Rabulea <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: Shawn Guo <[email protected]>
Cc: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: [email protected]
Signed-off-by: Kees Cook <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The find_csis_format() may return NULL in case supported format is not
found, check the return value of find_csis_format() before using the
result to avoid NULL pointer dereference.
Fixes: 11927d0fd0d0 ("media: imx-mipi-csis: Use V4L2 subdev active state")
Signed-off-by: Marek Vasut <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Replace the readl and writel with regmap to ease debugging the registers
from userspace.
Suggested-by: Alexander Stein <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
There's no need to use threaded IRQs with the PXP, as the interrupt
handler doesn't need to sleep and doesn't perform any time-consuming
operation. Switch to regular IRQ handler.
This fixes lockups of the PXP noticed on i.MX7, where the PXP would stop
generating interrupts after a variable number of frames (from a few
dozens to a few hundreds). The root cause is however unknown.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Add pxp_read() and pxp_write() functions to wrap readl() and writel()
respectively. This can be useful for debugging register accesses.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Implement support for the VIDIOC_ENUM_FRAMESIZES ioctl.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The find_format() function looks up format information for a given pixel
format. It takes a v4l2_format pointer, but only uses the contained
pixel format value. To prepare it for being used by callers that don't
have v4l2_format, modify it to take the pixel format value directly.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Register a media device for the PXP, using the v4l2-mem2mem MC
infrastructure to populate the media graph. No media device operation is
implemented, the main use of the MC API is to allow consistent discovery
of media devices for userspace.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The v4l2_capability.bus_info field is set by the V4L2 core when left
empty by the .querycap() handler. This is the recommended practice, in
order to ensure bus_info coherence between drivers. Don't set it
manually.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Sorting headers alphabetically helps locating duplicates, and make it
easier to figure out where to insert new headers.
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The i.MX7D needs a different data path configuration than the i.MX6ULL.
Configure the data path as close as possible to the data path on the
i.MX6ULL.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Unfortunately, the PXP_HW_VERSION register reports the PXP on the i.MX7D
and on the i.MX6ULL as version 3.0, although the PXP versions on these
SoCs have significant differences.
Use the compatible to configure the ctrl0 register as required dependent
on the platform.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The LUT block is always configured in bypass mode.
Take it entirely out of the pipeline by disabling it and routing the
data path around the LUT.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Various multiplexers in the pipeline are not used with the currently
configured data path. Disable all unused multiplexers by selecting the
"no output" (3) option.
The datasheet doesn't explicitly require this, but the PXP has been seen
to hang after processing a few hundreds of frames otherwise.
As at it, add documentation for the multiplexers that are actually
relevant for the data path.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The driver must configure the data path through the Pixel Pipeline.
Currently, the driver is using a fixed setup, but once there are
different pipeline configurations, it is helpful to have a dedicated
function for determining the register value for the data path.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Different versions of the Pixel Pipeline have different blocks and their
routing may be different. Read the PXP_HW_VERSION register to determine
the version of the PXP and print it to the log for debugging purposes.
Signed-off-by: Michael Tretter <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The .set_fmt() handler is responsible for adjusting the requested format
based on the device limitations. Implement .init_cfg() as a wrapper of
.set_fmt(), to ensure that the initial configuration always matches the
rules implemented in .set_fmt(), should they ever change.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mn-beacon
Acked-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Use the V4L2 subdev active state API to store the active format. This
simplifies the driver not only by dropping the mipi_csis_device csis_fmt
and format_mbus fields, but it also allows dropping the device lock,
replaced with the state lock.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mn-beacon
Acked-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
To prepare for usage of the subdev active state that will replace the
csis_fmt and format_mbus fields stored in the mipi_csis_device
structure, pass the format explicitly to the functions called when
starting streaming to avoid accessing those two fields. Not functional
change intended.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mn-beacon
Acked-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The runtime PM handlers don't need manual locking as
- they are serialized by the runtime PM core
- they can't race with other functions taking the same lock, as they
don't access any data protect by that lock
Drop the locking.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mn-beacon
Acked-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
It is customary to prefix error labels with 'err_' to make their purpose
clearer. Do so in the probe function.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mn-beacon
Acked-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Use the V4L2 subdev active state API to store the active format. This
simplifies the driver not only by dropping the state stored in the
imx7_csi structure, but also by replacing the manual lock with the state
lock.
The is_streaming field is now protected by the active state lock, either
explicitly in .s_stream(), where the active state is locked manually, or
implicitly in .set_fmt(), which is called with the state locked.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The v4l2_subdev_format structure passed to the .get_fmt() subdev
operation in imx7_csi_video_validate_fmt() isn't zeroed, which can cause
undefined behaviour. Fix it.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
It's good practice for functions to perform error cleanup internally
when they fail, in order to not leave the device in a half-initialized
state. Move the async notifier cleanup from the probe error path to the
imx7_csi_async_register(), and drop the v4l2_async_nf_unregister() call
as there is no error path after the async notifier gets registered.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The subdev core guarantees that the .set_fmt() operation is always
called with a valid pad. Drop the unneeded pad checks.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The .s_stream() and .link_validate() operations can't be called with a
NULL src_sd, as subdev nodes are not registered before the async
notifier completes. Remove the unneeded checks.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The .s_stream() operation is guaranteed not to be called to start/stop
an already started/stopped subdev. Remove the unneeded check.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The imx7_csi_video_init_format() function instantiates a
v4l2_subdev_format on the stack, to only use the .format field of that
structure. Replace it with a v4l2_mbus_framefmt instance.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The imx7_csi.cc field is set but never used. Drop it.
Signed-off-by: Laurent Pinchart <[email protected]>
Tested-by: Adam Ford <[email protected]> #imx8mm-beacon-kit
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Martin Kepplinger <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
using the api of clk_bulk can simplify the code.
and the clock of the jpeg codec may be changed,
the clk_bulk api can be compatible with the future change.
Fixes: 4c2e5156d9fa ("media: imx-jpeg: Add pm-runtime support for imx-jpeg")
Signed-off-by: Ming Qian <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
imx7_csi_init()
Add missing clk_disable_unprepare(), if imx7_csi_dma_setup() fails
in imx7_csi_init().
Fixes: ff43ca911978 ("media: imx: imx7-media-csi: Move CSI configuration before source start")
Signed-off-by: Yang Yingliang <[email protected]>
Reviewed-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
In case the IRQ allocation returns an error in dw100_probe(), the pm
runtime is not disabled before to return.
Add the missing unwind goto on the error handling path of the IRQ
allocation request.
Reported-by: kernel test robot <[email protected]>
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Xavier Roumegue <[email protected]>
Signed-off-by: Laurent Pinchart <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
The imx7-media-csi driver, currently in staging, is ready for
prime-time. The staging TODO file lists a few items specific to that
driver, that are already addressed (the "all of the above" part) or can
be addressed later:
- The frame interval monitoring support is a software mechanism to
monitor the device for unexpected stalls, and should be part of the
V4L2 core if desired.
- Restricting the support media bus formats based on the SoC integration
only aims at reducing userspace confusion by not enumerating options
that are known not to be possible, it won't cause regressions if
handled later.
Move the description of the media bus format restriction TODO item to
the driver, drop the other TODO items, and move the driver out of
staging.
Signed-off-by: Laurent Pinchart <[email protected]>
Reviewed-by: Paul Elder <[email protected]>
Acked-by: Rui Miguel Silva <[email protected]>
Tested-by: Alexander Stein <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Tag branch
* tag 'br-v6.2d' of git://linuxtv.org/hverkuil/media_tree: (35 commits)
media: saa7164: remove variable cnt
atomisp: fix potential NULL pointer dereferences
radio-terratec: Remove variable p
media: platform: s5p-mfc: Fix spelling mistake "mmaping" -> "mmapping"
media: platform: mtk-mdp3: remove unused VIDEO_MEDIATEK_VPU config
media: vivid: remove redundant assignment to variable checksum
media: cedrus: h264: Optimize mv col buffer allocation
media: cedrus: h265: Associate mv col buffers with buffer
media: mediatek: vcodec: fix h264 cavlc bitstream fail
media: cedrus: hevc: Fix offset adjustments
media: imx-jpeg: Fix Coverity issue in probe
media: v4l2-ioctl.c: Unify YCbCr/YUV terms in format descriptions
media: atomisp: Fix spelling mistake "mis-match" -> "mismatch"
media: c8sectpfe: Add missed header(s)
media: adv748x: afe: Select input port when initializing AFE
media: vimc: Update device configuration in the documentation
media: adv748x: Remove dead function declaration
media: mxl5005s: Make array RegAddr static const
media: atomisp: Fix spelling mistake "modee" -> "mode"
media: meson/vdec: always init coef_node_start
...
|
|
Possible dereference null return after of_match_node,
so check for NULL of_id.
Signed-off-by: Mirela Rabulea <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
the ioctl encoder/decoder cmd is under queue lock,
and buf_done is in the irq, it can't be locked with the mutex,
they are not synchronized.
when v4l2_update_last_buf_state is called to
handle the encoder/decoder stop cmd,
the last src buffer may be done at the same time.
so it's possible that last_src_buf is set,
but the output rdy_queue is empty,
then driver won't mark it stopped,
as v4l2_m2m_is_last_draining_src_buf() will always return false and
v4l2_m2m_dst_buf_is_last() return false too.
In this case, the drain will be blocked.
add the hw lock around the ioctl encoder/decoder cmd,
to synchronize with the buf_done.
Fixes: 4911c5acf935 ("media: imx-jpeg: Implement drain using v4l2-mem2mem helpers")
Signed-off-by: Ming Qian <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
mxc-jpeg supports non contiguous format nv12m,
and in order to compatible with the devices
that only support contiguous format nv12,
jpeg can support nv12 and nv12m in the same time.
Signed-off-by: Ming Qian <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
[hverkuil: document is_rgb variable]
|
|
The codec can support any image size WxH,
with arbitrary W (image width) and H (image height) dimensions.
But it requires buffer alignment,
so driver can report the aligned resolution through the g_fmt,
and report the actual resolution through the g_selection.
For encoder, it even support to encode a smaller jpeg
than the original picture through s_selection api.
For the decoder, we do not support cropping a portion smaller
than the original picture, due to hardware limitations (wrapper side).
Fixes: 9e7aa76cdb02 ("media: imx-jpeg: Align upwards buffer size")
Signed-off-by: Ming Qian <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|
|
There is a hardware bug that the interrupt STMBUF_HALF may be triggered
after or when disable interrupt.
It may led to unexpected kernel panic.
And interrupt STMBUF_HALF and STMBUF_RTND have no other effect.
So disable them and the unused interrupts.
meanwhile clear the interrupt status when disable interrupt.
Signed-off-by: Ming Qian <[email protected]>
Reviewed-by: Mirela Rabulea <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
|