- 26 Apr, 2024 30 commits
-
-
Hans de Goede authored
Change the single, unexpected user of CSI_PORT0_ID for (port = CSI_PORT0_ID; port < N_CSI_PORTS; port++) to for (port = 0; port < N_CSI_PORTS; port++) { matching all the other for-loops iterating over the ports in the same file. And remove the now fully unused input_port_ID_t enum type. 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>
-
Hans de Goede authored
The TPG support registers a v4l2-subdev for this, but this is not part of the media-controller graph in any way. After manually selecting the TPG as input using the s_input ioctl it does not work. Several supported sensors have their own working TPG and using the sensor's TPG means that the same data-flow is used as with actual sensors rather then the quite different data-flow from the ISP's builtin TPG. Remove the broken TPG support, when a test-pattern is needed for testing a sensor's TPG can be used. Using a sensor's TPG is actually better for testing since then the actual normal data-flow is being tested. Reviewed-by:
-
Hans de Goede authored
The atomisp driver now properly ensures s_power(1) is called before calling the set_fmt() callback, so this workaround is no longer necessary. Reviewed-by:
-
Hans de Goede authored
The media-controller core (__media_entity_setup_link()) already checks that the pads of the link are not streaming before calling the setup_link() pad-op so calling atomisp_pipe_check() is not necessary; and taking isp->mutex inside the setup_link() pad-op leads to a possible ABBA deadlock vs the media-device graph_mutex which in the case of the setup_link() pad-op is taken before calling the op, while in other scenarios the graph_mutex is taken after the isp->mutex. Reviewed-by:
-
Hans de Goede authored
Call media_pipeline_alloc_start() from atomisp_start_streaming() to mark all involved subdevs as busy so that the links cannot be changed through the media-controller APIs while streaming. Reviewed-by:
-
Hans de Goede authored
Most sensor drivers report V4L2_FIELD_NONE as v4l2_mbus_framefmt.field value. Propagate this to the v4l2_mbus_framefmt for the CSI port to pass the link validation done by media_pipeline_start(). Keep the v4l2_mbus_framefmt.field for the source pad of the CSI port as V4L2_FIELD_ANY to match the ISP sink pad. Reviewed-by:
-
Hans de Goede authored
So far the CSI port sink and source pads have not had their fmt set at all, leaving things at the default of SBGGR8_1X8/0x0. This means that pipeline validation such as e.g. done when calling media_pipeline_start() will fail since the links to / from the CSI ports have different fmts on each end. Store a pointer to the CSI port v4l2-subdev in struct atomisp_input_subdev, and use this in atomisp_set_sensor_crop_and_fmt() to propagate the sensors new fmt after a successful set_fmt to the CSI port it is connected too. The input->csi_port pointer also allows simplifying atomisp_link_setup(). Reviewed-by:
-
Hans de Goede authored
Add a new atomisp_setup_input_links() helper which ensures that the CSI-receiver -> ISP link for input_curr is marked as enabled and the other CSI-receiver -> ISP links are disabled. And call this helper from atomisp_register_device_nodes() for the initial setup and from atomisp_select_input() for runtime input_curr changes. Reviewed-by:
-
Hans de Goede authored
Simplify the checking of the input argument to atomisp_s_input(): - Remove the check for input >= ATOM_ISP_MAX_INPUTS, input_cnt never is > ATOM_ISP_MAX_INPUTS so checking only that is sufficient - Remove the unnecessary camera local variable - Move the 2 invalid argument checks to above the atomisp_pipe_check() which checks if the pipe is busy or not Reviewed-by:
-
Hans de Goede authored
When switching input/sensor the s_power() callback must be called for old sensor drivers to power on the new sensor and power off the previous sensor. atomisp_s_input() already does this but atomisp_link_setup() did not do this. Add a new atomisp_select_input() helper which does this and use this in both atomisp_s_input() and atomisp_link_setup() for consistent behavior. Also make atomisp_link_setup() turn the sensor back off when a link gets disabled. Reviewed-by:
-
Hans de Goede authored
ATM the atomisp driver does not call s_power() at all when no s_input ioctl() is done by the application. This breaks older sensor drivers which rely on s_power() for power-management. Some drivers have worked around this, e.g. commit c5fafbad ("media: atomisp: gc0310: Power on sensor from set_fmt() callback") and commit b3118a94 ("media: atomisp: ov2722: Power on sensor from set_fmt() callback"), but this really should be fixed in the atomisp driver itself, so that all old drivers can work. A logical place to call s_power() would be from atomisp_start_streaming() / atomisp_stop_streaming(). But some older drivers, e.g. the atomisp-ov2722 driver already write mode related registers on set_fmt() instead of waiting on stream on. So the s_power(1) needs to happen at the first set_fmt(). Add an atomisp_s_sensor_power(..., 1) call just before calling set_fmt() for this. If the power was already enabled through e.g. a s_input ioctl atomisp_s_sensor_power() will skip calling the s_power() v4l2-subdev-op a second time. Reviewed-by:
-
Hans de Goede authored
Since the atomisp driver only supports 1 stream / only has 1 asd now, there is no longer a need to track which stream owns the sensor. So the asd owner-pointer of struct atomisp_input_subdev can be dropped, replace this with a simple camera_on boolean and add a new atomisp_s_sensor_power() helper which takes care of avoiding unbalanced s_power() subdev calls as well as takes care of handling the special -ENOIOCTLCMD return for subdevs which don't implement s_power(). Reviewed-by:
-
Andy Shevchenko authored
Always check exact PCI ID match and drop hack checks. Link: https://lore.kernel.org/r/20240402204447.294280-1-andriy.shevchenko@linux.intel.comSigned-off-by:
Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by:
-
Andy Shevchenko authored
gmin_i2c_dev_exists() is using open-coded variant of i2c_find_device_by_fwnode(). Replace it with a corresponding call. Link: https://lore.kernel.org/r/20240326202813.1425431-3-andriy.shevchenko@linux.intel.comSigned-off-by:
-
Andy Shevchenko authored
We don't use the PMIC I²C client device after getting its address. Drop the reference to it. We do not expect device to disappear as it should be taken care by the OpRegion drivers. Link: https://lore.kernel.org/r/20240326202813.1425431-2-andriy.shevchenko@linux.intel.comSigned-off-by:
-
Andy Shevchenko authored
Entire defs.h can be killed, so do that. Also kill same macro definitions in hive_types.h Link: https://lore.kernel.org/r/20240326190903.1422069-1-andriy.shevchenko@linux.intel.comSigned-off-by:
-
Jonathan Bergh authored
This patch makes the following changes: * Removes spurious whitespace after the opening parentheses as per code style guidelines Link: https://lore.kernel.org/r/20240227163043.112162-2-bergh.jonathan@gmail.comSigned-off-by:
Jonathan Bergh <bergh.jonathan@gmail.com> Signed-off-by:
-
Jonathan Bergh authored
This patch makes the following fixes: * Reformats a number of multiline block comments to ensure * and */ align correctly Link: https://lore.kernel.org/r/20240227163043.112162-1-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch fixes the following minor code and code style issues: * Removes a block of commented out (unused) code from the src * Reformats various multiline block comments to meet the kernel code style guidelines for block comments Link: https://lore.kernel.org/r/20240226194023.69070-7-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch makes the following change: * Removes an unnecessary 'return' statement from a void function implementation Link: https://lore.kernel.org/r/20240226194023.69070-6-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch makes the following changes: * Removes additional, unnecessary parentheses from a conditional "if" statement Link: https://lore.kernel.org/r/20240226194023.69070-5-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
media: staging: media: atomisp: Ensure trailing statements are on a newline and remove spurious whitespaces This patch fixes the following minor code style issues: * Ensure trailing statements are located on a newline * Removes an instance of a spurious whitespace following a conditional statement Link: https://lore.kernel.org/r/20240226194023.69070-4-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
media: staging: media: atomisp: Fixed "unsigned int *" rather than "unsigned *" declaration in variable declaration This code fixes a code style issue where: * Checkpatch suggests using "unsigned int *" rather than a bare "unsigned *" declaration in the code Link: https://lore.kernel.org/r/20240226194023.69070-3-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch does the following things: * Tidies up code in several places where braces were used in conjunction with single line conditional statements Link: https://lore.kernel.org/r/20240226194023.69070-2-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch fixes the following code style and formatting issues: * Ensure multiline block comments are correctly formatted * Remove extra braces not required for single line conditional statements Link: https://lore.kernel.org/r/20240225155359.41435-1-bergh.jonathan@gmail.comSigned-off-by:
-
Jonathan Bergh authored
This patch fixes the following formatting issues: * Fix various misaligned * and */ in multiline block comments * Remove unnecessary braces from single line conditional statements * Remove repeated word "from" from comment Link: https://lore.kernel.org/r/20240225120350.31226-1-bergh.jonathan@gmail.comSigned-off-by:
-
Hans de Goede authored
When starting a new stream initialize the sequence counters to 0 instead of -1. This fixes the following libcamera warning: Zero sequence expected for first frame (got 4294967295) Signed-off-by:
-
Hans de Goede authored
Disable V4L2_PIX_FMT_RGBX32 support, because it is broken. Selecting V4L2_PIX_FMT_RGBX32 output shows vertical columns with random data. For each 128 pixels in a row the last 28 (32?) or so pixels contain random data. Signed-off-by:
-
Hans de Goede authored
V4L2_PIX_FMT_RGB32 is deprecated because its definition never made clear if the 4th byte was alpha data or unused. Instead either V4L2_PIX_FMT_ARGB32 or V4L2_PIX_FMT_RGBX32 should be used. V4L2_PIX_FMT_RGB32 internally is mapped to IA_CSS_FRAME_FORMAT_RGBA888 and ia_css_frame_format.h says that the alpha bits there are unused, so replace V4L2_PIX_FMT_RGB32 with V4L2_PIX_FMT_RGBX32. Signed-off-by:
-
Hans de Goede authored
Some userspace apps (e.g. libcamera) do not like it if the enum_fmts ioctl returns the same fmt twice. Drop the second "do not use" V4L2_PIX_FMT_UYVY entry for parallel sensors to fix this. The current atomisp code does not support anything other then raw bayer sensors anyways, so dropping this is not an issue. Signed-off-by:
-
- 24 Apr, 2024 6 commits
-
-
Dan Carpenter authored
The subtract in this condition is reversed. The ->length is the length of the buffer. The ->bytesused is how many bytes we have copied thus far. When the condition is reversed that means the result of the subtraction is always negative but since it's unsigned then the result is a very high positive value. That means the overflow check is never true. Additionally, the ->bytesused doesn't actually work for this purpose because we're not writing to "buf->mem + buf->bytesused". Instead, the math to calculate the destination where we are writing is a bit involved. You calculate the number of full lines already written, multiply by two, skip a line if necessary so that we start on an odd numbered line, and add the offset into the line. To fix this buffer overflow, just take the actual destination where we are writing, if the offset is already out of bounds print an error and return. Otherwise, write up to buf->length bytes. Fixes: 9cb2173e ("[media] media: Add stk1160 new driver (easycap replacement)") Signed-off-by:
Dan Carpenter <dan.carpenter@linaro.org> Reviewed-by:
Ricardo Ribalda <ribalda@chromium.org> Signed-off-by:
Hans Verkuil <hverkuil-cisco@xs4all.nl>
-
Fabio Estevam authored
The JPEG decoder/encoder present in iMX8QXP and iMX8QM SoCs need the PER and IPG clocks to be functional, so add the clock entries. This also fixes the following schema warning: imx8qm-apalis-eval.dtb: jpegdec@58400000: 'assigned-clock-rates', 'assigned-clocks', 'clock-names', 'clocks' do not match any of the regexes: 'pinctrl-[0-9]+' from schema $id: http://devicetree.org/schemas/media/nxp,imx8-jpeg.yaml#Signed-off-by:Fabio Estevam <festevam@denx.de> Signed-off-by:
Mirela Rabulea <mirela.rabulea@nxp.com> Reviewed-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by:
-
Hans Verkuil authored
Ensure that drivers set max_num_buffers to a value >= 32. For now there is no reason for drivers to request a lower limit and doing so might potentially cause userspace issues. Note that the old check of > MAX_BUFFER_INDEX was pointless since q->max_num_buffers was already limited to MAX_BUFFER_INDEX or less. Also add a sanity check in __vb2_init_fileio(), returning -ENOSPC if a driver returns more than 32 buffers from VIDIOC_REQBUFS with count = q->min_reqbufs_allocation. The vb2_fileio_data struct only supports up to 32 buffers, so we need a check there. Signed-off-by: -
Hans Verkuil authored
The error that V4L2 returns when an attempt is made to open an unregisted device node is ENODEV, no ENXIO. Update the documentation. Signed-off-by: -
Hans Verkuil authored
If the CEC device is unregistered, then an attempt to open the device node should result in an -ENODEV error instead of -ENXIO. Document this as well in cec-func-open.rst. This is consistent with the error code returned by other file operations such as ioctl. Signed-off-by: -
Hans Verkuil authored
After the new V4L2 device node was registered, some additional initialization was done before the device node was marked as 'registered'. During the time between creating the device node and marking it as 'registered' it was possible to open the device node, which would return -ENODEV since the 'registered' flag was not yet set. Hold the videodev_lock mutex from just before the device node is registered until the 'registered' flag is set. Since v4l2_open will take the same lock, it will wait until this registration process is finished. This resolves this race condition. Signed-off-by:
Sakari Ailus <sakari.ailus@linux.intel.com> Cc: <stable@vger.kernel.org> # for vi4.18 and up
-
- 22 Apr, 2024 4 commits
-
-
Hans Verkuil authored
Merge tag 'tags/media-next-uvc-20240419' of git://git.kernel.org/pub/scm/linux/kernel/git/pinchartl/linux.git into media_stage uvcvideo fixes: Ricardo Ribalda (3): media: uvcvideo: Add quirk for Logitech Rally Bar media: uvcvideo: Fix power line control for Shine-Optics Camera media: uvcvideo: Disable autosuspend for Insta360 Link Signed-off-by:
-
Umang Jain authored
Add description for 'struct regmap *cci' member of struct imx335. This will fix the following compile-time warning: warning: Function parameter or struct member 'cci' not described in 'imx335' Signed-off-by:
Umang Jain <umang.jain@ideasonboard.com> Reviewed-by:
Kieran Bingham <kieran.bingham@ideasonboard.com> Signed-off-by:
-
Hans de Goede authored
On some ACPI platforms, such as Chromebooks the ACPI methods to change the power-state (_PS0 and _PS3) fully take care of powering on/off the sensor. On other ACPI platforms, such as e.g. various HP models with IPU6 + hi556 sensor, the sensor driver must control the avdd regulator itself. Add support for having the driver control the sensor's avdd regulator. Note this relies on the regulator-core providing a dummy regulator (which it does by default) on platforms where Linux is not aware of the avdd regulator. Signed-off-by:
-
Hans de Goede authored
On some ACPI platforms, such as Chromebooks the ACPI methods to change the power-state (_PS0 and _PS3) fully take care of powering on/off the sensor. On other ACPI platforms, such as e.g. various HP models with IPU6 + hi556 sensor, the sensor driver must control the sensor's clock itself. Add support for having the driver control an optional clock. Signed-off-by:
-
