marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | https://asahilinux.org/2022/03/asahi-linux-alpha-release/ | General project discussion | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-alt #asahi-stream #asahi-offtopic | Keep things on topic | Logs: https://alx.sh/l/asahi
Brainium has quit [Quit: bye]
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
<chadmed_> did you recompile the DT and make a new stage 2 m1n1 with it?
<probie> Not yet. Given that there's a chance to break the speakers I'm trying to get a clear idea of what's involved before attempting anything
<probie> Is that harmless to do as long as I don't do anything else?
<chadmed_> ive done the procedure myself quite a few times and ive never busted anything but you must make absolutely certain that your machine will play no audio until youve had a chance to turn the levels down in alsamixer
bisko has quit [Read error: Connection reset by peer]
ciggi has quit [Read error: Connection reset by peer]
ciggi has joined #asahi
bisko has joined #asahi
phiologe has joined #asahi
PhilippvK has quit [Ping timeout: 480 seconds]
jx0 has joined #asahi
jx0 has quit []
jx0 has joined #asahi
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
pyropeter3 has joined #asahi
jx0 has quit [Quit: poof!]
pyropeter2 has quit [Ping timeout: 480 seconds]
marvin24_ has joined #asahi
marvin24 has quit [Ping timeout: 480 seconds]
krirogn[m] has quit [Quit: Client limit exceeded: 20000]
kov has quit [Quit: Coyote finally caught me]
gladiac has quit [Ping timeout: 480 seconds]
<falconhash[m]> Never depend on a single income to get financial freedom. I can teach you how you can earn from $5,000 to $6,000 weekly,without sending money to me or anyone, but you will pay me 10% commission when you receive your profit. if you are interested, ask me how via Whatsapp
<falconhash[m]> 👇👇👇
<falconhash[m]> +1 (570) 801-0862
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
falconhash[m] has quit [autokilled: Spambot. Mail support@oftc.net if you think this is in error. (2022-07-27 04:32:45)]
nesnass has joined #asahi
nesnass has quit [Remote host closed the connection]
gladiac has joined #asahi
ciggi_ has joined #asahi
ciggi has quit [Read error: Connection reset by peer]
the_lanetly_052__ has joined #asahi
gladiac is now known as Guest6166
gladiac has joined #asahi
Guest6166 has quit [Ping timeout: 480 seconds]
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
thevar1a- has quit [Ping timeout: 480 seconds]
thevar1able1 has joined #asahi
doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Lizard[m] has quit [Quit: Client limit exceeded: 20000]
circularsprojects[m] has joined #asahi
thevar1able1 has quit [Ping timeout: 480 seconds]
ptudor has quit [Quit: Strict-Transport-Security: max-age=48211200; preload]
thevar1able1 has joined #asahi
ptudor has joined #asahi
lucifer178[m] has quit [Quit: Client limit exceeded: 20000]
thevar1a- has joined #asahi
thevar1able1 has quit [Read error: Connection reset by peer]
bngs[m] has quit [Quit: Client limit exceeded: 20000]
<mps> anyone have issue that qemu 7 can't use more than 3GB of RAM on macos 12.5
MajorBiscuit has joined #asahi
<maz> mps: qemu + hvf?
<maz> do you have highmem=off on the command line?
<mps> maz: yes, with hvf
bisko has quit [Read error: Connection reset by peer]
<mps> I think highmem=off is set
<maz> then get rid of it.
<mps> have to reboot to macos to check
<mps> maz: thanks will check soon
bisko has joined #asahi
<Glanzmann> Fedora users: I updaed Fedora build with a Fresh kernel, u-boot, m1n1. Testers on m2 welcome. https://thomas.glanzmann.de/fedora/
<maz> mps: highmem=off was a workaround to deal with some internal qemu nonsense. I fixed it for both KVM and hvf. shout if it doesn't work with 7.0, and ping agraf in the process.
<mps> maz: thanks for info
<mps> will report results back
<maz> funnily, someone is shouting about the same thing on #qemu. I guess some package must have been updated for macos...
<mps> maz: yes. ncopa asked me on #alpine-infra do I have same issue
<mps> but I don't boot to macos often, about three weeks ago I tested qemu with disk partition and noticed this problem but didn't looked further because it is not so important to me
leitao has joined #asahi
neatsquirrel has quit [Ping timeout: 480 seconds]
<mps> maz: fyi ncopa is alpine linux creator/founder
KonradHoh has joined #asahi
MajorBiscuit has quit [Quit: WeeChat 3.5]
<KonradHoh> Hi, I was wondering if anyone with an M2 could help me identify which ARM PAC/Memory Tagging Features are enabled?
<KonradHoh> Should be relatively easy by running inline assembly that prints out the configuration registers
MajorBiscuit has joined #asahi
neatsquirrel has joined #asahi
neatsquirrel has quit []
<KonradHoh> Thank you so much!
<mps> I'm trying to select new machine to buy, what is better M1pro or M2 for linux (I don't use macos). I don't need much GPU or neural engine
ptudor_ has joined #asahi
ptudor has quit [Ping timeout: 480 seconds]
<_jannau_> mps: depends on what you want do do with it. 8/2 high perf/efficiency m1 cores are faster than 4/4 m2 cores
<mps> _jannau_: aha, thanks
<mps> I mostly need machine for fast compiling programs
<j`ey> 8/2 with more RAM then. m2 currently only goes up to 24G
leitao has quit [Ping timeout: 480 seconds]
<mps> j`ey: what 8/2 means. I don't see anything about this on reseller pages
<_jannau_> 24G should be sufficient for 8 cores, 16G for 10 cores probably not
<maz> 8 fast cores, 2 "slow" cores.
<_jannau_> mps: 8 high performnce 2 high efficiency cores
<maz> (they should call them red and blue, really...)
<mps> I'm thinking to buy m1pro with 16GB ram
<_jannau_> mps: I can look for kernel compilation benchmarks comparing m1 max vs m2 tonight
<Glanzmann> mps: You can see the performance / efficency cores when you go to apple.com, select the product and click on 'technical specification' and scroll down to the cpu.
<mps> _jannau_: I'm not in hurry and you don't need to waste precious time on this for me (anyway thanks)
<_jannau_> m1 max and m1 pro should not differ much wrt. cpu performance
<mps> Glanzmann: ah, will look. thanks
<_jannau_> mps: I did the benchmarks already I just need to look up the results
<mps> _jannau_: ah, ok then
fetsorn[m] has quit [Quit: Client limit exceeded: 20000]
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
the_lanetly_052___ has joined #asahi
the_lanetly_052__ has quit [Ping timeout: 480 seconds]
thevar1a- has quit []
thevar1able1 has joined #asahi
Brainium has joined #asahi
kov has joined #asahi
janrinze has joined #asahi
KonradHoh has quit [Quit: Connection closed]
lockna has quit [Quit: Client limit exceeded: 20000]
<janrinze> marcan: the USB3 firmware for the Mac Studio Ultra, is it a lot of work to get that extracted and working with the Asahi Linux kernel?
<janrinze> Glanzmann: grub attaches the 'quiet' parameter on the kernel command line i.i.r.c. so you don't see what's happening during boot, right?
<Glanzmann> janrinze: Yes, but of course can you change the grub config to remove the parameter. I do it most of the time.
<Glanzmann> janrinze: In Debian it is /etc/default/grub
<chadmed_> janrinze: usb3 isnt working at all on any machine yet. the problem isnt the firmware it's reversing apple's PHY IP block
<j`ey> chadmed_: usb2 doesnt even work on some ports though
<chadmed_> oh that sucks then lmao
<j`ey> on the ultra, and on the imac
<chadmed_> oh the ones on pcie?
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
<janrinze> USB2 works fine on the Ultra
<j`ey> janrinze: I thought that not all ports worked?
<janrinze> only the USBA ports don't work
<j`ey> janrinze: USB A is the plug, it can also support USB 3
<j`ey> A B C are the type of connector, 1 2 3 are the type of signals/protocol or whatever
<janrinze> USBA physical ports on the Ultra don't work.......
<janrinze> USBC physical ports on the Ultra all work in USB2 mode
<janrinze> On the mac mini the USBA ports work fine in USB3 mode.
<janrinze> The USBC ports on the mac mini will only do USB2
Brainium has quit [Quit: bye]
<janrinze> specifically for the mac mini i got a USBC to USBA converter so that the dock can be used with USB3. This allows me to use the SSD in the dock at full speed.
qyousef has joined #asahi
<maz> I've been using sven's patches on the Ultra (together with a bunch of u-boot and DT changes) for about 2 months now, and USB3 on the USBC connectors is doing fine. I wish these patches made it upstream.
<sven> I wish so as well, but the entire ATC thing is too entangled with the thunderbolt mess and I'm not confident enough the DT can stay this way :(
<maz> fair enough.
<janrinze> maz: where to get those patches?
<maz> I guess that this makes the FW loader for the PCIe USB slightly more urgent that we'd like it.
<maz> janrinze: in one of sven's branches on the Asahi repo.
<sven> I think jannau also has a branch that actually works on 5.19
<janrinze> maz: how can i see which branches are sven's?
<sven> there's one fix to the phy subsystem I submitted which got un-fixed during a merge conflict
<_jannau_> janrinze: https://github.com/jannau/linux/tree/atc-WIP-5.19 this is on top of asahi
<_jannau_> but does not include changes for m1 pro/max/ultra
<_jannau_> rquo has a PR for that https://github.com/AsahiLinux/linux/pull/22
* sven should finish bluetooth and finally figure out that thunderbolt mess
<_jannau_> I should send out the refix for that
<pyropeter3> but those commits are not trivial to rebase on top of the current asahi kernel
<sven> please only use these commits if you know what you're doing
<sven> they're called WIP and HACK for a reason
<_jannau_> yes, those have to be redone on top of asahi
<janrinze> sven: which part do you worry about?
<janrinze> _jannau_: best rout is to use the PR for testing a.t.m.?
<janrinze> *route
<sven> that stuff randomly breaks and/or is not backwards/forwards compatible at some point
<sven> *sigh*
<sven> i think i'll actually stop pushing WIP branches
<sven> nothing there is "ready for testing"
<sven> it's meant for developers
<janrinze> sven: please don't worry.
qyousef has quit [Ping timeout: 480 seconds]
<janrinze> I'm doing kernel stuff on ARM Linux since 1996
<sven> fair enough, have fun then :)
<janrinze> That's the whole purpose of this excercise isn't it ? ;-)
<sven> yeah. i've just had a few people without kernel development experience trying to use WIP branches
<sven> and i'd really like to avoid that
<janrinze> sven: fair enough.
<janrinze> maz: you're running the kernel built from jannau's branch?
<janrinze> sven: there are very few people with an Ultra so it's probably going to generate useful info on the state of affairs on the Ultra (whatever the outcome may be)
<sven> not really. that branch isn't ready to be used so any information will become obsolete once I start changing it again
<janrinze> sven: okay.
<_jannau_> janrinze: no, pushed that branch just last weekend, not really compatible with using it for 2 months, also that branch works only on m1 (non pro/max/ultra) devices
<_jannau_> sigh, why didn't use more distinct marketing names for their tablet and laptop/desktop chips
qyousef has joined #asahi
<sven> just wait for M2 Ultra Max :D
<janrinze> _jannau_: the atc-WIP-5.19 branch looks like a set of patches atop ahasi. Are you worried that it won't work with the Ultra?
<_jannau_> it does not contain the necessary device tree changes for any t600x device
<janrinze> _jannau_: thanks for the heads-up there. saves me a lot of time.
<janrinze> simplest route to take a.t.m. is to test the PR. right?
<_jannau_> no, that PR is not compatible with with the way the t600x devicetree looks after adding m1 ultra support
<maz> janrinze: not really. I cherry-picked what I needed, and hacked the rest myself.
<maz> plus I'm still running some old 5.18-rc.
Brainium has joined #asahi
Brainium has quit []
<janrinze> Spagetti :-) thanks for the pointers and info. The biggest step for most of us will be when GPU support arrives.
<janrinze> marcan surely doesn't have the time to do everything himself. Is there some basic overview of who works on what?
<maz> (this is starting to sound like project management... ;-)
<janrinze> LOL.. i've been accused of that before ;-)
<sven> let's not introduce any project management. i have to deal with that at $work already ;)
<janrinze> :-D
qyousef has quit [Ping timeout: 480 seconds]
Moprius has joined #asahi
neatsquirrel has joined #asahi
bisko has quit [Read error: Connection reset by peer]
bisko has joined #asahi
MajorBiscuit has quit [Ping timeout: 480 seconds]
<janrinze> sven: how about small incremental changes that will be acceptable upstream to prepare for the USB3 changes?
<_jannau_> janrinze: from earlier: "the entire ATC thing is too entangled with the thunderbolt mess and I'm not confident enough the DT can stay this way"
<janrinze> yeah. complicated.
<sven> please don't try to be my manager or project manager or whatever else
<sven> thanks.
<janrinze> oops.. sorry about that.
<janrinze> to be honest, compared to a huge amount of SBCs from the past asahi linux is doing a great job. I have far too many paperweights on the shelves that were promised to be useable by their respective manufacturers.. So no complaints here! I'm very grateful for all the work done already by you people!
the_lanetly_052___ has quit [Ping timeout: 480 seconds]
mini0n has joined #asahi
Glanzmann has quit [Quit: leaving]
<bocekcombatbow[m]> <opticron> "and I fixed the wiki feature..." <- Thank you! I mentioned this and got no reponse
<opticron> 👍
jasper0 has joined #asahi
jasper0 has left #asahi [#asahi]
<kettenis> Glanzmann: I suspect this is a particular issue with the u-boot usb code not playing well with one of your usb devices
Moprius has quit [Quit: bye]
<kettenis> scanning the usb bus seems to take a long time for you
<kettenis> I'm unlikely to look into this unless I run into this issue myself
jluthra has quit [Remote host closed the connection]
jluthra has joined #asahi
qyousef has joined #asahi
doggkruse has joined #asahi
qyousef has quit [Ping timeout: 480 seconds]
bisko has quit [Read error: Connection reset by peer]
MajorBiscuit has joined #asahi
bisko has joined #asahi
doggkruse has quit [Read error: Connection reset by peer]
doggkruse_ has joined #asahi
MajorBiscuit has quit []
zv has quit [Ping timeout: 480 seconds]
qyousef has joined #asahi
<bocekcombatbow[m]> Honestly, I am very suprised with what is working already. So many things already work! And they work well.
<bocekcombatbow[m]> *Cries in changable screenbrightness though*
<bocekcombatbow[m]> Might be daily driving this soon
zv has joined #asahi
ptudor_ is now known as ptudor
MrGigio has joined #asahi
MrGigio91 has joined #asahi
thevar1able1 has quit [Ping timeout: 480 seconds]
MrGigiuz has joined #asahi
MrGigio has quit [Ping timeout: 480 seconds]
bluetail has joined #asahi
<bluetail> hey
<bluetail> I think I'm in love with Asahilinux
<bluetail> it runs so well
thevar1able1 has joined #asahi
<pimeys[m]> is it already good enough to get a macbook air, wipe the macos completely and start working with your favorite distro?
<bluetail> Also what the heck. Asahi Lina did a 6h20 stream
MrGigio91 has quit [Ping timeout: 480 seconds]
<j`ey> bluetail: thats a short stream :p
<bluetail> pimeys[m], depends on what you mean with that. It is certainly not as user friendly as macos. Without Docker I wouldnt love it
<pimeys[m]> I mean I'm a sway + emacs
<pimeys[m]> + firefox guy
<qyliss> you need macos for firmware updates and things I think
<j`ey> pimeys[m]: it's recommended to keep macOS for now
<pimeys[m]> using anything else is not really an option for me anymore
<bluetail> you cant run asahilinux standalone yet
<pimeys[m]> mm, so I'll wait a few years then :)
<bluetail> Perhaps you could but you shouldnt
<bluetail> I just discovered Asahilina wears the LGBT thingy
<pimeys[m]> I want the hardware but none of the apple software
<bluetail> understandable
<pimeys[m]> what would also be really awesome is linux base and macos virtual machines, then we could virtualize much more easily some CI builds
<mps> pimeys[m]: I run linux on m1 from november as daily driver
<bluetail> pimeys[m], I dont think thats impossible.
<mps> rarely boot to macos, for upgrades, setting brightness
<bluetail> pimeys[m], theres ONE dealbreaker as of why I wouldnt want to use a m1 device as main rig
<pimeys[m]> we have a pile of mac studios in the office, we'd like to use them to speed up some github actions, but making them to act as servers with macos is just so much pain
<mps> pimeys[m]: so it is quite good
<bluetail> It's the actual I/O performance
<pimeys[m]> yeah the SSD trick they do
<pimeys[m]> I read about that
<bluetail> call me weird but I still believe the image is different...
thevar1a- has joined #asahi
<bluetail> I could reproduce that weird difference on windows tho
<bluetail> what did I do? Add 4 monitors, then one would look as awkward as on the m1
<bluetail> It seems that if a certain bandwidth is taken, the display is altered. I know that occams razor stuff...
<bluetail> It's just that I'm so sensitive
<bluetail> Btw
MrGigiuz has quit [Ping timeout: 480 seconds]
<bluetail> all of that was solved by... shorter cables, less active monitors and thats it
<bluetail> at least for the PC
thevar1able1 has quit [Ping timeout: 480 seconds]
<bluetail> I didnt solve it for m1 -> monitor. Perhaps its just one peripheral of mine causing it to look different.
<pimeys[m]> it's not that I miss some CPU power currently, having a 3950x and 5950x and they are fast, the 5950x in the office uses exactly the same amount of time to compile our big rust project than the mac studio with 20 cores
<bluetail> I use a 5800x3d+6950xt
bisko has quit [Ping timeout: 480 seconds]
MrGigio has joined #asahi
thevar1a- has quit [Ping timeout: 480 seconds]
ptudor_ has joined #asahi
thevar1able1 has joined #asahi
artemist has quit [Read error: Connection reset by peer]
ptudor has quit [Ping timeout: 480 seconds]
MrGigio91 has joined #asahi
Glanzmann has joined #asahi
MrGigio has quit [Ping timeout: 480 seconds]
<Glanzmann> kettenis: I'll try if I can reproduce it with another usb stick and let you know.
thevar1able1 has quit [Ping timeout: 480 seconds]
bluetail has quit [Quit: Leaving]
thevar1able1 has joined #asahi
MrGigio has joined #asahi
MrGigio91 has quit [Ping timeout: 480 seconds]
thevar1able1 has quit []
thevar1able1 has joined #asahi
artemist has joined #asahi
chir4gm has quit []
chir4gm has joined #asahi
___nick___ has joined #asahi
___nick___ has quit []
<Glanzmann> Leif, Linus and I try to get Fedora running on m2 air. What is strange is, on a m1 the software does not crash (asahi branch), but on the m2 xorg does not start. Any idea what the issue is? I'm currently reproducing the issue on a friends m2 air laptop who has installed fedora and I'm remotely logged in using ssh.
<Glanzmann> So kernel boots, wifi works. xorg does not start (on m1 it does with the very same binary).
___nick___ has joined #asahi
<j`ey> Glanzmann: can you try fedora 36?
<Glanzmann> This is on fedora 36.
<j`ey> oh
<Glanzmann> So I have updated my fedora build with asahi upstream tree and u-boot/m1n1 and it is currently runnong m2 air of a friend. I'm logged on to that system using a reverse ssh tunnel.
<Glanzmann> I have this very same fedora running on m1 mini without any issues whatsoever.
<j`ey> Glanzmann: your webpage, and the log sayd fedora 35
<j`ey> "I downloaded Fedora-Workstation-35-1.2.aarch64.raw.xz"
<Glanzmann> Oh sorry, I mean fedora 35.
<Glanzmann> j`ey: Can I upgrade Fedora 35 to Fedora 36 without reinstalling or do I have to reinstall?
<j`ey> no idea, dont use fedora, just remember reading about ^
<jannau> maybe libunwind but that should affect m1 as well
<Glanzmann> jannau: Exactly, we thought that at first, too.
<Glanzmann> But I think I found the issue.
<jannau> maybe the trackpad due to not loaded firmware?
<Glanzmann> Might be the issue, that on the m2 it loads efifb instead of simplefb? https://tg.st/u/bla2.txt
<Glanzmann> jannau: Full dmesg of the m2 air is here: /dev/input/mice is there.
<jannau> Glanzmann: possibly efifb but I do not understand why that only happens on m2
<mps> Glanzmann: try alpine :)
<Glanzmann> Neither do I, but I had the issue before.
<kettenis> but even efifb should just work...
<Glanzmann> Nope, I had this issue before, I don't remeber the details, but x was not working for me. I spent several days finding out why ...
<jannau> efifb might not work due to r10b10g10x2
<Glanzmann> Exactly, thanks.
<Glanzmann> That was it.
m6wiq has quit [Quit: ZNC - https://znc.in]
<jannau> asahi kernel has efifb disabled
<Glanzmann> Okay, than we might have multiple issues. I used my kernel config, but Linus installed Lifes Fedora which used the asahi kernel config and he has also issues.
m6wiq has joined #asahi
<kettenis> don't immediately see why efifb wouldn't work with r10b10g10x2
<kettenis> would be good if someone could check it isn't an issue with u-boot
<jannau> the xorg modeset driver has issues r10b10g10x2
<mps> hm, I have CONFIG_FB_EFI=y for alpine kernel
<Glanzmann> So, I blacklisted the efifb module, but still have issues.
<mps> but I don't have M2
<Glanzmann> mps: If simpledrm initialized the fb before efifb does, than you don't have issues.
<mps> Glanzmann: `dmesg | grep "efifb\|simpledrm"` => https://tpaste.us/5k1B
riker77 has quit [Quit: Quitting IRC - gone for good...]
<Glanzmann> mps: [ 0.456691] efi-framebuffer efi-framebuffer.0: efifb: a framebuffer is already registered
<mps> that is on M1
<Glanzmann> THat is also on m1.
<mps> hm, iirc with 5.18 I couldn't switch from X to console without enabling efifb
<mps> because this i enabled it
___nick___ has quit [Ping timeout: 480 seconds]
<jannau> Glanzmann: reproduced, Xorg dies with SIGILL. maybe an issue with a new CPU instruction extension
riker77 has joined #asahi
<Glanzmann> jannau: I see, how did you reproduce it?
<Glanzmann> jannau: I tried blackilisting efifb, did not help, than I tried to force fbdev that starts x but not cpu output.
<Glanzmann> s/cpu/monitor output/
<jannau> installing fedora on the m2 macbook pro
<j`ey> gdb it? should tell you if it's a weird instruction, but not sure what that would be..
<Glanzmann> I see, cool. ;-)
<jannau> it fails on the very unusual "stp x1, x0 [sp, #-16]!"
thevar1a- has joined #asahi
thevar1able1 has quit [Ping timeout: 480 seconds]
<sven> unusual?
<jannau> that was sarcastic
<sven> ah :D
<sven> that makes much more sense
<kettenis> SIGILL?
<jannau> might be an issue with pointer authentication
<sven> oh, because it’s the first instruction of some function it just jumped to?
<jannau> yes
<jannau> and gdb shows that the modesetting driver uses pointer authentication but libGL doesn't
<sven> looks like SIGILL is triggered on pac failure in userspace
<jannau> modesetting driver tries to use glamor
<j`ey> m1 has PAC too.. doesnt it?
fny[m] has left #asahi [#asahi]
<jannau> I thought so too
<sven> pretty sure it does :/
<Glanzmann> Tried Debian live on m2, xorg works.
<j`ey> jannau: try 'arm64.nopauth' on the command line
<j`ey> *kernel command line
<j`ey> (if you want..)
<jannau> thanks, I was looking for that
<j`ey> f8da5752fd1b25f1ecf78a79013e2dfd2b860589 added it
<jannau> gah, no keyboard in grub
<j`ey> not even usb?
<Glanzmann> usb keyboard works in grub.
<Glanzmann> jannau: grubby --update-kernel=ALL --args="video=efifb:off"
<jannau> I had to fetch/connect it it
<Glanzmann> Than reboot, to put the kernel command line ...
<Glanzmann> but of course with: grubby --update-kernel=ALL --args="arm64.nopauth"
<Glanzmann> m2 xorg and dmesg output from debian: https://tg.st/u/dmesg.txt https://tg.st/u/Xorg.0.log
LaurentiuCosma[m] has quit [Quit: Client limit exceeded: 20000]
<jannau> it was not pauth, same crash but no more PAC in the backtrace
<Glanzmann> jannau: We also tried it, also doesn't work for us.
<j`ey> sven: btw the first instruction thing would be BTI, not PAC
<sven> oh, true
<j`ey> (which m2 does have!)
<Glanzmann> I try the 4k patch and let you know the results. Takes a lot of time, because I need to compile it.
thevar1able1 has joined #asahi
thevar1a- has quit []
joske has joined #asahi
joske has quit []
MrGigio has quit [Read error: Connection reset by peer]
mini0n has quit [Read error: Connection reset by peer]
<jannau> with glamor disabled (https://wiki.raptorcs.com/wiki/Troubleshooting/GPU#Xorg_crashes_or_is_laggy_with_the_AST_VGA_GPU) Xorg starts but I get a "oh no! something has gone wrong" screen
<sven> I’m still confused about the SIGILL on that stp
<jannau> gnome-shell segfaults now in mesa
<j`ey> try arm64.nobti too?
<j`ey> looks like that also ends up with SIGILL do_bti() arch/arm64/kernel/traps.c
<jannau> me, only thing which looked fishy to me is that prvious frame made the call via 'blr x1' but x1 was not anywhere near pc
<j`ey> hm
<jannau> 'arm64.nobti arm64.nopauth' in the kernel command line works
<sven> maybe this is finally a good excuse to actually understand how this pointer authentication business works
<j`ey> jannau: just to confirm.. remove nopauth?
<sven> so far my extremely superficial understanding has been enough
<Glanzmann> jannau: Trying to reproduce.
<Glanzmann> jannau: Did you also disable 'glamor'?
<Glanzmann> btw. kernel with 4k-iommu patch did not work, but that was probably be expected.
<Glanzmann> jannau: You fixed it.
<Glanzmann> It works for us.
qyousef has quit [Quit: WeeChat 2.8]
<jannau> j`ey fixed it
<Glanzmann> Debian users: m2 work out of the box. Fedora users: grubby --update-kernel=ALL 'arm64.nobti arm64.nopauth'
<j`ey> jannau: and did you reenable glamour? (whatever that is!)
<sven> this also sounds more like a workaround than a fix
<j`ey> but cool, it's a BTI issue then
<Glanzmann> jannau: Works without disabling glamour.
<j`ey> Glanzmann: ok
<j`ey> sven: yes definitely, but at least they can continue!
<sven> sure
<sven> i just don’t like calling something a fix when we don’t even truly understand the root cause yet
<j`ey> ah right yeah
<jannau> Glanzmann: arm64.nopauth is not necessary
<Glanzmann> Okay, I'll update my documentation. Thanks.
<jannau> not is disabling glamor
<kettenis> so fedora is shipping xorg compiled with bti/pauth enabled but mesa with bti/pauth disabled?
<kettenis> that is just broken
<jannau> pauth seems to be there, I saw a pac* instruction in mesa
Lucy[m]1 has quit [Quit: Client limit exceeded: 20000]
<Glanzmann> jannau / j`ey: Thank you for finding the workaround.
<j`ey> if it helps linus just merge asahi right away.. :P
akira128 has joined #asahi
<Glanzmann> Exactly.
akira128 has quit []
<Glanzmann> No more upstreaming issues, upstream will upstream itself or shame that it is not already upstreamed will upstream it.
<Glanzmann> Now I have to read up on what BTI is about.
leif has joined #asahi
<Glanzmann> Hey there.
<leif> great work guys!
<j`ey> branch target identification, to help kickstart your reading
<Glanzmann> j`ey: THanks.
<jannau> shouldn't bti be reported in cpuinfo's flags on a BTI enabled kernel. it doesn't on m1 or m2
lemonsus[m] has quit [Quit: Client limit exceeded: 20000]
<j`ey> jannau: it should, that's weird
<j`ey> (if you mean /proc/cpuinfo)
<jannau> yes
<j`ey> it was in Glanzman's dmesg "[ 0.000000] CPU features: detected: Branch Target Identification
<jannau> with arm64.nobti it's of course missing. if m1 doesn't has it that would at least explain why m1 machines where working
<leif> ahhh...so that's why it affects the m2 and not the m1?
<j`ey> jannau: oh heh
<Glanzmann> jannau: Yep, I can't find it in my dmesg on m1 systems.
<jannau> yes, but I was under the impression that m1 has it
<Glanzmann> leif: Yep, it appears to be. And xorg is horrible broken on fedora.
<Glanzmann> and maybe in general.
<jannau> nothing to do with xorg
<Glanzmann> But Debian started xorg on m2 without any issue.
<jannau> debian doesn't use BTI
<Glanzmann> Ahhh.
<jannau> bti is reported in /proc/cpuinfo if I don't disable it
<Glanzmann> jannau: So BTI must be compiled in the binary. But xorg jumps somewhere it should branch, doesn't it?
<Glanzmann> jannau: Ah.
<jannau> on m2
<jannau> the issue is in mesa and not in xorg
<Glanzmann> There is another feature which only m2 has: +CPU features: detected: Enhanced Counter Virtualization
<kettenis> unlikely to be relevant
<Glanzmann> Just mentioning, while I was diffing the cpu features of m1 and m2.
<j`ey> jannau: the question is, if it's a build bug.. or an actual bug in mesa
<j`ey> but im not sure if that's the right package
<jannau> https://kojipkgs.fedoraproject.org/packages/mesa/22.1.4/2.fc36/data/logs/aarch64/build.log should be correct package which has -mbranch-protection=standard as well
<j`ey> ok thanks, I used https://packages.fedoraproject.org/search?query=mesa but it didnt show that one
<Glanzmann> Time to hit bed. jannau/j'ey thanks again for getting to the bottom of this.
<j`ey> not the bottom yet!
<Glanzmann> That is true, but enough to onboard upstream. And now that we know what triggers it, easier to isolate and fix. I'm eager to hear the end of the story.
<Glanzmann> Oh my god, I sent upstream an incorrect workaround. Shame on me. Time to hit bed.
<Glanzmann> n8
<jannau> are there already other CPU cores with bti? should be every armv8.5 or armv9 core, I guess nobody is running fedora on cortex-x2/a710/a510 phones
<Glanzmann> Probably not.
<j`ey> could test with QEMU
<Glanzmann> jannau: One last question, before I go sleeping. My friend with the m2 air tried to do a right click but was unable to do so. Have you managed on a m2 device to do a right-click (two fingers) or middle-click (three fingers). We connected an external usb mouse as workaround.
<Glanzmann> leif: Did you see that I screwed up the command line?
qyousef has joined #asahi
<jannau> Glanzmann: should work in the same way as on m1 can be configred through libinput
<Glanzmann> jannau: This was on Debian testing with m2 air and a 4k-iommu kernel. x worked. We used a minimal environment: apt-get install lightdm blackbox xterm firefox-esr big-cursor; startx
leif has quit [Quit: Leaving]
<Glanzmann> jannau: I see, let me check on m1 air using the same setup if right click works for me, but I already know the answer because I use all the time, but will double check.
<jannau> default in gnome/fedora seems to be 2 finger click == right mouse button
<Glanzmann> We tried two and three finders. and shift-click no dice. But that was on Debian, not fedora.
<Glanzmann> fingers*
<jannau> alternative behaviour is click on the right half of the touchpad is right button
<Glanzmann> I see. Will try it on Saturday.
<mps> I switched to mtrack from libinput
<Glanzmann> jannau: Just tested with the exact same config on the m1, there the right-click works. On the m2 it did not. Will verify myself once I'm in front of the machine on Saturday.
<Glanzmann> My friends lives in Stuttgart, I'm in Nuremberg. So we did a phone and reverse tunnel ssh session ...
<Glanzmann> Everyone sleep well and have sweet dreams.
<mps> Glanzmann: you too
<Glanzmann> thanks.
qyousef has quit [Quit: Updating details, brb]
qyousef has joined #asahi