heads
heads copied to clipboard
linuxboot
WiP: galp5 build based on librem_11 (GOP based iGPU)
@loftlifter31 this is my take helping you get the base to make this forward.
Superseeds https://github.com/linuxboot/heads/pull/1766
As per DCO CI warning https://github.com/linuxboot/heads/pull/1774/checks?check_run_id=29617868402, resigning all comits, making us co-authors.
Please sign commits form local git repo. Steal reuse my commit content and force push to your branch under #1766 (currently your master branch, otherwise ropen another PR, closing #1766 as per your choosing)
Rebase+squash+signoff detailed under https://github.com/linuxboot/heads/pull/1774/commits/4418282b366f11bac781d9f0000b8dbc0e49aa6a commit message
Here you go, commented in code @loftlifter31. CircleCI will produce ROM to be tested, as per https://osresearch.net/Downloading#downloading-heads-from-circleci but from this PR built artifacts.
WARNING: this is just first iteration, I do not own the hardware and cannot test. This is first iteration, where GOP/vbt/EFIFB will most probably need some additional work. If this doesn't boot, we can activate debbuging traces to be saved in SPI, and taking a backup there, increasing debug levels and extracting coreboot logs will push this forward.
@loftlifter31 did you test galp5 built rom artifact from this PR?
Here you go, commented in code @loftlifter31. CircleCI will produce ROM to be tested, as per https://osresearch.net/Downloading#downloading-heads-from-circleci but from this PR built artifacts.
WARNING: this is just first iteration, I do not own the hardware and cannot test. This is first iteration, where GOP/vbt/EFIFB will most probably need some additional work. If this doesn't boot, we can activate debbuging traces to be saved in SPI, and taking a backup there, increasing debug levels and extracting coreboot logs will push this forward.
@loftlifter31 did you test galp5 built rom artifact from this PR?
@loftlifter31 after following referred downloading doc page, you should have landed to https://output.circle-artifacts.com/output/job/5c445ab2-7bfa-4ed5-8b35-1e3993322e0f/artifacts/0/build/x86/galp5/heads-galp5-v0.2.0-2303-g9638cdb.zip
Oh. Librem 11 is not a board config with tpm2 dependencies. So above rom might boot with FB and console output or not, but t+heads rom toolstack won't work. Board config needs to be adapted from another board that is tpm2, ie nv41.
Oh. Librem 11 is not a board config with tpm2 dependencies. So above rom might boot with FB and console output or not, but t+heads rom toolstack won't work. Board config needs to be adapted from another board that is tpm2, ie nv41.
fixed under 2db9de9
From https://github.com/linuxboot/heads/issues/1765#issuecomment-2338646515
I got it working. Just needed to update vbt.
That was relevant to https://github.com/system76/firmware-open/issues/575#issuecomment-2327326963, not Heads issue.
Can you create a branch with everything tested working on your side? Like I said, HEAP size bumped to 40000 should work even with bootsplash if GOP + vbt+ efifb works. Some output would be great. Will rebase this branch on master and force push.
After that, I will let this PR bitrot until picked up to be upstreamed.
Tested equivalent under https://github.com/linuxboot/heads/pull/1795 under qemu: works without needing to change HEAP_SIZE. See commit details.
Tested local build: patches applies.
artifacts for 37f809e galp board should drive FB and be working, outside of on-demand ethernet enabling as previously discussed.
@loftlifter31 you should get bootsplash from this PR artifacts since you said only bootsplash was missing. So EFIFB drives FB. Missing etherent needed realtek drivers which might already be under modules/linux. If so, change board config for config name, make sure galp5 linux config compiles those modules as kernel and deactivate e1000e, unneeded (?)
Confused on your last statement saying its supposed to come with some ethernet but came with realtek. Makes me question their HCL (and changes of internal hardware at will: confusing).
I won't be able to do much more on this before october. Keep me posted.
Can you create a branch with everything tested working on your side? Like I said, HEAP size bumped to 40000 should work even with bootsplash if GOP + vbt+ efifb works. Some output would be great. Will rebase this branch on master and force push.
Create a branch? Where? On my local fork? At this point everything on my fork is tested, though bootsplash and NIC are not working.
Can you create a branch with everything tested working on your side? Like I said, HEAP size bumped to 40000 should work even with bootsplash if GOP + vbt+ efifb works. Some output would be great. Will rebase this branch on master and force push.
Create a branch? Where? On my local fork? At this point everything on my fork is tested, though bootsplash and NIC are not working.
37f809e is supposed to fix bootsplash. This PR includes fixes for your reported issues, leaving ethernet support to be figured out (you said card present is not as from upstream HCL, so didn't fixed this here, but showed ropes on how to do this (add drivers under modules/linux, add those compiled as modules, review modules loading scripts etc: all can be done later, this PR if made to work with blobs injected in would be useable by other galp5 owners desiring Heads. But work needs to be done to arrive there.
@loftlifter31 Putting this as draft, seems we are not aligned to make this first candidate and make this work for others as all other boards. When you are ready, let me know. Letting this bitrot as draft.