Martin Bugge [Tue, 10 Dec 2013 15:00:06 +0000 (12:00 -0300)]
[media] adv7842: restart STDI once if format is not found
The STDI block may measure wrong values, especially for lcvs and lcf.
If the driver can not find any valid timing, the STDI block is restarted
to measure the video timings again. The function will return an error,
but the restart of STDI will generate a new STDI interrupt and the format
detection process will restart.
Copied from adv7604.
Signed-off-by: Martin Bugge <marbugge@cisco.com> Cc: Mats Randgaard <matrandg@cisco.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Martin Bugge [Thu, 5 Dec 2013 14:52:39 +0000 (11:52 -0300)]
[media] adv7842: 625/525 line standard jitter fix
Both the PAL and NTSC standards are interlaced where a
frame consist of two fields. Total number of lines in a frame in both systems
are an odd number so the two fields will have different length.
In the 625 line standard ("PAL") the odd field of the frame is transmitted first,
while in the 525 standard ("NTSC") the even field is transmitted first.
This adds the possibility to change output config between the fields and standards.
This setting will reduce the "format-jitter" on the signal sent by the pixelport
moving the difference between the fields to vertical front/back-porch only.
Signed-off-by: Martin Bugge <marbugge@cisco.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Hans Verkuil [Fri, 20 Dec 2013 08:44:27 +0000 (05:44 -0300)]
[media] adv7842: support YCrCb analog input, receive CEA formats as RGB on VGA input
Added support for YCrCb analog input.
If input is ADV7842_MODE_RGB and RGB quantization range is set to
V4L2_DV_RGB_RANGE_AUTO, then video with CEA timings will be received
as RGB. For ADV7842_MODE_COMP, automatic CSC mode will be selected.
See table 48 on page 281 in "ADV7842 Hardware Manual, Rev. 0, January 2011"
for details.
Make sure that when switching inputs the RGB quantization range is
updated as well.
Also updated the platform_data in ezkit to ensure that what was the old
default value is now explicitly specified, so the behavior for that board
is unchanged.
Signed-off-by: Mats Randgaard <matrandg@cisco.com> Signed-off-by: Martin Bugge <marbugge@cisco.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Cc: Scott Jiang <scott.jiang.linux@gmail.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Martin Bugge [Tue, 10 Dec 2013 14:14:26 +0000 (11:14 -0300)]
[media] adv7842: properly enable/disable the irqs
The method of disabling the irq-output pin caused many "empty"
interrupts. Instead, actually disable/enable the interrupts by
changing the interrupt masks.
Also enable STORE_MASKED_IRQ in INT1 configuration, otherwise when HDMI
events happen while the interrupt is masked those events will be ignored
when the interrupt is unmasked.
Signed-off-by: Martin Bugge <marbugge@cisco.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Mats Randgaard [Thu, 5 Dec 2013 13:33:41 +0000 (10:33 -0300)]
[media] adv7604: improve HDMI audio handling
- Mute audio before switching inputs to avoid noise/pops
- Mute audio if audio FIFO over-/underflows (AD Recommended setting)
- Reset FIFO if it over-/underflows (AD Recommended setting)
Mats Randgaard [Thu, 5 Dec 2013 13:26:11 +0000 (10:26 -0300)]
[media] adv7604: remove debouncing of ADV7604_FMT_CHANGE events
ADV7604_FMT_CHANGE events was debounced in adv7604_isr() to avoid
that a receiver with a unstable input signal would block the event
handling for other inputs. This solution was prone to errors.
A better protection agains interrupt blocking is to delay the call
of the interrupt service routine in the adv7604 driver if too many
interrupts are received within a given time.
Mats Randgaard [Tue, 10 Dec 2013 12:57:09 +0000 (09:57 -0300)]
[media] adv7604: improve EDID handling
- split edid_write_block()
- do not use edid->edid before the validity check
- Return -EINVAL if edid->pad is invalid
- Save both registers for SPA port A
- Set SPA location to default value if it is not found
Mats Randgaard [Thu, 5 Dec 2013 13:05:58 +0000 (10:05 -0300)]
[media] adv7604: Receive CEA formats as RGB on VGA (RGB) input
If the input is ADV7604_INPUT_VGA_RGB and RGB quantization range is
set to V4L2_DV_RGB_RANGE_AUTO, video with CEA timings will be
received as RGB. For ADV7604_INPUT_VGA_COMP, automatic CSC mode
will be selected.
See table 44 on page 205 in "ADV7604 Hardware Manual, Rev. F, August 2010"
for details.
Martin Bugge [Thu, 5 Dec 2013 12:02:20 +0000 (09:02 -0300)]
[media] adv7511: disable register reset by HPD
Whenever the hotplug pin is pulled low the chip resets a whole bunch
of registers. It turns out that this can be turned off on the adv7511.
Do so, as this 'feature' introduces race conditions in setting up
registers, particular when the hotplug pin bounces a lot.
Signed-off-by: Martin Bugge <marbugge@cisco.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
[media] em28xx: prevent registering wrong interfaces for audio-only
A few devices (em2860) use a separate interface for audio only
Audio Vendor Class USB. That interface should not be used by
Remote Controller, Analog TV or Digital TV.
Prevents initializing all non-audio extensions for the audio
only interface.
[media] em28xx: move analog-specific init to em28xx-video
There are several init code inside em28xx-cards that are actually
part of analog initialization. Move the code to em28x-video, in
order to remove part of the mess.
In thesis, no functional changes so far.
Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Instead of using two I2C operations between write and read,
use just one i2c_transfer. That allows I2C mutexes to not
let any other I2C transfer between the two.
Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com> Tested-by: Frank Schäfer <fschaefer.oss@googlemail.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Dan Carpenter [Wed, 18 Dec 2013 07:29:24 +0000 (04:29 -0300)]
[media] v4l: omap4iss: Restore irq flags correctly in omap4iss_video_buffer_next()
The spin_lock_irqsave() macro is not nestable. The second call will
overwrite the first record of "flags" so the IRQs will not be enabled
correctly at the end of the function.
In the current code, this function is always called from the IRQ handler
so everything works fine and this fix doesn't change anything.
Dan Carpenter [Mon, 16 Dec 2013 20:19:50 +0000 (17:19 -0300)]
[media] v4l: omap4iss: use snprintf() to make smatch happy
Smatch complains here because name is a 32 character buffer and we
adding the "OMAP4 ISS " prefix as well for a total of 42 characters.
The sd->name buffer can only hold 32 characters. I've changed it to use
snprintf() to silence the overflow warning.
Also I have removed the call to strlcpy() which is a no-op.
[media] exynos4-is: Remove dependency on PM_RUNTIME from Kconfig
Now when the sub-drivers are fixed to work with runtime PM disabled
this erroneous dependency can be removed.
The CAM and ISP power domains should be left in active state by the
platform if runtime PM is not used.
[media] exynos4-is: Correct clean up sequence on error path in fimc_is_probe()
The memory allocator is being initialized before registering the subdevs
so reverse the cleanup sequence to avoid trying unregister not registered
subdevs.
[media] exynos4-is: Activate mipi-csis in probe() if runtime PM is disabled
Devices should also operate normally when runtime PM is not enabled.
In case runtime PM is disabled activate the device already in probe().
Any related power domain needs to be then left permanently in active
state by the platform.
Andrzej Hajda [Thu, 5 Dec 2013 11:38:39 +0000 (08:38 -0300)]
[media] Add driver for Samsung S5K5BAF camera sensor
Driver for Samsung S5K5BAF UXGA 1/5" 2M CMOS Image Sensor with
embedded SoC ISP. The device is exposed as two V4L2 subdevices:
- S5K5BAF-CIS - the image sensor matrix, fixed 1600x1200 format,
no controls.
- S5K5BAF-ISP - the Image Signal Processor, formats up to 1600x1200,
pre/post ISP cropping, downscaling via selection API, controls.
[m.chehab@samsung.com: Whitespace cleanups] Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Andrzej Hajda <a.hajda@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
[media] subdev autoselect only works if I2C and I2C_MUX is selected
As reported by the kbuild test robot <fengguang.wu@intel.com>:
> warning: (VIDEO_EM28XX_DVB) selects DVB_M88DS3103 which has unmet direct dependencies (MEDIA_SUPPORT && DVB_CORE && I2C && I2C_MUX)
> drivers/built-in.o: In function `m88ds3103_release':
> >> m88ds3103.c:(.text+0x1ab1af): undefined reference to `i2c_del_mux_adapter'
> drivers/built-in.o: In function `m88ds3103_attach':
> >> (.text+0x1ab342): undefined reference to `i2c_add_mux_adapter'
There are 3 possible ways to fix it:
1) make em28xx dependent on I2C_MUX.
That sounds wrong, as the em28xx bridge doesn't have i2c muxes on it,
and just one frontend has.
Well, subdevs could eventually be converted to, instead of using dvb
i2c gate control, to use i2c mux support.
That makes sense, but it takes time and lots of effort. Not sure if
this will happen anytime soon.
2) MEDIA_SUBDRV_AUTOSELECT can be dependent of I2C and I2C_MUX.
That means that users will need to manually enable I2C_MUX on some
distributions. Not sure about others, but, on Fedora, this option is
disabled.
So, it can end by generating a number of complains from users
that their devices suddenly stopped working after a Kernel upgrade,
at least until all distros that ship Kernels with I2C_MUX enabled.
3) if MEDIA_SUBDRV_AUTOSELECT is selected, it will select I2C and I2C_MUX.
Of course, MEDIA_SUBDRV_AUTOSELECT will need to inherit all dependencies
that I2C and I2C_MUX have (only HAS_IOMEM).
The disadvantage is that, if new dependencies are added on I2C, they'll
also need to be added here.
As the hole idea of autoselect is to let the user not bother about whatever
frontend/tuner is used by a driver, IMHO, (3) is the better solution.
Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Hans Verkuil [Thu, 12 Dec 2013 12:04:44 +0000 (09:04 -0300)]
[media] omap24xx/tcm825x: move to staging for future removal
The omap24xx driver and the tcm825x sensor driver are the only two
remaining drivers to still use the old deprecated v4l2-int-device API.
Nobody maintains these drivers anymore. But unfortunately the v4l2-int-device
API is used by out-of-tree drivers (MXC platform). This is a very bad situation
since as long as this deprecated API stays in the kernel there is no reason for
those out-of-tree drivers to convert.
This patch moves v4l2-int-device and the two drivers that depend on it to
staging in preparation for their removal.
If someone would be interested in getting these drivers to work, then start with
this since it's not very far from the state where they used to work:
Hans Verkuil [Thu, 12 Dec 2013 10:35:22 +0000 (07:35 -0300)]
[media] sn9c102: prepare for removal by moving it to staging
During the last media summit meeting it was decided to move this driver to
staging as the first step to removing it altogether.
Most webcams covered by this driver are now supported by gspca. Nobody has the
hardware to convert the remaining devices to gspca.
This driver needs a major overhaul to have it conform to the latest frameworks
and compliancy tests.
Without hardware, however, this is next to impossible. Given the fact that
this driver seems to be pretty much unused (it has been removed from Fedora
several versions ago and nobody complained about that), we decided to drop
this driver.
This patch moves it to staging. Some time in 2014 we will drop it completely.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Cc: Hans de Goede <hdegoede@redhat.com> Cc: Luca Risolia <luca.risolia@studio.unibo.it> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Antti Palosaari [Mon, 2 Dec 2013 16:11:21 +0000 (13:11 -0300)]
[media] m88ds3103: remove unneeded AGC from inittab
Optimal AGC is highly depended on used RF tuner and due to that
it is already included to chip configuration. However, inittab
has default AGC value, which was later replaced by one from config.
Add also comment to all chip configuration options about default
values and if those are needed to set or not.
Reported-by: David Howells <dhowells@redhat.com> Signed-off-by: Antti Palosaari <crope@iki.fi> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Used synthesizer is very typical integer-N PLL, with configurable
reference frequency divider, output frequency divider and of
course N itself. Most common method to calculate values is first
select output divider, then calculate VCO frequency and finally
calculate PLL N from VCO frequency. Do it that way.
Also make some cleanups for filter logic and signal strength.
Signed-off-by: Antti Palosaari <crope@iki.fi> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Antti Palosaari [Tue, 19 Nov 2013 23:32:42 +0000 (20:32 -0300)]
[media] m88ds3103: use I2C mux for tuner I2C adapter
Switch standard I2C adapter to muxed I2C adapter.
David reported that I2C adapter implementation caused deadlock.
I discussed with Jean and he suggested to implement it as a
multiplexed i2c adapter because tuner I2C bus could be seen like
own I2C segment.
Reported-by: David Howells <dhowells@redhat.com> Cc: Jean Delvare <khali@linux-fr.org> Signed-off-by: Antti Palosaari <crope@iki.fi> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
On dib8000, the block error count is a monotonic 32 bits register.
With DVBv5 stats, we use a 64 bits counter, that it is reset
when a new channel is tuned.
Change the UCB counting start from 0 and to be returned with
64 bits, just like the API requests.
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> Acked-by: Patrick Boettcher <pboettcher@kernellabs.com>
Better to have Signal strength in dB.
This takes a very rough estimation for the signal strength,
that was calibrated using a Dektec DTA-2111 Gold RF generator
and a Pixelview dib8076 stick.
It estimates the signal strength using a linear equation where:
- the max is -22.5 dBm, with returns 55953
- the min is -35.0 dBm, with returns 50110
With -22dBm, the signal strengh is returned as 65535.
Unfortunately, the min strength generated with DTA-2111 is
-35dBm.
It should be noticed that approximating it by a linear equation
is not right. I should probably be splitting it into 0.5 dB
linear segments, in order to get a higher precision, just like
it is done on mb86a20s, but that would force me to add some
attenuators, in order to get dB levels below -35dBm, which is,
btw, strong enough to get signal lock.
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> Acked-by: Patrick Boettcher <pboettcher@kernellabs.com>
On ISDB-T, the valid values for interleaving are 0, 1, 2 and 4.
While the first 3 are properly reported, the last one is reported
as 3 instead. Fix it.
Tested with a Dektec DTA-2111 RF generator.
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> Acked-by: Patrick Boettcher <pboettcher@kernellabs.com>
[media] dib8000: improves the auto search mode check logic
The logic that detects if auto search mode should be used is too
complex.
Also, it doesn't cover all cases, as the dib8000_tune logic
requires either auto mode or a fully specified manual mode.
So, move it to a separate function and add some extra debug
data to help identifying when it falled back to auto mode,
because the manual settings are invalid.
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> Acked-by: Patrick Boettcher <pboettcher@kernellabs.com>
[media] dib8000: Don't let tuner hang due to a call to get_frontend()
Both dvbv5-scan and dvbv5-zap tools call FE_GET_PROPERTY inside the
loop that checks for stats. If the frontend doesn't support DVBv5, it
falls back to call the DVBv5 stats APIs(FE_READ_BER, FE_READ_SIGNAL,
FE_READ_SNR and FE_READ_UNCORRECTED_BLOCKS).
A call to FE_GET_PROPERTY makes dvb-frontend core to call get_frontend().
However, due to a race condition on dib8000 between dib8000_get_frontend
and dib8000_tune, if get_frontend occurs too early, it causes the
tune state machine to fail and not get any lock.
This patch adds a workaround code that makes get_frontend() to just
return if none of the frontends have a SYNC. This change fixed the issue
with dvbv5-scan/dvbv5-zap, but a fine-tuned logic might be needed in
the future, when we implement DVBv5 stats on this frontend.
The procedure to test the bug and the fix is the one below:
As the dvb-frontend kthread can be called anytime, it can race
with some get status ioctl. So, it seems better to avoid one to
race with the other while reading a 32 bits register.
I can't see any other reason for having a mutex there at I2C, except
to provide such kind of protection, as the I2C core already has a
mutex to protect I2C transfers.
Note: instead of this approach, it could eventually remove the dib8000
specific mutex for it, and either group the 4 ops into one xfer or
to manually control the I2C mutex. The main advantage of the current
approach is that the changes are smaller and more puntual.
projects / linux-beck.git / log