ASoC: Fix for v5.17
This is rather late and at this point I'm expecting it to get merged in
the merge window rather than as a fix but if we get a -rc8 it's a small,
driver specific fix which should be fine to send.
ASoC: Updates for v5.18
Quite a quiet release for ASoC, lots of work on drivers and platforms
but nothing too groundbreaking but not much on the core itself:
- Start of moving SoF to support multiple IPC mechanisms.
- Use of NHLT ACPI table to reduce the amount of quirking required for
Intel systems.
- Some building blocks for use in forthcoming Intel AVS driver for
legacy Intel DSP firmwares.
- Support for AMD PDM, Atmel PDMC, Awinic AW8738, i.MX cards with
TLV320AIC31xx, Intel machines with CS35L41 and ESSX8336, Mediatek
MT8181 wideband bluetooth, nVidia Tegra234, Qualcomm SC7280, Renesas
RZ/V2L, Texas Instruments TAS585M
The parts supported by this driver can have product-specific
firmware and tunings files. Typically these have been used on
embedded systems where the manufacturer is responsible for
installing the correct product-specific firmware files into
/lib/firmware. However, the linux-firmware repository places all
available firmwares into /lib/firmware and it is up to the driver to
select the correct product-specific firmware from that directory.
For example a product containing four smart amplifiers may provide
firmware specific for that product and each of the amplifiers may
have coefficient files containing tunings for their placement in the
mechanical design.
This change extends firmware (wmfw) and coefficient (bin) filenames
to be of the general form:
<cirrus/>part-dspN-fwtype<-system_name<-asoc_component_prefix>>.type
Where the cirrus subdirectory, system_name and asoc_component_prefix
are optional.
New files will be placed in the cirrus subdirectory to avoid
polluting the main /lib/firmware/ location. The generic name must be
searched in /lib/firmware before /lib/firmware/cirrus so that a
generic file in the new location does not override existing
product-specific files in the legacy location.
The search order for firmware files is:
- cirrus/part-dspN-fwtype-system_name-asoc_component_prefix.wmfw
- cirrus/part-dspN-fwtype-system_name.wmfw
- part-dspN-fwtype.wmfw
- cirrus/part-dspN-fwtype.wmfw
- Qualifications are added to the filename so that rightwards is more
specific.
- The system_name is provided by the codec driver.
- The asoc_component_prefix is used to identify tunings for individual
parts because it would already exist to disambiguate the controls
and it makes it obvious which firmware file applies to which device.
The optional coefficient file must have the same filename
construction as the discovered wmfw except:
- where the wmfw has only system_name then the bin file can
optionally include the asoc_component_prefix. This is to allow a
common wmfw for all amps but separate tunings per amp.
Signed-off-by: Simon Trimmer <simont@opensource.cirrus.com>
Acked-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20220303155016.122125-1-simont@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Newer firmwares will support compressed buffers that may or may not
exist, for example debugging streams. Update the driver to make a
compressed stream optional. A warning will still be generated at DSP
boot time and opening the stream will fail if the compressed buffer in
question does not exist, however the DSP can still be booted and other
features used.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20220210172053.22782-2-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
When parsing the compressed stream the whole buffer descriptor is
now read in a single cs_dsp_coeff_read_ctrl; on older firmwares
this descriptor is just 4 bytes but on more modern firmwares it is
24 bytes. The current code reads the full 24 bytes regardless, this
was working but reading junk for the last 20 bytes. However commit
f444da38ac ("firmware: cs_dsp: Add offset to cs_dsp read/write")
added a size check into cs_dsp_coeff_read_ctrl, causing the older
firmwares to now return an error.
Update the code to only read the amount of data appropriate for
the firmware loaded.
Fixes: 04ae085967 ("ASoC: wm_adsp: Switch to using wm_coeff_read_ctrl for compressed buffers")
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20220210172053.22782-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
In the case a device can support retaining the firmware memory across
low power states it is useful for the preloader widget to only power up
whilst actually loading/unloading the core, as opposed to the normal
operation where the widget is powered for the entire time a firmware is
preloaded onto the core. Add support for this mode and a flag to enable
it.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20220105113026.18955-7-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
wm_adsp originally provided firmware loading on some audio DSP and was
implemented as an ASoC codec driver. However, the firmware loading now
covers a wider range of DSP cores and peripherals containing them,
beyond just audio. So it needs to be available to non-audio drivers. All
the core firmware loading support has been moved into a new driver
cs_dsp, leaving only the ASoC-specific parts in wm_adsp.
Signed-off-by: Simon Trimmer <simont@opensource.cirrus.com>
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210913160057.103842-17-simont@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
When parsing a compressed buffer from the firmware the driver currently
open codes reading the firmware coefficient containing the buffer
description. Improve this slightly by using the coefficient read
functions already provided by the wm_adsp driver. It is worth noting
this change requires the running variable to be set before
wm_adsp_buffer_init is called, however this is safe, since its all still
under the power lock and nothing in the compressed code gates itself on
running.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Signed-off-by: Simon Trimmer <simont@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210913160057.103842-4-simont@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch reverts commit acbf58e530 ("ASoC: wm_adsp: Let
soc_cleanup_component_debugfs remove debugfs"), and adds an
alternate solution to the issue. That patch removes the call to
debugfs_remove_recursive, which cleans up the DSPs debugfs. The
intention was to avoid an unbinding issue on an out of tree
driver/platform.
The issue with the patch is it means the driver no longer cleans up
its own debugfs, instead relying on ASoC to remove recurive on the
parent debugfs node. This is conceptually rather unclean, but also it
would prevent DSPs being added/removed independently of ASoC and soon
we are going to be upstreaming some non-audio parts with these DSPs,
which will require this.
Finally, it seems the issue on the platform is a result of the
wm_adsp2_cleanup_debugfs getting called twice. This is very likely a
problem on the platform side and will be resolved there. But in the mean
time make the code a little more robust to such issues, and again
conceptually a bit nicer, but clearing the debugfs_root variable in the
DSP structure when the debugfs is removed.
Fixes: acbf58e530 ("ASoC: wm_adsp: Let soc_cleanup_component_debugfs remove debugfs"
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210824101552.1119-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
When wm_coeff_tlv_get was updated it was accidentally switch to the _raw
version of the helper causing it to ignore the current DSP state it
should be checking. Switch the code back to the correct helper so that
users can't read the controls when they arn't available.
Fixes: 73ecf1a673 ("ASoC: wm_adsp: Correct cache handling of new kernel control API")
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210626155941.12251-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
wm_adsp_read_data_word() used if (ret) to check for an error from
wm_adsp_read_raw_data_block(). While this is perfectly valid,
wm_adsp_read_raw_data_block() itself uses if (ret < 0) and three
calls to wm_adsp_read_data_word() also use if (ret < 0).
This creates an error check chain like this:
1st) if (ret < 0) return ret;
2nd) if (ret) return ret;
3rd) if (ret < 0) ...
This can confuse the compiler into thinking that there are possible
returns > 0 from the middle if() that are not handled by the final
if(). If this was true it would lead to using uninitialized variables
later in the outer function.
Fix this by changing the test in wm_adsp_read_data_word() to be
if (ret < 0).
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210111133825.8758-1-rf@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Sparse will complain about trying to convert between values declared
as snd_ctl_elem_type_t and other types. This patch converts to
consistently use snd_ctl_elem_type_t for control type values. A __force
cast is needed in a couple of cases where the control type value is
parsed out of a DSP data block.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20201230172427.13865-2-rf@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This fixes some minor cases where u32 or unsigned int types were used
to store big-endian data, and __be32 types used to store both big-endian
and cpu-endian data. This was producing sparse warnings.
Most cases resulted from using the same variable to hold the big-endian
value and its converted cpu-endian value. These can be simply fixed by
introducing another local variable, or avoiding storing the intermediate
value back into the original variable.
One special case is the raw_buf used in the compressed streams to transfer
data from DSP to user-side. The endian conversion happens in-place (as
there's no point introducing another buffer) so a cast to (__be32 *) is
added when passing it to wm_adsp_read_raw_data_block().
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20201230172427.13865-1-rf@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
As the register map is 16-bit or 32-bit big-endian, the 24-bit
DSP words appear padded and with the bytes swapped. When reading a
raw stream of bytes, the pad bytes must be removed and the data bytes
swapped back to their original order.
The previous implementation of this assumed that the be32_to_cpu() in
wm_adsp_read_data_block() would swap back to little-endian. But this is
obviously only true on a little-endian CPU. It also made two walks
through the data, once to endian-swap and again to strip the pad bytes.
This patch re-works the code so that the endian-swap and unpad are done
together in a single walk, and it is not dependent on the endianness of
the CPU. The data_word_size argument to wm_adsp_remove_padding() has been
dropped because currently this is always 3.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20201216112512.26503-1-rf@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
A call to wm_adsp_write_ctl() could cause a kernel crash if it
does not retrieve a valid kcontrol from snd_soc_card_get_kcontrol().
This can happen due to a missing control name prefix. Then,
snd_ctl_notify() crashes when it tries to use the id field.
Modified wm_adsp_write_ctl() to incorporate the name_prefix (if applicable)
such that it is able to retrieve a valid id field from the kcontrol
once the platform has booted.
Fixes: eb65ccdb08 ("ASoC: wm_adsp: Expose mixer control API")
Signed-off-by: Adam Brickman <Adam.Brickman@cirrus.com>
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20201001152425.8590-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
