ASoC: amd: vangogh: Add check for acp config flags in vangogh platform
We have SOF and generic ACP support enabled for Vangogh platform
on some machines. Since we have same PCI id used for probing,
add check for machine configuration flag to avoid conflict with
newer pci drivers. Such machine flag has been initialized via
dmi match on few Vangogh based machines. If no flag is
specified probe and register older platform device.
Series removes the pm_runtime.h inclusion in files where
APIs exported though pm_runtime.h are not used. In case
of files that make use of pm.h which comes form pm_runtime.h
added patch 2/2.
Trevor Wu [Mon, 8 May 2023 07:15:32 +0000 (15:15 +0800)]
ASoC: mediatek: mt6359: update route for lineout mux
Originally, lineout playback source can only be DAC_3RD. Some SoC
masters only support stereo MTKAIF outputs, so lineout path can't be
used in such case.
MTKAIF connections are as follows.
MOSI0 -> DAC_L
MOSI1 -> DAC_R
MOSI2 -> DAC_3rd
In the patch, lineout playback source can be chosen between DAC_L and
DAC_3rd, so sound can be outputted via lineout even though SoC only
supports stereo MTKAIF outputs.
Trevor Wu [Mon, 8 May 2023 07:15:30 +0000 (15:15 +0800)]
ASoC: mediatek: mt6359: fix kselftest error of playback gain
kselftest tries to read/write the default value. The default register
value of playback gain is 0x1F(mute), but max gain we specified is 0x12.
The range of the control is 0x0~0x12 and mute(0x1F) is only used in the
driver internally. To solve the problem, implement a new callback
mt6359_get_playback_volsw to report user configured volume instead of
the register value.
In addition, update max of "Headset Volume" to 0x12, so it can match the
maximum seen on latest data sheet.
Trevor Wu [Mon, 8 May 2023 07:15:29 +0000 (15:15 +0800)]
ASoC: mediatek: mt6359: add supply for MTKAIF
There are three output data pins MISO0, MISO1 and MISO2 for mt6359.
UL_SRC should be enabled when MISO0 or MISO1 is used, and UL_SRC_34
should be enabled when MISO2 is used.
Paul Olaru [Wed, 3 May 2023 08:10:49 +0000 (11:10 +0300)]
ASoC: sof: Improve sof_ipc4_bytes_ext_put function
The function is improved in the way that if the firmware returns a
validation error on the newly sent bytes, then the kernel will
automatically restore to the old bytes value for a given kcontrol.
This way, if the firmware rejects a data blob then the kernel will also
reject it, instead of saving it for the next suspend/resume cycle. The
old behaviour is that the kernel would save it anyway and on next
firmware boot it would apply the previously-rejected configuration,
leading to errors during playback.
Additionally, the function also saves previously validated
configurations, so that if the firmware does end up rejecting a new
bytes value the kernel can send an old, previously-valid configuration.
Paul Olaru [Wed, 3 May 2023 08:10:48 +0000 (11:10 +0300)]
ASoC: sof: Improve sof_ipc3_bytes_ext_put function
The function is improved in the way that if the firmware returns a
validation error on the newly sent bytes, then the kernel will
automatically restore to the old bytes value for a given kcontrol.
This way, if the firmware rejects a data blob then the kernel will also
reject it, instead of saving it for the next suspend/resume cycle. The
old behaviour is that the kernel would save it anyway and on next
firmware boot it would apply the previously-rejected configuration,
leading to errors during playback.
Additionally, the function also saves previously validated
configurations, so that if the firmware does end up rejecting a new
bytes value the kernel can send an old, previously-valid configuration.
ASoC: SOF: Intel: hda-dai: Fix locking in hda_ipc4_pre_trigger()
hda_ipc4_pre_trigger() has two issues:
1. In the default case, we are returning without unlocking the mutex.
2. In case SNDRV_PCM_TRIGGER_STOP: when ret is less than zero it goes
to out, unlocks but returns zero instead of a negative value.
Fix this by changing the final return value to 'ret' instead of zero,
and initialize 'ret' to zero in the start of the function.
Yang Li [Tue, 16 May 2023 08:11:16 +0000 (16:11 +0800)]
ASoC: SOF: ipc4-topology: Fix an unsigned comparison which can never be negative
The return value from the call to sof_ipc4_get_valid_bits() is int.
However, the return value is being assigned to an unsigned
int variable 'out_ref_valid_bits', so making it an int.
Eliminate the following warning:
./sound/soc/sof/ipc4-topology.c:1537:6-24: WARNING: Unsigned expression compared with zero: out_ref_valid_bits < 0
Firmware may do context saving before powering off primary core, so driver
needs to send ipc msg by set_core_state. In IPC4 path, firmware needs to
save current context to IMR before powering off primary core. Firmware
does nothing for set_core_state message in IPC3 path. So IPC4 and IPC3
can share the same operation sequence.
Rander Wang [Tue, 23 May 2023 10:32:16 +0000 (13:32 +0300)]
ASoC: SOF: Intel: mtl: add core_get & put support on MeterLake platforms
In core_get case, driver can power up primary core and don't need to send
ipc message to fw. Non-primary core should be powered up by fw with ipc
message.
In core_put case, driver should first send ipc message to fw to disable dsp
core then power down primary core if the target is primary core.
Claudiu Beznea [Wed, 17 May 2023 09:49:03 +0000 (12:49 +0300)]
ASoC: use pm.h instead of runtime_pm.h
Do not include pm_runtime.h header in files where runtime PM support is
not implemented. Use pm.h instead as suspend to RAM specific
implementation is available.
David Lin [Tue, 23 May 2023 08:33:04 +0000 (16:33 +0800)]
ASoC: nau8825: Add pre-charge actions for input
Adding pre-charge actions to make FEPGA power stable faster. It
improve the recording quality at the beginning. Thus, it is also
meaningfully to decrease the final adc delay time.
Series of patches cleaning up error messages when loading topology. In
few places instead of logging in place of failure message is logged in
caller. Additionally there are places where both caller and failing
function log error, leading to unnecessary logs. Clean all of the above
up.
The MOD_INIT_INSTANCE message contains a CPC (Cycles Per Chunk/processing unit)
parameter.
This CPC value is used by the firmware to calculate the total cycles needed by
the enabled module instances and based on this it can decide to set the
frequency of the DSP core(s).
The manifest section of the firmware image contains a module configuration
section, where a per module table of configurations are listed with measured
CPC values as triplet of IBS/IBS/CPC (Input/Output buffer size - corresponding
to the selected audio format).
In case the CPC value is 0 (missing from the manifest or the
configuration cannot be matched) the firmware will force the DSP cores
to maximum speed to avoid audio glitches due to starvation. In these
cases the kernel will print a warning message to let the SOF developers
know about the gap and provide information to correct it with a firmware
update.
ASoC: topology: Do not split message string on multiple lines
Kernel coding guidelines recommend to not split string unnecessarily.
While at it adapt the other print present in the function to 100
characters line limit.
Peter Ujfalusi [Mon, 22 May 2023 10:13:12 +0000 (13:13 +0300)]
ASoC: SOF: ipc4-topology: Do not use the CPC value from topology
Stop parsing the CPC value from topology to module_base_cfg.
The CPC value is only set for few modules in topology which makes the CPC
handling inconsistent.
Peter Ujfalusi [Mon, 22 May 2023 10:13:11 +0000 (13:13 +0300)]
ASoC: SOF: ipc4-topology: Rename sof_ipc4_update_pipeline_mem_usage() to be generic
Rename sof_ipc4_update_pipeline_mem_usage() to
sof_ipc4_update_resource_usage() in order to be re-usable for generic
resource storage, calculation of a module, like CPC adjustment.
First two patches are bugfixes.
Third patch skips the overhead of rebooting the amp after applying
firmware files when we know that it isn't necessary.
Simon Trimmer [Thu, 18 May 2023 15:02:50 +0000 (16:02 +0100)]
ASoC: cs35l56: In secure mode skip SHUTDOWN and RESET around fw download
If the device is in secure mode it's unnecessary to send a SHUTDOWN and
SYSTEM_RESET around the firmware download. It could only be patching
insecure tunings. A tuning patch doesn't need a SHUTDOWN and only needs
a REINIT afterwards. This will reduce the overhead of exiting system
suspend in secure mode.
Simon Trimmer [Thu, 18 May 2023 15:02:49 +0000 (16:02 +0100)]
ASoC: cs35l56: sdw_write_no_pm() should be performed under a pm_runtime request
SoundWire bus accesses must be performed under the guard of a pm_runtime
request, in this case the write was being performed just after the
request had been put() and so the bus could not be guaranteed to be
available.
Simon Trimmer [Thu, 18 May 2023 15:02:48 +0000 (16:02 +0100)]
ASoC: cs35l56: Move DSP part string generation so that it is done only once
Each time we go through dsp_work() it does a devm_kasprintf() to
allocate memory to hold the part name string. It's not strictly a memory
leak because devm will free it all if the driver is removed. But we keep
allocating more and more memory to hold the same string.
Move the allocation so that it is performed after the version and
secured state information is gathered and handle allocation errors.
The Google Chameleon v3 is a device made for testing audio and video
paths of other devices. This patchset adds support for ASoC audio on
this device. It has two audio sources: HDMI audio from the it68051 chip
(RX only), and analog audio from the ssm2603 chip (RX and TX).
The patchset adds the ASoC platform and codec drivers.
Many SoundWire CODEC drivers store the device status in a member
variable in the driver data but never reference this, and in any case
the SoundWire core stores this information for drivers so it would be
redundant even if used.
ASoC: mediatek: mt8192-mt6359: Remove " Jack" from Headphone pin name
Function jack_kctl_name_gen() will remove the redundant " Jack" from
the name, if present, and then it will add it back, so that all of
the controls are named "(pin-name) Jack".
Remove " Jack" from the Headphone pin name to spare some CPU cycles.
Improve the logic to account for the current restrictions in topology while
making it extensible for future topology modifications.
The current topology definitions assume that input/output formats come in pairs.
For example even if there's only 1 output format for a module, we add 3 output
formats to match that of the input format count with the same parameters.
This is unnecessary but we have to deal with it until the topologies are
modified. Additionally, choosing the input/output audio format should
depend only on the pipeline params or the runtime FE hw_params depending
on where the module is in the pipeline.
This series modifies the logic for selection based on this and removes
unnecessary dependencies between the input and output formats.
The series of patches consists of four major changes.
First, remove redundant supply for ADDA DAI dirver. Second, revise ETDM
control including APLL dynamic switch via DAPM, so APLL can be enabled
when it is really required. Third, update AFE probe function. Bus
protection change was dropped at the previous patch because the dependent
change was not accepted at that time. Finally, correct some binding errors
and add required clocks.
Do not reset pipelines during the stop/suspend triggers in the BE DAI
ops as the BE DAI pipeline needs to be left in the PAUSED state. It should
only be reset during hw_free. This simplification is already done for
the FE pipelines and the DAI trigger only toggles the states between
PAUSED and RUNNING.
The HDAudio support has not changed much since CometLake/2019: the
code was reshuffled for IPC4 support and we used hdac_hda for external
and Display Audio codec, but the hda-dai code was only used for
HDAudio codecs.
In the LunarLake architecture, all endpoints (HDaudio, SSP, DMIC,
SoundWire) are handled with the HDaudio DMA, which requires us to
revisit the definitions of HDA_LINK, and remove the mutual exclusion
between NOCODEC and HDA_LINK: we do want the ability to test SSP/DMIC
in NOCODEC mode even with an HDA DMA.
This code change exposed a number of issues, with a useless .prepare
callback, a DAI number mismatch and the need to support SoundWire
which is handled by a different component in
drivers/soundwire/intel_ace2.c.
ASoC: SOF: ipc4-pcm: reset all pipelines during FE DAI hw_free
Do not reset pipelines during the stop/suspend triggers in the BE DAI
ops as the BE DAI pipeline needs to be left in the PAUSED state. It should
only be reset during hw_free. This simplification is already done for
the FE pipelines and the DAI trigger only toggles the states between
PAUSED and RUNNING.
But because the FE DAI hw_free is invoked first and all the pipelines are
freed during this op, we need to make sure that the BE DAI pipeline also
gets reset before it is freed. So do not skip the pipelines that have the
skip_during_fe_trigger flag set when resetting pipelines.
Also, because the pipeline state changes are split between the FE and BE
DAI ops now, protect the BE DAI pipeline state changes with the
pipeline_state_mutex as well.
ASoC: SOF: Intel: hda-dai-ops: Split the get_hext_stream() op for IPC4
Introduce a separate op implementation for get_hext_stream() for IPC4.
This op will also be used to set the skip_during_fe_trigger flag for the
BE DAI pipeline. With this change, we can remove the flag setting in
sof_ipc4_dai_config() which will further simplify support for
DMIC/SSP/Soundwire in the LunarLake platform.
ASoC: SOF: Intel: remove mutual exclusion between NOCODEC and HDA_LINK
The nocodec mode served two purposes so far
a) generate a test driver for DMIC/SSP without any codec connected
b) make sure the use of snd_hdac_ libraries was contained
b) is no longer an option for LunarLake, the HDaudio DMA is used for
DMIC/SSP and the HDA_LINK option needs to be enabled.
ASoC: SOF: Intel: hda-dai: use HDA_LINK instead of HDA_AUDIO_CODEC
For LunarLake support, we will have to use HDAudio DMA-based DAIs even
for SSP/DMIC/SoundWire. That's completely different to the
HDA_AUDIO_CODEC, the DAI ops deal with DMA configuration and that can
happen in the absence of any HDAudio codec.
ASoC: SOF: Intel: hda-dai: mark functions as __maybe_unused
hda_dai_hw_params, hda_dai_trigger(), hda_dai_hw_free are currently
only used for HDaudio codec support, but will be reused for
SSP/DMIC/SoundWire in the LunarLake/ACE2.x case. To avoid 'defined but
not used' errors or added complexity in Kconfig, mark all these
functions as __maybe_used.
When SSP/DMIC/SoundWire are added, some of these changes may be
reverted. For now this avoids compilation warnings.
ASoC: SOF: Intel: clarify initialization when HDA_AUDIO_CODEC is not used
For LunarLake support, we need to enable HDA_LINK but we also want the
ability to remove HDaudio codec support, e.g. for 'nocodec'
tests. This requires a small change in the bus initialization without
any codec-specific callbacks provided.
ASoC: SOF: Intel: hda-dai: remove use of cpu_dai->component drvdata
The existing code relies on conversions from cpu_dai to the sdev
structure pointer based on the cpu_dai component. This works fine for
HDaudio but will not work for SoundWire DAIs which are registered by a
different component. That's a problem preventing reuse of the HDaudio
DMA stream allocation for SoundWire DAIs starting with the LunarLake
platform.
This patch introduces a set of helpers to perform the conversion, and
an indirect way of retrieving the sdev pointer based on the
swidget->comp intermediate pointer.
The code has been cleaned-up multiple times, but while adding the new
abstractions for DMIC/SSP/SoundWire it appears that we don't really
need a specific sequence for .prepare, and we can reuse what
.hw_params already does.
ASoC: SOF: ipc4-topology: Modify input audio format selection logic
The current selection logic assumes that input and output formats always
come in pairs in topology. Handle this special case by checking if all
input formats are the same. And for the case where there are multiple
supported input audio formats, modify the selection logic to pick the
audio formats based on the reference params which is either the FE
hw_params or the pipeline params based on the type of module.
ASoC: SOF: ipc4-topology: Modify the output format selection logic
Modify the output format selection when there are multiple output
formats available to choose the one that matches the reference params.
The reference params depend on the type of module. In the case of
processing modules, the reference params are based on the selected input
audio format. This would be the case when a processing module does not
perform any format conversion during processing.
The only special case is the copier module. The copier module is capable
of format conversion but it is only used in the case when the output
is fixed to a single format. In the case of a module copier, when there are
multiple formats, the reference params is based on the selected input
params and the output format must match that of the selected input
format. In the case of host copier, the reference params should be
based on the input audio format for playback and the FE hw_params for
capture. In the case DAI copier, the reference params should be based on
the input audio format for capture and the FE hw_params for playback
when there is no format conversion in the pipeline from the host to the
DAI.
ASoC: SOF: ipc4-topology: Handle output format special case
The current topologies have input/output formats in pairs and even
though there are multiple output formats, they are all the same. Handle
this case as if there were only one format in topology. Also, add a check
for the number of output formats and reports errors where applicable.
ASoC: SOF: ipc4-topology: Move the call to init output format
In preparation for changing the logic for input/output format selection,
move the call to sof_ipc4_init_output_audio_fmt() into the individual
widget prepare ops.
projects / linux.git / log