Fix the position of arguments so camif->colorfx_cb, camif->colorfx_cr
are passed in proper order to the camif_hw_set_effect() function.
Addresses-Coverity-ID: 1248800
Addresses-Coverity-ID: 1269141
[s.nawrocki@samsung.com: edited commit message ]
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
As this is not in atomic context and it does not seem like a critical
timing setting a range of 1ms allows the timer subsystem to optimize
the hrtimer here.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Acked-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Use colorspace provided by the user as we are only doing scaling and
color encoding conversion, we won't be able to transform the colorspace
itself and the colorspace won't mater in that operation.
Also always use output colorspace on the capture side.
If the user does not provide a colorspace do not make it up, we might
later while processing need to figure out the colorspace, which
is possible depending on the frame size but do not ever guess and
leak that guess to the userspace.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Thibault Saunier <thibault.saunier@osg.samsung.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Trivial fix to spelling mistake in mfc_err error messages
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
v4l2_m2m_job_finish(), which is called from the interrupt handler with
slock acquired, can call the device_run() hook immediately if another
context was in the queue. This hook also acquires slock, resulting in
a deadlock for this scenario.
Fix this by releasing slock right before calling v4l2_m2m_job_finish().
This is safe to do as the state of the hardware cannot change before
v4l2_m2m_job_finish() is called anyway.
Signed-off-by: Alexandre Courbot <acourbot@chromium.org>
Acked-by: Jacek Anaszewski <jacek.anaszewski@gmail.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Now that rc_register_device() is reorganised, the dev->initialized
hack can be removed. Any driver which calls rc_register_device()
must be prepared for the device to go live immediately.
The dev->initialized commits that are relevant are commit c73bbaa4ec
("[media] rc-core: don't lock device at rc_register_device()") and
commit 08aeb7c9a4 ("[media] rc: add locking to fix register/show race").
The original problem was that show_protocols() would access
dev->rc_map.* and various other bits which are now properly
initialized before device_add() is called.
At the same time, remove the bogus "device is being removed" check.
Signed-off-by: David Härdeman <david@hardeman.nu>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The device core infrastructure is based on the presumption that
once a driver calls device_add(), it must be ready to accept
userspace interaction.
This requires splitting rc_setup_rx_device() into two functions
and reorganizing rc_register_device() so that as much work
as possible is performed before calling device_add().
Signed-off-by: David Härdeman <david@hardeman.nu>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
This module was merged after commit 5a8fc6a3ce ("Annotate hardware
config module parameters in drivers/media/"), so add add the missing
hardware annotations.
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
This patch add cec driver for STM32 platforms.
cec hardware block isn't not always used with hdmi so
cec notifier is not implemented. That will be done later
when STM32 DSI driver will be available.
Driver compliance has been tested with cec-ctl and cec-compliance
tools.
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Yannick Fertre <yannick.fertre@st.com>
[hans.verkuil@cisco.com: modified platform/Makefile to use obj-y]
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The buffer cache should be synchronised in buffer preparation, not when
the buffer is queued to the device. Fix this.
Mmap buffers do not need cache synchronisation since they are always
coherent.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Rename __qbuf_*() functions which are specific to a buffer type as
__prepare_*() which matches with what they do. The naming was there for
historical reasons; the purpose of the functions was changed without
renaming them.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
It is 'R-Car', not 'RCar'. No code or binding changes, only descriptive text.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Acked-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The async sub-device was not unregistered in ad5820_remove() as it should
have been; do it now. Also remove the now-redundant
v4l2_device_unregister_subdev().
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Depending on arch configs to include dirs is evil, and makes
harder to change drivers to work with COMPILE_TEST.
Replace them by obj-y.
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The DaVinci VPIF capture driver V4L2 OF support was added after the V4L2
OF framework got removed. Switch VPIF capture driver to V4L2 fwnode.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
The 2nd check of cnt > 8 is redundant as cnt is already checked
and thresholded to a maximum of 8 a few statements earlier.
Remove this redundant 2nd check.
Detected by CoverityScan, CID#114281 ("Logically dead code")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Some EXPORT_SYMBOL() on this file don't match the name of functions
that preceeds them.
Also, some kernel-doc markups also have a wrong name. Fix it.
Signed-off-by: Jasmin Jessich <jasmin@anw.at>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Make checkpatch happier by fixing this warning:
ERROR: "foo * bar" should be "foo *bar"
Signed-off-by: Jasmin Jessich <jasmin@anw.at>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
A driver error message is shown without DEBUG definition
to find an error and debug easily.
Signed-off-by: Hirokazu Honda <hiroh@chromium.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
If the mdp_* nodes are under an mdp sub-node, their corresponding
platform device does not automatically get its iommu assigned properly.
Fix this by moving the mdp component nodes up a level such that they are
siblings of mdp and all other SoC subsystems. This also simplifies the
device tree.
Although it fixes iommu assignment issue, it also break compatibility
with old device tree. So, the patch in driver is needed to iterate over
sibling mdp device nodes, not child ones, to keep driver work properly.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Signed-off-by: Minghsiu Tsai <minghsiu.tsai@mediatek.com>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
That way we don't have to rely on userspace to inject the headers on IDR
requests, and there is always enough information to start decoding at an
I-frame.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Implement the V4L2_CID_MPEG_VIDEO_FORCE_KEY_FRAME control to force IDR
frames. This is useful to implement VFU (Video Fast Update) on RTP
transmissions.
We already force an IDR frame at the beginning of each GOP to work
around a firmware bug on i.MX27, use the same mechanism to service IDR
requests from userspace.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The driver may sleep under a spin lock, and the function call path is:
snd_cx18_pcm_hw_free (acquire the lock by spin_lock_irqsave)
vfree --> may sleep
To fix it, the "substream->runtime->dma_area" is passed to a temporary
value, and mark it NULL when holding the lock. The memory is freed by
vfree through the temporary value outside the lock holding.
Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
[hans.verkuil@cisco.com: removed unnecessary 'if (dma_area)']
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The driver may sleep under a spin lock, and the function call path is:
snd_ivtv_pcm_hw_free (acquire the lock by spin_lock_irqsave)
vfree --> may sleep
To fix it, the "substream->runtime->dma_area" is passed to a temporary
value, and mark it NULL when holding the lock. The memory is freed by
vfree through the temporary value outside the lock holding.
Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
[hans.verkuil@cisco.com: removed unnecessary 'if (dma_area)']
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The current driver has a handful of hard-coded assumptions based on its
primary use for capture of video signals. Cleanup those assumptions,
and also query the subdev for format information and use that if
available.
Tested with 10-bit raw bayer input (SGRBG10) using the aptina,mt9v032
sensor, and also tested that composite video input still works from
ti,tvp514x decoder. Both tests done on the da850-evm board with the
add-on UI board.
NOTE: Will need further testing for other sensors with different bus
formats.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Enable getting of subdevs from DT ports and endpoints.
The _get_pdata() function was larely inspired by (i.e. stolen from)
am437x-vpfe.c
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Capture driver silently overrides pixel format with a hack (according to
the comments) to pass v4l2 compliance tests. This isn't needed for
normal functionality, and works for composite video and raw camera capture
without.
In addition, the hack assumes that it only supports raw capture with a
single format (SBGGR8) which isn't true. VPIF can also capture 10- and
12-bit raw formats as well. Forthcoming patches will enable VPIF
input with raw-camera support and has been tested with 10-bit format
from the aptina,mt9v032 sensor.
Any compliance failures should be fixed with a real fix.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The ir-spi driver has 2 issues which prevents it from working with
lirc:
1. The ir-spi driver uses 16 bits of SPI data to create one cycle of
the waveform. As such our SPI clock needs to be 16x faster than the
carrier frequency.
The driver is inconsistent in how it currently handles this. It
initializes it to the carrier frequency:
But the commit message has some example code which initialises it
to 16x the carrier frequency:
val = 608000;
ret = ioctl(fd, LIRC_SET_SEND_CARRIER, &val);
To maintain compatibility with lirc, always do the frequency adjustment
in the driver.
2. lirc presents pulses in microseconds, but the ir-spi driver treats
them as cycles of the carrier. Similar to other lirc drivers, do the
conversion with DIV_ROUND_CLOSEST().
Fixes: fe052da492 ("[media] rc: add support for IR LEDs driven through SPI")
Cc: stable@vger.kernel.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Drop a since commit e1159cb357 ("[media] mceusb: remove pointless
mce_flush_rx_buffer function") redundant reinitialisation of two urb
fields immediately after they have been initialised.
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Fix urb and transfer-buffer leaks in an urb-submission error path which
may be hit when a device is disconnected.
Fixes: 66e89522af ("V4L/DVB: IR: add mceusb IR receiver driver")
Cc: stable <stable@vger.kernel.org> # 2.6.36
Cc: Jarod Wilson <jarod@redhat.com>
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
interrupts is listed as an optional property in the DT
binding, but in reality the driver didn't work without it.
The existing driver relied on having the interrupt line
connected to the SoC to trigger handling various events.
Add the option to poll the interrupt status register via
a timer if no interrupt source is defined.
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Previously the mbus_fmt_code was set after the device was
registered. If a connected sub-device called tc358743_get_fmt
prior to that point it would get an invalid code back.
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
There was no way to query the supported mbus formats from this
driver. enum_mbus_code is the function to expose that, so
implement it.
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
- use __func__ instead of writing the full function name
- drop debug message in cec_config_log_addr since the same information
will be reported later
- use debug level 1 for errors and infrequent events, use level 2 for
debugging CEC message traffic
- log when a transmit is retried, very useful to know when debugging
- debug messages now all start with lower case
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
clk_prepare_enable() can fail here and we must check its return value.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The mutex that was initialised in v4l2_ctrl_handler_init_class() was not
destroyed in v4l2_ctrl_handler_free(). Do that.
Additionally, explicitly refer to the ctrl handler's mutex in mutex
initialisation for clarity.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Sometimes the caller is already holding the control handler mutex and
using it to serialise something. Provide an unlocked variant of the same
function to be used in those cases.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Ensure that ISI is clocked before starting sensor sub device.
Remove un-needed type check in try_fmt().
Use clamp() macro for hardware capabilities.
Fix wrong tabulation to space.
Signed-off-by: Hugues Fruchet <hugues.fruchet@st.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
-----BEGIN PGP SIGNATURE-----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=+jBV
-----END PGP SIGNATURE-----
Merge tag 'media/v4.12-2' into patchwork
media fixes for v4.12-rc4
* tag 'media/v4.12-2': (598 commits)
[media] rc-core: race condition during ir_raw_event_register()
[media] cec: drop MEDIA_CEC_DEBUG
[media] cec: rename MEDIA_CEC_NOTIFIER to CEC_NOTIFIER
[media] cec: select CEC_CORE instead of depend on it
[media] rainshadow-cec: ensure exit_loop is intialized
[media] atomisp: don't treat warnings as errors
Linux 4.12-rc3
x86/ftrace: Make sure that ftrace trampolines are not RWX
x86/mm/ftrace: Do not bug in early boot on irqs_disabled in cpu_flush_range()
selftests/ftrace: Add a testcase for many kprobe events
kprobes/x86: Fix to set RWX bits correctly before releasing trampoline
ftrace: Fix memory leak in ftrace_graph_release()
ipv4: add reference counting to metrics
net: ethernet: ax88796: don't call free_irq without request_irq first
ip6_tunnel, ip6_gre: fix setting of DSCP on encapsulated packets
sctp: fix ICMP processing if skb is non-linear
net: llc: add lock_sock in llc_ui_bind to avoid a race condition
PCI/msi: fix the pci_alloc_irq_vectors_affinity stub
blk-mq: Only register debugfs attributes for blk-mq queues
x86/timers: Move simple_udelay_calibration past init_hypervisor_platform
...
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Avoid a double fetch by reusing the values from the prior transfer.
Originally reported via https://bugzilla.kernel.org/show_bug.cgi?id=195559
Thanks to Pengfei Wang <wpengfeinudt@gmail.com> for reporting.
Signed-off-by: Steven Toth <stoth@kernellabs.com>
Reported-by: Pengfei Wang <wpengfeinudt@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
For both BT.656 and BT.1120 video, the pixel format
used by VPIF is Y/CbCr 4:2:2 in semi-planar format
(Luma in one plane and Chroma in another). This
corresponds to NV16 pixel format.
This is documented in section 36.2.3 of OMAP-L138
Technical Reference Manual, SPRUH77A.
The VPIF driver incorrectly sets the default format
to V4L2_PIX_FMT_YUV422P. Fix it.
Reported-by: Alejandro Hernandez <ajhernandez@ti.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Acked-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The hardware codec is not colorspace aware. We should trust userspace to
set the correct colorimetry information on the OUTPUT queue and mirror
the exact same setting on the CAPTURE queue.
There is no reason to restrict colorspace to JPEG or REC709 only. Also,
set the default colorspace, as returned by calling VIDIOC_TRY/S_FMT with
V4L2_COLORSPACE_DEFAULT, initially.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Used to differentiate between models with 3 and 6 inputs.
Signed-off-by: Ulrich Hecht <ulrich.hecht+renesas@gmail.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
If the requested pixelformat is not supported fallback to the default
format, do not revert the entire format.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The return value of __rvin_try_format_source is not checked, add a check
and propagate the error.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
There is only one subdevice registered with the async framework so there
is no need for the driver to check which subdevice is bound or unbound.
Remove these checks since the async framework preforms this.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
[hans.verkuil@cisco.com: fix typo: surce -> source]
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
With the driver stopping and starting the stream each time the driver is
stalled rvin_capture_off() can be folded in to the only caller.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
If userspace can't feed the driver with buffers as fast as the driver
consumes them the driver will stop video capturing and wait for more
buffers from userspace, the driver is stalled. Once it have been feed
one or more free buffers it will recover from the stall and resume
capturing.
Instead of continue to capture using the same capture mode as before the
stall allow the driver to choose between single and continuous mode
based on free buffer availability. Do this by stopping capturing when
the driver becomes stalled and restart capturing once it continues. By
doing
this the capture mode will be evaluated each time the driver is
recovering from a stall.
This behavior is needed to fix a bug where continuous capturing mode is
used, userspace is about to stop the stream and is waiting for the last
buffers to be returned from the driver and is not queuing any new
buffers. In this case the driver becomes stalled when there are only 3
buffers remaining streaming will never resume since the driver is
waiting for userspace to feed it more buffers before it can continue
streaming. With this fix the driver will then switch to single capture
mode for the last 3 buffers and a deadlock is avoided. The issue can be
demonstrated using yavta.
$ yavta -f RGB565 -s 640x480 -n 4 --capture=10 /dev/video22
Device /dev/video22 opened.
Device `R_Car_VIN' on `platform:e6ef1000.video' (driver 'rcar_vin') supports video, capture, without mplanes.
Video format set: RGB565 (50424752) 640x480 (stride 1280) field interlaced buffer size 614400
Video format: RGB565 (50424752) 640x480 (stride 1280) field interlaced buffer size 614400
4 buffers requested.
length: 614400 offset: 0 timestamp type/source: mono/EoF
Buffer 0/0 mapped at address 0xb6cc7000.
length: 614400 offset: 614400 timestamp type/source: mono/EoF
Buffer 1/0 mapped at address 0xb6c31000.
length: 614400 offset: 1228800 timestamp type/source: mono/EoF
Buffer 2/0 mapped at address 0xb6b9b000.
length: 614400 offset: 1843200 timestamp type/source: mono/EoF
Buffer 3/0 mapped at address 0xb6b05000.
0 (0) [-] interlaced 0 614400 B 38.240285 38.240303 12.421 fps ts mono/EoF
1 (1) [-] interlaced 1 614400 B 38.282329 38.282346 23.785 fps ts mono/EoF
2 (2) [-] interlaced 2 614400 B 38.322324 38.322338 25.003 fps ts mono/EoF
3 (3) [-] interlaced 3 614400 B 38.362318 38.362333 25.004 fps ts mono/EoF
4 (0) [-] interlaced 4 614400 B 38.402313 38.402328 25.003 fps ts mono/EoF
5 (1) [-] interlaced 5 614400 B 38.442307 38.442321 25.004 fps ts mono/EoF
6 (2) [-] interlaced 6 614400 B 38.482301 38.482316 25.004 fps ts mono/EoF
7 (3) [-] interlaced 7 614400 B 38.522295 38.522312 25.004 fps ts mono/EoF
8 (0) [-] interlaced 8 614400 B 38.562290 38.562306 25.003 fps ts mono/EoF
<blocks forever, waiting for the last buffer>
This fix also allow the driver to switch to single capture mode if
userspace doesn't feed it buffers fast enough. Or the other way around,
if userspace suddenly feeds the driver buffers faster it can switch to
continues capturing mode.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Instead of selecting single or continuous capture mode based on how many
buffers userspace intends to give us select capture mode based on number
of free buffers we can allocate to hardware when the stream is started.
This change is a prerequisite to enable the driver to switch from
continuous to single capture mode (or the other way around) when the
driver is stalled by userspace not feeding it buffers as fast as it
consumes it.
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
