[media] drxk: remove the option to load firmware asynchronously

The option to load firmware asynchronously were added due to
a requirement with a few versions of udev. It turns that this was
a bad idea and caused regressions on drxk-based devices.

So, we end by only letting the firmware to be loaded syncronously
everywhere.

So, let's remove the bad code.

This patch partially reverts the changeset 8e30783b0b3.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: adjust period size at runtime

While the current hardcoded period is ok for the current values,
we may latter change the driver to work with different bit rates
or with different latencies than 64ms.

So, adust the period size at runtime.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: push mutex down to extensions on .fini callback

Avoid circular mutex lock by pushing the dev->lock to the .fini
callback on each extension.

As em28xx-dvb, em28xx-alsa and em28xx-rc have their own data
structures, and don't touch at the common structure during .fini,
only em28xx-v4l needs to be locked.

[   90.994317] ======================================================
[   90.994356] [ INFO: possible circular locking dependency detected ]
[   90.994395] 3.13.0-rc1+ #24 Not tainted
[   90.994427] -------------------------------------------------------
[   90.994458] khubd/54 is trying to acquire lock:
[   90.994490]  (&card->controls_rwsem){++++.+}, at: [<ffffffffa0177b08>] snd_ctl_dev_free+0x28/0x60 [snd]
[   90.994656]
[   90.994656] but task is already holding lock:
[   90.994688]  (&dev->lock){+.+.+.}, at: [<ffffffffa040db81>] em28xx_close_extension+0x31/0x90 [em28xx]
[   90.994843]
[   90.994843] which lock already depends on the new lock.
[   90.994843]
[   90.994874]
[   90.994874] the existing dependency chain (in reverse order) is:
[   90.994905]
-> #1 (&dev->lock){+.+.+.}:
[   90.995057]        [<ffffffff810b8fa3>] __lock_acquire+0xb43/0x1330
[   90.995121]        [<ffffffff810b9f82>] lock_acquire+0xa2/0x120
[   90.995182]        [<ffffffff816a5b6c>] mutex_lock_nested+0x5c/0x3c0
[   90.995245]        [<ffffffffa0422cca>] em28xx_vol_put_mute+0x1ba/0x1d0 [em28xx_alsa]
[   90.995309]        [<ffffffffa017813d>] snd_ctl_elem_write+0xfd/0x140 [snd]
[   90.995376]        [<ffffffffa01791c2>] snd_ctl_ioctl+0xe2/0x810 [snd]
[   90.995442]        [<ffffffff811db8b0>] do_vfs_ioctl+0x300/0x520
[   90.995504]        [<ffffffff811dbb51>] SyS_ioctl+0x81/0xa0
[   90.995568]        [<ffffffff816b1929>] system_call_fastpath+0x16/0x1b
[   90.995630]
-> #0 (&card->controls_rwsem){++++.+}:
[   90.995780]        [<ffffffff810b7a47>] check_prevs_add+0x947/0x950
[   90.995841]        [<ffffffff810b8fa3>] __lock_acquire+0xb43/0x1330
[   90.995901]        [<ffffffff810b9f82>] lock_acquire+0xa2/0x120
[   90.995962]        [<ffffffff816a762b>] down_write+0x3b/0xa0
[   90.996022]        [<ffffffffa0177b08>] snd_ctl_dev_free+0x28/0x60 [snd]
[   90.996088]        [<ffffffffa017a255>] snd_device_free+0x65/0x140 [snd]
[   90.996154]        [<ffffffffa017a751>] snd_device_free_all+0x61/0xa0 [snd]
[   90.996219]        [<ffffffffa0173af4>] snd_card_do_free+0x14/0x130 [snd]
[   90.996283]        [<ffffffffa0173f14>] snd_card_free+0x84/0x90 [snd]
[   90.996349]        [<ffffffffa0423397>] em28xx_audio_fini+0x97/0xb0 [em28xx_alsa]
[   90.996411]        [<ffffffffa040dba6>] em28xx_close_extension+0x56/0x90 [em28xx]
[   90.996475]        [<ffffffffa040f639>] em28xx_usb_disconnect+0x79/0x90 [em28xx]
[   90.996539]        [<ffffffff814a06e7>] usb_unbind_interface+0x67/0x1d0
[   90.996620]        [<ffffffff8142920f>] __device_release_driver+0x7f/0xf0
[   90.996682]        [<ffffffff814292a5>] device_release_driver+0x25/0x40
[   90.996742]        [<ffffffff81428b0c>] bus_remove_device+0x11c/0x1a0
[   90.996801]        [<ffffffff81425536>] device_del+0x136/0x1d0
[   90.996863]        [<ffffffff8149e0c0>] usb_disable_device+0xb0/0x290
[   90.996923]        [<ffffffff814930c5>] usb_disconnect+0xb5/0x1d0
[   90.996984]        [<ffffffff81495ab6>] hub_port_connect_change+0xd6/0xad0
[   90.997044]        [<ffffffff814967c3>] hub_events+0x313/0x9b0
[   90.997105]        [<ffffffff81496e95>] hub_thread+0x35/0x170
[   90.997165]        [<ffffffff8108ea2f>] kthread+0xff/0x120
[   90.997226]        [<ffffffff816b187c>] ret_from_fork+0x7c/0xb0
[   90.997287]
[   90.997287] other info that might help us debug this:
[   90.997287]
[   90.997318]  Possible unsafe locking scenario:
[   90.997318]
[   90.997348]        CPU0                    CPU1
[   90.997378]        ----                    ----
[   90.997408]   lock(&dev->lock);
[   90.997497]                                lock(&card->controls_rwsem);
[   90.997607]                                lock(&dev->lock);
[   90.997697]   lock(&card->controls_rwsem);
[   90.997786]
[   90.997786]  *** DEADLOCK ***
[   90.997786]
[   90.997817] 5 locks held by khubd/54:
[   90.997847]  #0:  (&__lockdep_no_validate__){......}, at: [<ffffffff81496564>] hub_events+0xb4/0x9b0
[   90.998025]  #1:  (&__lockdep_no_validate__){......}, at: [<ffffffff81493076>] usb_disconnect+0x66/0x1d0
[   90.998204]  #2:  (&__lockdep_no_validate__){......}, at: [<ffffffff8142929d>] device_release_driver+0x1d/0x40
[   90.998383]  #3:  (em28xx_devlist_mutex){+.+.+.}, at: [<ffffffffa040db77>] em28xx_close_extension+0x27/0x90 [em28xx]
[   90.998567]  #4:  (&dev->lock){+.+.+.}, at: [<ffffffffa040db81>] em28xx_close_extension+0x31/0x90 [em28xx]

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: Fix usb diconnect logic

Now that everything is extension, the usb disconnect logic should
be the same.

While here, fix the device name.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: print a message at disconnect

That helps to identify if something fails and explain why em28xx
struct is not freed (if it ever happens).

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: disconnect before freeing URBs

URBs might be in usage. Disconnect the device before freeing
them.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: simplify error handling

Cleanup the error handling code at em28xx-audio init.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: fix return code on device disconnect

Alsa has an special non-negative return code to indicate device removal
at snd_em28xx_capture_pointer(). Use it, instead of an error code.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-v4l: fix the freeing of the video devices memory

Remove some dead code from em28xx_v4l2_fini() and fix the leaking of the video,
vbi and radio video_device struct memories.

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: always call em28xx_release_resources() in the usb disconnect handler

When the usb device is disconnected, the resources are no longer available,
so there is no reason to keep them registered.

This will also fix the various sysfs group removal warnings which we can see
since kernel 3.13.

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-v4l: move v4l2_ctrl_handler freeing and v4l2_device unregistration to em28xx_v4l2_fini

v4l2_ctrl_handler_free() and v4l2_device_unregister() are currently only called
when the last user closes the device and the device is already disconnected.
But that's wrong, we need to call these functions whenever the em28xx-v4l
extension is closed and we can already do this if the device is still opened
by some users.

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: move v4l2 dummy clock deregistration from the core to the v4l extension

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: move v4l2_device_disconnect() call from the core to the v4l extension

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: move usb transfer uninit on device disconnect from the core to the v4l-extension

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: move usb buffer pre-allocation and transfer uninit from the core to the dvb extension

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-v4l: fix device initialization in em28xx_v4l2_open() for radio and VBI mode

- bail out on unsupported VFL_TYPE
- em28xx_set_mode() needs to be called for VBI and radio mode, too
- em28xx_wake_i2c() needs to be called for VBI and radio mode, too
- em28xx_resolution_set() also needs to be called for VBI

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] tea575x: Fix build with ARCH=c6x

In file included from /devel/v4l/temp/include/asm-generic/page.h:23:0,
                 from /devel/v4l/temp/arch/c6x/include/asm/page.h:9,
                 from /devel/v4l/temp/include/asm-generic/io.h:14,
                 from arch/c6x/include/generated/asm/io.h:1,
                 from /devel/v4l/temp/drivers/media/radio/tea575x.c:23:
/devel/v4l/temp/arch/c6x/include/asm/setup.h:17:27: error: unknown type name ‘phys_addr_t’
 extern int c6x_add_memory(phys_addr_t start, unsigned long size);

It seems that, on such arch, the includes from asm/ should be
after the ones from linux/.

The proper fix would be to patch the arch files, but, as
this fix is trivial, apply it. Also, we generally put the
asm includes after the linux ones, anyway.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] lirc_parallel: avoid name conflict on mn10300 arch

The "irq_handler" name is already defined there on a header
file:

/devel/v4l/temp/drivers/staging/media/lirc/lirc_parallel.c:223:13: error: conflicting types for ‘irq_handler’
 static void irq_handler(void *blah)
             ^
In file included from /devel/v4l/temp/arch/mn10300/include/asm/reset-regs.h:16:0,
                 from /devel/v4l/temp/arch/mn10300/include/asm/irq.h:18,
                 from /devel/v4l/temp/include/linux/irq.h:24,
                 from /devel/v4l/temp/arch/mn10300/include/asm/hardirq.h:16,
                 from /devel/v4l/temp/include/linux/preempt_mask.h:5,
                 from /devel/v4l/temp/include/linux/sched.h:25,
                 from /devel/v4l/temp/include/linux/utsname.h:5,
                 from /devel/v4l/temp/arch/mn10300/include/asm/elf.h:15,
                 from /devel/v4l/temp/include/linux/elf.h:4,
                 from /devel/v4l/temp/include/linux/module.h:14,
                 from /devel/v4l/temp/drivers/staging/media/lirc/lirc_parallel.c:29:
/devel/v4l/temp/arch/mn10300/include/asm/exceptions.h:107:24: note: previous declaration of ‘irq_handler’ was here
 extern asmlinkage void irq_handler(void);

/devel/v4l/patchwork/drivers/staging/media/lirc/lirc_serial.c:653:20: error: conflicting types for ‘irq_handler’
 static irqreturn_t irq_handler(int i, void *blah)
                    ^
In file included from /devel/v4l/patchwork/arch/mn10300/include/asm/reset-regs.h:16:0,
                 from /devel/v4l/patchwork/arch/mn10300/include/asm/irq.h:18,
                 from /devel/v4l/patchwork/include/linux/irq.h:24,
                 from /devel/v4l/patchwork/arch/mn10300/include/asm/hardirq.h:16,
                 from /devel/v4l/patchwork/include/linux/preempt_mask.h:5,
                 from /devel/v4l/patchwork/include/linux/sched.h:25,
                 from /devel/v4l/patchwork/include/linux/utsname.h:5,
                 from /devel/v4l/patchwork/arch/mn10300/include/asm/elf.h:15,
                 from /devel/v4l/patchwork/include/linux/elf.h:4,
                 from /devel/v4l/patchwork/include/linux/module.h:14,
                 from /devel/v4l/patchwork/drivers/staging/media/lirc/lirc_serial.c:53:
/devel/v4l/patchwork/arch/mn10300/include/asm/exceptions.h:107:24: note: previous declaration of ‘irq_handler’ was here
 extern asmlinkage void irq_handler(void);

So, rename it, to avoid namespace conflicts.

This patch fixes building media drivers with allyesconfig/almodconfig on
mn10300 arch.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] go7007-usb: only use go->dev after allocated

Fixes those warnings:
drivers/staging/media/go7007/go7007-usb.c: In function 'go7007_usb_probe':
drivers/staging/media/go7007/go7007-usb.c:1060: warning: 'go' is used uninitialized in this function

While here, comment a code that will never run.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] dib8000: Fix a few warnings when compiled for avr32

drivers/media/dvb-frontends/dib8000.c: In function 'dib8000_get_time_us':
drivers/media/dvb-frontends/dib8000.c:3957: warning: 'interleaving' may be used uninitialized in this function
drivers/media/dvb-frontends/dib8000.c:3956: warning: 'rate_denum' may be used uninitialized in this function

Those are actually false positives, but it doesn't hurt cleaning them.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] dib8000: Properly represent long long integers

When compiling with avr32, it gets those errors:

drivers/media/dvb-frontends/dib8000.c: In function 'dib8000_get_stats':
drivers/media/dvb-frontends/dib8000.c:4121: warning: integer constant is too large for 'long' type

Fix integer representation to avoid overflow.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] radio-usb-si4713: make si4713_register_i2c_adapter static

This function isn't used nowhere outside the same .c file.
Fixes this warning:

drivers/media/radio/si4713/radio-usb-si4713.c:418:5: warning: no previous prototype for 'si4713_register_i2c_adapter' [-Wmissing-prototypes]
 int si4713_register_i2c_adapter(struct si4713_usb_device *radio)
     ^

Cc: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] sh_vou: comment unused vars

Fix two warns below, by commenting the unused code:

drivers/media/platform/sh_vou.c: In function 'sh_vou_configure_geometry':
drivers/media/platform/sh_vou.c:446:49: warning: variable 'height_max' set but not used [-Wunused-but-set-variable]
  unsigned int black_left, black_top, width_max, height_max,
                                                 ^
drivers/media/platform/sh_vou.c: In function 'sh_vou_isr':
drivers/media/platform/sh_vou.c:1056:13: warning: variable 'side' set but not used [-Wunused-but-set-variable]
  static int side;
             ^

Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] solo6x10: fix broken PAL support

The video_type was never set correctly for PAL: it's not a bool, instead
it is a register value.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Tested-by: Hans Verkuil <hans.verkuil@cisco.com>
Reported-by: tomdev@freenet.de
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] au0828: Add option to preallocate digital transfer buffers

Added command line parameter preallocate_big_buffers so that the digital
transfer buffers can be allocated when the driver is registered. They
do not have to be allocated every time a feed is started.

Signed-off-by: Tim Mester <tmester@ieee.org>
Acked-by: Devin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] au8028: Fix cleanup on kzalloc fail

Free what was allocated if there is a failure allocating
transfer buffers.

Stop the feed on a start feed error.  The stop feed is not always called
if start feed fails.  If the feed is not stopped on error, then the driver
will be stuck so that it can never start feeding again.

[m.chehab@samsung.com: CodingStyle cleanup]
Signed-off-by: Tim Mester <tmester@ieee.org>
Acked-by: Devin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] s5p-mfc: Add controls to set vp8 enc profile

Add v4l2 controls to set desired profile for VP8 encoder.
Acceptable levels for VP8 encoder are
0: Version 0
1: Version 1
2: Version 2
3: Version 3

Signed-off-by: Kiran AVND <avnd.kiran@samsung.com>
Signed-off-by: Pawel Osciak <posciak@chromium.org>
Signed-off-by: Arun Kumar K <arun.kk@samsung.com>
Signed-off-by: Kamil Debski <k.debski@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] media: s5p_mfc: remove s5p_mfc_get_node_type() function

s5p_mfc_get_node_type() relies on get_index() helper function, which in
turn relies on video_device index numbers assigned on driver
registration. All this code is not really needed, because there is
already access to respective video_device structures via common
s5p_mfc_dev structure. This fixes the issues introduced by patch
1056e4388b0454917a512618c8416a98628fc9ce ("v4l2-dev: Fix race condition
on __video_register_device"), which has been merged in v3.12-rc1.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: stable@vger.kernel.org
Signed-off-by: Kamil Debski <k.debski@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] s5p-mfc: Add QP setting support for vp8 encoder

Adds v4l2 controls to set MIN, MAX QP values and
I, P frame QP for vp8 encoder.

Signed-off-by: Kiran AVND <avnd.kiran@samsung.com>
Signed-off-by: Arun Kumar K <arun.kk@samsung.com>
Signed-off-by: Kamil Debski <k.debski@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: fix xc3028 demod and firmware setup on DVB

Now that em28xx can be compiled without V4L support, we should
call em28xx_setup_xc3028() on both em28xx-v4l and em28xx-dvb
modules.

Reported-by: Chris Lee <updatelee@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] nxt200x: increase write buffer size

The buffer size on nxt200x is not enough:

...
> Dec 20 10:52:04 rich kernel: [   31.747949] nxt200x: nxt200x_writebytes: i2c wr reg=002c: len=255 is too big!
...

Increase it to 256 bytes.

Reported-by: Rich Freeman <rich0@gentoo.org>
Cc: stable@vger.kernel.org
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: return -ENODEV when the device is disconnected

If em28xx is disconnected, return -ENODEV to all PCM callbacks.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: split URB initialization code

The URB calculus code may eventually be moved to some other
place, like at pcm open, if it ends by needing more setups, like
working with different bit rates, or different audio latency.

So, move it into a separate routine. That also makes the code
more readable.

No functional changes.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: don't wait for lock in non-block mode

Pulseaudio has the bad habit of stopping a streaming audio if
a device, opened in non-block mode, waits.

It is impossible to avoid em28xx to wait, as it will send commands
via I2C, and other I2C operations may be happening (firmware
transfers, Remote Controller polling, etc). Yet, as each em28xx
subdriver locks em28xx-dev to protect the access to the hardware,
it is possible to minimize the audio glitches by returning -EAGAIN
to pulseaudio, if the lock is already taken by another subdriver.

Reported-by: Antti Palosaari <crope@iki.fi>
Tested-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: fix the period size in bytes

If the period size is wrong, userspace will assume a wrong delay
any may negociate an inadequate value.

The em28xx devices use 8 for URB interval, in microframes,
and the driver programs it to have 64 packets.

That means that the IRQ sampling period is 125 * 8 * 64,
with is equal to 64 ms.

So, that's the minimal latency with the current settings. It is
possible to program a lower latency, by using less than 64 packets,
but that increases the amount of bandwitdh used, and the number of
IRQ events per second.

In any case, in order to support it, the driver logic should be
changed to fill those parameters in realtime.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: don't hardcode audio URB calculus

The current code hardcodes the number of audio URBs, the number
of packets per URB and the maximum URB size.

This is not a good idea, as it:

- wastes more bandwidth than necessary, by using a very
  large number of packets;

- those constants are bound to an specific scenario, with
  a bandwidth of 48 kHz;

- don't take the maximum endpoint size into account;

- with urb->interval = 1 on xHCI, those constraints cause a "funny"
  setup: URBs with 64 packets inside, with only 24 bytes total. E. g.
  a complete waste of space.

Change the code to do dynamic URB audio calculus and allocation.

For now, use the same constraints as used before this patch, to
avoid regressions.

A good scenario (tested) seems to use those defines, instead:

#define EM28XX_MAX_AUDIO_BUFS          8
#define EM28XX_MIN_AUDIO_PACKETS       2

But let's not do such change here, letting the optimization to
happen on latter patches, after more tests.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: Fix error path

De-allocate memory and free sound if an error happens.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: use bInterval on em28xx-audio

Just filling urb->interval with 1 is wrong, and causes a different
behaviour with xHCI.

With EHCI, the URB size is typically 192 bytes. However, as
xHCI specifies intervals in microframes, the URB size becomes
too short (24 bytes).

With this patch, the interval will be properly initialized, and
the device will behave the same if connected into a xHCI or an
EHCI device port.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-i2c: add timeout debug information if i2c_debug enabled

If i2c_debug is enabled, we splicitly want to know when a
device fails with timeout.

If i2c_debug==2, this is already provided, for each I2C transfer
that fails.

However, most of the time, we don't need to go that far. We just
want to know that I2C transfers fail.

So, add such errors for normal (ret == 0x10) I2C aborted timeouts.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-i2c: cleanup I2C debug messages

The I2C output messages is too polluted. Clean it a little
bit, by:
- use the proper core support for memory dumps;
- hide most stuff under the i2c_debug umbrella;
- add the missing KERN_CONT where needed;
- use 2 levels or verbosity. Only the second one
  will show the I2C transfer data.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-i2c: Fix error code for I2C error transfers

Follow the error codes for I2C as described at Documentation/i2c/fault-codes.

In the case of the I2C status register (0x05), this is mapped into:

- ENXIO - when reg 05 returns 0x10
- ETIMEDOUT - when the device is not temporarily not responding
      (e. g. reg 05 returning something not 0x10 or 0x00)
- EIO - for generic I/O errors that don't fit into the above.

In the specific case of 0-byte reads, used only during I2C device
probing, it keeps returning -ENODEV.

TODO: return EBUSY when reg 05 returns 0x20 on em2874 and upper.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: use a better value for I2C timeouts

In the lack of a better spec, let's assume the timeout
values compatible with SMBus spec:
http://smbus.org/specs/smbus110.pdf

at chapter 8 - Electrical Characteristics of SMBus devices

Ok, SMBus is a subset of I2C, and not all devices will be
following it, but the timeout value before this patch was not
even following the spec.

So, while we don't have a better guess for it, use 35 + 1
ms as the timeout.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: rename I2C timeout to EM28XX_I2C_XFER_TIMEOUT

This macro is used by all em28xx devices, and not just em2800.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: convert i2c wait completion logic to use jiffies

The I2C wait completion/timeout logic currently assumes that
msleep(5) will wait exaclty 5 ms. This is not true at all,
as it depends on CONFIG_HZ.

Convert it to use jiffies, in order to not wait for more time
than needed.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: properly implement AC97 wait code

Instead of assuming that msleep() is precise, use a jiffies
based code to wait for AC97 to be available.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] tuner-xc2028: Don't read status if device is powered down

That removes those timeout errors:

[ 3675.930940] xc2028 19-0061: Device is Xceive 3028 version 1.0, firmware version 2.7
[ 3676.060487] xc2028 19-0061: divisor= 00 00 8d d0 (freq=567.250)
[ 3676.349449] xc2028 19-0061: Putting xc2028/3028 into poweroff mode.
[ 3698.247645] xc2028 19-0061: xc2028_get_reg 0002 called
[ 3698.253276] em2860 #0: I2C transfer timeout on writing to addr 0xc2
[ 3698.253301] xc2028 19-0061: i2c input error: rc = -121 (should be 2)
[ 3698.253327] xc2028 19-0061: xc2028_signal called
[ 3698.253339] xc2028 19-0061: xc2028_get_reg 0002 called
[ 3698.259283] em2860 #0: I2C transfer timeout on writing to addr 0xc2
[ 3698.259312] xc2028 19-0061: i2c input error: rc = -121 (should be 2)

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] tuner-xc2028: Don't try to sleep twice

Only send a power down command for the device if it is not already
in power down state. That prevents a timeout when trying to talk
with the device.

Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx-audio: allocate URBs at device driver init

Instead of allocating/deallocating URBs and transfer buffers
every time stream is started/stopped, just do it once.

That reduces the memory allocation pressure and makes the
code that start/stop streaming a way simpler.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: use usb_alloc_coherent() for audio

Instead of allocating transfer buffers with kmalloc() use
usb_alloc_coherent().

This patch should make it work also with ARM CPUs.

Reviewed-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] cx231xx: add support for a CX23103 Video Grabber USB

Add a new USB ID to the driver.

Signed-off-by: Links (Markus) <help.markus+git@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: Fix comment in __qbuf_dmabuf

The comment incorrectly explains that the code verifies information
provided by userspace, while verification has been performed earlier in
reality. Fix it.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] m88rs2000: Correct m88rs2000_get_fec

Value of fec is achieved by the upper nibble bits 6,7 & 8.

Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] m88rs2000: Correct m88rs2000_set_fec settings

Register 0x70 is used to set fec, register 0x76 is used to get fec

Register 0x76 is set to 0x8.

Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] m88rs2000: correct read status lock value

The correct lock values is when bits of the value 0xee are set.

Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] m88rs2000: set symbol rate accurately

Current setting of symbol rate is not very actuate causing
loss of lock.

Covert temp to u64 and use mclk to calculate from big number.

Calculate symbol rate by dividing symbol rate by 1000 times
1 << 24 and dividing sum by mclk.

Add other symbol rate settings to function registers 0xa0-0xa3.

In set_frontend add changes to register 0xf1 this must be done
prior call to fe_reset. Register 0x00 doesn't need a second
write of 0x1

Applied after patch
m88rs2000: add m88rs2000_set_carrieroffset

Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>
Cc: stable@vger.kernel.org # v3.9+
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] m88rs2000: add m88rs2000_set_carrieroffset

Set the carrier offset correctly using the default mclk values.

Add function m88rs2000_get_mclk to calculate the mclk value
against crystal frequency which will later be used for
other functions.

Add function m88rs2000_set_carrieroffset to calculate
and set the offset value.

variable offset becomes a signed value.

Register 0x86 is set the appropriate value according to
remainder value of frequency % 192857 calculation as
shown.

Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>
Cc: stable@vger.kernel.org # v3.9+
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] e4000: fix error return code

Set the return variable to an error code as done elsewhere in the function.

A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
(
if@p1 (\(ret < 0\|ret != 0\))
 { ... return ret; }
|
ret@p1 = 0
)
... when != ret = e1
    when != &ret
*if(...)
{
  ... when != ret = e2
      when forall
 return ret;
}

// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] ec168: fix error return code

The rest of the function uses ret to store the return value, even setting
ret to i a few lines before this, so return ret instead of i.

A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
(
if@p1 (\(ret < 0\|ret != 0\))
 { ... return ret; }
|
ret@p1 = 0
)
... when != ret = e1
    when != &ret
*if(...)
{
  ... when != ret = e2
      when forall
 return ret;
}

// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] as102: fix leaks at failure paths in as102_usb_probe()

Failure handling is incomplete in as102_usb_probe().
The patch implements proper resource deallocations.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] [for,v3.14] sn9c102: fix build dependency

This driver should only build if MEDIA_USB_SUPPORT is set.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reported-by: Jim Davis <jim.epost@gmail.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] s5k5baf: Fix potential NULL pointer dereferencing

Dereference 'fw' after the NULL check.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] s5k5baf: Fix checkpatch error

Fixes the following error:
ERROR: return is not a function, parentheses are not required
FILE: drivers/media/i2c/s5k5baf.c:1353:

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] s5k5baf: Fix build warning

Fixes the following warnings:
drivers/media/i2c/s5k5baf.c: In function 's5k5baf_fw_parse':
drivers/media/i2c/s5k5baf.c:362:3: warning:
format '%d' expects argument of type 'int', but argument 3 has type 'size_t' [-Wformat=]
drivers/media/i2c/s5k5baf.c:383:4: warning:
format '%d' expects argument of type 'int', but argument 4 has type 'size_t' [-Wformat=]

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] cx231xx: Add missing KERN_CONT to i2c debug prints

Fix continuation lines.

[m.chehab@samsung.com: This was actually part of a v2 patch meant to
 fix i2c debug prints. As version 1 was already applied, I'm applying
 here the diff and fixing the patch subject/description]

Signed-off-by: Matthias Schwarzott <zzam@gentoo.org>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] em28xx: fix I2S audio sample rate definitions and info output

The audio configuration in chip config register 0x00 and eeprom are always
consistent. But currently the audio configuration #defines for the chip config
register say 0x20 means 3 sample rates and 0x30 5 sample rates, while the eeprom
info output says 0x20 means 1 sample rate and 0x30 3 sample rates.

I've checked the datasheet excerpts I have and it seems that the meaning of
these bits is different for em2820/40 (1 and 3 sample rates) and em2860+
(3 and 5 smaple rates).
I have also checked my Hauppauge WinTV USB 2 (em2840) and the chip/eeprom
audio config 0x20 matches the sample rates reproted by the USB device
descriptor (32k only).

Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] media: soc_camera: rcar_vin: Add preliminary R-Car M2 support

This adds R-Car M2 (R8A7791) VIN support. Both H2 and M2
variants look the same from the driver's point of view,
so use GEN2 id for both.

Signed-off-by: Valentine Barshak <valentine.barshak@cogentembedded.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
[g.liakhovetski@gmx.de: removed changelog from commit message]
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] omap3isp: ccdc: Don't hang when the SBL fails to become idle

Under abnormal conditions (such as glitches on the HSYNC/VSYNC signals)
the CCDC output SBL can fail to become idle. The driver currently logs
this condition to the kernel log and doesn't restart the CCDC. This
results in CCDC video capture hanging without any notification to
userspace.

Cancel the pipeline and mark the CCDC as crashed instead of hanging.
Userspace will be notified of the problem and will then be able to close
and reopen the device to trigger a reset of the ISP.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] omap3isp: Refactor modules stop failure handling

Modules failing to stop are fatal errors for the preview engine only.
Flag that condition separately from the other stop failures to prepare
support for more fatal errors.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] omap3isp: Cancel streaming when a fatal error occurs

When a fatal error that prevents any further video streaming occurs in a
pipeline, all queued buffers must be marked as erroneous and new buffers
must be prevented from being queued. Implement this behaviour with a new
omap3isp_pipeline_cancel_stream() function that can be used by
submodules to cancel streaming.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] export em28xx_release_resources() symbol

As reported by the kbuild test robot:

All error/warnings:

>> ERROR: "em28xx_release_resources" [drivers/media/usb/em28xx/em28xx-v4l.ko] undefined!

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] davinci-vpfe: fix compile error

The include for delay.h was missing.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Cc: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] videodev2: Set vb2_rect's width and height as unsigned

As discussed on the media summit 2013, there is no reason for the width
and height to be signed.

Therefore this patch is an attempt to convert those fields from __s32 to
__u32.

Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Acked-by: Sakari Ailus <sakari.ailus@iki.fi> (documentation and smiapp)
Acked-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] media: Remove OOM message after input_allocate_device

Emitting an OOM message isn't necessary after input_allocate_device
as there's a generic OOM and a dump_stack already done.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] fix coccinelle warnings

drivers/staging/media/bcm2048/radio-bcm2048.c:2255:3-4: Unneeded semicolon

 Removes unneeded semicolon.

Generated by: coccinelle/misc/semicolon.cocci

CC: Hans Verkuil <hans.verkuil@cisco.com>
CC: Mauro Carvalho Chehab <m.chehab@samsung.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] fix coccinelle warnings

drivers/staging/media/bcm2048/radio-bcm2048.c:2632:1-7: Replace memcpy with struct assignment
/c/kernel-tests/src/cocci/drivers/staging/media/bcm2048/radio-bcm2048.c:744:1-7: Replace memcpy with struct assignment
/c/kernel-tests/src/cocci/drivers/staging/media/bcm2048/radio-bcm2048.c:2360:3-9: Replace memcpy with struct assignment

Generated by: coccinelle/misc/memcpy-assign.cocci

CC: Hans Verkuil <hans.verkuil@cisco.com>
CC: Mauro Carvalho Chehab <m.chehab@samsung.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: don't set vfd->debug

You can set this through sysfs, so don't mix the two.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6588: add support for non-blocking mode

saa6588 always blocked while waiting for data, even if the filehandle
was in non-blocking mode.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6588: remove unused CMD_OPEN

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6588: after calling CMD_CLOSE, CMD_POLL is broken

CMD_CLOSE sets data_available_for_read to 1, which is necessary to do the
wakeup call, but it is never reset to 0.

Because of this calling CMD_POLL afterwards will always return that data is
available, even if there isn't any.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: drop log_status for radio

There are no controls for the radio node, so just drop support for this ioctl.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6752hs.h: drop empty header

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6752hs: move to media/i2c

This driver is independent from saa7134, so there is no reason why this
shouldn't be in media/i2c like all other i2c media drivers.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa6752hs: drop compat control code

The saa7134 driver is now converted to the control framework, so drop the
control compat code in saa6752hs.c.

Also add 'const' to several static arrays.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: use V4L2_IN_ST_NO_SIGNAL instead of NO_SYNC

NO_SYNC was meant for DVB and shouldn't be used anymore.

In this case NO_SIGNAL is a good alternative.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: add support for control events

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: share resource management between normal and empress nodes

The empress video node can share resource management with the normal
video nodes, thus allowing for code sharing and making the empress node
non-exclusive.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: remove dev from saa7134_fh, use saa7134_fh for empress node

Use the saa7134_fh struct for the empress video node as well, drop the dev
pointer from that struct since we can use drvdata for that.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: cleanup radio/video/empress ioctl handling

The video and empress nodes can share various ioctls.

Drop the input/std ioctls from the radio node (out of spec).

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: convert to the control framework

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] saa7134: move the queue data from saa7134_fh to saa7134_dev

These fields are global, not per-filehandle.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] DocBook: drop the word 'only'

There are already video output drivers that allow STREAMON without
any buffers queued, and with the change in vb2 there are now more
drivers like that.

So saying "The ioctl will succeed only when at least one output
buffer is in the incoming queue." isn't true. Just drop the word
'only'. We cannot say that it will also work if no output buffers are
queued as long as not all drivers are converted to vb2.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: Improve file I/O emulation to handle buffers in any order

videobuf2 file I/O emulation assumed that buffers dequeued from the
driver would return in the order they were enqueued in the driver.

Improve the file I/O emulator's book-keeping to remove this assumption.

Also set the buf->size properly if a write() dequeues a buffer and the
VB2_FILEIO_WRITE_IMMEDIATELY flag is set.

Based on an initial patch by Andy Walls.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: return ENOBUFS in start_streaming in case of too few buffers

This works together with the retry_start_streaming mechanism to allow userspace
to start streaming even if not all required buffers have been queued.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Cc: Tomasz Stanislawski <t.stanislaws@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Acked-by: Kamil Debski <k.debski@samsung.com>
Acked-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: don't set index, don't start streaming for write()

Two fixes:

- there is no need to set the index when calling dqbuf: dqbuf will
  overwrite it.
- __vb2_init_fileio already starts streaming for write(), so there is
  no need to do it again in __vb2_perform_fileio. It can never have
  worked anyway: either __vb2_init_fileio succeeds in starting streaming
  or it is never going to happen.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: retry start_streaming in case of insufficient buffers

If start_streaming returns -ENOBUFS, then it will be retried the next time
a buffer is queued. This means applications no longer need to know how many
buffers need to be queued before STREAMON can be called. This is particularly
useful for output stream I/O.

If a DMA engine needs at least X buffers before it can start streaming, then
for applications to get a buffer out as soon as possible they need to know
the minimum number of buffers to queue before STREAMON can be called. You can't
just try STREAMON after every buffer since on failure STREAMON will dequeue
all your buffers. (Is that a bug or a feature? Frankly, I'm not sure).

This patch simplifies applications substantially: they can just call STREAMON
at the beginning and then start queuing buffers and the DMA engine will
kick in automagically once enough buffers are available.

This also fixes using write() to stream video: the fileio implementation
calls streamon without having any queued buffers, which will fail today for
any driver that requires a minimum number of buffers.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: remove the 'fileio = NULL' hack

The read/write implementation in vb2 reuses existing vb2 functions, but
it sets q->fileio to NULL before calling them in order to skip the
'q->fileio != NULL' check.

This works today due to the synchronous nature of read/write, but it
1) is ugly, and 2) will fail in an asynchronous use-case such as a
thread queuing and dequeuing buffers. This last example will be necessary
in order to implement vb2 DVB support.

This patch removes the hack by splitting up the dqbuf/qbuf/streamon/streamoff
functions into an external and an internal version. The external version
checks q->fileio and then calls the internal version. The read/write
implementation now just uses the internal version, removing the need to
set q->fileio to NULL.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: fix race condition between REQBUFS and QBUF/PREPARE_BUF

When preparing a buffer the queue lock is released for a short while
if the memory mode is USERPTR (see __buf_prepare for the details), which
would allow a race with a REQBUFS which can free the buffers. Removing the
buffers from underneath __buf_prepare is obviously a bad idea, so we
check if any of the buffers is in the state PREPARING, and if so we
just return -EAGAIN.

If this happens, then the application does something really strange. The
REQBUFS call can be retried safely, since this situation is transient.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: simplify qbuf/prepare_buf by removing callback

The callback used to merge the common code of the qbuf/prepare_buf
code can be removed now that the mmap_sem handling is pushed down to
__buf_prepare(). This makes the code more readable.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] videobuf2: Fix CodingStyle

Changeset b18a8ff29d80 added a comment violating the 80cols max size,
with no good reason.

Fix it.

Cc: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>

[media] vb2: push the mmap semaphore down to __buf_prepare()

Rather than taking the mmap semaphore at a relatively high-level function,
push it down to the place where it is really needed.

It was placed in vb2_queue_or_prepare_buf() to prevent racing with other
vb2 calls. The only way I can see that a race can happen is when two
threads queue the same buffer. The solution for that it to introduce
a PREPARING state.

Moving it down offers opportunities to simplify the code.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>