17:27 <jmr2> alyssa: you mentionned a while ago that you might be interested in merging j_ey's version of Corellium's spi+keyboard patches. I have the keyboard working on the Air, on top of your 20211022 branch. Let me know. j_ey: thanks for the work, and kettenis: thanks for yesterday's DT commit that gave me the clues I was missing.

17:28 <jmr2> *merging in your tree, not upstream...

17:32 <j_ey> jix: might interest you ^

17:32 <jix> is there a tree with a separate DT for the air already? I'm trying to get an overview of things (just got my HW yesterday) and I noticed that I need to adjust the DT (remove 2 pcie darts) to get 20211022 to boot on my air

17:33 <jix> and yeah I am very much interested :)

17:33 <jmr2> No separate DT that I'm aware of. 20211022 boots successfully on my Air as-is.

17:34 <jix> huh interesting I was getting a panic when initializing the 2nd pcie dart, and I checked the ADT on the device and it indeed just has one for me... is that not expected?

17:43 <jix> https://paste.gg/p/anonymous/95d9a9c8e35e425f9922dc536a371b4d << that's what I get with the DT from the 20211022 tree and the dart@682008000 and dart@683008000 are not in the ADT I dumped

17:49 <kov> jmr2, nice! where can I find your branch? =)

17:49 <kov> I spent a few hours trying to get that to work as well, but I'm a noob, only managed to get some deadlocks ;D

17:49 <jix> this is my adt https://paste.gg/p/anonymous/d9cc00fd6da240a390055a887fb9da83

17:50 <j_ey> jix: that has some personal info in

17:50 <jix> j_ey: does it?

17:50 <jix> (I think I removed that part, but I might have screwed that up)

17:51 <jmr2> jix: difference might be that you're running the HV and I'm not. Not sure.

17:51 <j_ey> jix: so you did

17:51 <jmr2> kov: no branch, but here's a summary: http://paste.debian.net/plainh/86b3bf6c

17:52 <jix> j_ey: remove it or screwed it up? ^^

17:52 <j_ey> jix: remove

17:54 <kov> jmr2, great, thanks, will give it a go in a bit! I think my dts looked almost like that, but I did not have the last bit, I assume the second to last is essentially cleanup right?

17:55 <j_ey> yes

17:55 <j_ey> if you mean the interrupts-extended bit

17:55 <kov> exactly

17:56 <jix> jmr2: I think I also tried it without the HV, but let me double check... in any case the DT wouldn't match my ADT

17:58 <jix> yeah even without the HV I do get a panic in apple_dart_probe while probing the one at 682008000 which isn't in my ADT... so are ther different HW revisions? or what is happening here?

18:05 <jmr2> kov: second to last was required. Othewise, it was using aic as the parent, and failing to initialize, trying to read a 3rd interrupt cell. No idea why - I don't pretend to know what I'm doing...

18:06 <jmr2> jix: no idea

18:10 <j_ey> probably related to the last change.. but the -extended is the way we should do it anyway

18:22 <jix> ok, the macbook air ADT marcan posted a while back also only has the one pcie dart, so I asssume jmr2 might have used a config without CONFIG_APPLE_DART?

18:24 <jix> probably would be a good idea to maintain different DTs at this point then and not put everything in the common t8103.dtsi

18:25 <marcan> jix: the .dtsi is the SoC, the bits are supposed to be specialized in the .dts files that include it

18:25 <marcan> the parent file can enable/disable blocks and such, and override things

18:25 <jix> ah so all darts would be in the dtsi but the dts for the air would disable 2 of them?

18:25 <marcan> yes

18:25 <marcan> or possibly the converse

18:26 <marcan> note how the dts already enables the UARTs selectively

18:27 <marcan> meanwhile today I, uh, spent the whole day doing panic data recovery on a friends' iPhone :/

18:27 <jix> well then let me rephrase: it'd be a good idea to maintain the DTs in a way that doesn't require manually patching things when not using a mac mini ;)

18:27 <marcan> not exactly how I planned to spend the day. also the entire ecosystem of jailbreak/etc tooling is a gian mess outside of standard use cases

18:28 <marcan> I think I was 20 levels deep in the problem stack and about to give up when I had a really lucky break and something fixed itself enough to let me continue and make it work again

18:29 <marcan> jix: that's called we need mba/mbp DTs

18:29 <marcan> m1n1 already has the ability to filter DTs by machine, I added that recently (it won't even boot a mini DT on an air any more)

18:30 <marcan> so you just cat all of them together and it finds the right one

18:31 <jix> I'm still a bit lost regarding what happens in which trees

18:31 <marcan> yeah... after 5.16 I'm going to finally put together a bit of a canonical branch with this stuff

18:31 <marcan> it's just all over the place right now

18:31 <marcan> but a lot of things are getting mainlined

18:32 <marcan> there were a bunch of weirdo dependency chains that are getting simplified now

18:32 <j_ey> still very dependent on mailbox/rtkit

18:32 <marcan> but feel free to send me some patches to add a DT for the air (and/or mbp); non-schema DT changes like these are easy to review, just CC the usual folks/MLs so everyone gets a chance to see it :)

18:33 <marcan> j_ey: yes, that's the next big blocker, for nvme and soon smc and gfx and more

18:33 <marcan> also power states, but you can kind of ignore that the way the APIs work and just enable stuff in m1n1 prior to booting, kinda

20:39 <j_ey> marcan: jix https://termbin.com/8wip (nice, wip in the url) wip j313 dts

20:39 <jix> oh I also just started that

20:40 <j_ey> I think it might be better to have status="disabled" in t8103.dtsi.. than haveit this way around

20:40 <jix> I did disable all darts in the t8103.dtsi and then enabled all three in 274... yeah exactly

20:42 <jix> https://github.com/jix/linux/commit/13693c7dcce728b98f4cccd231bee195c5629f21 is what I did

20:43 <j_ey> makes sense to me I think

20:45 <jix> I completely lack the big picture regarding everything, but listing stuff that is present and not stuff that is absent does seem clearer to me

20:46 <jix> if I have a working /dev/input/event1 (using the patches jmr2 collected) can I attach it to the console? can I do that automatically? I fail at finding the relevant documentation for this stuff

20:47 <j_ey> set console=tty0 in your commandline args

20:48 <j_ey> im not sure of how to make it the default, other than that

20:50 <jix> unrelated, but something freezes for me sometimes if I leave linux running in the HV for longer ... virtual uart and hypervisor ^C don't respond anymore, but the devices are still there until I hard reboot

20:50 <jix> console=tty0 did the trick, thanks :)

20:51 <j_ey> I get that sometimes, dunno why

20:51 <jix> although now I'm puzzled how that ends up doing this, given that without it, the keyboard input didn't end up in /dev/tty0 AFAICT

20:54 <jmr2> j_ey, kov: you're correct - it's the added interrupt-cells that's important. interrupt-parent+interrupts and interrupts-extended both work. However, using interrupts-extended give a "Missing property '#interrupt-cells'" warning when compiling the DTB, which is not present when using the two-lines syntax. I only tested with interrupts-extended after fixing that warning. Sorry for the confusion.

20:55 <j_ey> jmr2: no worries

20:56 <jix> jmr2: do you have CONFIG_APPLE_DART set? if not that would probably explain why you didn't have the issues I had with the DT

20:57 <jmr2> jix: yes, it's set. I started from alyssa's config, available here: www.rosenzweig.io/.config

20:58 <jmr2> But given that you have a workaround, I'm not sure it's worth digging that much.

20:58 <jmr2> Could be because I'm using a version of m1n1 from a little while ago. Given marcan's comments, updating is on my list.

21:03 <kettenis> there are a few different ways how we can handle the different models

21:04 <kettenis> the DTs for the different models I mean

21:04 <kettenis> one question I have is what we should do with the nodes for the root ports

21:05 <kettenis> my current thinking is that the mba/mbp should only have one root port node and the mini should have all three

21:28 <jmr2> marcan: just FYI, m1n1 2736f0d (Sept. 24th, on stub partition) chainloading 73180e2 (latest) and then running linux.py still successfully boots an air with a j274 DT.