ASoC: Fixes for v5.13
A collection of fixes that have come in since the merge window, mainly
device specific things. The fixes to the generic cards from
Morimoto-san are handling regressions that were introduced in the merge
window on at least the Kontron sl28-var3-ads2.
NXP's TFA98xx (now part of Goodix) are fairly popular speaker amplifiers
used in many smartphones and tablets. Most of them are sold as "smart
amplifiers" with built-in "CoolFlux DSP" that is used for volume control,
plus a "sophisticated speaker-boost and protection algorithm".
Unfortunately, they are also almost entirely undocumented. The short
datasheets (e.g. [1] for TFA9897) describe the available features,
but do not provide any information about the registers or how to use
the "CoolFlux DSP".
The amplifiers are most often configured through proprietary userspace
libraries. There are also some (rather complex) kernel drivers (e.g. [2])
but even those rely on obscure firmware blobs for configuration (so-called
"containers"). They seem to contain different "profiles" with tuned speaker
settings, sample rates and volume steps (which would be better exposed
as separate ALSA mixers).
The format of the firmware files seems to have changed a lot over the time,
so it's not even possible to simply re-use the firmware originally provided
by the vendor.
Overall, it seems close to impossible to develop a proper mainline driver
for these amplifiers that could make proper use of the built-in DSP.
This commit implements a compromise: At least the TFA1 family of the
TFA98xx amplifiers (usually called TFA989x) provide a way to *bypass*
the DSP using a special register sequence. The register sequence can be
found in similar variations in the kernel drivers from lots of vendors
e.g. in [3] and was probably mainly used for factory testing.
With the DSP bypassed, the amplifier acts mostly like a dumb standard
speaker amplifier, without (hardware) volume control. However, the setup
is much simpler and it works without any obscure firmware.
This driver implements the DSP bypass combined with chip-specific
initialization sequences adapted from [2]. Only TFA9895 is supported in
this initial commit. Except for the lack of volume control I can not hear
any difference with or without the DSP, it works just fine.
This driver allows the speaker to work on mainline Linux running on the
Samsung Galaxy A3/A5 (2015) [TFA9895] and Alcatel Idol 3 [TFA9897].
TFA9897 support will be added in separate patch set later.
[1]: https://product.goodix.com/en/docview/TFA9897%20SDS_Rev.3.1?objectId=47&objectType=document&version=78
[2]: https://source.codeaurora.org/external/mas/tfa98xx
[3]: 57b5050e34/sound/soc/codecs/tfa98xx.c (L1422-L1462)
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Link: https://lore.kernel.org/r/20210513104129.36583-2-stephan@gerhold.net
Signed-off-by: Mark Brown <broonie@kernel.org>
cs43130_probe() does not do any valid error checking of things it
initializes, OR what it does, it does not unwind properly if there are
errors.
Fix this up by moving the sysfs files to an attribute group so the
driver core will correctly add/remove them all at once and handle errors
with them, and correctly check for creating a new workqueue and
unwinding if that fails.
Cc: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20210503115736.2104747-58-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This reverts commit a2be42f18d.
Because of recent interactions with developers from @umn.edu, all
commits from them have been recently re-reviewed to ensure if they were
correct or not.
Upon review, this commit was found to be incorrect for the reasons
below, so it must be reverted. It will be fixed up "correctly" in a
later kernel change.
The original patch here is not correct, sysfs files that were created
are not unwound.
Cc: Kangjie Lu <kjlu@umn.edu>
Cc: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20210503115736.2104747-57-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Check for return value from various snd_soc_dapm_* calls, as many of
them can return errors and this should be handled. Also, reintroduce
the allocation failure check for rt5645->eq_param as well. Make all
areas where return values are checked lead to the end of the function
in the case of an error. Finally, introduce a comment explaining how
resources here are actually eventually cleaned up by the caller.
Cc: Mark Brown <broonie@kernel.org>
Signed-off-by: Phillip Potter <phil@philpotter.co.uk>
Link: https://lore.kernel.org/r/20210503115736.2104747-56-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This reverts commit 51dd97d1df.
Because of recent interactions with developers from @umn.edu, all
commits from them have been recently re-reviewed to ensure if they were
correct or not.
Upon review, this commit was found to be incorrect for the reasons
below, so it must be reverted. It will be fixed up "correctly" in a
later kernel change.
Lots of things seem to be still allocated here and must be properly
cleaned up if an error happens here.
Cc: Kangjie Lu <kjlu@umn.edu>
Cc: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20210503115736.2104747-55-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
HSBIAS_SENSE_EN configures HSBIAS output current sense through
the external 2.21-k resistor. HSBIAS_SENSE is hardware feature to reduce
the potential pop noise during the headset plug out slowly. But on some
platforms ESD voltage will affect it causing test to fail, especially
with CTIA headset type. For different hardware setups, a designer might
want to tweak default behavior.
Signed-off-by: Vitaly Rodionov <vitalyr@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210511145220.125760-1-vitalyr@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
cs42l42 does not support standard burst transfers so the use_single_read
and use_single_write flags must be set in the regmap config.
Because of this bug, the patch:
commit 0a0eb567e1 ("ASoC: cs42l42: Minor error paths fixups")
broke cs42l42 probe() because without the use_single_* flags it causes
regmap to issue a burst read.
However, the missing use_single_* could cause problems anyway because the
regmap cache can attempt burst transfers if these flags are not set.
Fixes: 2c394ca796 ("ASoC: Add support for CS42L42 codec")
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Acked-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210511132855.27159-1-rf@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Pierre requested I have a look at some cppcheck warnings in the cs42l42
driver, since it is reassigning the ret variable without ever checking
the result. Looking a bit more broadly this happens in quite a few
legacy Cirrus parts, as they all use the same process to read the ID,
factor out a small helper so they can all share the same code. Whilst
in there fix up a couple of other trivial error path issues as well.
Thanks,
Charles
Charles Keepax (10):
ASoC: cirrus: Add helper function for reading the device ID
ASoC: cs35l32: Minor error paths fixups
ASoC: cs35l33: Minor error paths fixups
ASoC: cs35l34: Minor error paths fixups
ASoC: cs35l35: Minor error paths fixups
ASoC: cs35l35: Correct errata handling
ASoC: cs42l42: Minor error paths fixups
ASoC: cs42l73: Minor error paths fixups
ASoC: cs43130: Minor error paths fixups
ASoC: cs53l30: Minor error paths fixups
sound/soc/codecs/cirrus_legacy.h | 21 +++++++++++++++++++++
sound/soc/codecs/cs35l32.c | 34 ++++++++++++++++++----------------
sound/soc/codecs/cs35l33.c | 15 +++++++++------
sound/soc/codecs/cs35l34.c | 39 ++++++++++++++++++++++-----------------
sound/soc/codecs/cs35l35.c | 21 ++++++++++-----------
sound/soc/codecs/cs35l35.h | 1 +
sound/soc/codecs/cs42l42.c | 18 ++++++++----------
sound/soc/codecs/cs42l73.c | 30 +++++++++++++++++-------------
sound/soc/codecs/cs43130.c | 31 +++++++++++++++++++------------
sound/soc/codecs/cs53l30.c | 22 +++++++++++-----------
10 files changed, 136 insertions(+), 96 deletions(-)
create mode 100644 sound/soc/codecs/cirrus_legacy.h
--
2.11.0
Currently the check of errata_chk will always evaluate to false since
the values tested don't come under the mask used. A shift of the field
is missing, add this. Also there is an error in the values tested, they
don't match the comment and the value 0x3 is not a valid value for the
field in question. Update the value to match the comment.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210510131357.17170-7-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Many of the older Cirrus devices share very similar code for reading the
device ID, and frequently this code is generating cppcheck warnings such
as:
sound/soc/codecs/cs42l42.c:1886:6: style: Variable 'ret' is reassigned
a value before the old one has been used. [redundantAssignment]
ret = regmap_read(cs42l42->regmap, CS42L42_DEVID_CD, ®);
Add a small helper function that older Cirrus devices can use to read
the device ID, which should help correct these issues.
Reported-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210510131357.17170-2-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Static analysis reports this problem
lpass-wsa-macro.c:1732:6: warning: Array subscript is undefined
if (wsa->ec_hq[ec_tx]) {
^~~~~~~~~~~~~~~~~
The happens because 'ec_tx' is never initialized and there is
no default in switch statement that sets ec_tx. Add a default
case that returns an error before the array is accessed.
Signed-off-by: Tom Rix <trix@redhat.com>
Reviewed-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Link: https://lore.kernel.org/r/20210430142117.3272772-1-trix@redhat.com
Signed-off-by: Mark Brown <broonie@kernel.org>
