Remove set_crop soc-camera device method and switch to s_crop from v4l2-subdev
video operations. Also extend non-i2c drivers to also hold a pointer to their
v4l2-subdev instance in control device driver-data, i.e., in
dev_get_drvdata((struct device *)to_soc_camera_control(icd))
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Switch to using struct v4l2_rect in struct soc_camera_device for uniformity and
simplicity.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Convert the soc-camera framework to use the v4l2-(sub)dev API. Start using
v4l2-subdev operations. Only a part of the interface between the
soc_camera core, soc_camera host drivers on one side and soc_camera device
drivers on the other side is replaced so far. The rest of the interface
will be replaced in incremental steps, and will require extensions and,
possibly, modifications to the v4l2-subdev code.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Convert soc-camera core and all drivers to platform device API. We already
converted platforms to register a platform device for each soc-camera client,
now we remove the compatibility code and switch completely to the new scheme.
This is a preparatory step for the v4l2-subdev conversion.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

The .buf_queue() V4L2 driver method is called under
spinlock_irqsave(q->irqlock,...), don't take the lock again inside the
function.
Reported-by: NAntonio Ospite <ospite@studenti.unina.it>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

mclk_get_divisor uses pcdev->soc_host.dev, make sure it is initialized.
Signed-off-by: NAntonio Ospite <ospite@studenti.unina.it>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

The v4l function has a better algorithm for aligning image size.

For instance the old code would change 159x243 into 156x240 to meet the
alignment requirements.  The new function will use 160x243, which is a lot
closer to what was asked for originally.

Cc: Robert Jarzmik <robert.jarzmik@free.fr>
Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NTrent Piepho <xyzzy@speakeasy.org>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Add missing __devexit_p() to several drivers. Also add a few missing
__init, __devinit and __exit markers. These errors could result in
build failures depending on the kernel configuration.
Signed-off-by: NJean Delvare <khali@linux-fr.org>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Make camera devices direct children of host platform devices, move the
inheritance management into the soc_camera.c core driver.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Embed struct soc_camera_host in platform-specific per host instance objects
instead of allocating them statically in drivers, use platform_[gs]et_drvdata
consistently, use resource_size().
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

pxa_dma_update_sg_tail() is called only once, runs exactly the same loop as the
caller and has to recalculate the last element in an sg-list, that the caller
has already calculated. Eliminate redundancy by merging the two loops and
re-using the calculated pointer. This also saves a bit of performance which is
always good during video-capture.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Acked-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

The last buffer queued will often overrun, as the DMA chain
is finished, and the time the dma irq handler is activated,
the QCI fifos are filled by the sensor.

The fix is to ignore the overrun condition on the last
queued buffer, and restart the capture only on intermediate
buffers of the chain.

Moreover, a fix was added to the very unlikely condition
where in YUV422P mode, one channel overruns while another
completes at the very same time. The capture is restarted
after the overrun as before, but the other channel
completion is now ignored.
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

The DMA transfers in pxa_camera showed some weaknesses in
multiple queued buffers context :
 - poll/select problem
   The bug shows up with capture_example tool from v4l2 hg
   tree. The process just "stalls" on a "select timeout".

 - multiple buffers DMA starting
   When multiple buffers were queued, the DMA channels were
   always started right away. This is not optimal, as a
   special case appears when the first EOF was not yet
   reached, and the DMA channels were prematurely started.

 - Maintainability
   DMA code was a bit obfuscated. Rationalize the code to be
   easily maintainable by anyone.

 - DMA hot chaining
   DMA is not stopped anymore to queue a buffer, the buffer
   is queued with DMA running. As a tribute, a corner case
   exists where chaining happens while DMA finishes the
   chain, and the capture is restarted to deal with the
   missed link buffer.

This patch attemps to address these issues / improvements.

 create mode 100644 Documentation/video4linux/pxa_camera.txt
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

All planes were PAGE aligned (ie. 4096 bytes aligned). This
is not consistent with YUV422 format, which requires Y, U
and V planes glued together.  The new implementation forces
the alignement on 8 bytes (DMA requirement), which is almost
always the case (granted by width x height being a multiple
of 8).

The test cases include tests in both YUV422 and RGB565 :
 - a picture of size 111 x 111 (cross RAM pages example)
 - a picture of size 1023 x 4 in (under 1 RAM page)
 - a picture of size 1024 x 4 in (exactly 1 RAM page)
 - a picture of size 1025 x 4 in (over 1 RAM page)
 - a picture of size 1280 x 1024 (many RAM pages)
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Due to DMA constraints, the DMA chain always transfers bytes
from the QCI fifos to memory in 8 bytes units. In planar
formats, that could mean 0 padding between Y and U plane
(and between U and V plane), which is against YUV422P
standard.

Therefore, a frame size is required to be a multiple of 16
(so U plane size is a multiple of 8). It is enforced in
try_fmt() and set_fmt() primitives, be aligning height then
width on 4 multiples as need be, to reach a 16 multiple.
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

As host and camera drivers become more complex, differences between S_FMT and
S_CROP functionality grow, this patch separates them.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

All soc-camera camera and host drivers must specify supported data signal
polarity, after all drivers are fixed, we'll add a suitable test to
soc_camera_bus_param_compatible().
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

1. Driver code where pxa_request_dma() is called will most likely
   reference DMA registers as well,  and it is really unnecessary
   to include pxa-regs.h just for this. Move the definitions into
   <mach/dma.h> and make relevant drivers include it instead of
   <mach/pxa-regs.h>.

2. Introduce DMAC_REGS_VIRT as the virtual address base for these
   DMA registers. This allows later processors to re-use the same
   IP while registers may start at different I/O address.
Signed-off-by: NEric Miao <eric.miao@marvell.com>

Recent format-negotiation patches caused S_CROP breakage in pxa_camera.c
and sh_mobile_ceu_camera.c drivers, fix it.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

1. now pxa_camera.c uses ioremap() for register access, pxa_camera.h is
   totally useless. Remove it.

2. <asm/dma.h> does no longer include <mach/dma.h>, include the latter
   file explicitly

 delete mode 100644 drivers/media/video/pxa_camera.h
Signed-off-by: NEric Miao <eric.miao@marvell.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

sh_mobile_ceu_camera.c is already prepared to support interlaced format, this
patch moves the choice of a field type down to host and / or camera drivers.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Move mutex from host drivers to camera device object, take into account
videobuf locking.
Signed-off-by: NGuennadi Liakhovetski <lg@denx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

With the introduction of the format conversion support in soc-camera, we now
also have to take care to pass the correct pixel format to the camera driver
when calling its try_fmt() method.
Signed-off-by: NGuennadi Liakhovetski <lg@denx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Simplify multiple drivers by replacing f->fmt.pix.* with a single pointer
dereference, merge some needlessly broken lines, verify host and camera
operations pointers on registration.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Using PXA270's FIFO inactivity time-out register (CITOR) reduces FIFO overruns.
The time-out is calculated in CICLK / LCDCLK ticks and has to be longer than
one pixel time. For this we have to know the pixel clock frequency, which
usually is provided by the camera. We use the struct soc_camera_sense to
request PCLK frequency from the camera driver upon each data format change.
Tested-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Signed-off-by: NEric Miao <eric.miao@marvell.com>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Use the new format-negotiation infrastructure, support all four YUV422
packed and the planar formats.

The new translation structure enables to build the format
list with buswidth, depth, host format and camera format
checked, so that it's not done anymore on try_fmt nor
set_fmt.
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

We anyway don't follow the s_fmt_vid_cap / g_fmt_vid_cap / try_fmt_vid_cap
naming, and soc-camera is so far only about video capture, let's simplify
operation names a bit further. set_fmt_cap / try_fmt_cap wasn't a very good
choice too.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Pixel format requested by the user is not necessarily the same, as what
a sensor driver provides. There are situations, when a camera host driver
provides the required format, but requires a different format from the
sensor. Further, the list of formats, supported by sensors is pretty static
and can be pretty good described with a constant list of structures. Whereas
decisions, made by camera host drivers to support requested formats can be
quite complex, therefore it is better to let the host driver do the work.
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

.try_bus_param() method from struct soc_camera_host_ops is only called at one
location immediately before .try_fmt_cap(), there is no value in keeping these
two methods separate, merge them.
Signed-off-by: NGuennadi Liakhovetski <lg@denx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Signed-off-by: NEric Miao <eric.miao@marvell.com>
Acked-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>

Where devices only have one consumer, passing a consumer clock ID
has no real benefit.  Remove it.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Unsigned dma_chans[] cannot be negative

Also the third time dma_chans[0] < 0 was tested instead of dma_chans[2]
Signed-off-by: NRoel Kluin <roel.kluin@gmail.com>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Make .power and .reset callbacks per camera instead of per host, also move
their invocation to camera drivers.

.arch/arm/mach-pxa/include/mach/camera.h    |    2 -
Signed-off-by: NStefan Herbrechtsmeier <hbmeier@hni.uni-paderborn.de>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

1. DRCMRxx is no longer recommended, use DRCMR(xx) instead, and
   pass DRCMR index by "struct resource" if possible

2. DCSRxx, DDADRxx, DSADRxx, DTADRxx, DCMDxx is never used, use
   DCSR(), DDADR(), DSADR(), DTADR(), DCMD() instead
Signed-off-by: NEric Miao <eric.miao@marvell.com>
Acked-by: NNicolas Pitre <nico@cam.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

This just leaves include/asm-arm/plat-* to deal with.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

PXA suspend switches off DMA core, which loses all context
of previously assigned descriptors. As pxa_camera driver
relies on DMA transfers, setup the lost descriptors on
resume and retrigger frame acquisition if needed.
Signed-off-by: NRobert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@infradead.org>

So, platform_driver_unregister() doesn't actually have a return value,
nor do any of the void __exit routines. It's reassuring to know that
people copy and paste blindly. This completely blew up my compiler.
Signed-off-by: NPaul Mundt <lethal@linux-sh.org>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@infradead.org>

This patch moves the spinlock handling from soc_camera.c to the actual
camera host driver. The spinlock_alloc/free callbacks are replaced with
code in init_videobuf(). So far all camera host drivers implement their
own spinlock_alloc/free methods anyway, and videobuf_queue_core_init()
BUGs on a NULL spinlock argument, so, new camera host drivers will not
forget to provide a spinlock when initialising their videobuf queues.
Signed-off-by: NMagnus Damm <damm@igel.co.jp>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@pengutronix.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@infradead.org>

Makes SoC camera videobuf independent. Includes all necessary changes for
PXA camera driver (currently the only driver using soc_camera in the mainline).
These changes are important for the future soc_camera based drivers.
Signed-off-by: NPaulius Zaleckas <paulius.zaleckas@teltonika.lt>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@pengutronix.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@infradead.org>