#asahi-dev · 2025-05-24 07:19
svenbut the atcphy side of tbt is included #asahi-dev · 2025-05-24 07:18
sventhat would be another 20+ commits :D#asahi-dev · 2025-05-24 07:18
sventhat branch doesn’t contain any tbt work fwiw#asahi-dev · 2025-05-24 07:18
sven*then #asahi-dev · 2025-05-24 07:18
svenmy plan is to get this upstream, the a useless version of tbt that’s only the native host interface and then see about PCIe tunneling #asahi-dev · 2025-05-24 07:17
svenit’s the combination of what I did at 38c3 and marcan then fixed up in early January #asahi-dev · 2025-05-24 07:17
sventhat’s the plan once that’s submitted and accepted, yes#asahi-dev · 2025-05-24 07:05
svenhttps://git.kernel.org/pub/scm/linux/kernel/git/sven/linux.git/log/?h=wip/atcphy commit descriptions needs some work, i have no idea what pipehandler_workaround is about and i still need to track down that suspend issue. happy with that thing otherwise though#asahi-dev · 2025-05-24 05:38
svenhttps://github.com/AsahiLinux/m1n1/pull/466#issuecomment-2877481496#asahi-dev · 2025-05-24 05:37
svenit still causes issues #asahi-dev · 2025-05-24 05:37
svenSee that gpu initdata PR#asahi-dev · 2025-05-23 18:16
svenokay, let's see. i still need to adjust them to ensure they work with the smc that's on lkml anyway#asahi-dev · 2025-05-23 18:16
svenoh, right, the power button. that needs some smc sub-device, doesn't it?#asahi-dev · 2025-05-23 17:57
sven*a few seconds later#asahi-dev · 2025-05-23 17:57
svendumb question: what's the easiest way to test suspend on a headless m1 mini? ideally something that goes to suspend and then comes back up a few seconds again. there's some comment that atcphy suspend is broken and i want to investigate#asahi-dev · 2025-05-22 15:32
svenyeah, we really don't want to corrupt that thing#asahi-dev · 2025-05-22 13:22
svenbecause if it does you lose access to e.g. all encrypted data on the macos volumes#asahi-dev · 2025-05-22 13:22
svenis there a apfs-fuse driver? we'll have to test that one very well so make sure it doesn't corrupt that xARTs partition in any way#asahi-dev · 2025-05-22 13:19
svenprobably#asahi-dev · 2025-05-22 10:22
svenlooks like smc will be merged after -rc1 🥳#asahi-dev · 2025-05-22 08:45
svenlet me put it differently: I don't care much about apfs support in the kernel honestly but I'm also not going to complain if someone else manages to get it upstream #asahi-dev · 2025-05-22 08:43
svenbut as long as we don't have to maintain it ;P#asahi-dev · 2025-05-22 08:42
svenyup#asahi-dev · 2025-05-22 08:42
svenit would just make it a bit more convenient and less hacky#asahi-dev · 2025-05-22 08:39
svenso unless we want to parse that protocol in kernel space which doesn't seem like a good idea it's all userspace access#asahi-dev · 2025-05-22 08:38
svengigalocker is a key/value store and SEP just requests reads/write from there through its protocol#asahi-dev · 2025-05-22 08:38
svenj`ey: so, uh, open(/dev/nvme0ns0p3) and then just read/write essentially#asahi-dev · 2025-05-22 08:37
sventhe access would happen from asahi-sepd from userspace anyway#asahi-dev · 2025-05-22 08:36
svenso we can kinda bypass all the apfs crap once we know where it is#asahi-dev · 2025-05-22 08:36
svenbut it's stored contiguously on the disk#asahi-dev · 2025-05-22 08:35
svenruntime#asahi-dev · 2025-05-22 08:35
sventhe gigalocker thing we can also hack: figure out the offset of the file from m1n1 inside that disk image and then just use raw disk r/w to modify it#asahi-dev · 2025-05-22 08:35
svenwhat might be interesting for users is that apple container(?) feature where they could have a volume for macos and one for linux that share the available space. not sure if the apfs driver supports this or if that's even something users want#asahi-dev · 2025-05-18 13:01
svenyeah, my point is simply that every method has its issues and that I can live with email thanks to b4#asahi-dev · 2025-05-18 11:14
svenyup, those bots are quite useful. at some point asahilinux/linux got added to their list of repos to pull from and now I get stuff like e.g. https://lore.kernel.org/oe-kbuild-all/202505181857.NVLANqSD-lkp@intel.com/T/#u#asahi-dev · 2025-05-18 11:04
svenoh yeah, it has some issues. but I e.g. also prefer to read the changeset in a text editor to the github web UI#asahi-dev · 2025-05-18 10:59
svenb4 already takes away much of the pain with email imho#asahi-dev · 2025-05-18 10:55
sveni keep forgetting that various CI bots pick up all branches from our github repo and then point out issues in sven/atcphy that I likely would've only found after v1 otherwise #asahi-dev · 2025-05-18 10:26
svenchadmed: mca looks to me now#asahi-dev · 2025-05-18 10:03
svenit certainly wasn't there when I originally brought nvme up#asahi-dev · 2025-05-17 16:56
svenhuh, so there was a mux that allowed to switch the phy between dwc2 and dwc3?#asahi-dev · 2025-05-17 16:54
sventhey switched at some point from dwc2 to dwc3 as well, didn't they?#asahi-dev · 2025-05-17 16:44
sventhat thing is the next series i'll send upstream. atcphy will already have thunderbolt and dp-altmode support but both won't work due to other missing drivers#asahi-dev · 2025-05-17 16:44
svenbut that's the goal#asahi-dev · 2025-05-17 16:42
svenhmm?#asahi-dev · 2025-05-17 16:39
svenor, maybe i should've phrased this more clearly: it's not even ready for review yet#asahi-dev · 2025-05-17 16:30
svenyeah, that's probably why it says "WIP"#asahi-dev · 2025-05-17 15:39
svenhttps://github.com/AsahiLinux/linux/commits/sven/atcphy/ mostly happy with the split into commits now I think. still needs a bit more cleanup though#asahi-dev · 2025-05-17 10:24
svenfwiw, i'll only start picking stuff up again after the merge window. can't hurt to send it out to already get reviews etc. ofc#asahi-dev · 2025-05-17 08:51
svenchadmed_: left some comments, the earlier commits might benefit from Fixes: tags as well