Bug 215664

Summary:	Discrepancy between ACPI DebounceTimeout and Intel Pinctrl driver
Product:	Drivers	Reporter:	Kai-Heng Feng (kai.heng.feng)
Component:	GPIO/Pin Control	Assignee:	Andy Shevchenko (andy.shevchenko)
Status:	RESOLVED PATCH_ALREADY_AVAILABLE
Severity:	normal	CC:	andy.shevchenko, jwrdegoede, mika.westerberg
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	mainline, linux-next	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	dmesg acpidump

Description Kai-Heng Feng 2022-03-07 04:16:08 UTC

Cirrus codec amplifier doesn't work:
[    4.695135] Serial bus multi instantiate pseudo device driver CSC3551:00: error -EINVAL: Error requesting irq at index 0: -22
[    4.696659] Serial bus multi instantiate pseudo device driver: probe of CSC3551:00 failed with error -22

Cirrus folks identified the root cause is that the debounce unit between ACPI and Intel pinctrl driver are different.

ACPI spec [1] states "DebounceTimeout is an optional argument specifying the hardware debounce wait time, in hundredths of milliseconds. The bit field name _DBT is automatically created to refer to this portion of the resource descriptor."

However, intel_config_set_debounce() is using "debounce * NSEC_PER_USEC / DEBOUNCE_PERIOD_NSEC". To match the value with ACPI, NSEC_PER_USEC should be 10 * NSEC_PER_USEC instead.

[1] https://uefi.org/specs/ACPI/6.4/19_ASL_Reference/ACPI_Source_Language_Reference.html#gpioio-gpio-connection-io-resource-descriptor-macro

Comment 1 Kai-Heng Feng 2022-03-07 04:16:29 UTC

Created attachment 300541 [details]
dmesg

Comment 2 Kai-Heng Feng 2022-03-07 04:16:47 UTC

Created attachment 300542 [details]
acpidump

Comment 3 Andy Shevchenko 2022-03-07 10:39:46 UTC

Thanks for the report! Lemme check this and come up with something.

Comment 4 Andy Shevchenko 2022-03-07 11:31:46 UTC

Hans, this might explain why debounce setting wasn't working for you (I don't remember the details, though).

Comment 5 Hans de Goede 2022-03-07 11:45:19 UTC

(In reply to Andy Shevchenko from comment #4)
> Hans, this might explain why debounce setting wasn't working for you (I
> don't remember the details, though).

My issue was on Bay Trail, which has the issue of there being a single debounce timeout for all pins in a single GPIO island. Since this is about the new laptops with Cirrus Logic amplifiers directly exposed to the OS, this is likely Tiger Lake, or maybe even Alder Lake.

As discussed on the list the plan is to just disable setting the debounce timeout on Bay Trail since there is no way to get it to work and it has not even been hooked up for most of Bay Trail's life-cycle, so clearly not having support for it is not an issue, also see: https://lore.kernel.org/linux-gpio/051a9e42-02e3-4db0-bdbd-f8d7855b5db0@redhat.com/

Comment 6 Andy Shevchenko 2022-03-07 13:33:59 UTC

The fix has been submitted:

https://lore.kernel.org/linux-gpio/CAAd53p6UvOqGkhGuwtZ0hvEAQFb4DR-RUd7R1uZuR+Duhf72EQ@mail.gmail.com/T/#u

P.S. Hans, thank you for clarifications!