marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | Not ready for end users / self contained install yet. Soon. | General project discussion | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-stream #asahi-offtopic | Keep things on topic | Logs: https://alx.sh/l/asahi
PhilippvK has quit [Ping timeout: 480 seconds]
yuyichao has quit [Ping timeout: 480 seconds]
espo has quit [Ping timeout: 480 seconds]
tpw_rules has joined #asahi
<tpw_rules>
this new DCP code doesn't seem to work for me and my monitors
<tpw_rules>
here is what m1n1 says: pastebin.com/jT9137MR
<tpw_rules>
my stub partition should be 12.0.1
<tpw_rules>
my monitor sees something happen when m1n1 starts but then it just goes to no signal mode again
<tpw_rules>
it also doesn't seem to do anything with EDID? I have a non-1080p monitor and it made a 1080p resolution anyway
<tpw_rules>
okay i take that back it is still generating a weird signal. so it seems exactly as compatible as the 11.x firmware was
<tpw_rules>
is there a way to improve that? i don't know if my monitors have weird timings or what
pavan_ has joined #asahi
yuyichao has joined #asahi
<tpw_rules>
hm, i think pixel timing is defined by the hdmi spec though, right?
<tpw_rules>
like there's only one way to send 1080p60 down an hdmi cable
<tpw_rules>
modulo color
pavan_ has quit [Ping timeout: 480 seconds]
pavan has quit [Ping timeout: 480 seconds]
PhilippvK has joined #asahi
kov has quit [Quit: Coyote finally caught me]
jbowen has quit [Ping timeout: 480 seconds]
c10l has quit [Quit: Bye o/]
c10l has joined #asahi
jbowen has joined #asahi
jbowen has quit [Ping timeout: 480 seconds]
the_lanetly_052__ has quit [Ping timeout: 480 seconds]
jbowen has joined #asahi
jbowen has quit [Ping timeout: 480 seconds]
jbowen has joined #asahi
roxfan2 has quit [Ping timeout: 480 seconds]
roxfan has joined #asahi
jbowen has quit [Ping timeout: 480 seconds]
jbowen has joined #asahi
jbowen has quit [Ping timeout: 480 seconds]
espo has joined #asahi
eroux has joined #asahi
<Glanzmann>
tpw_rules: That is currently to be expected. You can try the alyssas dcp driver that was rebased by jannau. But for me it does not work for two monitors (one one I have what appears to be crash on the other one a coprocessor crashes) - in both cases no screen output.
<espo>
morning :)
<Glanzmann>
Good morning.
espo_ has joined #asahi
espo__ has joined #asahi
espo has quit [Ping timeout: 480 seconds]
espo has joined #asahi
PhilippvK has quit [Ping timeout: 480 seconds]
espo_ has quit [Ping timeout: 480 seconds]
espo__ has quit [Ping timeout: 480 seconds]
MajorBiscuit has joined #asahi
grange_c4 has joined #asahi
mort_6 has quit []
grange_c4 is now known as grange_c
grange_c has quit [Quit: Ping timeout (120 seconds)]
mort_6 has joined #asahi
Major_Biscuit has joined #asahi
jbowen has joined #asahi
MajorBiscuit has quit [Ping timeout: 480 seconds]
jbowen has quit [Ping timeout: 480 seconds]
chadmed has quit [Read error: Connection reset by peer]
chadmed has joined #asahi
dianshi has quit []
chadmed has quit [Read error: Connection reset by peer]
chadmed has joined #asahi
aleasto has joined #asahi
eroux_ has joined #asahi
eroux has quit [Ping timeout: 480 seconds]
jbowen has joined #asahi
eroux_ has quit []
<_jannau_>
tpw_rules: there's nothing in the iboot dcp interface (used by m1n1) which indicates supported or preferred modes. so m1n1 currently selects the mode with the highest resolution/refresh rate limited to 1920x1080@60. hoping that that would be a mode most monitors support somehow
jbowen has quit [Ping timeout: 480 seconds]
eroux has joined #asahi
<_jannau_>
tpw_rules: the linux dcp driver is mostly the same. I rebased it to the same base as asahi (which apparently broke hotplug and maybe modesetting) and did minimal changes to adapt it to the macos 12 dcp firmware
le0n_ has quit [Remote host closed the connection]
aleasto has quit [Remote host closed the connection]
bit_shifts has joined #asahi
agraf_ is now known as agraf
mort_6 has quit []
mort_ has joined #asahi
bit_shifts has quit []
bit_shifts has joined #asahi
Gaspare has joined #asahi
the_lanetly_052__ has joined #asahi
bit_shif_ has joined #asahi
bit_shif_ has quit []
Gaspare has quit [Read error: Connection reset by peer]
bit_shifts has quit [Ping timeout: 480 seconds]
jbowen has joined #asahi
<tpw_rules>
_jannau_: i thought macOS used a somehow more advanced dcp protocol than iboot
<tpw_rules>
if linux uses the iboot one, that doesn't bode well for my monitors...
<_jannau_>
the linux driver uses the same protocol as macos, m1n1 uses the iboot one
yuyichao has quit [Quit: Konversation terminated!]
jbowen has quit [Ping timeout: 480 seconds]
<_jannau_>
the standard protocol has scores for the modes and setting the mode with the hightest score works a full HD and a 1600x1200 display I tested so far
<_jannau_>
the full hd has two modes with the same highest score, macos doesn't use the same mode
<_jannau_>
so iy might be worth checkling if using the last mode with the max score instead of the first one helps
the_lanetly_052__ has quit [Remote host closed the connection]
espo_ has joined #asahi
jbowen has joined #asahi
<Glanzmann>
I updated https://tg.st/u/asahi-debian-live.txt it now has a newer kernel, can do wifi (firmware needs to be brought though) and more utilities. I also automated the whole build will upload the git repository soon.
<Glanzmann>
Still need to add the instructions on how to install debian with it.
espo has quit [Ping timeout: 480 seconds]
espo_ is now known as espo
<tpw_rules>
_jannau_: ahh, i haven't tried the linux drivers yet
<tpw_rules>
Glanzmann: maybe worth putting to use 12.0.1 in the installer
<tpw_rules>
not sure we necessarily want it to be the newest version?
<Glanzmann>
tpw_rules: Marcan said he will soon update it to 12.1 and we will use that as soon as it is there.
<tpw_rules>
what on earth does the program "eatmydata" do
<Glanzmann>
tpw_rules: It eats 'syncs' so the nvme in the apple becomes fast.
<Glanzmann>
dpkg does a sync after each operation which can really take time if you wait for the disk flushes.
<tpw_rules>
hm, i thought he said we were going to stay on 12.0.1
<tpw_rules>
but i could have a bad memory
<Glanzmann>
tpw_rules: Not to my understanding, but maybe I misunderstood.
<Glanzmann>
tpw_rules: Jannau and marcan had a conversation yesterday or the day before IIRC. That they will soon bump it to 12.1 and deprecate the 11.x versions.
pg12 has joined #asahi
pg12_ has quit [Ping timeout: 480 seconds]
<espo>
the current u-boot/m1n1 also works with 12.(x) not sure which number it was ( i applied it the first time to my dev install which run´s the latest beta of osx )
<Glanzmann>
espo: So you're not using the asahi installer?
<sven>
i think the only reason to stay on 12.0 instead of 12.1 is that the installer doesn't support 12.1 yet
<espo>
now i use the asahi installer
<Glanzmann>
espo: I see, got it.
<sven>
but once that's fixed I don't see any reason to stay on 12,0
<tpw_rules>
ah okay
<espo>
That was on my first totally noobish try to get something booted :D ( and i didn´t know what i´m was doing, didn´t changed much but i use the stub install from the installer now :) )
<Glanzmann>
espo: In the old days you had to it that way. But the Asahi installer was not promoted enought and also had no good explanation that you had to click Restart and so on, so only the people who read this channel probably used it.
<Glanzmann>
do*
<espo>
ah okay :)
tpw_rules has joined #asahi
<Glanzmann>
But marcan did a very fine job with the installer because now you can install Linux in minutes, what took hours before.
<espo>
thats true! If i got it working nearly everyone can do so ;P
Gaspare has joined #asahi
<Glanzmann>
espo: I was yesterday recording a video how to install Linux on the mac mini but I hit a major road block because I created the esp parition wrong, I'll try today again.
<espo>
ah cool! :)
<Glanzmann>
I hope to be able to cut the video to be something like 10 minutes, but we will see.
VinDuv has quit [Quit: ZNC 1.8.2+deb2+b1 - https://znc.in]
VinDuv has joined #asahi
<tpw_rules>
Glanzmann: do you have stable links for any of this stuff yet? i might put links on my nixos guide for people who want something else
<Glanzmann>
tpw_rules: Not yet, but soon.
<Glanzmann>
tpw_rules: I'll host the git repository and put the rest in the asahi wiki.
<Glanzmann>
tpw_rules: I'll ping you when I'm ready.
<tpw_rules>
okay. maybe there could be a guide section of the wiki so i can link mine
<Glanzmann>
That is the idea. I want to write a way how to install Debian on m1. Make a video and also link the repository if you don't want third party build artefacts to use in the process.
<tpw_rules>
i plan on hosting build artifacts using github's release feature. waiting on a couple upstream fixes so they are properly reproducible though
<Glanzmann>
I see.
the_lanetly_052 has joined #asahi
<tpw_rules>
i guess you have your own git server?
<Glanzmann>
tpw_rules: Yep, I thought about putting it on github, but I'll put it on mine, I guess.
the_lanetly_052 has quit [Ping timeout: 480 seconds]
aleasto has joined #asahi
aleasto has quit [Remote host closed the connection]
<j`ey>
(but that's after you've resized the partition, run the installer and kmutil'd m1n1)
<jannau>
you dont have to do anything after running step2.sh from the installer
<j`ey>
oh yeah sorry, step2.sh kmutils m1n1 for you
<sven>
once step2.sh is done you probably want to build your kernel and use cat build/m1n1.macho Image.gz devicetree.dtb > payload.macho and then python3 proxyclient/tools/run_guest.py -S payload.macho to run it under a small hypervisor which will expose the serial console on the second emulated CDC ACM interface
<Glanzmann>
axboe: I uploaded the old version which is close of what you need: https://tg.st/u/asahi.txt
<Glanzmann>
jannau: From the asahi branch is still the second spi patch missing, isn't it?
<Glanzmann>
jannau: If I can give dcp another try, let me know.
<Glanzmann>
I also have three more monitors that I could try.
<Glanzmann>
jannau: Are you using the dcp also on the laptop?
<tpw_rules>
Glanzmann: how do your monitors work in m1n1?
<Glanzmann>
tpw_rules: Good, after marcan fixed it. But only with the standard resolution. I have three eizos that I bought in the last 17 years and two 4K TVs which I did not try.
<tpw_rules>
"standard resolution"
<tpw_rules>
?
<VinDuv>
1080p probably, this is a hard coded limit in m1n1
<Glanzmann>
tpw_rules: 1080p
<tpw_rules>
i have an HP 1680x1050 monitor which half works but it gets sent 1080p
<tpw_rules>
and an acer 1080p monitor which does not work
<Glanzmann>
tpw_rules: I see. If you want I can later try all the monitors that I have in my house and let you know.
<VinDuv>
but AFAIK iboot in < 12.0 did the same so it’s not sure whether larger resolutions would work reliably
<tpw_rules>
i am not sure if that would be helpful
<tpw_rules>
you said you already tried three monitors?
<Glanzmann>
tpw_rules: Yes.
<jannau>
Glanzmann: no need for further tests, I haven't tried dcp on the t600x, no need to create more problems while there is still enough to do on the mini
<newchima[m]>
Glanzmann: was just looking at your script for your build and you are referencing your config file 2022-01-19 rather than 2022-01-24 (which has the module for mini sound card)
<Glanzmann>
newchima[m]: Which one?
<Glanzmann>
newchima[m]: I made a quickstart video to install Debian: https://tg.st/u/m1debian.mp4 rough around the edges.
<mps>
Glanzmann: FS to dd could be good to have, though I would prefer installer on usb
<jannau>
Glanzmann: asahi-dcp is at 35dbff7da3d61bf25e9f69674f384c22452c98f9
hwatwasthat[m] has joined #asahi
<j`ey>
Glanzmann: lol 'image' in grub everytime
<Glanzmann>
j`ey: I know ...
<Glanzmann>
j`ey: I just checked my knowhow file, a older version of grub called 'linux' 'kernel'. I typed it to often. So if it does not work, I get confused. Late hour and not much sleep last night does not help. :-)
<Glanzmann>
I'm looking forward when marcan finishes the installer and does all what I do with a single command.
yuyichao has joined #asahi
<j`ey>
Glanzmann: was it meant to just randomly stop?
<Glanzmann>
j`ey: I paused at some point when I rebooted, sorry.
<j`ey>
Glanzmann: you were waiting for it to get a login prompt, then the video stopped
<Glanzmann>
When I was fighting with grub, but the resolution is all there. :-)
<Glanzmann>
jannau: I compiled it, and it does crash for me when I try to start xorg, I think.
<Glanzmann>
jannau: Is there a way I can get you debug output?
<Glanzmann>
I compiled with =y
<Glanzmann>
Have to rebuild. It was switched to =m.
<j`ey>
Glanzmann: the recording stops a few mins after you disable ethernet and try to use wifi only
yuyichao_ has quit [Ping timeout: 480 seconds]
<Glanzmann>
j`ey: No, it was not supposed to do that.
<Glanzmann>
Probably forgot to resume the video. I did not resolve the wlan thing and threw it out again.
yuyichao_ has joined #asahi
<Glanzmann>
jannau: Can you pleae send me your config?
<Glanzmann>
jannau: Are using xorg or wayland with the driver?
<Glanzmann>
newchima[m]: I updated all the build artefacts (live, partition image) with the right kernel config. Thank you for spotting.
<Glanzmann>
I also had to throw out firmware-linux because it is no longer in testing ...
<jannau>
both, Xorg for sddm and then kwin/wayland
yuyichao has quit [Ping timeout: 480 seconds]
<Glanzmann>
okay, trying with your config.
<jannau>
I think there is nothing except DRM_APPLE to make it work
Gaspare has quit [Quit: Gaspare]
<Glanzmann>
jannau: Okay. So I have now booted your kernel. I autostart xorg (lightdm) but simpledrm is loaded.
<Glanzmann>
Should I stop x and than unload simpledrm and load 'apple' or can I do it while x is running?
<Glanzmann>
Okay with your kernel config it is compiled in, so no loading necessary.
<jannau>
Glanzmann: have you changed DRM_APPLE to a module? it's built-in in my config. are you sure you have the right device tree? is there anything match dcp in dmesg?
<Glanzmann>
No dcp in dmesg.
<Glanzmann>
I'm booting via u-boot.
<Glanzmann>
Should I boot via m1n1?
<Glanzmann>
And use the device tree of the kernel, right?
<jannau>
you need the device tree of the kernel
<Glanzmann>
Should the device tree of the kernel work with u-boot?
<Glanzmann>
Or do I have to modify it to work with u-boot?
<Glanzmann>
For now I'll use m1n1.
<jannau>
yes, it should work. the only downside is the annoying installation of m1n1+dtb+u-boot via recovery
<Glanzmann>
no issue, I see.
<jannau>
for testing booting via proxyclient is more convenient. if you don't have it already you could dual boot between two stub installs. one with m1n1 + u-boot payload and one without payload
<Glanzmann>
jannau: It works like a charm.
<Glanzmann>
Thank you for this great piece of software. ;-)
ciggi has joined #asahi
<Glanzmann>
jannau: Does your tree have everything someone needs or should I merge it in another tree?
<Glanzmann>
jannau: So I tried rotation works, but the scrolling of a fullscreen xterm appears to be slower than simpledrm. Trying a movie now.
<Glanzmann>
jannau: Browser scrolling is fast.
<jannau>
it is asahi + the 2 additional spi-hid patches + dcp by Alyssa (with rebasing, porting and debugging by me)
<Glanzmann>
Got it.
<Glanzmann>
So only thing is missing is povics sound patch.
jx0 has joined #asahi
<Glanzmann>
jannau: The videoperformance also seems to be good for me. Best results for me is with mpv -vo x11 ...
<Glanzmann>
jannau: HDMI hotplug also works for me.
<Glanzmann>
ciggi: I would make, space, use marcans installer, forget about the wifi, you have ethernet, don't you? Than install u-boot and either with the instructions for the video or the usb live.
<jannau>
ciggi: arm64 defconfig builds builds a kernel with 4k page size. the iommu for PCIe works currently only with 16k page size. Page size can be changed under "Kernel features" in Linux' menuconfig
<ciggi>
Thank you for that video still watching :)
<Glanzmann>
ciggi: And from what you're saying you probably have the wrong dtb or you're missing simpledrm config.
<Glanzmann>
ciggi: The funny thing is the testrun for the video went smoother than the video itself. :-)
<jannau>
the usb-a ports on the mac mini are connected via pcie, the usb-c ports should work, at usb 2.0 speed though
<tpw_rules>
the usb-c ports do work
<ciggi>
I missed whole efi partition means I run installer and all even setup2.sh but somehow I forget I thought .macho will care :headbang
<ciggi>
yes I have ethernet but this guide says upload wifi firmware but idk where should I uplaod (https://tg.st/u/asahi.txt)
<ciggi>
upload*
<ciggi>
@jannau are you using 16k page size? is chromium working? last time I tried it failed
jx0 has quit [Quit: poof!]
jx0 has joined #asahi
<chadmed>
the kernel needs to be built with 4k pages for x86 translation layers to work, correct?
<chadmed>
its pretty useless until we get a kernel driver for the gpu going in any case. really the only thing youd need to translate on linux these days is games since virtually everything else is available for aarch64 in any good distro
<chadmed>
and games need 3d accel, so for fex to be useful to the masses on these machines we need a gpu driver
<mps>
chadmed: it is quite useful for me in current state (I don't play games but use machine for work and $day_job development)
<j`ey>
"its pretty useless" meant FEXemu, not the m1s themselves
<mps>
ah, sorry then
<mps>
first time see fexemu term
jbowen has quit [Ping timeout: 480 seconds]
<chadmed>
fex is a fast x86/amd64 to aarch64 user space binary translator
<chadmed>
after a couple of days of testing and sifting through logs, the current bottleneck for the games i tested (mostly source engine) is the 4 translation layer deep graphics stack, not the binary translation
<chadmed>
the source engine build for *nixes does not actually have a native opengl render pipeline, instead it integrates toGL, which translates the engine's dx9 draw calls to opengl transparently. under macos this then needs to be translated to vulkan and then to metal
<chadmed>
the bottleneck is going from vulkan to metal, and under linux you would obviously not have to worry about this
<chadmed>
i predict that linux performance for such workloads will initially be marginally worse just because the agx support in mesa isnt mature but it will be an interesting experiment nonetheless
<chadmed>
plus, having these sorts of workloads at the very least working will make it easier to find where exactly the agx compiler could be improved