Quoting Matthias Kaehlcke (2023-05-16 09:46:30)
The rt5682 driver switches its regmap to cache-only when the device suspends and back to regular mode on resume. When the jack detect interrupt fires rt5682_irq() schedules the jack detect work. This can result in invalid reads from the regmap in cache-only mode if the work runs before the device has resumed:
I was wondering why we can't resume the device before device irqs are enabled (and similarly suspend it after irqs are disabled). It looks like snd_soc_component_driver is not an actual 'struct driver' instance so it reimplements PM hooks like suspend and resume for the components and the suspend hooks run from snd_soc_suspend() at the struct dev_pm_ops::suspend() path. Adding a noirq variant looks like it isn't worth it.
[ 56.245502] rt5682 9-001a: ASoC: error at soc_component_read_no_lock on rt5682.9-001a for register: [0x000000f0] -16
Disable the jack detection interrupt during suspend and re-enable it on resume. The driver already schedules the jack detection work on resume, so any state change during suspend is still handled.
This is essentially the same as commit f7d00a9be147 ("SoC: rt5682s: Disable jack detection interrupt during suspend") for the rt5682s.
Cc: stable@kernel.org Signed-off-by: Matthias Kaehlcke mka@chromium.org
Reviewed-by: Stephen Boyd swboyd@chromium.org
diff --git a/sound/soc/codecs/rt5682-i2c.c b/sound/soc/codecs/rt5682-i2c.c index 2935c1bb81f3..5bc46b041786 100644 --- a/sound/soc/codecs/rt5682-i2c.c +++ b/sound/soc/codecs/rt5682-i2c.c @@ -267,7 +267,9 @@ static int rt5682_i2c_probe(struct i2c_client *i2c) ret = devm_request_threaded_irq(&i2c->dev, i2c->irq, NULL, rt5682_irq, IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING | IRQF_ONESHOT, "rt5682", rt5682);
if (ret)
if (!ret)rt5682->irq = i2c->irq;else dev_err(&i2c->dev, "Failed to reguest IRQ: %d\n", ret);
Not in this patch, but
s/reguest/request/