SSICR has SWSP bit (= Serial WS Polarity) which decides WS pin 1st
channel polarity (low or hi). This bit shouldn't exchange after running.
Current SSI "parent" doesn't care SSICR, just controls clock only.
Because of this behavior, if platform uses SSI0 as playback,
SSI1 as capture, and if user starts capture -> playback order,
SSI0 SSICR::SWSP bit exchanged 0 -> 1 during captureing, and it makes
capture noise.
This patch cares SSICR on SSI parent, too.
Special thanks to Yokoyama-san
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
If SSIs are sharing WS pin, it should has .symmetric_rates.
This patch sets it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
It is waiting udelay(50) x 1024 (= 50ms) for status check
in worst case, but it is overkill.
And we shouldn't use udelay() for 50us
(linux/Documentation/timers/timers-howto.txt)
Waiting maximum udelay(5) x 1024 (= 5ms) is very enough
for status check.
This patch fixes these issue.
Reported-by: Hiromitsu Yamasaki <hiromitsu.yamasaki.ym@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
commit 4821d914fe ("ASoC: rsnd: use dma_sync_single_for_xxx() for
IOMMU") (= v1) which have been already reverted had supported IPMMU
support on rsnd driver.
Because memory allocating timing and DMAEngine access timing were
different, it used continuous memory and called dma map function by
itself.
OTOH, it is using DMA descriptor mode (= DMA cyclic mode), thus, there
was timing conflict between DMA sync/unsync and DMA transfer starting,
and it maked sound noise.
This patch supports IPMMU with coherent memory, and, it uses Audio DMAC
dev for allocating memory by snd_pcm_lib_preallocate_pages_for_all() to
indicate memory area to IPMMU.
One note is that Playback/Capture need each paired Audio DMAC dev.
Because of this, we need to keep each paired Audio DMAC dev when probing,
and use it when allocating each memory for IPMMU.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current ssi.c is assuming below 2 things to probing HDMI node.
1) remote node is including "hdmi0" or "hdmi1" in node name
2) remote_ep->full_name is including full path name
But, these assumptions are broken by below
1) Node names should not use numerical suffixes
commit 6b5ac2f1cb ("arm64: dts: renesas: r8a7795: Drop bogus HDMI
node names suffixes")
2) node full_name no longer include full path name
commit a7e4cfb0a7 ("of/fdt: only store the device node basename
in full_name")
Because of these reasons, ssi.c can't probe HDMI on current kernel.
This patch probes HDMI0/1 by using its address.
Note is that we need to keep updating for this address for future
generation chip.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Nguyen Viet Dung <dung.nguyen.aj@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
rsnd driver is supporting Gen3. The difference between Gen1 and Gen2
were very big, but, between Gen2 and Gen3 are not so much.
Thus, it is assuming Gen2 and Gen3 have compatible, therefore,
there is no RSND_GEN3 and rsnd_is_gen3() macro.
But in the future, it will need Gen2 and Gen3 different operation,
and for Gen4.
This patch adds missing RSND_GEN3 and rsnd_is_gen3() macro.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Nguyen Viet Dung <dung.nguyen.aj@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
rsnd_cmd_mod_get() is used from cmd.c only.
Let's makes it static function
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The suspend/resume callbacks are now optional, leading to a warning
when they are unused:
sound/soc/sh/rcar/core.c:1548:12: error: 'rsnd_resume' defined but not used [-Werror=unused-function]
static int rsnd_resume(struct device *dev)
^~~~~~~~~~~
sound/soc/sh/rcar/core.c:1539:12: error: 'rsnd_suspend' defined but not used [-Werror=unused-function]
static int rsnd_suspend(struct device *dev)
This marks the as __maybe_unused to avoid the warning.
Fixes: f8a9a29c4f ("ASoC: rsnd: set pm_ops in hibernate-compatible way")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
Use SET_SYSTEM_SLEEP_PM_OPS() macro instead of direct assignment to
.suspend and .resume fields.
This makes driver working after restore from hibernation.
Signed-off-by: Nikita Yushchenko <nikita.yoush@cogentembedded.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
rsnd_dai_call() is using dev_dbg(), but its message is sometimes
blocks nessesary other messages. If RSND_DEBUG_NO_DAI_CALL was
defined it will be suppressed by this patch.
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
SSI/SRC have under/over flow error handling, and its status is useful
for debuging. But sometimes it might be too much message,
and it might blocks necessity other information.
To avoid such situation, basically this patch indicates interrupt
status debug message if DEBUG was defined, but it will be suppressed
if RSND_DEBUG_NO_IRQ_STATUS was defined.
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Now platform can be replaced to component, let's do it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
BRGCKR should use 0x80770000, instead of 0x80FF0000.
R-Car Gen2 xxx_TIMSEL should use 0x0F1F,
R-Car Gen3 xxx_TIMSEL should use 0x1F1F.
Here, Gen3 doesn't support AVD, thus, both case can use 0x0F1F.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
SSI had shared counting pointer position method between PIO/DMA mode
before. But now DMA mode is using DMAEngine feature to get it.
Thus, this counting pointer position method is needed for only PIO mode.
We don't need to share code anymore.
This patch names PIO related functions as rsnd_ssi_pio_xxx(), and
merged/cleanuped each feature.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current rsnd driver is judging 16bit/24bit data by using
runtime->sample_bits, but it is indicating physical size,
not format size. This is confusable code.
This patch uses snd_pcm_format_width() to be more correct code.
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
On Renesas sound device, DALIGN which exchanges channel position
is needed because SW and HW are using defferent data order if
16bit data. It is not needed when 24bit data.
rsnd_get_dalign() returns necessary value, but it was confusable
code. This patch makes it more simple.
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
period_pos can always be calculated by byte_pos and
byte_per_period, there is no reason to maintain this
variable in rsnd_dai_stream.
This patch removes period_pos from rsnd_ssi and calculates
next_period_byte with consideration of actual byte_pos value.
Signed-off-by: Jiada Wang <jiada_wang@mentor.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Currently there is race condition between set of byte_pos and wrap
it around when new buffer starts. If .pointer is called in-between
it will result in inconsistent pointer position be returned
from .pointer callback.
This patch increments buffer pointer atomically to avoid this issue.
Signed-off-by: Jiada Wang <jiada_wang@mentor.com>
Reviewed-by: Takashi Sakamoto <takashi.sakamoto@miraclelinux.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
DMA handler had needed to calculate pointer before, but it doesn't
need now. Thus, we can remove unnecessary spin lock from DMAC handler.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Renesas sound needs 8ch clock if TDM 6ch mode, and needs 2ch clock for
6ch or 8ch sound if Multi SSI mode. And these are related to before/after
CTU (= Channel Transfer Unit).
To calculate these we already has rsnd_runtime_channel_for_ssi() which
returns runtime necessary channels.
But, it based on runtime->channels which is not yet set when hw refine.
We need to use hw_params instead of runtime->xxx when hw refine,
and it is not needed after runtime was set.
This patch adds new hw_params on rsnd_dai_stream, and it will be removed
on rsnd_hw_params().
This is very temporary durty code, but it seems no choice at this point.
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Renesas R-Car sound driver should be stopped if unbinded during
playbacking/capturing. Otherwise clock open/close counter mismatch
happen.
One note is that we can't skip from remove function (= return -Exxx)
in such case if user used unbind. Because unbind function doesn't
check return value from each driver's remove function.
This means we must to stop and remove driver in remove function.
Now ASoC has snd_soc_disconnect_sync() for this purpose.
Let's use it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
register SSI_MODE is set when SSI works in TDM Extended,
but it isn't reset when SSI starts to work in other modes,
thus causes issues.
This patch clearss SSI_MODE register when SSI works in modes
other than TDM Extended.
Fixes: 186fadc132 ("ASoC: rsnd: add TDM Extend Mode support")
Signed-off-by: Jiada Wang <jiada_wang@mentor.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
commit 4821d914fe ("ASoC: rsnd: use dma_sync_single_for_xxx() for
IOMMU") had supported IOMMU, but it breaks normal sound "recorde"
and both PulseAudio's "playback/recorde". The sound will be noisy.
That commit was using dma_sync_single_for_xxx(), and driver should
make sure memory is protected during CPU or Device are using it.
But if driver returns current "residue" data size correctly on pointer
function, player/recorder will access to protected memory.
IOMMU feature should be supported, but I don't know how to handle it
without memory cache problem at this point.
Thus, this patch simply revert it to avoid current noisy sound.
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Tested-by: Ryo Kodama <ryo.kodama.vz@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
PTR_ERR(NULL) is success. Normally when a function returns both NULL
and error pointers, it means that NULL is not a error.
But, rsnd_dmaen_request_channel() returns NULL if requested resource
was failed.
Let's return -EIO if rsnd_dmaen_request_channel() was failed on
rsnd_dmaen_nolock_start().
This patch fixes commit edce5c496c ("ASoC: rsnd: Request/Release DMA
channel eachtime")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current rsnd driver has rsnd_mod_id() which returns mod ID,
and it returns -1 if mod was NULL.
In the same time, this driver has rsnd_mod_name() which returns mod
name, and it returns "unknown" if mod or mod->ops was NULL.
Basically these "mod" never be NULL, but the reason why rsnd driver
has such behavior is that DMA path finder is assuming memory as
"mod == NULL".
Thus, current DMA path debug code prints like below.
Here "unknown[-1]" means it was memory.
...
rcar_sound ec500000.sound: unknown[-1] from
rcar_sound ec500000.sound: src[0] to
rcar_sound ec500000.sound: ctu[2]
rcar_sound ec500000.sound: mix[0]
rcar_sound ec500000.sound: dvc[0]
rcar_sound ec500000.sound: ssi[0]
rcar_sound ec500000.sound: audmac[0] unknown[-1] -> src[0]
...
1st issue is that it is confusable for user.
2nd issue is rsnd driver has something like below code.
mask |= 1 << rsnd_mod_id(mod);
Because of this kind of code, some statically code checker will
reports "Shifting by a negative value is undefined behaviour".
But this "mod" never be NULL, thus negative shift never happen.
To avoid these issues, this patch adds new dummy "mem" to
indicate memory, and use it to indicate debug information,
and, remove unneeded "NULL mod" behavior from rsnd_mod_id() and
rsnd_mod_name().
The debug information will be like below by this patch
...
rcar_sound ec500000.sound: mem[0] from
rcar_sound ec500000.sound: src[0] to
rcar_sound ec500000.sound: ctu[2]
rcar_sound ec500000.sound: mix[0]
rcar_sound ec500000.sound: dvc[0]
rcar_sound ec500000.sound: ssi[0]
rcar_sound ec500000.sound: audmac[0] mem[0] -> src[0]
...
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
SSI parent mod might be NULL. ssi_parent_mod() needs to care
about it. Otherwise, it uses negative shift.
This patch fixes it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
When stop case, it was Playback, it need to check all data were
completely sent. But in Capture case, it might not receive data
anymore. SSISR::DIRQ check is not need for Capture case.
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Let's use more common style to checking running/working
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
clk_unprepare() is checking parameter by IS_ERR_OR_NULL().
clk NULL check is not needed on rsnd_mod_quit()
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
We have rsnd_io_to_mod() macro. Let's use it
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
ADG inputs clock from CLK{A,B,C,I} and outputs clock from
CLKOUT{0,1,2,3} which is selected by BRG{A,B}.
Now, ADG is assuming BRGA is for 44100Hz related clocks,
BRGB is for 48000Hz related clocks.
Clock related debug is very difficult/confusable.
This patch cleanups clock related debug info.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current CTU/MIX/DVC are directly using rsnd_kctrl_cfg_m/s to control
val etc, but it is difficult to read/understand.
And there was no uniformity in access method.
This patch adds new rsnd_kctrl_xxx() and implements uniformed access
method.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Both DVC/MIX have Volume Ramp Control. This patch supprts MIX
Volume Ramp. One note is that main purpose of MIX Volume Ramp
is to reduce noise, thus, MIX Ramp range is very few if you
compare to DVC Volume Ramp (DVC = 5bit, MIX = 4bit).
You can use MIX Volume Ranp like below
amixer set "MIX Ramp Up Rate" "0.125 dB/1 step"
amixer set "MIX Ramp Down Rate" "0.125 dB/1 step"
amixer set "MIX Ramp" on
aplay xxx.wav &
amixer set "MIX",0 80% // DAI0 Volume Down
amixer set "MIX",1 100% // DAI1 Volume Up
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
DVC is supporting Volume Ramp Rate, and MIX has Volume Ramp
but not yet supported. To support MIX Volume Ramp, we want to
share Rate List since DVC/MIX are using almost same list.
This patch move DVC specific Volume Ramp Rate List to core.c.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Same CTU might be used few times if system/platform is using MIX.
For example below case.
DAI0 playback = <&src0 &ctu02 &mix0 &dvc0 &ssi0>;
DAI1 playback = <&src2 &ctu03 &mix0 &dvc0 &ssi0>;
This case, ALSA will have CTU,0 and CTU,1 kcontrol interfaces,
but these are same CTU. This is confusing.
This patch adds new flags and avoid such case.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Same DVC might be used few times if system/platform is using MIX.
For example below case.
DAI0 playback = <&src0 &ctu02 &mix0 &dvc0 &ssi0>;
DAI1 playback = <&src2 &ctu03 &mix0 &dvc0 &ssi0>;
This case, ALSA will have DVC,0 and DVC,1 kcontrol interfaces,
but these are same DVC. This is confusing.
This patch adds new flags and avoid such case.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
SSI is using rsnd_ssi_flags_xxx() macro to control flags.
But it is useful macro not only for SSI. This patch replace it
to more generic rsnd_flags_xxx().
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The current device tree representation of the R-Car SSI assumes that they
are numbered consecutively, starting from 0. Alas, this is not the case
with the R8A77995 (D3) SoC which SSI1/SSI2 aren't present. In order to
keep the existing device trees working, I'm suggesting to use a disabled
node for SSI0/SSI1. Teach the SSI probe to just skip disabled nodes.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
R-Car sound DMA will be used from SSI/SRC.
dma.c doesn't alloc DMA handler in .probe timing, because we don't
know what kind of DMA transfer will be used then.
Thus, SSI/SRC have *rsnd_mod for DMA. rsnd_dma_attach() will allocate
it and attach it to system.
It will be PIO mode if it can't alloc DMA handler.
In case of MIX is used, rsnd_dma_attach() will be called twice from SSI.
To avoid duplicate allocation, current rsnd_dma_attach() is checking
allocated DMA handler. This DMA related operation is a little bit
difficult to understand.
This patch adds new rsnd_dma_alloc() and separates allocation and attach
for readable code.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current ADG driver is over-writing flags. This patch fixes it.
Reported-by: Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
