| 09:09 | agraf | Artox: that depends on how you configure grub |
| 09:10 | agraf | Artox: so grub as it's configured for you there doesn't know where it should load anything from |
| 09:10 | agraf | Artox: which i guess is why it's not even adding a partition number |
| 09:11 | agraf | Artox: what we usually have is a grub.cfg in the same directory as grub.efi for the first boot |
| 09:11 | agraf | Artox: so either the installation disk (iso) or the kiwi first boot step |
| 09:12 | agraf | Artox: and for second boot, we generate a grub.efi that knows where it can find its real config |
| 12:23 | Artox | agraf: I figured out the magic |
| 12:23 | Artox | grub-install did not replace /boot/efi/boot/bootarm.efi |
| 12:23 | Artox | instead it installed /boot/efi/efi/debian/grubarm.efi |
| 12:24 | Artox | it has only 82kB, but with that it works as expected |
| 12:24 | Artox | menu shows up |
| 12:24 | Artox | however it looks like suse has a few impotant arm patches that debian still misses |
| 12:46 | Artox | I guess mkconfig should be able to identify the devicetree file |
| 12:47 | Artox | so this ask firmware for fdt is onlya fallback |
| 13:02 | agraf | Artox: yeah, grub upstreaming is hell |
| 13:02 | agraf | Artox: takes about 1 year for anyone to even remotely look at your patches |
| 13:03 | agraf | Artox: the "ask firmware for dtb" is the preferred path |
| 13:03 | agraf | Artox: because ideally the device tree comes from firmware |
| 13:03 | agraf | Artox: OSs shouldn't have any business in providing them |
| 13:03 | agraf | Artox: look at how PowerPC boots |
| 13:03 | agraf | (the real server ones, or the macs) |
| 13:04 | Artox | my colleague heartedly disagrees |
| 13:04 | Artox | because in reality DTBs change |
| 13:05 | Artox | and because u-boot carries just one dtb |
| 13:05 | Artox | how would one handle board variants and dtb variants? |
| 13:27 | agraf | Artox: that's the crux of the problem |
| 13:27 | agraf | Artox: dtbs should not change |
| 13:27 | agraf | Artox: the mere fact that they do change means the model is broken |
| 13:27 | agraf | Artox: dtbs describe hardware, not Linux |
| 13:28 | Artox | I guess I can live with patching teh firmware occasionally |
| 13:28 | agraf | Artox: in a way having AIX and Mac OS X on PowerPC helped a lot with that |
| 13:28 | Artox | any remark on variants? |
| 13:28 | agraf | Artox: because they made sure that you don't randomly get a new device tree |
| 13:29 | agraf | Artox: grub2 can override the device tree |
| 13:29 | Artox | true |
| 13:29 | agraf | Artox: and loic did a patch a while back to have grub2 automatically add that to grub.cfg |
| 13:29 | agraf | Artox: not sure how far that is upstream |
| 13:30 | agraf | Artox: you can always use that to hack up something that overrides the firmware dtb |
| 13:30 | agraf | Artox: but there's a good chance that breaks too |
| 13:30 | agraf | Artox: spin table location for example is part of dtb |
| 13:31 | agraf | Artox: and needs to be adjusted by firmware depending on where it puts the spin tables |
| 13:31 | agraf | Artox: unless you're on a PSCI enabled system of course ;) |
| 13:31 | Artox | now to an easier topic |
| 13:32 | Artox | I mentioned that debian likes to not install bootarm.efi |
| 13:32 | Artox | is there such a thing as a chainload.efi? |
| 13:32 | Artox | that I can put as bootarm.efi to load debian/grubarm.efi? |
| 13:34 | agraf | Artox: sure, it's called shim |
| 13:34 | agraf | Artox: but you don't need that |
| 13:34 | agraf | Artox: if you call the grub-install thing with --removable it should do that automatically |
| 13:34 | Artox | thats what --removable does? |
| 13:34 | Artox | someone should document the output of --help |
| 13:35 | agraf | :) |
| 13:35 | agraf | Artox: https://github.com/openSUSE/perl-bootloader/blob/master/grub2-efi/install |
| 13:37 | Artox | that works so much better! |
| 13:37 | Artox | thanks a lot |
| 13:37 | agraf | np |
| 13:37 | agraf | always happy to move people to the efi side ;) |
| 13:37 | agraf | it makes life easier eventually |
| 13:45 | Artox | the path looks stony though |
| 16:24 | agraf | Artox: well, more layers is always harder to get right if you're doing the plumbing |
| 16:24 | agraf | Artox: but it means you move plumbing requirements away from others ;) |
| 18:32 | Artox | agraf: the transition itself is hard too |
| 18:32 | Artox | due to requiring the firmware to provide the right thing |
| 19:04 | vpeter | Artox: good news - got lede running. But have a question: what steps are required to set ethernet interfaces like in openwrt? uci-defaults/02_network script would be enough? |
| 19:04 | vpeter | It is little hard because building lede at my workplace but running clearfog at home :( |
| 19:05 | vpeter | Meaning no internet when sd card switched. |
| 19:59 | agraf | Artox: yeah, it's a mental shift ;) |
| 20:00 | agraf | Artox: but you can always create hybrids |