Smatch produces two warnings when building this file:
	./arch/x86/include/asm/bitops.h:433:22: warning: asm output is not an lvalue
	./arch/x86/include/asm/bitops.h:433:22: warning: asm output is not an lvalue

On some asm instructions.

I suspect that those asm instructions might not be producing the
right code, so, better to use two intermediate vars, get rid of
the warnings and of the risk of producing a wrong code.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Now that af9015 requires I2C_MUX, all drivers that select
it should also depend on it.

   drivers/media/dvb-frontends/af9013.o: In function `af9013_remove':
>> drivers/media/dvb-frontends/af9013.c:1560: undefined reference to `i2c_mux_del_adapters'
   drivers/media/dvb-frontends/af9013.o: In function `af9013_probe':
>> drivers/media/dvb-frontends/af9013.c:1488: undefined reference to `i2c_mux_alloc'
>> drivers/media/dvb-frontends/af9013.c:1495: undefined reference to `i2c_mux_add_adapter'
   drivers/media/dvb-frontends/af9013.c:1544: undefined reference to `i2c_mux_del_adapters'
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

While the logic there is correct, it tricks both humans and
machines, a the check if "i" var is not zero is actually to
validate if the "frames" var was initialized when the loop
ran for the first time.

That produces the following warning:
	drivers/media/usb/uvc/uvc_v4l2.c:1192 uvc_ioctl_enum_framesizes() error: potentially dereferencing uninitialized 'frame'.

Change the logic to do the right test instead.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

If the adapter doesn't have error_inj_parse_line() ops, the
write() logic won't return -EINVAL, but, instead, it will keep
looping, because "count" is a non-negative number.
Reviewed-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

There's a false positive warning there:
	drivers/media/i2c/tda9840.c:79 tda9840_status() error: uninitialized symbol 'byte'.

Change the code to match our coding style, in order to fix it.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Instead of casting, just use %p.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The current I2C error handling logic makes static analyzers
confused:

	drivers/media/usb/em28xx/em28xx-input.c:96 em28xx_get_key_terratec() error: uninitialized symbol 'b'.

Change it to match the coding style we're using elsewhere.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Instead of casting, just use %p.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The logic at vivid_radio_rx_g_tuner() is producint an overflow
warning:

	drivers/media/platform/vivid/vivid-radio-rx.c:250 vivid_radio_rx_g_tuner() warn: potential negative subtraction from max '65535 - (__builtin_choose_expr( ==  ||  == , , __builtin_choose_expr( ==  ||  == , , __builtin_choose_expr( ==  ||  == , , __builtin_choose_expr( ==  ||  == , , __builtin_choose_expr( ==  ||  == , , __builtin_choose_expr( == , , (0))))))) * 65535) / delta'

Add a cast to prevent that.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Change the logic there to avoid casting, solving this warning:
	drivers/media/pci/saa7134/saa7134-alsa.c:276 saa7134_alsa_dma_init() warn: argument 3 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Instead of using an ancillary function to avoid duplicating
a small portion of code that copies data either to kernelspace
or between userspace-kernelspace, duplicate the code,
as it prevents static analyzers to complain about it:

	drivers/media/pci/solo6x10/solo6x10-g723.c:260:46: warning: cast removes address space of expression

The hole idea of using __user is to make sure that the code is
doing the right thing with address space, so there's no
sense on use casting.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/pci/zoran/zoran_driver.c:242 v4l_fbuffer_alloc() warn: argument 5 to %lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

While the code is correct, it produces this warning:
	drivers/media/i2c/ir-kbd-i2c.c:593 zilog_ir_format() error: buffer overflow 'code_block->codes' 61 <= 173

As static analyzers may be tricked by arithmetic expressions on
comparisions. So, change the order, in order to shut up this
false-positive warning.

That also makes easier for humans to understand that it won't
be trying to go past buffer size.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The current I2C error handling logic makes static analyzers
confused, and it doesn't follow the coding style we're using:

	drivers/media/i2c/ir-kbd-i2c.c:180 get_key_pixelview() error: uninitialized symbol 'b'.
	drivers/media/i2c/ir-kbd-i2c.c:224 get_key_knc1() error: uninitialized symbol 'b'.
	drivers/media/i2c/ir-kbd-i2c.c:226 get_key_knc1() error: uninitialized symbol 'b'.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Currently, the code produces those false-positives:
	drivers/media/pci/saa7134/saa7134-input.c:203 get_key_msi_tvanywhere_plus() error: uninitialized symbol 'b'.
	drivers/media/pci/saa7134/saa7134-input.c:251 get_key_kworld_pc150u() error: uninitialized symbol 'b'.
	drivers/media/pci/saa7134/saa7134-input.c:275 get_key_purpletv() error: uninitialized symbol 'b'.

Improve the error handling code, making it to look like our
coding style.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/usb/s2255/s2255drv.c:651 s2255_fillbuff() warn: argument 3 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/pci/ivtv/ivtvfb.c:349 ivtvfb_prep_frame() warn: argument 3 to %08lx specifier is cast from pointer
drivers/media/pci/ivtv/ivtvfb.c:360 ivtvfb_prep_frame() warn: argument 3 to %08lx specifier is cast from pointer
drivers/media/pci/ivtv/ivtvfb.c:363 ivtvfb_prep_frame() warn: argument 4 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Just use %p. Fixes this warning:
	drivers/media/v4l2-core/videobuf-dma-sg.c:247 videobuf_dma_init_kernel() warn: argument 2 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/platform/soc_camera/soc_camera.c:790 soc_camera_mmap() warn: argument 4 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/pci/cx23885/cx23885-alsa.c:92 cx23885_alsa_dma_init() warn: argument 3 to %08lx specifier is cast from pointer
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/pci/cx88/cx88-alsa.c:295 cx88_alsa_dma_init() warn: argument 3 to %08lx specifier is cast from pointer
drivers/media/pci/cx88/cx88-alsa.c:669 snd_cx88_wm8775_volume_put() warn: potential negative subtraction from max '65535 - (32768 * left) / right'
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The logic with gets a p2m_id is more complex than needed,
causing false positives with static analyzers:

	drivers/media/pci/solo6x10/solo6x10-p2m.c:81 solo_p2m_dma_desc() error: buffer overflow 'solo_dev->p2m_dev' 4 <= s32max

Make it simpler and use unsigned int.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The error handling logic at get_key_pv951() is a little bit
akward, with produces this false positive warning:

	drivers/media/pci/bt8xx/bttv-input.c:344 get_key_pv951() error: uninitialized symbol 'b'.

Do a cleanup. As a side effect, it also improves its coding
style.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The error handling logic at tvaudio is broken on several ways,
as it doesn't really check right when an error occurs.

Change it to return the proper error code from read/write
routines and fix the errors on reads.

Shuts up the following warnings:
	drivers/media/i2c/tvaudio.c:222 chip_read() error: uninitialized symbol 'buffer'.
	drivers/media/i2c/tvaudio.c:223 chip_read() error: uninitialized symbol 'buffer'.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

drivers/media/dvb-frontends/sp887x.c:179 sp887x_initial_setup() error: memcpy() '&buf[2]' too small (30 vs 16384)

This is actually a false alarm, but reverting the check order
makes not only for humans to review the code, but also cleans
the warning.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

While the code there is right, it produces three false positives:
	drivers/media/v4l2-core/v4l2-ioctl.c:2868 video_usercopy() error: copy_from_user() 'parg' too small (128 vs 16383)
	drivers/media/v4l2-core/v4l2-ioctl.c:2868 video_usercopy() error: copy_from_user() 'parg' too small (128 vs 16383)
	drivers/media/v4l2-core/v4l2-ioctl.c:2876 video_usercopy() error: memset() 'parg' too small (128 vs 16383)

Store the ioctl size on a cache var, in order to suppress those.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Fix the following warnings:
	drivers/media/common/v4l2-tpg/v4l2-tpg-core.c:1146 gen_twopix() error: buffer overflow 'buf[1]' 8 <= 8
	drivers/media/common/v4l2-tpg/v4l2-tpg-core.c:1152 gen_twopix() error: buffer overflow 'buf[1]' 8 <= 8
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

There are some troubles on this driver with respect to the usage
of __be16 and __b32 macros:

	drivers/media/i2c/ov5670.c:1857:27: warning: incorrect type in initializer (different base types)
	drivers/media/i2c/ov5670.c:1857:27:    expected unsigned short [unsigned] [usertype] reg_addr_be
	drivers/media/i2c/ov5670.c:1857:27:    got restricted __be16 [usertype] <noident>
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1880:16: warning: cast to restricted __be32
	drivers/media/i2c/ov5670.c:1901:13: warning: incorrect type in assignment (different base types)
	drivers/media/i2c/ov5670.c:1901:13:    expected unsigned int [unsigned] [usertype] val
	drivers/media/i2c/ov5670.c:1901:13:    got restricted __be32 [usertype] <noident>

Fix them.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

As warned:
	drivers/media/platform/rockchip/rga/rga.c:210:16: warning: symbol 'formats' was not declared. Should it be static?
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Fix those warnings:

	drivers/media/platform/davinci/vpss.c:510:25: warning: incorrect type in argument 1 (different address spaces)
	drivers/media/platform/davinci/vpss.c:510:25:    expected void volatile [noderef] <asn:2>*addr
	drivers/media/platform/davinci/vpss.c:510:25:    got unsigned int [usertype] *static [toplevel] [assigned] vpss_regs_base2
	drivers/media/platform/davinci/vpss.c:520:34: warning: incorrect type in assignment (different address spaces)
	drivers/media/platform/davinci/vpss.c:520:34:    expected unsigned int [usertype] *static [toplevel] [assigned] vpss_regs_base2
	drivers/media/platform/davinci/vpss.c:520:34:    got void [noderef] <asn:2>*
	drivers/media/platform/davinci/vpss.c:522:54: warning: incorrect type in argument 2 (different address spaces)
	drivers/media/platform/davinci/vpss.c:522:54:    expected void volatile [noderef] <asn:2>*addr
	drivers/media/platform/davinci/vpss.c:522:54:    got unsigned int [usertype] *static [toplevel] [assigned] vpss_regs_base2

Weird enough, vpss_regs_base0 and vpss_regs_base1 were
properly annotated.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Solves those warnings:
	drivers/media/dvb-core/dvb_frontend.c:2297:39: warning: incorrect type in argument 1 (different address spaces)
	drivers/media/dvb-core/dvb_frontend.c:2297:39:    expected void const [noderef] <asn:1>*<noident>
	drivers/media/dvb-core/dvb_frontend.c:2297:39:    got struct dtv_property *props
	drivers/media/dvb-core/dvb_frontend.c:2331:39: warning: incorrect type in argument 1 (different address spaces)
	drivers/media/dvb-core/dvb_frontend.c:2331:39:    expected void const [noderef] <asn:1>*<noident>
	drivers/media/dvb-core/dvb_frontend.c:2331:39:    got struct dtv_property *props

No functional changes.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The logic at find_format() is a little bit confusing even for
humans, and it tricks static code analyzers:

	drivers/staging/media/imx/imx-media-utils.c:259 find_format() error: buffer overflow 'array' 14 <= 20

Rewrite the logic in a way that it makes it clearer to understand,
while prevent static analyzers to produce false positives.
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

If allocation of struct board_info fails, return NULL from
dvb_module_probe().

Fix this warning:
	drivers/media/dvb-core/dvbdev.c:958 dvb_module_probe() error: potential null dereference 'board_info'.  (kzalloc returns null)
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Document the core of the debugfs CEC error injection ABI.

The driver specific commands are documented elsewhere and
this file points to that documentation.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

The CEC Pin framework adds support for Error Injection.

Document all the error injections commands and how to use it.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

It turns out that the struct cec_fh event buffer size of 64 events
(64 for CEC_EVENT_PIN_CEC_LOW and 64 for _HIGH) is too small. It's
about 160 ms worth of events and if the Raspberry Pi is busy, then it
might take too long for the application to be scheduled so that it can
drain the pending events. Increase these buffers to 800 events which
is at least 2 seconds worth of events.

There is also a FIFO in between the interrupt and the cec-pin thread.
The thread passes the events on to the CEC core. It is important that
should this FIFO fill up the cec core will be informed that events
have been lost so this can be communicated to the user by setting
CEC_EVENT_FL_DROPPED_EVENTS.

It is very hard to debug CEC problems if events were lost without
informing the user of that fact.

If events were dropped due to the FIFO filling up, then the debugfs
status file will let you know how many events were dropped.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Keep track of the number of short or long start bits, the number
of short or long data bits and the number of initiated or detected
low drive conditions.

Show this information in the status debugfs log.

Helpful when debugging, particularly when doing error injection
as well.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Implement all the error injection commands.

The state machine gets new states for the various error situations,
helper functions are added to detect whether an error injection is
active and the actual error injections are implemented.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

Add support to the CEC Pin framework to parse error injection commands
and to show them.

The next patch will do the actual implementation of this.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>

This function will be needed for injecting a custom pulse.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NMauro Carvalho Chehab <mchehab@s-opensource.com>