SoundWire: fix port_ready[] allocation and remove it when unnecessary

[plbossart]

The use of port_ready[] is broken for ports that are not contiguous, and none of the existing codec devices use this functionality anyways. Let's fix the core and remove unnecessary allocations in codec drivers.

No obvious regressions seen on CML-U w/ SoundWire, but additional checks might not hurt.

@shumingfan@ryans-lee@oder-chiou@jack-cy-yu Can you do me a favor and double-check that the use of the simplified channel prepare state machine is indeed a feature of your hardware, not a copy/paste error or omission. If we do need to enable the more complicated state machine then the last patch needs to be redone. Thanks!

[bardliao]

@plbossart Now slave->port_ready is only allocated and initialized at sdw_slave_read_prop(). But I don't see sdw_slave_read_prop() is called anywhere. Will it cause problem? Can we move the allocation and initialization to sdw_slave_add() since it is a fixed array now?

[RanderWang]

@plbossart Now slave->port_ready is only allocated and initialized at sdw_slave_read_prop(). But I don't see sdw_slave_read_prop() is called anywhere. Will it cause problem? Can we move the allocation and initialization to sdw_slave_add() since it is a fixed array now?

I also found this. I added log in sdw driver but found port_ready was not checked by complete(&slave->port_ready[port]) on TGL RVP for no port irq was triggered.

[plbossart]

@plbossart Now slave->port_ready is only allocated and initialized at sdw_slave_read_prop(). But I don't see sdw_slave_read_prop() is called anywhere. Will it cause problem? Can we move the allocation and initialization to sdw_slave_add() since it is a fixed array now?

I also found this. I added log in sdw driver but found port_ready was not checked by complete(&slave->port_ready[port]) on TGL RVP for no port irq was triggered.

Sorry I don't get your point @RanderWang. If the codecs uses the simplified channel prepare state machine, there will not be any port_ready interrupts.

[plbossart]

@plbossart Now slave->port_ready is only allocated and initialized at sdw_slave_read_prop(). But I don't see sdw_slave_read_prop() is called anywhere. Will it cause problem? Can we move the allocation and initialization to sdw_slave_add() since it is a fixed array now?

we could indeed move to a common place, good suggestion, so that codec drivers don't need to worry about this.

[RanderWang]

@plbossart Now slave->port_ready is only allocated and initialized at sdw_slave_read_prop(). But I don't see sdw_slave_read_prop() is called anywhere. Will it cause problem? Can we move the allocation and initialization to sdw_slave_add() since it is a fixed array now?

I also found this. I added log in sdw driver but found port_ready was not checked by complete(&slave->port_ready[port]) on TGL RVP for no port irq was triggered.

Sorry I don't get your point @RanderWang. If the codecs uses the simplified channel prepare state machine, there will not be any port_ready interrupts.

I meant not error in original PR for the port_ready was not checked by slave, or it would lead to kernel panic.
That was why Bard said it would cause problem.

[RanderWang]

Tested on TGL RVP. no regression issue was found.

[plbossart]

@bardliao@RanderWang@lyakh can you take a look?

@bardliao this would need to be squashed with another patch 13e6f3d ('ASoC: codecs: rt*-sdw.c: fix off-by-one port count') since it's really the same issue.

[bardliao]

@bardliao this would need to be squashed with another patch 13e6f3d ('ASoC: codecs: rt*-sdw.c: fix off-by-one port count') since it's really the same issue.

acked.

[RanderWang]

@plbossart a CI compiling error

drivers/soundwire/mipi_disco.c: Infunction ‘sdw_slave_read_prop’:
drivers/soundwire/mipi_disco.c:292:12: error: unusedvariable ‘i’ [-Werror=unused-variable]
292 | intnval, i, dp0=0;
| ^
cc1: allwarningsbeingtreatedaserrorsscripts/Makefile.build:280: recipefortarget'drivers/soundwire/mipi_disco.o'failed

[plbossart]

@plbossart a CI compiling error

drivers/soundwire/mipi_disco.c: Infunction ‘sdw_slave_read_prop’:
drivers/soundwire/mipi_disco.c:292:12: error: unusedvariable ‘i’ [-Werror=unused-variable]
292 | intnval, i, dp0=0;
| ^
cc1: allwarningsbeingtreatedaserrorsscripts/Makefile.build:280: recipefortarget'drivers/soundwire/mipi_disco.o'failed

fixed now