Renesas R-Car sound (= rsnd) needs 2 DMAC which are called as
Audio DMAC (= 1st DMAC) and Audio DMAC peri peri (2nd DMAC).
And rsnd had assumed that 1st / 2nd DMACs are implemented as DMAEngine.
But, in result of DMA ML discussion, 2nd DMAC was concluded that it is
not a general purpose DMAC (2nd DMAC is for Device to Device inside
sound system). Additionally, current DMAEngine can't support Device to
Device, and we don't have correct DT bindings for it at this point.
So the easiest solution for it is that move it from DMAEngine to rsnd
driver.
Audio DMAC peri peri is controlled from sound driver without DMAEngine
by this patch.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Renesas sound driver had been assumed that Audio DMAC /
Audio DMAC peri peri are implemented by DMAEngine. But, result of DMA
ML discussion, it was concluded that it is not a general purpose DMAC.
And it should be moved from current DMAEngine to rsnd driver.
So, Audio DMAC peri peri become non DMAEngine.

OTOH, ALSA SoC has soc-generic-dmaengine-pcm implementation. but it
seems difficult to use this generic implementation on rsnd driver at
this point, since it needs to fallback to PIO mode if DMA can't use.
and additionally it needs 2 DMAC (= Audio DMAC / Audio DMAC peri peri).
These are not "generic" feature.

Of course I will try to use this generic dmaengine in the future somehow,
but just use current style at this point until it can formally use 2 DMACs.
This patch adds new dma.c and moves current dma code to dma.c from core.c.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

priv is not used on rsnd_dma_quit()
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

rsnd_dma_available() is not used. Let's remove it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

It can get DMA direction via rsnd_dai_stream.
Remove un-needed is_play from rsnd_dma_init().
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Renesas R-Car sound had SH-DMA-BASE specific implementation before,
but, now, it is no longer needed. Let's remove it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current Renesas R-Car sound driver doesn't call
snd_soc_unregiter_component/platform() in .remove.
This patch call these functions.
Reported-by: NNguyen Viet Dung <nv-dung@jinso.co.jp>
Reported-by: NBui Duc Phuc <bd-phuc@jinso.co.jp>
Reported-by: NCao Minh Hiep <cm-hiep@jinso.co.jp>
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current Renesas R-Car sound driver is using snd_ctl_xxx()
functions, but it didn't call snd_ctl free_one() / snd_ctl_remove().
This patch call these functions.
Reported-by: NNguyen Viet Dung <nv-dung@jinso.co.jp>
Reported-by: NBui Duc Phuc <bd-phuc@jinso.co.jp>
Reported-by: NCao Minh Hiep <cm-hiep@jinso.co.jp>
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver has rsnd_mod_to_priv() macro, and struct rsnd_mod
has struct rsnd_priv pointer. But, it is waste of memory from data
structure point of view.
Today we can link mod <-> io <-> rdai each other, so saving priv in
rdai is very reasonable.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver's function requests rdai, but it is not used.
Adding priv is more reasonable. Let's replace it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver's many functions requests struct rsnd_dai (= rdai),
but, 1) we can get rdai from rsnd_io_to_rsnd(), 2) almost all rdai
is not used. Let's remove these rdai.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver can use rsnd_io_to_rdai()
we can get play/capture direction via io now.
Let's replace rsnd_dai_is_play() to rsnd_io_is_play()
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver request rdai in everywhere, but each mod should
reach to rdai by itself. this patch adds missing rsnd_io_to_rdai().

io   = rsnd_mod_to_io(mod);
rdai = rsnd_io_to_rdai(io);
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

rsnd_dai_get_platform_info() is not used. Let's remove it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

rsnd_dai_id() was not used. Let's remove it
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

struct rsnd_dai is called as "rdai", and struct snd_soc_dai is called
as "dai" on this driver. but macro/function have confusable naming.
This patch rename rsnd_dai_get() to rsnd_rdai_get()
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current rsnd driver's mod (= ssi/src/dvc) have each own clk,
but, these are needed to start/stop each mod.
This patch adds struct clk in rsnd_mod, and start/stop these
via rsnd_mod_hw_start/stop() macro
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The ALSA core takes care that all preallocated memory is freed when the PCM
itself is freed. There is no need to do this manually in the driver.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current DVC is using snd_kcontrol_new functions, but, SRC will need
same method. Move common functions to core.c
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

R-Car sound can use SSI/SRC/DVC modules, and these are controlled as
rsnd_mod in rsnd driver. These rsnd_mod has each own function as
callback. Basically these callback function has pair like probe/remove,
start/stop, etc. And, these functions are called by order to each stage
like below.
 1. src->probe
 2. ssi->probe
 3. dvc->probe
 4. src->start
 5. ssi->start
 6. dvc->start
 7. src->stop
 8. ssi->stop
 9. dvc->stop
10. src->remove
11. ssi->remove
12. dvc->remove

But, current rsnd driver doesn't care about its status which indicates
which function is called.
For example, if 5) returns error, 6) is not called. In such case,
9) should not be called. This patch care about each modules status.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current R-Car sound has PIO fallback support if it couldn't use DMA.
This fallback is done in .remove callback, but, it should have
.fallback callback. Otherwise, normal .remove callback will have
strange behavior. This patch adds .fallback callback.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current Renesas R-Car sound driver probe will be failed
if it try to use DMA mode and it couldn't use for some reasons.
But PIO mode works even though in such case.
This patch try to fallback to PIO mode if DMA mode probing was failed.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current Renesas R-Car sound driver debug message
is using random format
(ex "ssi0: xxx" / "SSI0 xxx" / "ssi[0]: xxx")
and confusable timing
("xxx probe failed" and "xxx probed" are shown in same time)
This patch fixes these
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

R-Car sound doesn't support PAUSE.
Remove SNDRV_PCM_INFO_PAUSE flags from snd_pcm_hardware info
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org

Of_node_put supports NULL as its argument, so the initial test is not
necessary.

Suggested by Uwe Kleine-König.

The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression e;
@@

-if (e)
   of_node_put(e);
// </smpl>
Signed-off-by: NJulia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: NMark Brown <broonie@linaro.org>

In current R-Car rsnd driver,
the SND_SOC_DAIFMT_xB_xF flags are used to HW default behavior,
but, it should be used to specific format.
The waveforms of LEFT_J/RIGHT_J format with
SND_SOC_DAIFMT_NB_NF flag will be
started from "falling edge" without this patch.
But, it should be started from "rising edge".
Reported-by: NJun Watanabe <jun.watanabe.ue@renesas.com>
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

bff58ea4
(ASoC: rsnd: add DVC support) added DVC support,
and it added pcm_new callback feature for it.
Then it called all DAI's pcm_new callback, and it was wrong.
This patch fixup it and call correct callback.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

rsnd driver is using SSI/SRC/DVC which are
using "mod" base operation.
These "mod" are supporting "probe" and "remove" callbacks.

Current rsnd_probe should call "remove" if "probe" was failed,
since "probe" might be having DMAEngine handle.
Some mod's "remove" callback might be called without calling
"probe", but it is no problem. because "remove" do nothing
in such case.

So, all mod's "remove" should be called when error case
of rsnd_probe() and rsnd_remove().
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Renesas new DMAEngine driver requests src/dst_addr_width
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Acked-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Now, DVC can use DT
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

R-Car sound uses Audio DMAC and Audio DMAC peri peri.
Audio DMAC peri peri transfers data inside circuit.

DMA transfer needs source / destination address,
and destination address can be set via dmaengine_slave_config().
The source address can be set when starting DMAEngine.
Because Audio DMAC peri peri always ignores its value,
current driver always used same source address for
Audio DMAC / Audio DMAC peri peri
(Audio DMAC peri peri source / destination address
is always fixed value)
But, This is not good match for DT booting.
This patch properly uses DMA start address
for Audio DMAC / Audio DMAC peri peri.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Current R-Car sound SSI/SRC/DVC selection has feature limit.
(It is assuming that SSI/SRC are using same index number)

So that enabling SSI/SRC flexible selection,
this patch modifies DMA settings.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Current R-Car sound driver is using DMAEngine directly,
but, ASoC is requesting to use common DMA transfer method,
like snd_dmaengine_pcm_trigger() or dmaengine_pcm_ops.
It is difficult to switch at this point, since Renesas
driver is also supporting PIO transfer.
This patch uses dmaengine_prep_dma_cyclic() instead
of dmaengine_prep_slave_single().
It is used in requested method,
and is good first step to switch over.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Sound data needs to be sent to R-Car sound SSI when playback.
But, there are 2 interfaces for it.
1st is SSITDR/SSIRDR which are mapped on SSI.
2nd is SSIn_BUSIF which are mapped on SSIU.

2nd SSIn_BUSIF is used when DMA transfer,
and it is always used if sound data came from via SRC.
But, we can use it when SSI+DMA case too.
(Current driver is assuming 1st SSITDR/SSIRDR for it)

2nd SSIn_BUSIF can be used as FIFO.
This is very helpful/useful for SSI+DMA.

But DMA address / DMA ID are not same between 1st/2nd cases.
This patch care about these settings.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

dmaengine_prep_slave_single() expects a enum dma_transfer_direction and not a
enum dma_data_direction. Since the integer representations of both DMA_TO_DEVICE
and DMA_MEM_TO_DEV aswell as DMA_FROM_DEVICE and DMA_DEV_TO_MEM have the same
value the code worked fine even though it was using the wrong type.

Fixes the following warning from sparse:
	sound/soc/sh/rcar/core.c:227:49: warning: mixing different enum types
	sound/soc/sh/rcar/core.c:227:49:     int enum dma_data_direction  versus
	sound/soc/sh/rcar/core.c:227:49:     int enum dma_transfer_direction
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Acked-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Current dma name search loop didn't care about SSI index
This patch fixes it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Index of dma name should use -1, not +1 when capture case.
Thank you Dan.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

The DMAC src/dst addr needs to be set from driver when DT case.
(It was set from SoC/DMAEngine code when non-DT case)
This patch adds rsnd_gen_dma_addr() to set DMAC src/dst addr.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Renesas sound driver is supporting to use DMAEngine.
But, DMA slave channel name "tx", "rx" is not enough
in DT case.
Becuase, it has many ports and path combination.

This patch adds rsnd_dma_of_name() to find
DMA channel name, for example
memory to SSI0 is "mem_ssi0",
SSI0 to memory is "ssi0_mem",
SSI0 to SRC0   is "ssi0_src0",
SRC0 to SSI0   is "src0_ssi0",
SRC0 to DVC0   is "src0_dvc0"...

Renesas sound want to use PIO transfer mode for some reasons.
It will be PIO tranfer mode if device node doesn't have
DMA settings.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

DT DMA support needs struct platform_device pointer,
and it can get struct device pointer from platform_device.
Save platform_device instead of device.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@linaro.org>