Hi Jerome, >> Thank you for the instructions, I managed to successfully boot the latest >> mainline tree on the Hikey960. Full log in attachment. >> And even the poweroff sequence looks fine to me. > > Nice. > >> Could you please try to follow the following steps and verify that it works for >> you as well? >> >> 1. Download op-tee version 3.7.0. >> 2. Enter the linux directory and add the remote >> git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git >> 3. Rebase the 3.7.0 branch kernel of linux master. >> 4. Build and Flash op-tee. >> >> It should boot and have no issues on init. > > [Unfortunately, it doesn't work for me. The board still hangs (or > rather, loops) on /init. I'm puzzled at this point... What could be > wrong with my setup? Do we have different board revisions maybe? > So my board is an Hikey960 ver.B and I am running the latest UEFI firmware. Some useful links: [1] https://snapshots.linaro.org/reference-platform/components/uefi-staging/ [2] https://github.com/96boards-hikey/tools-images-hikey960/blob/master/README-flash-uefi.md > ...] > > Now that's embarrassing :-/ I decided to re-build everything once again, > but *without ccache* this time. Guess what? Kernel v5.4 boots fine :-O > and master too. > > I used to trust ccache blindly, until I discovered a bug which has > caused me some trouble for quite some time [1]. After upgrading I > thought all would be good again, but apparently not... That is quite > annoying! :-/ > > Would you recommend against using ccache for kernel development? > I never use ccache for kernel development. > Anyway, thank you guys for your valuable help. I feel ashamed for > wasting your time troubleshooting a non-bug... but at least it wasn't in > vain. > Nothing to be embarrassed of, I am curious as well now to know what is going on with your setup :) > [1] https://github.com/ccache/ccache/issues/136 > >> What I noticed is that the optee >> provided kernel is not a 5.1 but a 5.1 plus some patches. > > True, but nothing really important. A vanilla kernel is supposed to work > fine including most of the OP-TEE stuff (except for some very specific > "secure data path" code which depends on the presence of a special ION > heap). > Ok, I thought it was worth mentioning. > Thanks, > -- Regards, Vincenzo