ChanServ changed the topic of #asahi-alt to: Asahi Linux: porting Linux to Apple Silicon macs | User-contributed/unofficial distribution ports | Logs: https://alx.sh/l/asahi-alt
kidplayer666 has quit [Quit: Connection closed for inactivity]
<u154ss>
@tobhe thanks. Glanzmann originally reported this in December 2022. I am told that even though libunwind appears in the xorg.log, it is NOT in fact libunwind, it is something else...
<u154ss>
@all most likely. You are right, it is the xorg executable. The second last line of the xorg log states that I should look for a log, I am not sure where though. As stated already, Fedora 39/KDE Plasma works with the M2 Mini. All of the Mesa stuf is release 24.x I my (Debian) environment, the Mesa stuf is at release level 23.x
<u154ss>
I am not sure whether that has an influence on things though.
<j`ey>
yeah because KDE plasma is wayland :)
<u154ss>
j'ey I know.
<j`ey>
u154ss: are you using the exact same versions for mesa on the m1 and m2 machine?
<j`ey>
it seems that OsLookupColor is a place that crashes a lot, according to google
<u154ss>
Again, on the M1 Mini everything works (Debian/XFCE). Using the (exact) same installation procedure on the M2 Mini - see above :(
<leio>
u154ss: I presume it's on systemd, so journalctl should have stuff. I don't know under which unit file for you to filter for it, but you can scroll through the whole log of the boot with journalctl -b0 or just reproduce and look immediately
<leio>
(xorg log)
<leio>
but I think you had something already shared to that effect?
<j`ey>
you might have to try run it in gdb or.. file a bug on GH, in case it is a GPU driver issue
<j`ey>
but not sure if anyone will look at if it's only affecting X11..
<u154ss>
@j'ey understand, but I am loath to change to fedora and/or kde. Not my thing. Is the instruction set so different between the M1 and M2?
<j`ey>
not much for the CPU, not sure about the GPU, there is definitely some changes
<u154ss>
@j'ey there we have it...
<j`ey>
problem is that there's not much to gone on really. I mean you could try boot with `nomodeset` that will confirm it's the GPU kinda
<j`ey>
(for CPU changes, that's why I suggested the arm64.nobti option)
<leio>
u154ss: as it's a segfault, have you checked with coredumpctl?
<leio>
you should be able to just get a gdb backtrace from there (and maybe confirm the stack is nuts and no useful backtrace there, but who knows)
<u154ss>
@all No, I am not a developer, I will have to get my head around these things. I will have to reproduce this again, will take some time..
<leio>
the coredump is probably still there and can check it out unless you had disk too full and it had to quickly clean it up or you changed binaries.
<leio>
coredumpctl
<leio>
note that last Xorg crash, then something like
<leio>
coredumpctl info <PID of last Xorg crash>
<leio>
or debug instead of info to enter debugger, though info should give the crashing threads backtrace too
<leio>
(or process name instead of PID, I just don't recall it after not using Xorg for 3+ years or know what it is for sure on debian)
<u154ss>
@leio, I scrubbed the installation and put fedora back on, subsequently wiped that too. Things are virginal at the moment. I will have to start from scratch again using Glanzmann's installer.
<leio>
ok, guess the coredump history is gone then :)
<j`ey>
janneg: I guess that `nomodeset` doesn't actually disable GPU rendering for us, just DCP..
<leio>
don't think GPU can draw on actual screen without DCP, just offscreen?
<u154ss>
This is what is loaded on the M1 - appledrm 49152 1 apple_dcp 114688 1 appledrm drm_dma_helper 49152 2 appledrm,apple_dcp root@debian:/Transit# lsmod |grep asahi asahi 1245184 0
<j`ey>
yeah
<u154ss>
@j'ey to reiterate blacklist asahi in /etc/modprobe.d/drm_asahi.conf?
<j`ey>
double check the module name
<j`ey>
if its just asahi or drm_asahi, not sure
<u154ss>
@j'ey no sign of drm_asahi anywhere.
<u154ss>
@j'ey methinks asahi.
<j`ey>
ok
<leio>
I would also install debug symbols package for Xorg
<u154ss>
@j'ey, the blacklist thing did not work. Still the same, segfault. Console inaccessible, cursor on top LH corner - reboot into Debian recovery (single mode) with limited possibilities.
<j`ey>
u154ss: can you check that the blacklist actually worked, and the module isnt loaded
<u154ss>
@j'ey, I an only check (lsmod) from Debian recovery. There is no asahi loaded.
<j`ey>
u154ss: and dmesg | grep gpu shows nothing?
<j`ey>
well that's weird then :) no idea why it's crashing, other reports from the internet seemed to be related to the GPU driver..
<j`ey>
might need the blacklist *and* nomodeset.. but im just randomly guessing :)
<u154ss>
@j'ey did so.
kidplayer666 has quit [Quit: Connection closed for inactivity]
<u154ss>
@all on Glanzmann's Debian initial image, the asahi module in built in-stream (CONFIG_DRM_ASAHI=y), so the blacklist exercise will not work anyway or?
<cy8aer>
it is in kernel.
<cy8aer>
except there is a boot option for the kernel to ignore it?
eiln has joined #asahi-alt
<cy8aer>
I just checked it, Glanzmann like kernels are able to work with "CONFIG_DRM_ASAHI=y".
<j`ey>
ok, so the blacklist isn't working / the module is built in
<j`ey>
so it could still be related to the GPU drive
<j`ey>
r
<u154ss>
@j'ey correct. Glanzmann's asahi is in-stream i,e, CONFIG_DRM_ASAHI=y. That is his latest and greatest (newest). Strangely, all of this works on the M1 Mini.
<j`ey>
u154ss: are you building the kernel yourself, or using a .deb?
<u154ss>
@j'ey on the M1 Mini, I roll my own kernels. On the M2 Mini, I am not in the position to do anything, as things die before I can even get started. Om my M1 Mini, I still have the same driver (old timestamp??) - Initialized asahi 0.0.0 20220831 for 206400000.gpu on minor 1, but things work.
<u154ss>
@j'ey, i.e. I am using Glanzmann's insteller (tg.st/d). It has always word, at least for the M1 Mini.
<j`ey>
you could build the kernel on another machne and copy it across
<j`ey>
or try install xfce on fedora, see if it repros there
<u154ss>
@j'ey could do, but a bit of work involved. As I stated earlier, the mesa stuff is up to date on Fedora (maybe taking the M2 requirements/specifics into account), not so with Debian testing.
<u154ss>
Again, speculation on my part..
<u154ss>
word = worked; insteller = installer. All thumbs today.
<leio>
u154ss: CONFIG_DRM_ASAHI is not supported, it doesn't bring up HDMI output for example.
<leio>
u154ss: sorry, I meant built-in is not supported; it needs to be a module
<j`ey>
leio: are you thinking of DRM_APPLE?
<leio>
I am, I'm referring to whatever CONFIG_DRM for apple was there and was just repeating from above. I wasn't aware there's two of them
<leio>
or am I?
<leio>
I'm scrolling back history of channel where it was stated, I should use the web logs instead :)
<j`ey>
leio: APPLE is DCP ASAHI is GPU
<u154ss>
@j'ey/leio no, CONFIG_DRM_ASAHI=y is what is delivered in Glanzmann's .config. I know things have changed. On this machine (M1), I have CONFIG_DRM_ASAHI=m. I believe Glanzmann built his latest offering in October '23.
<leio>
ok, DRM_APPLE is what has to be module indeed, unsure about DRM_ASAHI
<leio>
sorry for the mixup
<leio>
(can't we rename these to be clearer?)
<u154ss>
@leio Glamzmann's .config has CONFIG_DRM_APPLE=y
<u154ss>
On this machime (m1 Mini), I have CONFIG_DRM_APPLE=m
<leio>
ok, that's good, with DRM_APPLE=y at least on M2 Ultra there's no HDMI output, unsure if affects m1 mini
<u154ss>
As per j'ey's suggestion, I have installed (again :() fedora fc39 on the M2 Mini, also XFCE (4.19) - things work. What does that tell us?
<u154ss>
4.19 = 4.18
<j`ey>
I guess that tells us that there isn't a GPU driver bug?
<j`ey>
that it could be mesa
<j`ey>
I mean, at least it's a debian issue, not asahi directly?
<u154ss>
@j'ey it would seem so. As the M2 has not been as well tested as the M1 (Desktop Environments etc.), things will happen... Again, I prefer Debian and a very lean .config :)
<j`ey>
and Xorg doubly so
<j`ey>
(less tested)
<j`ey>
so yeah, next think would be to try upgrade mesa
<mps>
just note, xorg works fine with asahi kernel
<u154ss>
@mps depends on the machine, I would have thought.
<j`ey>
u154ss: could try another Xorg WM
<j`ey>
or or try start it from the terminal, not lightdm