Investigate wake up from suspend

[acostach]

This issue on nxp's website mentions impossibility of waking up the coral board from sleep, as there are no buttons available on the device, and uart gpios don't work for this.

Check if this is reproducible and see if some gpio or some other input could be used to wake the board up from sleep.

[acostach]

Reproduced the issue on the coral image, indeed with suspend2idle can be resumed via ttymxc2, and after deep suspend it cannot be resumed anymore via uart.

Initially suspected kernel/dtb differences, but after further investigation I could trace this down to the bootloader image.

Booting with mendel u-boot and Balena kernel, dtb and rootfs I'm able to resume suspend so the issue is likely in the uboot image, which in mendel contains a different imx-atf, and is compiled with a different imx-mkimage.

Below is Balena image booted with mendel u-boot, resuming correctly from suspend:

root@f6f77e9:~# echo mem > /sys/power/state 
[  522.069683] PM: suspend entry (deep)
[  522.073360] PM: Syncing filesystems ... done.
[  522.230673] Freezing user space processes ... (elapsed 0.002 seconds) done.
[  522.239838] OOM killer disabled.
[  522.243068] Freezing remaining freezable tasks ... (elapsed 0.024 seconds) done.
[  522.274709] Suspending console(s) (use no_console_suspend to debug)
[  522.409511] PM: suspend devices took 0.128 seconds
[  522.416264] Disabling non-boot CPUs ...
[  522.440070] IRQ 6: no longer affine to CPU1
[  522.447518] CPU1: shutdown
[  522.447522] psci: CPU1 killed.
[  522.479516] CPU2: shutdown
[  522.479520] psci: CPU2 killed.
[  522.503523] CPU3: shutdown
[  522.503527] psci: CPU3 killed.
[  522.504394] Enabling non-boot CPUs ...
[  522.504994] Detected VIPT I-cache on CPU1
[  522.505017] GICv3: CPU1: found redistributor 1 region 0:0x00000000388a0000
[  522.505064] CPU1: Booted secondary processor [410fd034]
[  522.505668]  cache: parent cpu1 should not be sleeping
[  522.505873] CPU1 is up
[  522.506419] Detected VIPT I-cache on CPU2
[  522.506436] GICv3: CPU2: found redistributor 2 region 0:0x00000000388c0000
[  522.506467] CPU2: Booted secondary processor [410fd034]
[  522.506826]  cache: parent cpu2 should not be sleeping
[  522.507003] CPU2 is up
[  522.507549] Detected VIPT I-cache on CPU3
[  522.507567] GICv3: CPU3: found redistributor 3 region 0:0x00000000388e0000
[  522.507599] CPU3: Booted secondary processor [410fd034]
[  522.507956]  cache: parent cpu3 should not be sleeping
[  522.508160] CPU3 is up
[  522.511288] imx6q-pcie 33800000.pcie: pcie phy pll is locked.
[  522.565978] imx6q-pcie 33c00000.pcie: pcie phy pll is locked.
[  522.653670] PM: resume devices took 0.032 seconds
[  522.775504] OOM killer enabled.
[  522.778647] Restarting tasks ... done.
[  522.784633] PM: suspend exit

Continuing investigation on u-boot/atf side.

[acostach]

Fixed by #18

[acostach]

Image containing the fix is now available in staging dashboard.

[acostach]

Production release will be available after the rootfs space problem is solved in meta-balena.

[acostach]

This fix is now available in production dashboard v2.52.7 for the Coral Dev

[acostach]

Fix merged but not longer in production (supervisor update necessary). To close in this ticket when a release newer than 2.52.7 is available in production.

[acostach]

Fix is back in production