[SDW] rt711-sdca: timeout on INT1 and INT2 mask write

[plbossart]

Intel internal test 1122 shows the following errors w/ rt711-sdca. There's obviously something wrong with SDCA interrupts.

[ 8446.911146] kernel: intel-sdw intel-sdw.0: IO transfer timed out, cmd 3 device 1 addr 5c len 1
[ 8446.911160] kernel: soundwire sdw-master-0: trf on Slave 1 failed:-110 write addr 5c count 0
[ 8447.414240] kernel: intel-sdw intel-sdw.0: IO transfer timed out, cmd 3 device 1 addr 5d len 1
[ 8447.414256] kernel: soundwire sdw-master-0: trf on Slave 1 failed:-110 write addr 5d count 0

#2622 suggests a first pass to revisit on SDCA interrupt handling.

[bardliao]

I can easily see the issue with the test_sdca_int.txt script

[bardliao]

The issue happens if regmap_cache_only is N when soc component is removed.
The call trace is:

[ 77.309990] regmap_write+0x39/0x60
[ 77.309994] rt711_sdca_jack_init+0x1b1/0x1f0 [snd_soc_rt711_sdca]
[ 77.309996] rt711_sdca_set_jack_detect+0x44/0x50 [snd_soc_rt711_sdca]
[ 77.310003] snd_soc_component_set_jack+0x22/0x40 [snd_soc_core]
[ 77.310007] soc_remove_component+0x1b/0x80 [snd_soc_core]
[ 77.310011] soc_cleanup_card_resources+0xae/0x250 [snd_soc_core]
[ 77.310016] snd_soc_unbind_card+0x65/0xe0 [snd_soc_core]
[ 77.310020] snd_soc_unregister_card+0x1f/0x60 [snd_soc_core]

The problem is that rt711_sdca_jack_init is called when sdw bus is suspended and we should not try to access sdw register at that moment.
#2629 has same root cause, too. The reason why #2627 has higher reproduce rate is that we set regcache_cache_only(rt711->regmap, false); in rt711_sdca_update_status.
Adding a snd_soc_component callback function and set cache only there should fix the issue.
I will cook a PR for that and @shumingfan will refine rt711-sdca codec driver for the usage of regmap.

[mengdonglin]

Fixed by PR #2640 and merged. No reproduced in recent daily test. Closing.