#asahi-dev · 2022-03-26 16:28
marcanwhen's that?#asahi-dev · 2022-03-26 16:28
marcanI should send another wifi series out, that's probably the only one likely to cause conflicts longer term and it's mostly reviewed already anyway#asahi-dev · 2022-03-26 16:28
marcanI'll give it a shot next time I rebase and see how it goes, it's getting more reasonable now that most patches to existing things are in#asahi-dev · 2022-03-26 16:27
marcanI mean I won't do X.Y.Z releases but if you want those hopefully those merges won't be painful#asahi-dev · 2022-03-26 16:27
marcanI can probably just say the base is stable tbh, though I'll only rebase on top of X.Y versions, I won't be maintaining minor merges#asahi-dev · 2022-03-26 16:26
marcanstable as in normal kernel releases, or as in longterm?#asahi-dev · 2022-03-26 16:25
marcanand hopefully not too much explodes#asahi-dev · 2022-03-26 16:25
marcanwhile you'd pull in cur too#asahi-dev · 2022-03-26 16:25
marcanand then I can have all the bits/ branches base off of an RC or whatever in general, and our main -next merge would happen between next and only the bits/ branches#asahi-dev · 2022-03-26 16:25
marcanhow about when something gets upstreamed, I pull that upstream tip into a bits-cur/ or something, or rebase it if it has too much unrelated history#asahi-dev · 2022-03-26 16:24
marcanbut basically it's not actually too bad for us to pick a different base point, or merge linux-next after or split up bits/ in a way to make that easier#asahi-dev · 2022-03-26 16:23
marcan(which is why a couple bits/ branches base on each other, it's always MAINTAINERS changes, hint hint sven)#asahi-dev · 2022-03-26 16:23
marcanI have some scripts to do that and the merge is supposed to be clean#asahi-dev · 2022-03-26 16:23
marcanfor us asahi is just an auto-merge of our bits/ branches, which are frequently rebased on top of whatever linux-next I'm going by#asahi-dev · 2022-03-26 16:22
marcanright#asahi-dev · 2022-03-26 16:22
marcanyeah, we should do that soon#asahi-dev · 2022-03-26 16:21
marcanbut we don't *have* to do that#asahi-dev · 2022-03-26 16:20
marcanthe main reason for tracking linux-next is so that we can pull in upstreamed changes ASAP, without waiting for the next merge window#asahi-dev · 2022-03-26 16:20
marcanI have some merge scripts now anyway, so it would be fairly easy to offer/test both anyway#asahi-dev · 2022-03-26 16:20
marcanok, then that's a good reason for us to try to avoid linux-next bases in the future, at least for our main branch#asahi-dev · 2022-03-26 16:20
marcanEighth_Doctor: so just torvalds master?#asahi-dev · 2022-03-26 16:20
marcanjannau: we still need to poke that#asahi-dev · 2022-03-26 16:19
marcanEighth_Doctor: re the kernel, what does fedora track for ARK?#asahi-dev · 2022-03-26 16:10
marcanbut I wanted to prove to myself we can do module autoloading "properly"#asahi-dev · 2022-03-26 16:10
marcanor if you explicitly modprobe it some other way#asahi-dev · 2022-03-26 16:10
marcanif you build the nvme driver and all its dependencies into the kernel you don't need it, of course#asahi-dev · 2022-03-26 16:10
marcanhttps://pagure.io/fedora-asahi/project/issue/4#asahi-dev · 2022-03-26 16:10
marcanwithout that you either end up with firmware not working on first boot / being out of step, or the firmware pull script blocking on nvme that won't come until it's done#asahi-dev · 2022-03-26 16:09
marcanEighth_Doctor: the udev trigger thing is to make udev trigger the nvme driver, *then* the firmware pull script runs, *then* normal udev trigger runs and loads the broadcom wifi driver#asahi-dev · 2022-03-26 08:42
marcansoon...#asahi-dev · 2022-03-26 07:22
marcanif you really start digging into the lower address bits you could probably map out the interleaving, but that wouldn't be useful for software anyway#asahi-dev · 2022-03-26 07:22
marcanmacOS can't do NUMA anyway#asahi-dev · 2022-03-26 07:21
marcansame as e.g. a threadripper in that mode#asahi-dev · 2022-03-26 07:21
marcanjannau: I expect it to be fully interleaved and thus UMA#asahi-dev · 2022-03-26 06:27
marcanthat tells me they taped out t6000 before finding the first one, then found it before taping out t6001, then found the second one, or something like that#asahi-dev · 2022-03-26 06:26
marcanwhile the unknown HID18_SPAREBIT17 got added in rev >0x10 on both#asahi-dev · 2022-03-26 06:25
marcanso for example, HID4_ENABLE_LFSR_STALL_LOAD_PIPE_2_ISSUE got added in t6000 rev >= 0x10, but base t6001 has it already#asahi-dev · 2022-03-26 06:24
marcanif you look at the chicken bits, you can see they can mostly be sorted chronologically by looking at what got fixed/added, and you end up with interleaved soc/revision pairs#asahi-dev · 2022-03-26 06:23
marcanI think apple forks their cores per SoC, then backports fixes and follows a different revision number lineage for each tape-out#asahi-dev · 2022-03-26 06:20
marcanmaz: t8103, t6000, t6001/2 cores are all slightly different revisions, that's why they have different MIDRs#asahi-dev · 2022-03-25 10:31
marcanIIRC there was always the chroot issue#asahi-dev · 2022-03-25 10:31
marcanI don't remember it ever working for me#asahi-dev · 2022-03-25 08:43
marcan(I only have the templates)#asahi-dev · 2022-03-25 08:43
marcanthere's runtime stuff in there I'm interested in#asahi-dev · 2022-03-25 08:43
marcanjannau: can you paste the t6002 adt somewhere?#asahi-dev · 2022-03-25 08:07
marcanmove to -re#asahi-dev · 2022-03-25 08:07
marcanit looks like it could've come from traces though, which implies there isn't really anything copyrightable in there#asahi-dev · 2022-03-25 08:06
marcanand the script is mostly a pile of writes, which follows from the docs and an HV trace#asahi-dev · 2022-03-25 08:06
marcanthe actual docs you have are just register docs and none of that worries me#asahi-dev · 2022-03-25 08:05
marcanrqou_: is that all from driver RE, or hypervisor traces, or both?