When we moved to 3.0, we found that the cx18 driver was oopsing on close with:
NULL pointer deref at:
[ 2290.461009] Call Trace:
[ 2290.461009] [<c046007b>] ? pm_qos_add_request+0xc/0x6e
[ 2290.461009] [<c082631c>] __mutex_lock_common+0x87/0x125
[ 2290.461009] [<f8970e92>] ? cx18_queue_flush+0x31/0x87 [cx18]
[ 2290.461009] [<c0436b85>] ? __might_sleep+0x29/0xe4
[ 2290.461009] [<c0826515>] __mutex_lock_slowpath+0x25/0x27
[ 2290.461009] [<c08264b2>] ? mutex_lock+0x2e/0x3b
[ 2290.461009] [<c08264b2>] mutex_lock+0x2e/0x3b
[ 2290.461009] [<f88d3137>] videobuf_queue_lock+0x13/0x15 [videobuf_core]
[ 2290.461009] [<f88d3f86>] __videobuf_free+0xfc/0x112 [videobuf_core]
[ 2290.461009] [<f89741e6>] cx18_v4l2_close+0x158/0x172 [cx18]
[ 2290.461009] [<c0507522>] ? cpumask_next+0x1a/0x1d
[ 2290.461009] [<f88a319d>] v4l2_release+0x35/0x52 [videodev]
[ 2290.461009] [<c04f5717>] fput+0x100/0x1a5
[ 2290.461009] [<c04f2e09>] filp_close+0x5c/0x64
[ 2290.461009] [<c04f2e70>] sys_close+0x5f/0x93
[ 2290.461009] [<c082cd5f>] sysenter_do_call+0x12/0x28
Some digging showed that a merge at some previous point partially
added broken mmap() support, causing this trace. Remove the broken
code completely.
On top of that, the calculation in place for "buffer full" depended on
UYUV instead of HM12, while our GStreamer code was picking HM12 in
some circumstances.
Finally, the V4L2_CAP_STREAMING capability was never exposed. Patch it
into the YUV encoder node only.
Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The initialisation of vb_type in serialized_open was preventing
REQBUFS from working reliably. Remove it, and move the spinlock into
stream_init for good measure - it's only used when we have a stream
that supports videobuf anyway.
Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The initial version of this patch (commit
d5976931639176bb6777755d96b9f8d959f79e9e) had some issues:
* It didn't correctly calculate the size of the YUV buffer for 4:2:2,
resulting in capture sometimes being offset by 1/3rd of a picture.
* There were a lot of variables duplicating information the driver
already knew, which have been removed.
* There was an in-kernel format conversion - libv4l can do this one,
and is the right place to do format conversions anyway.
* Some magic numbers weren't properly explained.
Fix all these issues, leaving just the move from videobuf to videobuf2
to do.
Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Acked-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Add support for mmap method streaming of raw YUV video on cx18-based
hardware, in addition to the existing support for read() streaming of
raw YUV and MPEG-2 encoded video.
[simon.farnsworth@onelan.co.uk: I forward-ported this from Steven's original work,
done under contract to ONELAN. The original code is at
http://www.kernellabs.com/hg/~stoth/cx18-videobuf]
Signed-off-by: Steven Toth <stoth@kernellabs.com>
Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
If allocating stream buffers for one of the primary streams (e.g. YUV)
failed during card probe, the error path clean up would try to manipulate
the yet unitialized IDX stream structures. This caused a BUG due to
unitialized list heads.
Detect that case and ignore the uninitialized IDX stream.
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Rather than guess which driver supports core priority handling, require drivers
that do to explicitly set the V4L2_FL_USE_FH_PRIO flag in video_device.
Updated the core prio handling accordingly and set the flag in the three
drivers that do.
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The cx18_stream struct contained a struct cx18_dvb for
every stream object, most of which were for analog capture.
Now we only allocate the cx18_dvb object for the DTV TS stream.
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Also added locking around snd_cx18_pcm_ioctl as a precaution
as requested by Andy Walls.
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
A trivial change to update my email address from my dead awalls@radix.net
address to my current awalls@md.metrocast.net address.
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Export a couple of more symbols required by the cx18-alsa module.
This work was sponsored by ONELAN Limited.
Signed-off-by: Devin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The CX18_CPU_SET_INDEXTABLE command was being called with the wrong number
of arguments causing the index table frame type selection mask to be set wrong.
Now the IDX stream properly sends entries for I, P, and B frames.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
According to the v4l2 spec, very old MPEG index entries needs to be discarded
in favor of newer index entries. This change ensures the firmware always has
buffers for index entries at the expense of the oldest unread buffers.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
This change starts the IDX stream along with the MPG stream as an internal
use (only) stream much like the VBI stream can be started as an internal use
stream for inserting sliced VBI packets.
The IDX stream is not started automatically with an MPEG strem if the IDX
stream is disabled (no buffers allocated) or if sliced VBI insertion is being
performed by the cx18 driver. The cx18 driver doing sliced VBI insertion
makes the offsets in the IDX stream inaccurate for the final MPEG stream
presented to user space. Since fixing the IDX offsets ourselves is not easy
and we cannot easily do what ivtv does to fix the offsets, we'll make sliced
VBI insertion and MPEG Index capture mutually exclusive for now.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
This change allows the IDX stream to be started and stopped as any other
stream even though it has no associated device node. This is needed for
cx18 driver internal use.
Also always tell the CX23418 to generate index entries when an analog capture
starts and the IDX stream has had buffers allocated (i.e. is enabled).
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Fix a long standing memory leak of stream buffers for streams that did not
have a struct video_device allocated: namely the TS and IDX streams.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Fix all device drivers to use the new video_device_node_name function.
This also strips kernel log messages from the "/dev/" prefix, has the device
node location is a userspace policy decision unknown to the kernel.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Fix YUV capture such that the encoder will pass one frame per transfer. This
will allow the application to maintain frame alignment when a transfer from
the encoder is missed due to high system latency in service the CX23418 IRQ.
Also force YUV buffer sizes to be specified in multiples of 33.75 kB, the
smalled amount of buffer sizes need to store a complete set of HM12 4:2:0
macroblocks specifying 32 lines of the frame. A full 60Hz/525 line
screen requires 15 * 33.75 kB per frame and a full 50Hz/625 line screen
requires 18 * 33.75 kB per frame so the default buffer size is 3 * 33.75 kB,
requiring exactly 5 or 6 buffers per MDL respectively. The bytes needed per
frame and hence MDL need not be the bytes in an integer number of buffers.
However, if frame artifacts are seen with scaled screen sizes, the YUV buffer
size can be set 34 kB (33.75 kB) to get rid of the artifacts at the cost of more
copies between the kernel and userspace.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Add a Memory Descriptor List (MDL) layer to buffer handling to implement
scatter-gather I/O. Currently there is still only 1 buffer per MDL.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Warn when the desired device node number is already in use, except when
the new video_register_device_no_warn function is called since in some
use-cases that warning is not relevant.
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Finish changes for sliced and raw VBI for 625 line systems. Tested with VPS
and WSS being emitted by a PVR-350 in field 1 lines 16 and 23.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Initial changes to get sliced VBI for 625 line system working. This is patch
is untested.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Simplify the way outgoing work handler gets scheduled to send empty buffers
back to the firmware for use. Also reduced the memory required for scheduling
this outgoing work, by using a single, per stream work object.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
To avoid sleeps in providing buffers to user space and in handling incoming
buffers from the capture unit, converted the per stream mutex for locking
queues to 3 spin locks. There is now a spin lock per queue
to increase concurrency when moving buffers around.
Also simplified queue manipulations and buffer handling of incoming buffers
of data from the capture unit.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
This change defers sending all CX18_CPU_DE_SET_MDL commands, for a stream with
an ongoing capture, by adding a work queue to handle sending such commands when
needed. This prevents any sleeps, caused by notifying the firmware of new
usable buffers, when a V4L2 application read() is being satisfied or when
an incoming buffer is processed by the cx18-NN-in work queue thread.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Convert GPIO line functions, such a audio routing and device resets, to
v4l2_subdevices. This essentially completes the conversion of cx18 to the
v4l2_device/v4l2_subdevice framework. No regression testing has taken place as
of yet. Also an ivtv legacy bug with GPIO mux routing and going to/from radio
mode was commented, but not fixed.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Added a new chip identifer to v4l2-chip-ident for the integrated A/V broadcast
decoder core internal to the CX23418. Completed separation and encapsulation
of the A/V decoder core interface as a v4l2_subdevice. The cx18 driver now
compiles and links again.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Convert card instance handling to a lighter weight mechanism like ivtv.
Also convert name strings and debug messages to use v4l2_device.name.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Sliced VBI, in the manner that ivtv implements it as a separate data stream,
now works for 525 line 60 Hz systems like NTSC-M. It may work for 625 line
50 Hz systems, but I have more engineering work to do, to verify it is operating
properly. Sliced data insertion into the MPEG PS should be working, but is
untested.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The driver was incorrectly setting 0 line counts in a firmware
API call to set the maximum amount of lines per field.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The Encoder will only allow the Raw VBI parameters, along with a number of
other API parameters, to take effect when no analog captures are in
progress. These parameters must be set before the first analog capture starts,
be it MPEG, VBI, YUV, etc., and cannot be changed until the last one stops. It
is not obvious to me what capture channel API parameters are shared and which
ones must be set per capture channel, so set them all for every analog
capture channel start up. This fixes the driver so that VBI capture can be
started up after the MPEG capture is going.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Renamed structure member name to be more specific to type in anticipation
of updating to the v4l2_device/v4l2_subdev framework. Too many objects named
"dev" and /v4l2_\{0,1\}dev/ would be to confusing.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Introduce a struct v4l2_file_operations for v4l2 drivers.
Remove the unnecessary inode argument.
Move compat32 handling (and llseek) into the v4l2-dev core: this is now
handled in the v4l2 core and no longer in the drivers themselves.
Note that this changeset reverts an earlier patch that changed the return
type of__video_ioctl2 from int to long. This change will be reinstated
later in a much improved version.
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
cx18: Use a known open task handle when setting stream CX2341x parameters
Sometimes, we might only have VBI or YUV stream open instead of an MPEG stream.
Let's make sure we use a valid task handle to perform the CX2341x control
settings.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Removed magic number that referred to firmware limit on the number of
MDLs the firmware can maintain for any stream at any one time.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
A combined authorship patch from Hans Verkuil and Andy Walls. Raw
VBI can now be captured but requires a video capture to be in progress as well.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
This is a port of the fixes Hans Verkuil made for ivtv/cx25840:
The service_set field was used to determine whether raw or sliced VBI was
desired. This is incorrect since it is perfectly valid to select sliced VBI
with a service_set of 0.
Instead the driver should check on VIDIOC_S_FMT whether the type
field matches the raw or sliced VBI type.
Updated the cx18 driver accordingly, including an additional check in
cx18_start_v4l2_encode_stream() that didn't exist in ivtv.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
cx18: Avoid making firmware API calls with the queue lock held. The source
of MPEG strem corruption when not holding the queue lock was found to be that
the MPEG buffer could be retrieved by the user app before it was sync'ed for
the host cpu. Incoming buffers are now sync'ed before being put on q_full and
releasing the queue lock. We can thus avoid the sometimes lengthy call to
the firmware for CPU_DE_SET_MDL while holding the queue lock, so we can get
better performance.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
cx18: Add module parameters for finer control over buffer allocations.
User now has the option of setting smaller buffers to get lower
latency transfers from the encoder. User can also now set the number
of buffers used for a stream explicitly.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
cx18: Allow more than 63 capture buffers in rotation per stream. Implement
q_busy to hold buffers the firmware has for use. q_free holds truly unused
buffers in a pool. New buffers are given to the firmware as soon as the
firmware returns one, if there are any to give to the firmware.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Eliminate q_io from stream buffer handling in anticipation of upcoming
changes in buffer handling. q_io was a holdover from ivtv and it's function
in cx18 was trivial and not necessary. We just push things back onto the
front of q_full now, instead of maintaining a 1 buffer q_io queue.
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Add copyright attribution for files modified by awalls in 2008
Signed-off-by: Andy Walls <awalls@radix.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
