Rename min_buffers_needed into min_queued_buffers and update
the documentation about it.
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil: Drop the change where min_queued_buffers + 1 buffers would be]
[hverkuil: allocated. Now this patch only renames this field instead of making]
[hverkuil: a functional change as well.]
[hverkuil: Renamed 3 remaining min_buffers_needed occurrences.]
Supported media bus codes on the resizer sink pad are identical to the
ISP source pad. The .enum_mbus_code() handler thus delegates the
enumeration to the ISP's operation. This is problematic for two
reasons:
- Format enumeration on the ISP source pad is dependent on the format
configured on the ISP sink pad for the same subdev state (TRY or
ACTIVE), while format enumeration on the resizer sink pad should
return all formats supported by the resizer subdev, regardless of the
ISP configuration.
- Delegating the operation involves creating a fake v4l2_subdev_state on
the stack to pass to the ISP .enum_mbus_code() handler. This gets in
the way of evolution of both the ISP enumeration handler and, more
generally, the V4L2 subdev state infrastructure.
Fix those two issues by implementing format enumeration manually for the
resizer.
Link: https://lore.kernel.org/r/20231126020948.2700-1-laurent.pinchart@ideasonboard.com
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
In rkisp1_isp_stop() and rkisp1_csi_disable() the driver masks the
interrupts and then apparently assumes that the interrupt handler won't
be running, and proceeds in the stop procedure. This is not the case, as
the interrupt handler can already be running, which would lead to the
ISP being disabled while the interrupt handler handling a captured
frame.
This brings up two issues: 1) the ISP could be powered off while the
interrupt handler is still running and accessing registers, leading to
board lockup, and 2) the interrupt handler code and the code that
disables the streaming might do things that conflict.
It is not clear to me if 2) causes a real issue, but 1) can be seen with
a suitable delay (or printk in my case) in the interrupt handler,
leading to board lockup.
Link: https://lore.kernel.org/r/20231207-rkisp-irq-fix-v3-4-358a2c871a3c@ideasonboard.com
Tested-by: Adam Ford <aford173@gmail.com> #imx8mp-beacon
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The subdev .init_cfg() operation is affected by two issues:
- It has long been extended to initialize a whole v4l2_subdev_state
instead of just a v4l2_subdev_pad_config, but its name has stuck
around.
- Despite operating on a whole subdev state and not being directly
exposed to the subdev users (either in-kernel or through the userspace
API), .init_cfg() is categorized as a subdev pad operation.
This participates in making the subdev API confusing for new developers.
Fix it by renaming the operation to .init_state(), and make it a subdev
internal operation.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Michael Riesch <michael.riesch@wolfvision.net> # for imx415
Acked-by: Shuah Khan <skhan@linuxfoundation.org> # for vimc
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
[Sakari Ailus: Resolved a conflict in Renesas vsp1 driver.]
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The resizer is used to scale the image, but also to change the
subsampling of YUV formats. Both the luma and chroma dimensions need to
be taken into account to decide whether or not to enable the resizer.
The current implementation disables the resizer if the chroma vertical
size isn't changed, which would be the case when scaling up by a factor
of 2 vertically while at the same time converting from YUV 4:2:2 to
4:2:0. Fix it by checking the luma sizes too.
While at it, reflow and clarify comments in the function.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The debug messages that show the resizer configuration are only printed
if the driver enables the resizer. This prevents checking the resizer
configuration when the driver believes it should be disabled. Fix it by
moving the dev_dbg() statements earlier.
Also, combine the two debug prints into a single one to gather all the
information in one place, which makes reading the log easier.
While at it, use %u instead of %d to print unsigned values.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The rkisp1_rsz_config() and rkisp1_rsz_config_regs() functions use a
v4l2_rect to pass frame sizes, leaving the top and left members unused
and uninitialized. Use v4l2_area instead.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The self path has a Y_PIC_SIZE register that needs to be programmed to
the total number of pixels, including the stride. This isn't done by the
driver, fix it.
While at it, reorder the register write order to sort them by address.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The dual crop register is the same for both the MP and SP channels. Drop
it from the rkisp1_rsz_config structure and use the
RKISP1_CIF_DUAL_CROP_CTRL macro directly in the code.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The line stride is expressed in the hardware as a number of pixels for
the first plane. The bytesperline must thus be a multiple of the first
plane's bpp value. Enforce this constraint.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Hex constants in the media subsystem are typically in lowercase. The
rkisp1 driver mostly follows that convention already, except in the
register definitions. Convert all hex constants to lowercase for
consistency.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the V4L2 subdev active state API to store the active format and crop
rectangle. This simplifies the driver not only by dropping the state
stored in the rkisp1_csi structure, but also by replacing the ops_lock
with the state lock.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the V4L2 subdev active state API to store the active format and crop
rectangle. This simplifies the driver not only by dropping the state
stored in the rkisp1_isp structure, but also by replacing the ops_lock
with the state lock.
The rkisp1_isp.sink_fmt field needs to be kept, as it is accessed from
the stats interrupt handler. To simplify the rkisp1_isp_set_sink_fmt()
implementation, the field is now set when starting the ISP, instead of
when setting the format.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the V4L2 subdev active state API to store the active format and crop
rectangle. This simplifies the driver not only by dropping the state
stored in the rkisp1_resizer structure, but also by replacing the
ops_lock with the state lock.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Set the v4l2_device already in async notifier init, so struct device
related to it will be available before the notifier is registered. This
requires separating notifier initialisation into two functions, one that
takes v4l2_device as its argument, v4l2_async_nf_init and
v4l2_async_subdev_nf_init, for sub-device notifiers. Registering the
notifier will use a single function, v4l2_async_nf_register.
This is done in order to make struct device available earlier, during
construction of the async connections, for sensible debug prints.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Add v4l2_async_connection_unique() function for obtaining a struct
v4l2_async_connection, typically allocated by drivers together with their
own information on an external sub-device.
The relation between connections and sub-devices still remains 1:1 but
this code becomes more complex when the relation soon changes.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Rename v4l2_async_subdev as v4l2_async_connection, in order to
differentiate between the sub-devices and their connections: one
sub-device can have many connections but the V4L2 async framework has so
far allowed just a single one. Connections in this context will later
translate into either MC ancillary or data links.
This patch prepares changing that relation by changing existing users of
v4l2_async_subdev to switch to v4l2_async_connection. Async sub-devices
themselves will not be needed anymore
Additionally, __v4l2_async_nf_add_subdev() has been renamed
__v4l2_async_nf_add_connection().
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The DT of_device.h and of_platform.h date back to the separate
of_platform_bus_type before it as merged into the regular platform bus.
As part of that merge prepping Arm DT support 13 years ago, they
"temporarily" include each other. They also include platform_device.h
and of.h. As a result, there's a pretty much random mix of those include
files used throughout the tree. In order to detangle these headers and
replace the implicit includes with struct declarations, users need to
explicitly include the correct includes.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Commit 0d346d2a6f ("media: v4l2-subdev: add subdev-wide state struct")
applied a large media tree-wide change produced by coccinelle. It was so
large that a set of identical indentation issues went unnoticed during
review. Fix them.
While at it, and because it's easy to review both changes together, add
a trailing comma for the last (and only) struct member initialization of
the related structures, to avoid future changes should new fields need
to be initialized.
Fixes: 0d346d2a6f ("media: v4l2-subdev: add subdev-wide state struct")
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Add support for NV16M and NV61M as output formats. As NV16, NV61, NV12M
and NV21M are already supported, the infrastructure is already in place
to support NV16M and NV61M, so it is sufficient to simply add relevant
entries to the list of output formats.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
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 <u.kleine-koenig@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Don't populate the const arrays on the stack, instead make them
static. Also makes the object code smaller.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Don't populate the const arrays on the stack, instead make them
static. Also makes the object code smaller.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The local sd_fmt variable in rkisp1_capture_link_validate() has
uninitialized fields, which causes random failures when calling the
subdev .get_fmt() operation. Fix it by initializing the variable when
declaring it, which zeros all other fields.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The ISP8000Nano v18.02 (found in the i.MX8MP) requires the ISP to be
enabled (as indicated by the ISP_CTRL.ISP_ENABLE bit) to configure the
lens shading table in internal RAM. The driver currently configures all
ISP initial parameters before enabling the ISP, which causes the LSC RAM
to not be initialized properly.
To fix this, split the rkisp1_params_configure() function into a
rkisp1_params_pre_configure() and a rkisp1_params_post_configure(). The
former configures all ISP parameters but LSC, while the latter
configures LSC. To implement this, the rkisp1_params_apply_params_cfg()
function is deconstructed, with two small helpers created to deal with
the parameters buffers, which are then used in rkisp1_params_isr(),
rkisp1_params_pre_configure() and rkisp1_params_post_configure().
While this initialization ordering is only needed for the ISP8000Nano
v18.02, it doesn't affect other ISP versions negatively, and can thus be
followed unconditionally.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The rkisp1_lsc_config() function incorrectly uses the
RKISP1_CIF_ISP_LSC_SECT_SIZE() macro for the gradient registers. Replace
it with the correct macro, and rename it from
RKISP1_CIF_ISP_LSC_GRAD_SIZE() to RKISP1_CIF_ISP_LSC_SECT_GRAD() as the
corresponding registers store the gradients for each sector, not a size.
This doesn't cause any functional change as the two macros are defined
identically (the size and gradient registers store fields in the same
number of bits at the same positions).
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The LSC module x/y size and grad configuration is stored in a set of 4
indexed registers each. The rkisp1-regs.h header defines all those
registers, but only the first one in each set is used, with manual
calculation of addresses of subsequent registers. Simplifies this by
merging all 4 register macros into one that takes the index as a
parameter. No functional change intended.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Use the u32 type instead of unsigned int to store register values in the
LSC configuration code, to make the variables' size more explicit. No
functional change intended.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Clean up the LSC configuration code to improve its readability by
shortening lines, using extra local variables and renaming long
variables. No functional change intended.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The resizer doesn't deal with color spaces, so it can accept any color
space on its input, and propagates it unchanged to its output. When
operating with a Bayer input format (in pass-through mode) further
restrict the YCbCr encoding and quantization to Rec 601 and full range
respectively, as for raw data the former ought to be ignored and the
latter is always full range.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Initialize the four color space fields on the sink and source video pads
of the resizer in the .init_cfg() operation. The resizer can't perform
any color space conversion, so set the sink and source color spaces to
the same defaults, which match the ISP source video pad default.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The driver currently only implements the Rec. 601 YCbCr encoding, extend
it with support for the other encodings defined by V4L2 (Rec. 709, Rec.
2020 and SMPTE240m). The coefficients have been calculated by rounding
the floating point values to the nearest Q1.7 fixed-point value,
adjusting the rounding to ensure that the sum of each line in the matrix
is preserved to avoid overflows.
At the hardware level, the RGB to YUV conversion matrix is fully
configurable, custom encoding could be supported by extending the ISP
parameters if desired.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The rkisp1_csm_config() function takes a pointer to the rkisp1_params
structure which contains the quantization value. There's no need to pass
it separately to the function. Drop it from the function parameters.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The rkisp1_config_isp() function uses the format on the sink pad of the
ISP to configure quantization at the output of the ISP. This is
incorrect, as hinted by the src_frm variable name that stores the
format. Fix it by using the source pad.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
The ISP output color space is configured through the ISP source pad. At
the moment, only the quantization can be set. Extend it to the three
other color space fields:
- The ycbcr_enc field will be used to configure the RGB to YUV matrix
(currently hardcoded to Rec. 601).
- The colorspace (which controls the color primaries) and xfer_func
fields will not be used to configure the ISP, as the corresponding
hardware blocks (the cross-talk RGB to RGB matrix and the tone mapping
curve) are programmed directly by applications through ISP parameters.
Nonetheless, those two fields should be set by applications to match
the ISP configuration, in order to propagate the correct color space
down the pipeline up to the capture video nodes.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Dafna Hirschfeld <dafna@fastmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
