#asahi · 2021-12-02 14:50
marcanif you *really* want to re-implement all that for Debian, I won't stop you... but I imagine most distros will be perfectly happy telling people to `curl alx.sh | sh` from recovery mode, follow the prompts, then plug in some EFI USB install media :-)#asahi · 2021-12-02 14:50
marcan... Apple's boot picker before it reboots because we don't have proper tools to do that programmatically yet, etc#asahi · 2021-12-02 14:50
marcanwouter: well, the process involves at least two mandatory reboots, and to make it as user friendly as possible you need to do streaming partial ZIP file fetches from Apple's CDN, firmware cutting and filename munging, manually creating a macOS stub partition with all the Preboot directory structure (which involves parsing and understanding Apple's manifest plists), also you need to race to kill ...#asahi · 2021-12-02 14:30
marcanthey run Safari in there, they obviously have all the core stuff#asahi · 2021-12-02 14:01
marcanbut then again we're already downloading 1GB for the installs anyway#asahi · 2021-12-02 14:01
marcanQt would be nice, since it has nice Python bindings and can make it easier to test on Linux, though it might be a bit heavy to ship#asahi · 2021-12-02 13:27
marcanall the firmwares will come from the chosen IPSW, we do not mix and match with the running system#asahi · 2021-12-02 13:27
marcanthat comes from the kernelcache that's already being downloaded#asahi · 2021-12-02 13:11
marcanalso choosing what distro/image to install (optionally)#asahi · 2021-12-02 13:11
marcanthe FW isn't copied from an existing macos, it's downloaded from Apple so it would be choosing a version from those we support / are usable on this machine#asahi · 2021-12-02 12:55
marcanyou just need to test it I guess#asahi · 2021-12-02 12:55
marcanthere are fewer tools/frameworks available, but obviously they have quite a bit#asahi · 2021-12-02 12:55
marcangood question :-)#asahi · 2021-12-02 12:06
marcanhonestly with all the rebasing and DT conflicts going on, patches seems painful right now#asahi · 2021-12-02 11:11
marcanI mean, compare that with what corellium did for NVMe...#asahi · 2021-12-02 11:11
marcansven: maybe, but at least it's on the way and built on a sane RTKit WIP#asahi · 2021-12-02 11:10
marcanI'm going to rename the t6k branch to something more useful#asahi · 2021-12-02 11:09
marcanseriously though, people testing known-bad code doesn't really help anyone...#asahi · 2021-12-02 11:09
marcanor just wait until we have a proper SPI driver :p#asahi · 2021-12-02 11:08
marcanwishes people wouldn't be so impatient that they insist on using known-bad-and-obsolete code...#asahi · 2021-12-02 11:07
marcanthe internal keyboard is not supported in u-boot#asahi · 2021-12-02 11:05
marcanmps: the devicetree is passed from m1n1, before u-boot; it is not something you specify later#asahi · 2021-12-02 11:05
marcankettenis: that's... not ideal :-)#asahi · 2021-12-02 11:03
marcanbut yes we can kick it with the reset bit too#asahi · 2021-12-02 11:03
marcankettenis: u-boot should perform the shutdown sequence though#asahi · 2021-12-02 10:52
marcandefinitely not going to work compiling for 32-bit arm :-)#asahi · 2021-12-01 19:23
marcanI've had... a few days full of external IRQs#asahi · 2021-12-01 19:23
marcanand hopefully catch up on reviews and send arnd_ some pulls tomorrow#asahi · 2021-12-01 19:22
marcangoing to get some sleep now :)#asahi · 2021-12-01 19:22
marcanremind me if I don't get to it, shouldn't be a big deal to add#asahi · 2021-12-01 19:22
marcanyeah, but that'd just walking up the tree#asahi · 2021-12-01 19:21
marcanwe already do a bunch of walking for things like disabling devices/DARTs on certain platforms, so...#asahi · 2021-12-01 19:21
marcanthat should work for all machines, right?#asahi · 2021-12-01 19:21
marcancould even do it in code in m1n1 though, it wouldn't be too hard to pick out the serial0 alias and walk the power domain reference adding that#asahi · 2021-12-01 19:20
marcanand yeah, I think we can just have some makefile voodoo to overlay those properties on top#asahi · 2021-12-01 19:18
marcanah, I see#asahi · 2021-12-01 19:17
marcanso given that DT tree tracks all official kernel releases, I'm thinking I can have some minor scripting to transfer changes from one or more of our asahi branches on top of that, and then we can have some includes/overlays if we need something extra#asahi · 2021-12-01 19:15
marcan(that would break earlycon too)#asahi · 2021-12-01 19:15
marcanfwiw the serial console should be powered on boot, m1n1 certainly uses it and won't be shutting it down#asahi · 2021-12-01 19:14
marcanah, yeah, this: https://git.kernel.org/pub/scm/linux/kernel/git/devicetree/devicetree-rebasing.git/#asahi · 2021-12-01 19:13
marcanthough there was already a tree for that IIRC#asahi · 2021-12-01 19:13
marcanah, sure, but I mean not just that, also just extracting the DTs out#asahi · 2021-12-01 19:13
marcanideally I want to automate syncing with linux somehow, while still being able to have patches on top#asahi · 2021-12-01 19:12
marcanI was actually thinking about how to maintain our out of tree devicetrees#asahi · 2021-12-01 19:12
marcankettenis: re power domains, sure#asahi · 2021-12-01 19:10
marcanand it's not just firmware; it's also potentially a devicetree thing (though we will try not to screw up compat there)#asahi · 2021-12-01 19:10
marcanand just track linux kernel release support#asahi · 2021-12-01 19:10
marcanthe tag *could* be macos version, but it's possible they release updates that don't change the ABI (they already have, dcp.py works across a few releases at least) so I feel it makes more sense to have the flexibility to decide what works and what doesn't at the installer#asahi · 2021-12-01 19:09
marcanperfect#asahi · 2021-12-01 19:08
marcanI assume openbsd will at least try to follow the same "blessed" firmware versions we pick for linux, otherwise things will get really confusing :-)