13:22 <aggi> greetings. short question: is it possible to launch an aarch32 linux kernel with an aarch64 u-boot loader?

13:47 <apritzel> aggi: theoretically yes, but not easily in practice

13:48 <aggi> thanks apritzel; confirms my findings with the test setup

13:50 <apritzel> aggi: U-Boot would need to support it. It would be relatively easy to allow entering the kernel in EL1, but PSCI (SMP support) might become a bit nasty

13:50 <aggi> packed an approprieate aarch32 kernel into a legacy uImage (which explicitely tags the architecture); then bootm, crash

13:51 <aggi> didn't at least reach the point when kernel could complain about PSCI

13:52 <aggi> CONFIG_ARM64_SUPPORT_AARCH32=y; doesn't seem to suffice for u-boot to switch execution state to aarch32

13:53 <apritzel> yeah, just checked that as well, that's just used to figure out if a FIT image wrapped kernel would be compatible

13:54 <aggi> fyi, this is what i got after two days of fiddling: https://dpaste.org/6ELV

13:55 <apritzel> aggi: this is the macro that would be the magic: armv8_switch_to_el1_m

13:56 <aggi> i digged into the sources, however i do not feel competent to hack this

13:56 <aggi> noticed already execution state must be explicitely switched

13:57 <apritzel> aggi: you might try a FIT image and set CONFIG_ARMV8_SWITCH_TO_EL1

13:58 <apritzel> you can only switch bitness when dropping an exception level, and since U-Boot starts in EL2, you cannot enter the ARM kernel in HYP/EL2

14:05 <aggi> apritzel: ok, i'll try with FIT and where and how CONFIG_ARMV8_SWITCH_TO_EL1 has to be passed.

14:19 <apritzel> aggi: U-Boot's config, for instance "make menuconfig", then search for that string

14:24 <aggi> ok, i'll tie CONFIG_ARMV8_SWITCH_TO_EL1 to CONFIG_ARM64_SUPPORT_AARCH32=y

14:25 <aggi> add some printf into bootm.c to see what happens; i'll notify what i found in a few hours. need to eat first. thanks for now.

14:35 <pcBob> Hello, someone where who is familiar with the F1C100S? I am having trouble getting USB working with a buildroot image I created with: https://github.com/unframework/licheepi-nano-buildroot

14:37 <pcBob> This is the startup log: https://pastebin.com/sWqY0DmK

14:37 <pcBob> "starting USB...No working controllers found" doesn't sound good :/

14:45 <pcBob> it's a Lichee Pi Nano

14:51 <pcBob> I plugged in a generic USB 2.0 mass storage device and it didn't show any sign of life and lsusb is empty

16:03 <apritzel> pcBob: the "starting USB ..." message is from U-Boot, that's independent from the kernel

16:04 <apritzel> pcBob: I don't see any attempt from the kernel to probe USB devices, do you have the proper nodes in the DT?

16:10 <pcBob> apritzel: How can I check that?

16:12 <apritzel> well, in the DT you are using/loading. Looking at the kernel tree DT, I don't see USB supported for that SoC at all

16:13 <apritzel> but I don't know if nobody bothered to try or there is some serious issue with the USB controllers or the PHY

16:14 <pcBob> apritzel: Someone else in #buildroot just told me that it seems to be work in progress according to this: https://linux-sunxi.org/Linux_mainlining_effort#Major_drivers

16:14 <pcBob> So I guess I can't use it for now

18:25 <aggi> anyway, CONFIG_ARMV8_SWITCH_TO_EL1=y locks up the system as soon as an aarch32 kernel is launched from aarch64 u-boot with bootm command

18:36 <aggi> added some printf into arch/arm/lib/bootm.c, and armv8_switch_to_el1 with ES_TO_AARCH32 paramter is reached as it should be

18:38 <aggi> nonetheless, the aarch32 kernel doesn't launch, nothing on serial console anymore (correctly configured)

18:39 <aggi> in any case, u-boot and aarch64 are not maintained to launch aarch32 linux kernel

20:15 <aggi> further testing, removed ATF bl31.elf from u-boot, because ATF supposedly causes u-boot being locked down from EL3 to EL2 which prevents switching from aarch64 to aarch32 to bootm an aarch32 kernel

20:15 <aggi> no luck