This patch transforms drivers/staging/dt3155v4l driver to use videobuf2 framework.
Tested and works with "xawtv -f".
Either streaming API or read method should be selected during kernel configuration.
If both are selected into the driver (not possible without another patching),
either due to my misunderstanding or problems in xawtv (or both), I get kernel panic
after some start/stop of xawtv (not strictly reproducible).
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
videobuf_waiton() must unlock and relock ext_lock if it has to wait.
For that to happen it needs the videobuf_queue pointer.
Don't attempt to unlock/relock q->ext_lock unless it was locked in the
first place.
vb->state has to be protected by a spinlock to be safe.
This patch is based on code from Mauro Carvalho Chehab <mchehab@redhat.com>.
[mchehab@redhat.com: add extra argument to a few missing places]
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
The third argument of dma_alloc_coherent() is of type dma_addr_t and is defined
as u64 for x86_64, while the member 'device_base' of struct dma_coherent_mem is
defined as u32.
While at it, clean up the code. Now if dev->dma_mem != NULL all the fields/members
are already properly initialized.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch removes the private memory allocator
and all conjugated functions. Now the driver uses
videobuf-dma-contig module for allocations.
Added reserving 4MB coherent memory for private
per device allocations at probe time.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch introduces a v4l2 state machine, so now
CONFIG_DT3155_STREAMING is no more needed.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The previous patch "use_local_copy_qtype_ops.patch"
http://lkml.org/lkml/2010/5/9/40
has introduced a new BUG.
This patch corrects it.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
dt3155v4l driver, as in -rc6-next-20100506 has a BUG.
When it modifies q->int_ops structure in videobuf-dma-contig
module the change is visible for all other modules using it.
Make a local copy of this structure and use its modification
to solve the bug.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
dt3155v4l driver, as in -rc6-next-20100506 compiles, but will
not run properly due to recent changes in the videobuf APIs.
This patch synchronizes some functions that have been copied
from drivers/media/video/videobuf-dma-contig.c (and modified)
with the recent (-rc6 -> -rc6-next-) changes in videobuf layer
especially drivers/media/video/videobuf-dma-contig.c
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The prototypes of udelay() and msleep() are in linux/delay.h,
so include it.
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Kernel module (device driver) for dt3155 frame grabber
video4linux2 compliant (finally). Works with "xawtv -f".
======================================================
This driver is written (almost) from scratch, using the
allocator developed for dt3155pci see bellow). The driver
uses videobuf-dma-contig interface modified to use the above
mentioned allocator instead of dma_alloc_coheren().
The first thing to do was to design a new allocator based
on allocating a configurable number of 4MB chunks of memory,
that latter are broken into frame buffers of 768x576 bytes
kept in different FIFOs (queues). As far as the driver autoloads
as a kernel module during kernel boot, the allocation of 4MB
chunks succeeds.
The driver keeps three FIFOs: one for 4MB chunks, one for free
buffers (available for allocations) and one for buffers already
allocated. Allocation/deallocation is done automatically though
the video4linux videobuf subsystem (some pointers to functions
are replaced by driver supplied functions).
Sure, there are problems:
1. The device tested to work with "xawtv -f" either via read()
method (DT3155_STREAMING not selected), or via mmap() method
(DT3155_STREAMING is selected) only. This coresponds to either
cap->capabilities = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_READWRITE;
or
cap->capabilities = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING;
but not when
cap->capabilities = V4L2_CAP_VIDEO_CAPTURE |
V4L2_CAP_STREAMING |
V4L2_CAP_READWRITE;
This is because xawtv calls poll() before starting streaming,
but videobuf_poll_stream() automatically starts reading if streaming
is not started.
This selection is made during kernel configuration (for now).
2. Works for CCIR, but should work for RS-170 (not tested)
This is made also during kernel configuration.
3. Could work for multiple dt3155 frame grabbers in a PC,
(private data is allocated during PCI probe() method), but
is not tested due to lack of a second board.
4. Not tested on a BIG ENDIAN architecture.
5. Many others you could find .... :-)
All critics, comments, suggestions are wellcome.
Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Andrew Morton