kernel/linux.git/include/sound/cs35l56.h, branch linux-7.0.y

ASoC: cs35l56: Only patch ASP registers if the DAI is part of a DAIlink

2026-03-01T23:48:07+00:00

Move the ASP register patches to a separate struct and apply this from the ASP DAI probe() function so that the registers are only patched if the DAI is part of a DAI link. Some systems use the ASP as a special-purpose interconnect and on these systems the ASP registers are configured by a third party (the firmware, the BIOS, or another device using the amp's secondary host control interface). If the machine driver does not hook up the ASP DAI then the ASP registers must be omitted from the patch to prevent overwriting the third party configuration. If the machine driver includes the ASP DAI in a DAI link, this implies that the machine driver and higher components (such as alsa-ucm) are taking ownership of the ASP. In this case the ASP registers are patched to known defaults and the machine driver should configure the ASP. Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20260226110137.1664562-1-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Support for reading speaker ID from on-chip GPIOs

2026-02-05T16:58:11+00:00

Add support for using the state of pins on the amplifier to indicate the type of speaker fitted. Previously, where there were alternate speaker vendors, this was indicated using host CPU GPIOs. Some new Dell models use spare pins on the CS35L63 as GPIOs for the speaker ID detection. Cirrus-specific SDCA Disco properties provide a list of the pins to be used, and pull-up/down settings for the pads. This list is ordered, MSbit to LSbit. The code to set the firmware filename has been modified to check for using chip pins for speaker ID. The entire block of code to set firmware name has been moved out of cs35l56_component_probe() into its own function to make it easier to KUnit test. Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20260205164838.1611295-2-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Log a message if firmware is missing

2025-11-28T11:48:02+00:00

If the amp is still reporting FIRMWARE_MISSING after cs35l56_patch() has completed it is helpful to log a warning. After a complete firmware download the FIRMWARE_MISSING flag will be clear. If this isn't the case, the driver should log a message to report this. The amp can produce basic audio output without firmware, as a fallback, so this wasn't originally logged as a warning condition because the amp is still in an operational state - just not with full functionality. However, it was not at all obvious to an end user that anything is unusual. Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20251128112520.40067-1-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Allow restoring factory calibration through ALSA control

2025-11-11T13:52:49+00:00

Add an ALSA control (CAL_DATA) that can be used to restore amp calibration, instead of using debugfs. A readback control (CAL_DATA_RB) is also added for factory testing. On ChromeOS the process that restores amp calibration from NVRAM has limited permissions and cannot access debugfs. It requires an ALSA control that it can write the calibration blob into. ChromeOS also restricts access to ALSA controls, which avoids the risk of accidental or malicious overwriting of good calibration data with bad data. As this control is not needed for normal Linux-based distros it is a Kconfig option. A separate control, CAL_DATA_RB, provides a readback of the current calibration data, which could be either from a write to CAL_DATA or the result of factory production-line calibration. The write and read are intentionally separate controls to defeat "dumb" save-and-restore tools like alsa-restore that assume it is safe to save all control values and write them back in any order at some undefined future time. Such behavior carries the risk of restoring stale or bad data over the top of good data. Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20251111130850.513969-3-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Add control to read CAL_SET_STATUS

2025-11-11T13:52:48+00:00

Create an ALSA control to read the value of the firmware CAL_SET_STATUS control. This reports whether the firmware is using a calibration blob or the default calibration from the .bin file. The firmware only reports a valid value in this register while audio is actually playing and the internal PLL is locked to the audio clock. Otherwise it returns a status of "unknown". Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20251111130850.513969-2-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Alter error codes for calibration routine

2025-11-10T12:26:32+00:00

Adjust the error codes returned by the calibration debugfs files to provide a wider range of errors and make them more consistent. There is a limited number of standard errors and it's not always easy to find an error code that accurately describes what happened. Additionally, user code often uses strerror() or something similar to report a generic error description. The original calibration code used a limited set of errors to attempt to avoid user error strings that would be confusing or unclear on a file read/write. However, this restricts the ability to provide informative errors. This limited error range didn't help very much with debugging so it has been expanded, rather than worrying about what strerror() would return. The errors are now more consistent: ENXIO Calibration is not supported by the driver. EOPNOTSUPP The given calibration command is not supported. EBUSY Cannot calibrate because the amp is playing audio. ERANGE Calibration result was out-of-range. ETIMEDOUT Calibration did not complete. EMSGSIZE Blob written to cal_data is the wrong size. ENODATA No calibration data available to read from cal_data, or Blob written to cal_data does not contain calibration, or No calibration data available to save to UEFI. EIO General failure to communicate with the firmware, mainly indicating that firmware controls are missing. EINVAL Has its normal meaning that an invalid argument was passed. Signed-off-by: Richard Fitzgerald Link: https://patch.msgid.link/20251110114327.84370-1-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Add calibration command to store into UEFI

2025-10-27T14:07:49+00:00

Add a new command 'store_uefi' to the calibrate debugfs file. Writing this command will call cs_amp_set_efi_calibration_data() to save the new data into a UEFI variable. This is intended to be used after a successful factory calibration. On systems without UEFI the write to the debugfs file will return an error. Signed-off-by: Richard Fitzgerald Reviewed-by: Takashi Iwai Link: https://patch.msgid.link/20251021105022.1013685-10-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Add common code for factory calibration

2025-10-27T14:07:43+00:00

Add core code to support factory calibration. This can be used by both the ASoC and HDA drivers. This code consists of implementations of debugfs handlers for three debugfs files used to start factory calibration and read the results. This is not a full implementation of debugfs files. There are some requirements to synchronize with the rest of the amp driver, and the way this is done is significantly different between ASoC and HDA. Therefore cs35l56-shared.c provides the main part of the file handlers, but the files themselves are defined in the ASoC and HDA drivers with suitable handling before calling into this shared code. The cal_data file allows the calibration to be read and also for a previous calibration to be written (for systems where the storage is not something directly accessible to drivers, such as on filesystems). Code outside the kernel should treat the content of cal_data as an opaque blob, so the struct definition is not exported as a user API. Signed-off-by: Richard Fitzgerald Reviewed-by: Takashi Iwai Link: https://patch.msgid.link/20251021105022.1013685-4-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Read silicon ID during initialization and save it

2025-10-27T14:07:41+00:00

Read the silicon ID from the amp during one-time cs35l56_hw_init() and store it in struct cs35l56_base, instead of reading it from registers every time it is needed. Note that marking it non-volatile without a default in regmap isn't a suitable alternative because this causes regcache_sync() to always write the cached value out to the registers. This could trigger a bus fault interrupt inside the amp, which we want to avoid. Signed-off-by: Richard Fitzgerald Reviewed-by: Takashi Iwai Link: https://patch.msgid.link/20251021105022.1013685-2-rf@opensource.cirrus.com Signed-off-by: Mark Brown

ASoC: cs35l56: Add support for CS35L56 B2 silicon

2025-09-28T06:27:15+00:00

This adds support for changed firmware addresses on the B2 revision of CS35L56 silicon. Signed-off-by: Richard Fitzgerald Signed-off-by: Takashi Iwai