soc-core sets the snd_soc_pcm_runtime->dev drvdata to
snd_soc_pcm_runtime in soc_post_component_init, and access
it in places like codec_reg_show.

hsw_pcm_open overwrites the drvdata to point to hsw_pcm_data,
confusing soc-core, and causing crashes when cat
/sys/devices/pci0000:00/INT3438:00/.../System PCM/codec_reg

This patch removes the set in hsw_pcm_open since it's no longer
used. commit 7ff9d671 ("ASoC: Intel: Split hsw_pcm_data for
playback and capture") already removed all calls to
snd_soc_pcm_get_drvdata(rtd).
Signed-off-by: NBen Zhang <benzh@chromium.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

Format number after 0x in hex.

Cc: Jie Yang <yang.jie@linux.intel.com>
Cc: Liam Girdwood <lgirdwood@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Jaroslav Kysela <perex@perex.cz>
Cc: Takashi Iwai <tiwai@suse.com>
Signed-off-by: NJoonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The recent fix left a variable declaration without usage.
  sound/soc/intel/haswell/sst-haswell-pcm.c:1349:18: warning: unused variable ‘hsw’ [-Wunused-variable]

Fixes: edd8ed49 ('ASoC: Intel: handle haswell pcm suspend including runtime modules freeing')
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

It needs free pcm runtime modules before unloading firmware, here
add hsw_pcm_suspend() to handle this procedure:
suspends firmware ==> frees runtime modules ==> unloads firmware.

This fixes the broadwell module unload failed issue.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Add check runtime module pointers before freeing them, and clear
them to NULL after freed.

With this implemented, we can avoid NULL pointer dereference or
double free errors.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

This reverts commit 01f202c7.

We shouldn't leave the device as suspended state after module freed,
it is not good to do runtime suspend at driver free, here revert
this fixing, and replace it with the procedure:
suspends firmware ==> frees runtime modules ==> unloads firmware.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

This reverts commit 506c148e.

We still need this hsw_pcm_free_modules(), we plan to remove the
runtime modules at both fw_unload(D0->D3) and snd_soc_sst_haswell_pcm
module removing.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

In haswell-pcm module unloading, we can't free runtime modules
directly, for they may be already freed in runtime suspend.

Here add executing suspend call to unload runtime modules, only
for status not equal to RPM_SUSPEND, to fix broadwell module
removing failed issue.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Remove the unused function hsw_pcm_free_modules() to fix the
compling warning:

sound/soc/intel/haswell/sst-haswell-pcm.c:923:13:
 warning: 'sw_pcm_free_modules' defined but not used
 [-Wunused-function]
 static void hsw_pcm_free_modules(struct hsw_priv_data *pdata)
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

0day robot reported a buffer overflow issue:

...
sound/soc/intel/haswell/sst-haswell-pcm.c:1107 hsw_pcm_probe() error: buffer\
overflow 'hsw_dais' 4 <= 4
sound/soc/intel/haswell/sst-haswell-pcm.c:1109 hsw_pcm_probe() error: buffer\
overflow 'hsw_dais' 4 <= 4
...

Fix it by initializing the index(i) to correct value.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Restructure the sound/soc/intel/ directory: create haswell folder, and
move haswell platform files here.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Reviewed-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The smatch tool report warning:
  ...
    CHECK   sound/soc/intel/sst-haswell-pcm.c
  sound/soc/intel/sst-haswell-pcm.c:1110 hsw_pcm_probe() error: buffer overflow\
'hsw_dais' 4 <= 4
  sound/soc/intel/sst-haswell-pcm.c:1112 hsw_pcm_probe() error: buffer overflow\
'hsw_dais' 4 <= 4
  ...
fix it by use its own struct member for post-process module, rather than sharing
unused pcm member.
Signed-off-by: NLu, Han <han.lu@intel.com>
Acked-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Check waves state before set parameter through ipc to prevent unexpected
operation. Also remove redundant check.
Signed-off-by: NLu, Han <han.lu@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Each kcontrol command includes a line of parameters up to 128 bytes.
kcontrol command to set param:
    cset "name='Waves Set Param' <0x01,0xff,...>"
or
    cset-bin-file "name='Waves Set Param' <path/to/binary/config/file>"

The parameter lines are stored in a buffer array, so can be read back from
buffer rather than from DSP, and be relaunched to DSP when resume from RTD3.
The buffer size is 160 parameter lines.
kcontrol command to reset the buffer:
    cset "name='Waves Set Param' 0xff"

alsa-lib v1.0.29 or commit 6ea14c36 and f47480af are required to support the
kcontrol commands.
Signed-off-by: NLu, Han <han.lu@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Add kcontrol to enable/disable module waves. IPC is valid only when module
is loaded. Also track module state over suspend so it's state can be restored
on resume.
Signed-off-by: NLu, Han <han.lu@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Try to load module waves and allocate runtime blocks for it if the firmware
image of module waves exists.
Signed-off-by: NLu, Han <han.lu@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Add SNDRV_PCM_INFO_DRAIN_TRIGGER in snd_pcm_hardware.info to
call sst_hsw_stream_set_silence_start() when PCM draining.
Signed-off-by: NLibin Yang <libin.yang@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

The BDW audio firmware DSP manages the DMA and the DMA cannot be
stopped exactly at the end of the playback stream. This means
stale samples may be played at PCM stop unless the driver copies
silence to the subsequent periods.
Signed-off-by: NLibin Yang <libin.yang@intel.com>
Reviewed-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

For runtime suspend/resume, it is some different with suspend/resume,
e.g. codec power supply won't be switch off, codec jack detection
still working(to wake up system from Jack event), won't call call
snd_soc_suspend/resume, etc.

So here, we add a platform PM status, HSW_PM_STATE_RTD3, to make
the status more clear, when in idle, it will enter this status, to
transfer from HSW_PM_STATE_RTD3 to HSW_PM_STATE_D3, we will do those
extra jobs, and vice versa for resuming.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

There may be 2 pcm streams for a same DAI at most, and these 2
streams should have different hsw_pcm_data, e.g. they have
different persistent data, so here we need split hsw_pcm_data
for playback and capture, to make sure they won't be mixed
and keep cleaned.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Reviewed-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

A DAI may have 2 streams(playback/capture) and different modules
may be needed for them respectively, so we need add a stream
direction here, the combination(dai_id + stream) can tell us
which module we really need here.
Signed-off-by: NJie Yang <yang.jie@intel.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>
Acked-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The sst-haswell-pcm driver is a new user of CONFIG_PM_RUNTIME.

However, after commit b2b49ccb (PM: Kconfig: Set PM_RUNTIME if
PM_SLEEP is selected) PM_RUNTIME is always set if PM is set, so
#ifdef blocks depending on CONFIG_PM_RUNTIME may now be changed to
depend on CONFIG_PM.

Replace CONFIG_PM_RUNTIME with CONFIG_PM in
sound/soc/intel/sst-haswell-pcm.c.
Signed-off-by: NRafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: NMark Brown <broonie@kernel.org>

Move capture PCM pin from PCM4 to PCM0 for Broadwell/Haswell.
This will allow us to integrate with pulseaudio better for
usually default device is set to 0.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The xmax volume should be corrected to ARRAY_SIZE(volume_map)-1, otherwise,
the xmax value will be mapped to 0 wrongly.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

On Broadwell, the ADSP FW don't support loopback record volume tuning,
so here remove this control.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The volume setting control for capture stream doesn't take effect on intel
Broadwell platform. Root cause it at 2 points:
1. set stream volume with channel=2 is wrongly bapassed;
2. the saved stream volume should be restored after stream is commit.

Here correct these 2 items to fix the stream volume set issue.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Fix the following build error when CONFIG_SLEEP is enabled and CONFIG_RUNTIME
is disabled. The BDW ADSP sleep PM functionality depends on the runtime pm
calls for context save/restore.

All error/warnings:

>> ERROR: "snd_soc_suspend" undefined!
>> ERROR: "snd_soc_resume" undefined!
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The priv_data is allocated again here wrongly, and it
is not set to the driver data after assignment. This
make the pdata->dev is NULL and oops occurs on the first
call to hsw_volume_put.

The resource has been allocated in driver probe callback
hsw_pcm_dev_probe, so here just remove this sencond
allocation is OK.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Fix the following errors:

All error/warnings:

>> sound/soc/intel/sst-haswell-pcm.c:1168:13: error: 'hsw_pcm_prepare' undeclared here (not in a function)
     .prepare = hsw_pcm_prepare,
                ^
>> sound/soc/intel/sst-haswell-pcm.c:1169:14: error: 'hsw_pcm_complete' undeclared here (not in a function)
     .complete = hsw_pcm_complete,
                 ^
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Add PM and RTD3 support to the HSW/BDW PCM driver. The PCM driver will
now save DSP context and then power off the DSP when it's not in use.
DSP power and context is then restored when it's next used.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current block allocation is tied to block type and requestor type. Make the
allocation more generic by removing the struct module parameter and adding
a generic block allocator structure. Also pass in the list that the blocks
have to be added too in order to remove dependence on block requestor type.

ASoC: Intel: update scratch allocator to use generic block allocator

Update the scratch allocator to use the generic block allocator and calculate
total scratch buffer size.

ASoC: Intel: Add call to calculate offsets internally within the DSP.

A call to calculate internal DSP memory addresses used to allocate persistent
and scartch buffers.

ASoC: Intel: Add runtime module support.

Add support for runtime module objects that can be created for every FW
module that is parsed from the FW file. This gives a 1:N mapping between
the FW module from file and the runtime instantiations of that module.

We also need to make sure we remove every module and runtime module when
we unload the FW.

ASoC: Intel: Add DMA firmware loading support

Add support for DMA to load firmware modules to the DSP memory blocks.
Two DMA engines are supported, DesignWare and Intel MID.

ASoC: Intel: Add runtime module lookup API call

Add an API to allow quick lookup of runtime modules based on ID.

ASoC: Intel: Provide streams with dynamic module information

Remove the hard coded module paramaters and provide each module with
dynamically generated buffer information for scratch and persistent
buffers.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

A platform_driver does not need to set an owner, it will be populated by the
driver core.
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Improve the debug SNR by making the positional pointer debug more verbose.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

DSP can now support 4 channels in certain use cases.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Fix driver with correct formats.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org

The sst-haswell-pcm driver registers both a snd_soc_component and a
snd_soc_platform and expects that the DAPM widgets for the DAIs registered by
component are added to the DAPM context of the platform. This requires us to
have a hack in the ASoC core which does so. Moving the DAPM elements over to
the component allows us to remove this hack.

While we are at it also move the controls over to the component. The controls
don't need the platform for anything other than snd_soc_platform_get_drvdata(),
this can easily be replaced by snd_soc_component_get_drvdata(). As the long
term goal is to register only a single component this is a step in the right
direction.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NMark Brown <broonie@linaro.org>

Resource allocations should happen in driver probe callback rather than in
snd_soc_platform probe functions. Especially if the resource is device
managed. The snd_soc_* probe/remove functions are mainly intended to be used
for things that require the component to be already bound to a card.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NMark Brown <broonie@linaro.org>

This change removes unsupported formats from System,
Capture and Loopback FE DAIs.
Also it fixes S24_LE support on all DAIs.
While at this fix 24 bit flag for BYT as well.
Signed-off-by: NJie Yang <yang.jie@intel.com>
Signed-off-by: NMark Brown <broonie@linaro.org>
Cc: stable@vger.kernel.org

A few files contain duplicate headers. This patch removes the second entry of
duplicate in each file under question.

There is no functional changes.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Cc: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@linaro.org>