The PLL calculator only searched even total divisor values apart from one,
but this is wrong: the total divisor is odd in cases where system divisor
is one. Fix this by including odd total PLL values where system divisor is
one to the search.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Make VT divisors 16-bit unsigned numbers. They don't need 32 bits after
all.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Refactor bounds checks so that the caller can decide what to check. This
allows doing the checks early, when the values are available.
This also adds front OP PLL configuration and limits.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Print information on PLL tree configuration based on the flags. This also
adds support for printing dual PLL trees, and better separates between OP
and VT PLL trees.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Better separate OP PLL branch calculation from VT branch calculation.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Some sensors support derating (VT domain speed faster than OP) or
overrating (VT domain speed slower than OP). While this was supported for
the driver, the hardware support for the feature was never verified. Do
that now, and for those devices without that support, VT and OP speeds
have to match.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Split off the VT sub clock tree calculation from the rest, into its own
function. Also call the op_pll_fr argument pll_fr, since soon these may
not be OP tree values.
This paves way for additional features in the future such as dual PLL
support.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Add C-PHY support for the CCS PLL calculator.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Add sanity checks for fields that could cause division by zero.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Flexible OP PLL pixel clock divider allows a higher OP pixel clock than
what the bus can transfer. This generally makes it easier to select pixel
clock dividers.
This changes how the pixel rate on the bus and minimum VT divisor are
calculated, as the pixel rate is no longer directly determined by the
OP pixel clock and the number of the lanes.
Also add a sanity check for sensors that do not support flexible OP PLL
pixel clock divider. This could have caused the PLL calculator to come up
with an invalid configuration for those devices.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The l parameter defines the number of clock cycles to process a single
pixel per OP lane. It is calculated based on a new register
op_bits_per_lane.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
CCS allows odd PLL dividers other than 1, granted that the corresponding
capability bit is set. Support this both in the PLL calculator and the CCS
driver.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Add support for decoupled OP domain clock calculation. This means that the
number of VT and OP domain clocks are no longer dependent on the number of
CSI-2 lanes in the lane speed mode.
The support also replaces the existing quirk flag to calculate OP domain
clocks per lane.
Also support decoupled OP domain calculation in the CCS driver.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Convey the relevant PLL flags to the PLL calculator. Also the lane speed
model affects how the link rate is calculated on the CSI-2 bus, as the
rate is total of all lanes.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
CCS PLL includes a capability to calculate the VT clocks on per-lane
basis. Add support for this feature.
Move calculation of the pixel rate on the CSI-2 bus early in the function
as everything needed to calculate it is already available.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Use uint32_t instead of unsigned int for a variable that contains
explicitly 32-bit numbers.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The additional multiplier (for higher VT timing) of the PLL multiplier was
checked against the upper limit but the result was rounded up, possibly
producing too high additional multiplier. Round down instead to keep
within hardware limits.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The comment is about minimum PLL multiplier but the related check really
deals with the maximum PLL multiplier.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The external clock frequency times the PLL multiplier may exceed the value
range of 32-bit unsigned integers. Instead perform the same calculation y
using two divisions.
The result has some potential to be different, but that's ok: this number
is used to limit the range of pre-PLL divisors to find optimal values. So
the effect of the rare case of a different result here would mean an
invalid pre-PLL divisor is tried. That will be found out a little later in
any case.
Also guard against dividing by zero if the external clock frequency is
higher than the maximum OP PLL output clock --- a rather improbable case.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The lower bound of the pre-PLL divider was calculated based on OP SYS
clock frequency which is also affected by the OP SYS clock divider. This
is wrong. The right clock frequency is that of the PLL output clock.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The OP system clock frequency defines the CSI-2 bus clock frequency, not
the PLL output clock frequency. Both values were overwritten in the end,
but the wrong limit value was used for the OP system clock frequency,
possibly leading to too high frequencies being used.
Also remove now duplicated calculation of OP system clock frequency later
in the PLL calculator.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Use the BIT macro for setting individual bits.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The CCS pll is used by the CCS driver at the moment, but documenting the
interface makes sense. It's non-trivial and the calculator could be used
elsewhere.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
This way the struct will use less memory, with better packing and no waste
due to unsigned long.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Differentiate between CSI-2 D-PHY and C-PHY. This does not yet include
support for C-PHY.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The parallel bus PLL calculation has no users. Remove it.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The VT divisor search can be ended if we've already found the value that
corresponds exactly the total divisor, as there are no better (lower)
values available.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Use the correct video timing divisor to calculate the SYS divisor. Instead
of the current value, the minimum was used. This could have resulted in a
too low SYS divisor.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The CCS spec supports a lot of variation in the PLL. Split the PLL in
front and back parts to better prepare for supporting it.
Also use CCS compliant naming for IP and OP PLL frequencies (i.e. include
"clk" in the name).
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
pll->pll_op_clk_freq is a 32-bit number. It does not need div_u64 to
divide it.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Fix a few new indentation warnings found with yamllint (now integrated
into the checks).
Cc: Ricardo Ribalda <ribalda@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Acked-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> # For adv7604
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The 'disconnect' error injection functionality suffered from bit rot.
New device nodes were added without updating vivid_user_gen_s_ctrl(), so
that function had to be updated for the new device nodes.
Also, vivid didn't check if specific device nodes were actually ever
created, so the vivid_is_last_user() would fail on that (it would return
true instead of false in that case).
Finally, selecting Disconnect, then unbind the vivid driver would fail
since the remove() would think that the device nodes were already
unregistered. Keep track of whether disconnect was pressed and re-register
the device nodes in remove() before doing the real unregister.
[hverkuil: unsigned uses -> unsigned int uses]
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The VNCSI_IFMD register controls the data expansion mode and the
channel routing between the CSI-2 receivers and VIN instances.
According to the chip manual revision 2.20 not all fields are available
for all the SoCs:
- V3M, V3H and E3 do not support the DES1 field has they do not feature
a CSI20 receiver.
- D3 only supports parallel input, and the whole register shall always
be written as 0.
Inspect the per-SoC channel routing table where the available CSI-2
instances are reported and configure VNCSI_IFMD accordingly.
This patch supports this BSP change commit:
https://github.com/renesas-rcar/linux-bsp/commit/f54697394457
("media: rcar-vin: Fix VnCSI_IFMD register access for r8a77990")
[hverkuil: replace BSP commit ID with BSP URL]
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Suggested-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The GE2D is a 2D accelerator with various features like configurable
blitter with alpha blending, frame rotation, scaling, format conversion
and colorspace conversion.
The driver implements a Memory2Memory VB2 V4L2 streaming device permitting:
- 0, 90, 180, 270deg rotation
- horizontal/vertical flipping
- source cropping
- destination compositing
- 32bit/24bit/16bit format conversion
This adds the support for the GE2D version found in the AXG SoCs Family.
The missing features are:
- Source scaling
- Colorspace conversion
- Advanced alpha blending & blitting options
Is passes v4l2-compliance:
SHA: ea16a7ef13a902793a5c2626b0cefc4d956147f3, 64 bits, 64-bit time_t
[hverkuil: add missing linux/bitfield.h include]
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The GE2D is a 2D accelerator with various features like configurable blitter
with alpha blending, frame rotation, scaling, format conversion and colorspace
conversion.
This adds the bindings for the GE2D version found in the AXG SoCs Family.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Commit 31d9b9ef8564 ("media: cedrus: Register all codecs as capability")
makes separate capability flags for each codec. However, VP8 codec was
merged at the same time as mentioned patch, so there is no capability
flag for it.
This patch adds capability flag for VP8 and enables it for all variants
except for V3s.
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The gspca driver leaks memory when a probe fails. gspca_dev_probe2()
calls v4l2_device_register(), which takes a reference to the
underlying device node (in this case, a USB interface). But the
failure pathway neglects to call v4l2_device_unregister(), the routine
responsible for dropping this reference. Consequently the memory for
the USB interface and its device never gets released.
This patch adds the missing function call.
Reported-and-tested-by: syzbot+44e64397bd81d5e84cba@syzkaller.appspotmail.com
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
CC: <stable@vger.kernel.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Planar pixel formats are documented in separate files. This duplicates
information, as those formats share comon traits. Consolidate them in a
single file and summarize their descriptions in a single table.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Semi-planar pixel formats are documented in separate files. This
duplicates information, as those formats share comon traits. Consolidate
them in a single file and summarize their descriptions in a single
table.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Luma-only pixel formats are documented in separate files. This
duplicates information, as those formats share comon traits. Consolidate
them in a single file and describe them in a single table.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Document the naming scheme for the existing packed YUV 4:4:4 formats, as
previously done for the RGB formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The 4:4:4 packed YUV formats are documented with a bit-level
representation, which creates a wide table. Switch to a byte-oriented
representation to make it more compact. This prepares for the addition
of formats with more than 8 bits per component, that would make the
table way too wide.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The padding bits are left blank, which look weird in the XYUV format,
and, worse, may lead to the VUYX format to be understand as consuming 3
bytes per pixel. Add 'X' for padding bits as we do for RGB formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The pixfmt-packed-yuv.rst file documents packed YUV 4:4:4 formats, but
is titled generically as "Packed YUV formats". 4:2:2 and 4:1:1 packed
YUV formats are documented in separate files, which can be confusing.
Group all packed YUV formats in pixfmt-packed-yuv.rst, which allows
documenting the 4:2:2 formats in a more concise way.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Document YUV subsampling, including chroma spatial siting, and replace
the siting examples in individual formats by references to the common
documentation.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Using '-' to represent padding bits and bytes make text and tables more
difficult to read. Use 'x' and 'X' instead.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
All formats using 8 bits per component can be described with a byte
granularity instead of a bit granularity without loss of precision. This
makes the corresponding table more compact and easier to read.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The naming scheme for the RGB pixel formats has been developed
organically, and isn't consistent between formats using less than 8 bits
per pixels (mostly stored in 1 or 2 bytes per pixel, except for RGB666
that uses 4 bytes per pixel) and formats with 8 bits per pixel (stored
in 3 or 4 bytes). For the latter category, the names use a components
order convention that is the opposite of the first category, and the
opposite of DRM pixel formats. This has led to lots of confusion in the
past, and would really benefit from being explained more precisely. Do
so, which also prepares for the addition of additional RGB pixels
formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Improve readability of the documentation by adding a section title for
the deprecated formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>