<tpw_rules>
yeah there's no gpu support yet, it's all cpu emulation
<JackClayton[m]>
Has anyone found a way to fix `Alacritty` or `Wezterm` going super slow with text editing? I'm running on i3 with the minimal installation so I'm not sure if this a well known problem or not. It goes fast if I use something like `xterm` but there are features in the other two I need
<chadmed>
like tpw_rules said, all rendering is currently performed in software so anything that relies on GL or Vulkan is going to suck and suck hard
<chadmed>
the fix is to wait for the GPU stuff to be done
<JackClayton[m]>
No worries I'll try find a way to force Alacritty/Wezterm to use CPU thanks
<tpw_rules>
i doubt you can
<chadmed>
it is kind of odd that it's slow enough to be noticeable, i use full fat plasma and it's buttery smooth considering its llvmpipe
<chadmed>
i can even watch 1080p youtube with no frame drops (the machine does get hot though)
Stroller has quit [Quit: Stroller]
duban has quit [Quit: I'm out]
randyzhu[m] has joined #asahi
<JackClayton[m]>
Yeah full hd video in firefox works with no issue so I thought there might be a setting somewhere to turn off trying to render with GPU or something like that
<chadmed>
well the "gpu" alacritty is rendering to is llvmpipe, which is cpu rendering
<chadmed>
same as with firefox
duban has joined #asahi
<chadmed>
so id say its just alacritty being poorly optimised or something
<tpw_rules>
the whole point of alacritty iirc is that it uses the gpu. idk why they would build a "gpu off" setting in
<randyzhu[m]>
hey folks, i was wondering whether there was any demand for a guide on installing asahi on external devices?
<randyzhu[m]>
also, it's just llvmpipe that handles rendering, it's not that alacritty has a "built in gpu off" setting, but instead, it just renders to a gpu that is technically a software renderer.
<randyzhu[m]>
well, """gpu"""
<JackClayton[m]>
Yep fair enough thanks for the insight all, looks like I'll have to find a non-gpu terminal emulator
imtsuki[m] has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
skipwich has quit [Quit: DISCONNECT]
skipwich has joined #asahi
skipwich has quit [Read error: Connection reset by peer]
skipwich has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<JackClayton[m]>
Sorry one more noob question, I see hdmi and thunderbolt-displayport don't work for external displays (i tried both), but will a usb-c to hdmi adapter or something like that work for macbook m1? Anyone have any recommendations for one they use?
<chadmed>
no currently the only thing you can get out of the USB ports is USB 2.0
<chadmed>
displaylink might work though i havent tried it
<chadmed>
USB3/4/Thunderbolt is being worked on currently though
surge9n has joined #asahi
qeeg has quit [Remote host closed the connection]
julio7359 has joined #asahi
qeeg has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
chadmed has quit [Remote host closed the connection]
chadmed has joined #asahi
the_lanetly_052__ has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
julio7359 has quit [Ping timeout: 480 seconds]
Ry_Darcy has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
rvalue has quit [Remote host closed the connection]
rvalue has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
Ry_Darcy has quit [Remote host closed the connection]
surge9n has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
<mps>
jannau: just for info, bootloop also happened with usb-c ssd without your patch. it boot fine from usb flash
surge9n has quit [Ping timeout: 480 seconds]
badt[m] has joined #asahi
surge9n has joined #asahi
ciggi_ has joined #asahi
ciggi has quit [Read error: Connection reset by peer]
surge9n has quit [Ping timeout: 480 seconds]
hm[m] has joined #asahi
<badt[m]>
Hey guys, would it perhaps be possible to boot M1/M2 computers from USB, namely using Asahi? I've heard that with the recent firmware Apple opened the door to this, and potentially it would be possible. Also, what's the extent to which Asahi can leverage Secure Enclave? Can we actually perform the cryptographic operations, and would it be possible to use TouchID/Yubikey PGP smartcard to authenticate it? In such a way that wouldn't allow
<badt[m]>
applications to access the secrets randomly.
<j`ey>
you can boot asahi from USB *if* you install a stub, thats 2.5G, on the main nvme
<j`ey>
not sure about the recent firmware thing, I haven't seen anything on that
<j`ey>
SEP is not currently usuable from asahi, needs work
bent2 has joined #asahi
bent3 has quit [Read error: Connection reset by peer]
bent3 has joined #asahi
<badt[m]>
Okay, so no plug-and-play like what you'd typically do with these Tails sticks? I've almost considered having a USB stick with both amd64 and arm64 distributions to be able and boot of potentially compromised computers if USB access is provided. A pity. However, I'm still considering a three-point security harness setup of GNU Guix on top of Asahi Linux with (1) SEP, (2) Hashicorp Vault, (3) Yubikey/Yubikey HSM to operate said vault. In this
<badt[m]>
context, it could be very advantageous to rely on SEP for extra point. This way you don't have to trust either Apple, Yubico, or the software. As long as at least one of said points is not compromised, the system as a whole would remain secure...
<badt[m]>
Food for thought.
<badt[m]>
How far is SEP from being usable, is it at least on the map?
<chadmed>
i mean youre still relying on apple to provide the root of trust in the SEP firmware
<j`ey>
badt[m]: no time frame
<chadmed>
some preliminary RE on it has been done but mostly just snooping on comms with macos, we still cant make it do anything useful ourselves
<badt[m]>
chadmed: Exactly, but in there you keep only a part of the secret so without compromising the other two it's of no use
<badt[m]>
Now if you could compomise Yubikey's firmware as well as SEP's, and override the kernel with intimate knowledge of the software piping in Vault, then you-the-attacker- could theoretically go through the secrets that you-the-user don't want to go through
<badt[m]>
I.e. for me it seems I'm decrypting secret A, while in fact I'm decrypting secrets A-Z
<dottedmag>
What prevents Apple from snooping the secrets off the memory, if they ship a backdoored update to iBoot?
bent2 has quit [Ping timeout: 480 seconds]
<dottedmag>
Or your threat model is evil maid, not evil manufacturer?
surge9n has joined #asahi
<chadmed>
you can (mostly) stop evil maid by turning filevault on, SEP will demand you assert presence with a filevault password before it even loads m1n1
<chadmed>
...i think
<chadmed>
what would be cool is if they rolled touchid into filevault
<dottedmag>
Didn't they? I think I'm booting macOS with FileVault enabled without entering my password.
<badt[m]>
<dottedmag> "What prevents Apple from..." <- Well, first of all that would be possible only if the computer is directly connected to the Internet, and it doesn't have to be (in the use-cases where you want to treat is as a HSM-enabled box of some sort! for the purpose of CA and cold wallet ops)
<badt[m]>
Moreover, Apple on their own would then be able to only snoop on what's been decrypted completely
<badt[m]>
And you can have the operation happen in an encrypted VM
<badt[m]>
Think homomorphic encryption and stuff like taht
<badt[m]>
So a security model can be constructed in which what's truly exposed is just the threshold fragments of the key material stored in the Enclave but not the other pieces stored on the Yubikey, and/or provided over the network
bluetail has joined #asahi
<bluetail>
hey
___nick___ has joined #asahi
<bluetail>
Somebody said its a waste to only use the m1 mac mini as a docker slave without ever using it as a PC
<chadmed>
that seems very overengineered for these machines tbh
surge9n has quit [Ping timeout: 480 seconds]
<chadmed>
its not like theres any piece of the machine where arbitrary data can be leaked out of memory by rogue firmware, everythings behind an IOMMU
<chadmed>
theres no ME or PSP or any crap like that
<chadmed>
so unless you live life on the edge and download gigabytes worth of crapware just to try your luck that solution is pretty ridiculous overkill
<bluetail>
chadmed, the m1 does its job but I cant ever use it as a PC cause I dont like the crosstalk or whatever it is that my display ends up flawed
<bluetail>
Just to note
<bluetail>
I had reproduced it on a normal x86 PC when I used 'too long' usb cables, letting me identify 'rectangles' on the monitor affected
zzywysm has quit [Read error: Connection reset by peer]
<bluetail>
so guessing the USB controller somehow affected the video signal
<bluetail>
dont ask me how
<bluetail>
once I used less-long cables the problem went away
<chadmed>
is this RE using displaylink?
<bluetail>
no
<bluetail>
displayport
<bluetail>
displaylink is a flawed technology by some weird vendor
zzywysm has joined #asahi
<bluetail>
I argue about how the signal seems flawed on the m1 no matter what
<bluetail>
and how I cannot perceive it on a higher bandwidth PC
<bluetail>
unless I mess up USB
___nick___ has quit []
<bluetail>
it almost feels like a duplicated display on windows but far worse
<chadmed>
is this in macos?
<bluetail>
OS doesnt play a role
<bluetail>
m1 does behave this way in both linux and macos
<badt[m]>
And as to GPU support is there a time table as to what's up with it? Can I donate money directly to speed up the development in this particular area?
<chadmed>
well it would because we dont have dp alt mode on linux yet
<bluetail>
no
<bluetail>
I use HDMI on the m1
<bluetail>
I have a different output
<j`ey>
badt[m]: you can donate money yes, that will help
<chadmed>
badt[m]: lina is starting work on the kernel driver this coming week! userspace is (almost) GLES2 compliant thanks to alyssa
___nick___ has joined #asahi
<bluetail>
I argue the issue appears on systems with the crosstalk or whatever it is
<bluetail>
The m1 just happens to be one of em
<chadmed>
well ive been using a thunderbolt dock with my 14" pro for months now and theres nothing wrong with it
___nick___ has quit []
<bluetail>
does the dock feed additional power into the display connection?
<chadmed>
likewise there are many hundreds of thousands of people with these machines who report no such issues so i can only assume you either have a faulty device or your setup is faulty elsewhere
<chadmed>
which would make sense since you said it was sensitive in windows too
<bluetail>
no
<bluetail>
your argument is not valid, why?
<chadmed>
yeah the dock is a lenovo tb4 workstation thingy
<bluetail>
cause I can only feel that issue if I for instance saturate the cpu USB chipset (red one) or whatever is the term while I use many displays on my dedicated GPU
<bluetail>
and it never happens when I use proper cable lengths
___nick___ has joined #asahi
<bluetail>
it doesnt happen on my lenovo laptop too
<bluetail>
but on the m1
<bluetail>
it happens even when HDMI is the only thing plugged in
<bluetail>
on the imac 2019 it happened only when I used my armada of USB devices
<bluetail>
When you look at it, you can see little rectangles, no?
<bluetail>
You also see little 'dots'
<bluetail>
In fact its just a matrix of rectangles
<bluetail>
But thats what I see in a less-visible way on the m1 or when my PC is acting up cause I messed with USB
Major_Biscuit has joined #asahi
surge9n has joined #asahi
Major_Biscuit has quit [Ping timeout: 480 seconds]
surge9n has quit [Ping timeout: 480 seconds]
<hm[m]>
Hi guys, just installed asahi today on M1 macbook air. I installed the minimal version and manually installed gnome, gdm, and other things. The setup is fine except there is no audio from the headphone jack, no matter from gui or from "speaker-test" command. Restarting doesn't make a difference. I saw on the website that headphone jack should be working, so I'm wondering what could go wrong... Any help would be great
<bluetail>
what device do you want to plug into the headphone jack?
<hm[m]>
bluetail: a headphone that also have a small mic on it
<bluetail>
I guess thats related to the GUI option
<bluetail>
perhaps you have that available when installing plasma
<bluetail>
but who knows
<bluetail>
lets wait till somebody writes back here
surge9n has quit [Ping timeout: 480 seconds]
<hm[m]>
I’ll give it a try, thanks for the help!
<hm[m]>
Wow that worked. Turns out I need to manually select the sound card in alsamixer and then change the audio volume. After that adjusting volume also “just works” in gnome
<mps>
bootloop with external sd card and ssd is fixed, and as bonus bigrer fonts in u-boot
<mps>
s/bigrer/bigger/
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
sri has quit [Quit: Leaving]
eroc1990 has quit [Remote host closed the connection]
Chainfire has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
<Chainfire>
Trying to qemu-user to x86_64, some things work fine, but some generate "Bail out! ERROR:../qemu-7.0.0/accel/tcg/cpu-exec.c:933:cpu_exec: assertion failed: (cpu == current_cpu)" and crash. Has anyone run into this? There's nothing I can find on Google that helps me.
<Chainfire>
This does not happen in Arch or Ubuntu under Parallels, only in Asahi, if that helps.
Stroller has joined #asahi
<j`ey>
people have mentioned that error before, but I cant remember if there was ever a resolution
<jannau>
maybe a big.LITTLE issue? can you try if the issue is reproducible with taskset -c X?
<Chainfire>
I did try by setting affinity on the container, is that not the same thing?
<maz>
it seems to be a consequence of a longjmp
<Chainfire>
Compiling with clang instead of gcc gets rid of this error but triggers another one :')
<Chainfire>
taskset doesn't help either
<dottedmag>
M1 seems to be pretty good at uncovering latent bugs in software.
<Chainfire>
Could it be TSO related, as that was a thing with Rosetta I think has been mentioned here? The exact same setup but emulating arm32 works fine, it's just x86_64 that borks.
<maz>
what if you taskset it on *one* CPU?
<Chainfire>
tried, no effect
<maz>
it'd be interesting to disassemble this function and work out where the 'current_cpu = cpu;' assignment actually occurs.
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
<maz>
nah, seems to happen long before the sigsetjmp(). interestingly, this is a per-thread variable, so relative to tpidr_el0. maybe worth checking if this gets clobbered one way or another.
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
Brainium has joined #asahi
surge9n has joined #asahi
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
the_lanetly_052___ has joined #asahi
the_lanetly_052___ has quit []
the_lanetly_052__ has quit [Ping timeout: 480 seconds]
julio7359 has joined #asahi
CME has joined #asahi
zzywysm_ has joined #asahi
CME_ has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
<jannau>
maz: do you have newer NV branch than v6 against 5.17-rc1? my curiosity to test it on m2 is not high enough to resolve the merge conflicts against 5.19
zzywysm has quit [Ping timeout: 480 seconds]
<jannau>
at least for the vmid change I do not know how resolve it yet
bluetail has quit [Read error: Connection reset by peer]
<jannau>
if you don't have anything newer do not rebase it on my account. It's just curiosity and sending a "Tested-by: <me> on Apple M2" mail
surge9n has quit [Ping timeout: 480 seconds]
eroc1990 has joined #asahi
<maz>
jannau: no, I don't at the moment. I'm hoping to find a bit of time for that after September, as I'm pretty busy until then...
<maz>
and hopefully, a M2-mini will appear shortly after, which should speed up the development (watching paint dry in the model is only so much fun...)
<maz>
there is also a huge number of comments from Alex that I absolutely need to address.
<jannau>
ok, no problem
<Chainfire>
<maz> nah, seems to happen long before the sigsetjmp(). <--- if I rebuild with clang, the original error is bypassed due to ifdefs, the error then ends up being "... include/qemu/rcu.h:102: void rcu_read_unlock(void): Assertion `p_rcu_reader->depth != 0' failed."
<Chainfire>
which also seems to be TLS related. I don't know enough about QEMU to even begin to look into this, though.
surge9n has joined #asahi
<Chainfire>
So far I've failed virtualizing x86_64 the way I need it with both Qemu and Box64. Guess I'm going to try Rosetta next... :')
surge9n has quit [Ping timeout: 480 seconds]
<mps>
ktz_[m]: upgraded kernel for alpine linux with CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y please test and join #asahi-alt and report results
doggkruse has joined #asahi
yoda_like_soda[m] has joined #asahi
surge9n has joined #asahi
yoda_like_soda[m] is now known as YODA_LIKE_SODA[m]
surge9n has quit [Ping timeout: 480 seconds]
babble has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
babble has quit []
Techcable has quit [Remote host closed the connection]
Techcable has joined #asahi
Moprius has joined #asahi
surge9n has joined #asahi
nela has joined #asahi
zzywysm_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
surge9n has quit [Ping timeout: 480 seconds]
zzywysm has joined #asahi
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
julio7359 has quit []
julio7359 has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
sri has joined #asahi
doggkruse has quit [Ping timeout: 480 seconds]
doggkruse has joined #asahi
zzywysm has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zzywysm has joined #asahi
zzywysm has quit []
off^ has joined #asahi
Moprius has quit [Quit: bye]
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
Stroller has quit [Quit: Stroller]
doggkruse_ has joined #asahi
off^ has quit [Ping timeout: 480 seconds]
doggkruse has quit [Ping timeout: 480 seconds]
doggkruse has joined #asahi
doggkruse_ has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
<Chainfire>
Filed a bug report with QEMU, maybe one of the big brains there has an idea
surge9n has quit [Ping timeout: 480 seconds]
___nick___ has quit [Ping timeout: 480 seconds]
doggkruse_ has joined #asahi
doggkruse has quit [Ping timeout: 480 seconds]
surge9n has joined #asahi
surge9n has quit [Ping timeout: 480 seconds]
Thomas___ has joined #asahi
<ktz_[m]>
<mps> "ktz_: upgraded kernel for alpine..." <- thank you sir, I'll report back asap