Merge series from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
This patchset improves the pm_runtime behavior in rare corner cases
identified by the Intel CI in the last 6 months.
a) in stress-tests, it's not uncommon to see the following type of
warnings when the codec reports as ATTACHED
"rt711 sdw:0:025d:0711:00: runtime PM trying to activate child device
sdw:0:025d:0711:00 but parent (sdw-master-0) is not active"
This warning was not correlated with any functional issue, but it
exposed a design issue on when to enable pm_runtime. The recommended
practice in the pm_runtime documentation is to keep the devices in
'suspended' mode and mark them as 'active' when they are really
functional.
b) enabling pm_runtime when the codec reports as ATTACHED also creates
a problematic case when the ASoC pm_runtime_get_sync() will silently
fail due to the -EACCESS error handling. This can happen when playback
starts before the codec is enumerated.
This patchset modifies the initial stages so that codecs are
pm_runtime enabled in the .probe() callback, but become pm_runtime
'active' only when they report present. This is better aligned with
the design of the pm_runtime helpers and improved CI results
significantly.
This patchset modifies all existing SoundWire codecs (except Qualcomm
ones), but the pattern of changes is exactly the same in all patches.
The maple tree register cache is based on a much more modern data structure
than the rbtree cache and makes optimisation choices which are probably
more appropriate for modern systems than those made by the rbtree cache. In
v6.5 it has also acquired the ability to generate multi-register writes in
sync operations, bringing performance up to parity with the rbtree cache
there.
Update the jz4740 driver to use the more modern data structure.
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20230712-asoc-jz-maple-v1-1-3f745adf96e5@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-17-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-16-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-15-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-14-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-13-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-12-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-11-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-10-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-9-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-8-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch suggests enabling pm_runtime during the probe, but marking the
device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
The changes are directly inspired by RT711 and RT711-sdca changes.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-7-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
In stress cases involving module insertion/removal followed by
playback/capture, it can happen that capture/playback is started
before the codec enumeration completes.
The codec driver registers its components with the ASoC framework
during the probe stage, so there is currently no way for the card
creation to wait for the codec enumeration/initialization to complete.
In addition, when the capture/playback starts, the ASoC framework uses
pm_runtime_get_sync() to properly refcount and power-manage
devices. This is problematic in the SoundWire case because pm_runtime
is enabled during the enumeration/initialization stage, so
pm_runtime_get_sync() will return -EACCESS which is
ignored. Additional errors will happen when setting the pm_runtime
status as 'active' because the parent is not properly resumed,
resulting in an error such as:
"rt711 sdw:0:025d:0711:00: runtime PM trying to activate child device
sdw:0:025d:0711:00 but parent (sdw-master-0) is not active"
This patch suggests enabling pm_runtime during the probe, but marking
the device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
Closes: https://github.com/thesofproject/linux/issues/4328
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-6-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
In stress cases involving module insertion/removal followed by
playback/capture, it can happen that capture/playback is started
before the codec enumeration completes.
The codec driver registers its components with the ASoC framework
during the probe stage, so there is currently no way for the card
creation to wait for the codec enumeration/initialization to complete.
In addition, when the capture/playback starts, the ASoC framework uses
pm_runtime_get_sync() to properly refcount and power-manage
devices. This is problematic in the SoundWire case because pm_runtime
is enabled during the enumeration/initialization stage, so
pm_runtime_get_sync() will return -EACCESS which is
ignored. Additional errors will happen when setting the pm_runtime
status as 'active' because the parent is not properly resumed,
resulting in an error such as:
"rt711 sdw:0:025d:0711:00: runtime PM trying to activate child device
sdw:0:025d:0711:00 but parent (sdw-master-0) is not active"
This patch suggests enabling pm_runtime during the probe, but marking
the device as 'active' only after it is enumerated. That will force a
dependency between the card and the codec, pm_runtime_get_sync() will
have to wait for the codec device to resume and hence implicitly wait
for the enumeration/initialization to be completed. In the nominal
case where the codec device is already active the get_sync() would
only perform a ref-count increase.
Closes: https://github.com/thesofproject/linux/issues/4328
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-5-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RT5682 needs specific attention: there are two regmap in
rt5682_priv struct, one is sdw_regmap which is for IO transfer, and
the other is used for registers control.
We need to set both regmaps when we set cache only. Because if we set
rt5682->sdw_regmap only, rt5682->regmap won't set/get the right value
when it call regmap_write/read(rt5682->sdw_regmap, ...). If we set
rt5682->regmap only, regmap_write(rt5682->sdw_regmap, ...) is used
in rt5682_clock_config which will be called by the ..bus_config ops.
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20230802153629.53576-4-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Merge series from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
Lots of small cleanups coming from Bard Liao and Charles Keepax for
SoundWire platforms, and minor additions for RVPs and Chromebooks.
Merge series from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
GCC11 provides an '-fanalyzer' static analysis option which does not
provide too many false-positives. This patch cleans-up known
problematic code paths to help enable this capability in CI. We've
used this for about a month already.
