03:17 <shenki> i'm loading the osx kernel under m1n1, but don't have a second mac or serial hardware

03:17 <shenki> so i'm relying on m1n1 staying in charge of the usb port

03:17 <shenki> under linux, i can set the xhci device and dart to disabled in the device tree

03:17 <shenki> is there a way to blacklist the driver for a usb port under osx?

05:57 <jannau> shenki: the m1n1 hypervisor keeps control of the usb port it is running on by removing it from the ADT

07:13 <VinDuv> sven: I can confirm that building a custom debug kernelcache without AppleSerialShim makes cu.debug-console works in 12.0.1. I think I’ll also try to rename "uart0" in the ADT with the hypervisor, given how AppleSerialShim matches devices it might also work…

07:26 <bgb> VinDuv: can you share the buiding guide?

08:01 <sven> yeah, that building guide would be very useful

08:01 <sven> i kinda gave up at that point because building a custom kernel sounded painful

08:23 <VinDuv_> bgb, sven: I basically followed the beginning of https://github.com/AsahiLinux/docs/wiki/SW:Hypervisor to create the dev kernel macho and then installed it with kmutil configure-boot as if it were m1n1

08:24 <sven> ohhh.... true. i totally forgot that custom kernel for the hv thing

08:24 <sven> i guess i should also teach the linux tipd driver to behave like macvdmtool as well

08:24 <sven> and then just switch to linux on my development machine

08:25 <VinDuv_> AIUI the kmutil inspect part of the command builds the list of extensions that will be included in the kernelcache; it takes all of them except SEPHiber. I just changed it to: kmutil inspect -V release --no-header | awk '!/SEPHiber|SerialShim/ {print " -b "$1; }'

08:26 <VinDuv_> I'm guessing there is a way to do this without installing the dev kernel though, and maybe without using configure-boot?

08:26 <sven> i've briefly looked into it and neither found a boot argument to specifically disable SerialShim nor another one to just block some kext from loading

08:27 <j_ey> I dont understand why *removing* AppleSerialShim makes it work

08:27 <sven> afaiu AppleSerialShim has a higher priority than the real serial driver

08:27 <j_ey> :<

08:27 <sven> so if that kext doesn't exist it'll fall back to AppleSamsungSerial or however it's called again

08:28 <sven> and AppleSerialShim is quite literally a shim. all it does is ack any serial irqs it gets

08:28 <VinDuv_> It seems to be "more specific" in its device matches than SamsungSerial so I guess that's why it takes priority

08:28 <sven> probably. i think there's also a probe priority but i didn't check if the serial shim has a higher value there

09:14 <amw> j_ey: Is this commit in linux-next a0f160ffcb83 pinctrl: add pinctrl/GPIO driver for Apple SoCs has a compile problem - see https://paste.debian.net/1218157

09:15 <amw> Sorry I meant do you have fix for this?

09:15 <j_ey> one sec

09:15 <j_ey> https://lore.kernel.org/linux-gpio/20211101151836.GA6547@e124191.cambridge.arm.com/T/#t

09:16 <amw> j_ey: ok - ta

10:08 <amw> ok - I got a message that some one got the linux-next tree working with USB keyboard and mass storage but even after fixing the above compile error I can't get USB keyboard to work???

10:08 <amw> Now trying with make mod2yesconfig to see if it helps as that was suggested too...

10:39 <sven> there are no usb nodes in linux-next yet

10:41 <amw> Ok - thanks I'll try a device tree from asahi linux then

10:45 <amw> sven: Thanks a million - that was it - now getting USB keyboard working again.... with linux-next build

10:47 <sven> sure. i should prepare the apple dwc3 bindings and then resubmit that usb reset quirk series. then we can also add those nodes to linux-next

10:49 <mps> what I could expect if install linux on M1 to work, and what does not work

10:49 <mps> ok, gpu still doesn't and iirc audio

10:50 <chadmed> usb4/thunderbolt, the gpu and builtin audio are the biggies for the mac mini

10:50 <mps> I have macbook pro

10:50 <chadmed> for the macbooks its the same plus no touchbar

10:50 <chadmed> i think there were issues with the display controller too

10:51 <mps> yes, forgot to mention that I know that touchbar doesn't work

10:51 <mps> does wifi works

10:52 <mps> for audio I could use external usb dongle so no big issue (for now)

10:53 <sven> usb3 also doesn't work on the type-c ports, only usb 2

10:53 <chadmed> i think wifi works with alyssa's collected patches. it definitely does on the mini, not sure if its the same on the laptops

10:53 <mps> sven: so speed is not so high

10:55 <mps> and laptops only have usb-c ports

10:56 <mps> what kernel to build? mainline or something else?

10:57 <chadmed> https://github.com/mu-one/linux/tree/20211022

10:58 <kettenis> wifi should work just fine on the laptops; you just need to install the right firmware files for your model

10:59 <mps> kettenis: thanks for info. hope will find proper one

11:01 <mps> probably it is somewhere in macos FS, but I don't have any experience with macos (except looking few times what it have on FS)

11:31 <amw> sven: Is there any work yet on adding the apple-spi master so the builtin keyboard can work?

11:33 <sven> don't think so

11:34 <sven> i believe marcan said he eventually wanted to work on writing that spi driver

11:38 <amw> ok - thanks - have to keep squeezing in the corellium one for the moment I guess

11:47 <arnd> amw: the corellium spi driver seems fine for the most part, someone could just submit that. This driver is fairly independent of everything else, so there are not a lot of things one could do differently. The main missing bits are documentation for the DT binding, and properly hooking it up to the clk and pinctrl drivers. In the corellium dts the spi controller links to two nodes handled by their clk driver, but that is not upstream

11:48 <kettenis> I should finish my proposal for the DT binding

11:48 <kettenis> there is some cruft in the corellium driver that doesn't seem to be necessart

11:55 <sven> The clocks are probably just power domains or maybe the ref clk if the driver also configures the bus frequency like i2c

11:56 <kettenis> it needs the ref clock

11:56 <kettenis> the controller has its own divider

11:57 <amw> It would be very nice to have that in there some how with USB it would make a USB installable M1 much more viable

11:57 <kettenis> and the ADT suggests there are indee power domains

11:57 <kettenis> but they all seem to be powered on already