Removal of child drivers and the bus driver was broken and would result in a slew of various errors.
Most of these were caused by the code shutting down in the wrong order, shutting down the bus driver first. The bus driver should be shut down after the child drivers have been removed (compare with the I2C and SPI subsystem for example).
These patches fix that.
A secondary problem was over the cleanup of child drivers. The removal functions were not the opposite of the probe function, and the ownership of struct sdw_slave is tricky because it mixes two separate usages and currently has to be "owned" by the bus driver.
Tested with 4 peripherals on 1 bus and 8 peripherals on 2 buses.
Richard Fitzgerald (7): soundwire: bus: Do not forcibly disable child pm_runtime soundwire: intel_init: Separate shutdown and cleanup ASoC: SOF: Intel: Don't disable Soundwire interrupt before the bus has shut down soundwire: bus: Add remove callback to struct sdw_master_ops soundwire: intel: Don't disable interrupt until children are removed soundwire: intel: Don't disable pm_runtime until children are removed soundwire: bus: Fix premature removal of sdw_slave objects
drivers/soundwire/bus.c | 37 ++++++++++++++++++++++++----- drivers/soundwire/intel.c | 13 ++++++++-- drivers/soundwire/intel_init.c | 25 +++++++++++++++---- drivers/soundwire/slave.c | 21 ++++++++++++---- include/linux/soundwire/sdw.h | 3 ++- include/linux/soundwire/sdw_intel.h | 2 ++ sound/soc/sof/intel/hda.c | 16 ++++++++++--- 7 files changed, 96 insertions(+), 21 deletions(-)