Intermittent RX truncation and loss of IR received data. This resulted
in receive stream synchronization errors where driver attempted to
incorrectly parse IR data (eg 0x90 below) as command response.

[ 3969.139898] mceusb 1-1.2:1.0: processed IR data
[ 3969.151315] mceusb 1-1.2:1.0: rx data: 00 90 (length=2)
[ 3969.151321] mceusb 1-1.2:1.0: Unknown command 0x00 0x90
[ 3969.151336] mceusb 1-1.2:1.0: rx data: 98 0a 8d 0a 8e 0a 8e 0a 8e 0a 8e 0a 9a 0a 8e 0a 0b 3a 8e 00 80 41 59 00 00 (length=25)
[ 3969.151341] mceusb 1-1.2:1.0: Raw IR data, 24 pulse/space samples
[ 3969.151348] mceusb 1-1.2:1.0: Storing space with duration 500000

Bug trigger appears to be normal, but heavy, IR receiver use.
Signed-off-by: NA Sun <as1033x@comcast.net>
Signed-off-by: NSean Young <sean@mess.org>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Trivial fix to spelling mistake in dev_err message
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Initialize scalar variables _pid_ and _ver_ to avoid a possible misbehavior.

Addresses-Coverity-ID: 1324239
Addresses-Coverity-ID: 1324240
Signed-off-by: NGustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

As of commit bb475230 ("reset: make optional functions really
optional"), the reset framework API calls use NULL pointers to
describe optional, non-present reset controls.

This allows to return errors from devm_reset_control_get_optional
without special cases and to call reset_control_reset unconditionally.
Signed-off-by: NPhilipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

This patch provide only digital support;
The device is based on 24C02N EEPROM, Panasonic MN88473 demodulator,
Rafael Micro R828D tuner and CX23102-11Z chipset;
USB id: 15f4:0135.

Status:
- DVB-T/T2 works fine;
- Composite works fine;
- Analog not implemented.
Signed-off-by: NOleh Kravchenko <oleg@kaa.org.ua>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

For reasons unclear, we intermittently see a case where the tune
is successful but the bulk stream fails to deliver any packets.

Add a timer to automatically stop/start the data pump if we
encounter such a case.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

We need to set the initial modulation on driver setup, or else any
calls to GET_FRONTEND prior to the first SET_FRONTEND call will get
back garbage.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The xc5000 driver should not be spinning waiting for an analog lock.
The ioctl() should be returning immediately and the application is
responsible for polling for lock status.

This behavior isn't very visible in cases where you tune to a valid
channel, since lock is usually achieved much faster than 400ms.
However it is highly visible where doing things like changing video
standards, which sends tuning request for a frequency that is
almost never going to have an actual channel on it.

Also fixup the return values to treat zero as success and an actual
error code on error (to be consistent with other functions).  Note
this change has no practical effect at this time as none of the
callers inspect the return value.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Only looking at the lock register causes the status to float
between locked and not locked when there is no signal.  So improve
the logic to also examine the state of the FSC PLL, which results
in the lock status being consistently reported.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The second highest bit in the register value is an indicator to do
a register read, so remove it since now au8522_regread() inserts
the bit automatically.

Also remove a stray instance where we were actually trying to write
to the I2C status register, which was actually a read.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The leading bit in register values is actually an indicator as to
whether to perform a read or write, so remove the bit from the
register values, since the au8522_writereg() is now responsible
for adding this bit automatically.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

I got this working a couple of years ago.  Remove it from the
list of known issues.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The original code was based on my reverse engineering of an I2C trace
of the Windows driver.  Now that I know what the registers actually do,
restructure the code a bit, removing some unneeded register programming
and fixing the sequencing of operations.

This reduces the time it takes to change inputs from 1300ms down to
600ms (as measured by "time v4l2-ctl -i 0")

Note this does not address outstanding issues related to the management
of the module clocks and power control for the various blocks, which
will be done in a separate patch.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Some stray lines got inserted into the driver when I reverse engineered
the I2C traffic (at the time I didn't know what the registers did).

It turns up these registers muck with the onboard I2C master, which
we don't use since we instead use the I2C gate.  Remove the lines
which can actually interfere with the operation of the bus.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Since we don't suppoort sliced VBI with the au0828/au8522, there is
no need to configure the au8522 VBI slicer, which because of the
coefficients requires a large amount of i2c traffic.

Remove the relevant code.  Note that this has no effect on raw VBI
support, which is currently the only supported way to access VBI on
this device.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Fixed spelling mistake of "successfully"
Signed-off-by: NRene Hickersberger <renehickersberger@gmx.net>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Minor fixes in comments
Signed-off-by: NBenjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Hardware could support up to 16 logical addresses which is more
than needed by CEC specifications.
Let use CEC_MAX_LOG_ADDRS instead of limited it on one.
stih_cec_adap_log_addr() function was alredy written to support
multiple addresses requests.
Signed-off-by: NBenjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Remove variable self assignment.

Addresses-Coverity-ID: 1408817
Signed-off-by: NGustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Add the missing endianness conversions to a debug call printing the
USB device-descriptor idVendor and idProduct fields during probe.
Signed-off-by: NJohan Hovold <johan@kernel.org>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The VIDIOC_G/S_EDID ioctls can return valid data even if an error is returned.

Mark those ioctls accordingly. Rather than using an explicit 'if' to check for the
ioctl (as was done until now for VIDIOC_QUERY_DV_TIMINGS) just set a new flag in the
v4l2_ioctls array.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Acked-by: NSakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

In 'commit 295ab497 ("[media] media: platform: pxa_camera: make
printk consistent")' a pointer to the device structure in
mclk_get_divisor() was changed to pcdev_to_dev(pcdev). The pointer used
by pcdev_to_dev() is still uninitialized during the call to
mclk_get_divisor() as it happens in v4l2_device_register() at the end
of the probe. The dev_warn and dev_dbg caused a line in the log:

	(NULL device *): Limiting master clock to 26000000

Fix this by using an initialized pointer from the platform_device
(as before the old patch).
Signed-off-by: NPetr Cvek <petr.cvek@tul.cz>
Acked-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The v4l2-compliance complains about nonexistent vidioc_subscribe_event
and vidioc_unsubscribe_event calls. Add them to fix the complaints.
Signed-off-by: NPetr Cvek <petr.cvek@tul.cz>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

During the transfer from the soc_camera a test in pxa_mbus_image_size()
got removed. Without it any PXA_MBUS_LAYOUT_PACKED format causes either
the return of a wrong value (PXA_MBUS_PACKING_2X8_PADHI doubles
the correct value) or EINVAL (PXA_MBUS_PACKING_NONE and
PXA_MBUS_PACKING_EXTEND16). This was observed in an error from the ffmpeg
(for some of the YUYV subvariants).

This patch re-adds the same test as in soc_camera version.
Signed-off-by: NPetr Cvek <petr.cvek@tul.cz>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

This patch adds Bayer 8 GBRG and RGGB support and move GRBG definition
close to BGGR (so all Bayer 8 variants are together). No other changes are
needed as the driver handles them as RAW data stream.

The RGGB variant was tested in a modified OV9640 driver.
Signed-off-by: NPetr Cvek <petr.cvek@tul.cz>
Acked-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Permits distinguishing between two /dev/videoX entries from the same
physical UVC device (that naturally share the same iProduct name).

This change matches current Windows behavior by prioritizing iFunction
over iInterface, but unlike Windows it displays both iProduct and
iFunction/iInterface strings when both are available.
Signed-off-by: NPeter Boström <pbos@google.com>
Reviewed-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Add the probe def quirk for the webcam found in the Apple MacBook Pro
2016, to get it working out of the box.
Signed-off-by: NDaniel Roschka <danielroschka@phoenitydawn.de>
Signed-off-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

- avoid duplicate debugging messages in em28xx_read_reg_req_len()
- do not describe successful usb transfers in em28xx_read_reg_len()
  as "failed"
- report errors in em28xx_write_regs_req(), too
- print the usb error numbers, too
Signed-off-by: NFrank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

trivial fix to spelling mistake and add in a white space in
a CX18_ERR error message
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

In functions cx25840_initialize(), cx231xx_initialize(), and
cx23885_initialize(), the return value of create_singlethread_workqueue()
is used without validation. This may result in NULL dereference and cause
kernel crash. This patch fixes it.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Function pci_find_ext_capability() may return 0, which is an invalid
address. In function cobalt_pcie_status_show(), its return value is used
without validation. This patch adds checks to validate the return
address.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Function kthread_create() returns an ERR_PTR on error. However, in
function s5k83a_start(), its return value is used without validation.
This may result in a bad memory access bug. This patch fixes the bug.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The recent changes in 4.9 to mandate USB buffers be heap allocated
broke this driver, which was allocating the buffers on the stack.
This resulted in the device failing at initialization.

Introduce dedicated send/receive buffers as part of the state
structure, and add a mutex to protect access to them.

Note: we also had to tweak the API to mxl111sf_ctrl_msg to pass
the pointer to the state struct rather than the device, since
we need it inside the function to access the buffers and the
mutex.  This patch adjusts the callers to match the API change.
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Reported-by: NDoug Lung <dlung0@gmail.com>
Cc: Michael Ira Krufky <mkrufky@linuxtv.org>
Cc: <stable@vger.kernel.org>      # for v4.5 and up
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Sometimes 'DMA single access' is not enough to transfer
a frame of image, '8-beat burst access' is set as the
default DMA memory burst size.
Signed-off-by: NSongjun Wu <songjun.wu@microchip.com>
Acked-by: NNicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Setting initial standard at the top of cx8800_initdev would cause the
first call to cx88_set_tvnorm() to return without programming any
registers (leaving the driver saying it's set to NTSC but the hardware
isn't programmed).  Even worse, any subsequent attempt to explicitly
set it to NTSC-M will return success but actually fail to program the
underlying registers unless first changing the standard to something
other than NTSC-M.

Set the initial standard later in the process, and make sure the field
is zero at the beginning to ensure that the call always goes through.

This regression was introduced in the following commit:

commit ccd6f1d4 ("[media] cx88: move width, height and field to core
struct")

Author: Hans Verkuil <hans.verkuil@cisco.com>

[media] cx88: move width, height and field to core struct
Signed-off-by: NDevin Heitmueller <dheitmueller@kernellabs.com>
Cc: <stable@vger.kernel.org>      # for v3.19 and up
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

This V4L2 subdev driver enables Digital Camera Memory Interface (DCMI)
of STMicroelectronics STM32 SoC series.
Signed-off-by: NYannick Fertre <yannick.fertre@st.com>
Signed-off-by: NHugues Fruchet <hugues.fruchet@st.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

This adds documentation of device tree bindings for the STM32 DCMI
(Digital Camera Memory Interface).
Acked-by: NRob Herring <robh@kernel.org>
Signed-off-by: NHugues Fruchet <hugues.fruchet@st.com>
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Without annotations, gcc 7.1 will complain.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

On this driver, it can fall through a switch. I tried to
annotate it, in order to shut up a gcc warning, but that
didn't work, as the logic there is somewhat complex.

So, instead, let's just repeat the code. gcc should likely
optimize it anyway, and this makes the code better readable,
IMHO.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Clearly, hsync and vsinc bool vars are part of the return
logic on the second case of the switch. Annotate that, in
order to shut up gcc warnings.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>