Hi Takashi,
On 04/07/2022 13:50, Takashi Iwai wrote:
On Thu, 30 Jun 2022 02:23:21 +0200,
Vitaly Rodionov wrote:
The CS35L41 Amplifier contains a DSP, capable of running firmware.
The firmware can run algorithms such as Speaker Protection, to ensure
that playback at high gains do not harm the speakers.
Adding support for CS35L41 firmware into the CS35L41 HDA driver also
allows us to support several extra features, such as hiberation
and interrupts.
The chain adds support in stages:
- General fixes to improve generalization and code re-use inside
the CS35L41 HDA driver.
- Add support for interrupts into the driver, which is required
for complete support of the firmware.
- Refactor ASoC CS35L41 code which deals with firmware to allow
for code re-use inside the CS35L41 HDA driver.
- Add support for loading firmware and tuning files from file system,
and creating alsa controls to control it.
- Support firmware load paths for different hardware systems.
- Support suspend/resume in the driver when using firmware. The firmware
supports hibernation, which allows the CS35L41 to drop into a low
power mode during suspend.
- Support the ability to unload firmware, swap and reload the firmware.
This is to allow different firmware to run during calibration.
The intended use-case is to load the firmware once on boot, and the driver
autmatically tries to load the firmware after it binds to the HDA driver.
This behaviour can be switched off using a kconfig, if desired.
changes since v7:
- Use private_data rather than private_value to save control info
- Clean up alsa control memory allocation/deallocation
- Remove unnecessary whitespace
- Get subsystem id from codec, rather than saving it separately
changes since v6:
- Fix warning by kernel test robot<lkp@xxxxxxxxx>
changes since v5:
- Fix warning by kernel test robot<lkp@xxxxxxxxx>
changes since v4:
- Fully remove tlv remnants from control add apis
- Remove unnecessary debug
- Rename variable to be more generic
- Remove redundent length check from read/write control apis
- Use SNDRV_CTL_ELEM_IFACE_CARD for firmware load controls
- Make kcontrol add/remove synchronous
- Load firmware asynchronous when loading via control
- Used cached controls when reloading firmware; only delete
controls when removing the driver itself
- Improve kcontrol remove
- Fix control write + notify
- Cleanup of unnecessary code
- Fix race condition when loading firmware before playback
- Ensure errors are properly propogated
- Fix include for Module parameters
Stefan Binding (13):
ALSA: hda: hda_cs_dsp_ctl: Add Library to support CS_DSP ALSA controls
ALSA: hda: hda_cs_dsp_ctl: Add apis to write the controls directly
ALSA: hda: cs35l41: Save codec object inside component struct
ALSA: hda: cs35l41: Save Subsystem ID inside CS35L41 Driver
ALSA: hda: cs35l41: Support reading subsystem id from ACPI
ALSA: hda: cs35l41: Support multiple load paths for firmware
ALSA: hda: cs35l41: Support Speaker ID for laptops
ALSA: hda: cs35l41: Support Hibernation during Suspend
ALSA: hda: cs35l41: Read Speaker Calibration data from UEFI variables
ALSA: hda: hda_cs_dsp_ctl: Add fw id strings
ALSA: hda: cs35l41: Add defaulted values into dsp bypass config
sequence
ALSA: hda: cs35l41: Support Firmware switching and reloading
ALSA: hda: cs35l41: Add module parameter to control firmware load
Vitaly Rodionov (1):
ALSA: hda: cs35l41: Add initial DSP support and firmware loading
Thanks, this version looks better than previous ones, and I'm fine to
apply as is, to make things going forward. But this seems requiring
the prerequisite in ASoC codec side.
Mark, could you send a PR to merge into my tree so that I can apply
those series?
And, one still uncertain thing is about the handling of system
suspend/resume, especially about the firmware loading work. e.g. what
happens if a f/w work is in run and the system goes to suspend at the
very same time? Or does the S3/S4 work in general at all?
Sorry for the delay, but your comments raised our attention and we
wanted to run extensive testing again to make sure there is no hidden
issue. No issues were found during tests. When we start FW download work
we call "resume", so the resume counter should be incremented and
suspend will not happen till we finish the download process.
Thanks,
Vitaly
Takashi
