imx6q Android 6.0.1_r22 Kernel 4.1.15 suspend/resume

[benhewson]

I have downloaded Android M6.0.1_2.1.0 from NXP and have this running on a custom board with an iMX6Q processor.

I am trying to suspend to memory and then resume. 

There is an external MCU ( Kinetis ) with RTC which generates an interrupt on GPIO6.14. I can see Android setting an alarm time (I have debug in my kernel driver) . With a debugger on the MCU I can see the interrupt happen (this all works in Android 4.4.0)

While awake, the RTC interrupts happen, and I can see the count incremented in /proc/interrupts.

The interrupt is listed as a wakeup source in /sys/kernel/debug/wakeup_sources.

The interrupt is not masked in GPC_IMR3

However once asleep, the CPU never wakes.

Has anyone managed to get Android 6.0.1_r22 suspending and resuming on any imx6 board ?

[JayTu]

You can check if you left suspend in suspend-imx6.S. If so, you can further trace where it blocks.

If no, it can't wakeup usually because some power are not recovered when resume, i.e., VDD_ARM_CAP/VDD_PU_CAP, ...

[jimmychan]

please check the gpio pin IOMUX and PAD setting in your board .dtsi file.

For your reference, this is the sabresd dtsi file

linux-2.6-imx.git - Freescale i.MX Linux Tree 

[benhewson]

the iomux/pad settings are correct. The interrupt is on gpio6_14 and I have an entry in my iomux section of

MX6QDL_PAD_NANDF_CS1__GPIO6_IO14 0x80000000

As I have said, when not sleeping the interrupts appear to work normally, that is the interrupt count in /proc/interrupts increments by 1 every time I set an alarm.

GPIO6_14 is interrupt 108, which is masked or not in GPC_IMR3 ( irq 96-127)

Directly before the cpu_suspend() call in pm-imx6.c, this register contains 0xffffefff. The zero bit by my counting is for irq 108, so the interrupt is not masked.

If I comment out the "wfi" instruction in suspend-imx6.S, the code goes through the shutdown/wakeup sequence OK, so it looks to me that for whatever reason, the interrupt is not waking the processor.