marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | General project discussion | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-offtopic | Keep things on topic | Logs: https://alx.sh/l/asahi
<modwizcode>
Interseting they talk about kernel support for Rosetta.
<modwizcode>
I used to be super interested in xnu because of it's microkernel origins and I think the way mach does some things is pretty clever. But like the codebase is a sort of impenetrable mess
<davidrysk[m]>
> CAVEAT: the kmutil configure-boot process will only work on the first macOS volume. If you have multiple bootable volumes, you will only be able to boot a custom kernel on the first installation.
<modwizcode>
Is that even right? I read that and that sounds wrong?
<davidrysk[m]>
it's definitely not intended behaior
<modwizcode>
Isn't marcan's setup using a second volume?
<davidrysk[m]>
I thought it was
raster has quit [Quit: Gettin' stinky!]
Namidairo has quit [Read error: Connection reset by peer]
Namidairo has joined #asahi
coderobe has quit [Ping timeout: 246 seconds]
DarthCloud has quit [Ping timeout: 268 seconds]
<davidrysk[m]>
also uhh still no kernel symbols for t8101 in this KDK
<modwizcode>
What's the t8101?
coderobe has joined #asahi
Glanzmann has quit [Ping timeout: 265 seconds]
<davidrysk[m]>
the t8103 is the M1, the t8101 is the A14 but Apple labels the kernel for the M1 "t8101"
<davidrysk[m]>
since it's compatible with both
odmir has joined #asahi
Tokamak has quit [Ping timeout: 240 seconds]
alvaro_fe has quit [Ping timeout: 265 seconds]
Glanzmann has joined #asahi
<marcan>
davidrysk[m]: that's a bug and it's not the "first"
<marcan>
it's "some order"
<marcan>
it works fine for me on the second volume on both machines; I think I got lucky
<marcan>
I'm not sure if it's fixed in 11.2 final yet; it might
<marcan>
I should try it
<davidrysk[m]>
marcan: so is Apple's engineer wrong? :D
<marcan>
they are :)
<marcan>
as to the specifics of the bug anyway
<marcan>
it might be the "first discovered" or "first in UUID order" or "first according to some random hashtable implementation"
<marcan>
it's definitely not the first on disk
odmir has quit []
odmir has joined #asahi
Tokamak has joined #asahi
odmir has quit [Remote host closed the connection]
odmir has joined #asahi
odmir has quit [Ping timeout: 246 seconds]
alvaro_fe has joined #asahi
PhilippvK has joined #asahi
phiologe has quit [Ping timeout: 264 seconds]
odmir has joined #asahi
odmir has quit [Ping timeout: 240 seconds]
odmir has joined #asahi
amw has quit [Ping timeout: 258 seconds]
odmir has quit [Ping timeout: 258 seconds]
amw has joined #asahi
amw has quit [Ping timeout: 264 seconds]
awesomebing1 has quit [Quit: ZNC 1.7.2+deb3 - https://znc.in]
awesomebing1 has joined #asahi
marvin24 has quit [Ping timeout: 260 seconds]
marvin24 has joined #asahi
amw has joined #asahi
Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
a3541 has joined #asahi
<modwizcode>
It seems like an odd bug to have.
odmir has joined #asahi
VinDuv has joined #asahi
<dhewg>
most bugs are
odmir has quit [Ping timeout: 240 seconds]
mndza has joined #asahi
VinDuv has quit [Quit: Leaving.]
ncason has quit [Ping timeout: 246 seconds]
ch has quit [Ping timeout: 260 seconds]
ncason has joined #asahi
jaXvi has quit [Quit: Leaving]
a3541 has quit [Remote host closed the connection]
jaXvi has joined #asahi
ephe_meral has joined #asahi
klaus has joined #asahi
furkan has quit [Read error: Connection reset by peer]
furkan has joined #asahi
furkan has quit [Ping timeout: 246 seconds]
furkan has joined #asahi
<klaus>
i understand asahi linux is at the very first stage of implementation, but i wanted to ask: in the long run, is something like rosetta2 planned for it ? how will it be possible to run x86_64 programs inside asahi linux ? (with reasonable speed and gfx passthrough)
<j`ey>
klaus: qemu!
<klaus>
j`ey : is qemu working the way i described ? (good speed and gfx passthrough)
<never_released>
qemu isn't a nice answer though...
<never_released>
klaus: no.
<klaus>
otherwise any emu would have been a valid answer, like bosch and others..
<never_released>
you will not have graphics working with a user-mode Linux x86_64 -> arm64 compatibility layer today
<never_released>
because you'll need a second ioctl32 interface for that
<never_released>
or a very heavy hack
<never_released>
don't worry, people are working on it
<klaus>
never_released : really ? that is good to hear
<never_released>
klaus: yeah, as part of the FEX project
<klaus>
i can see many cases where the ability of running x86_64 code from inside a native arm64 OS is desirable.
<klaus>
thanks for the link, i go read that
<klaus>
hmm so cpu instruction translation is being worked on, but you were saying no gfx passthrough ? does that mean programs that run through fex will be term only, or just no gfx hardware acceleration ? (software rendering )
<never_released>
(it translates x86 32-bit -> 32-bit arm)
<j`ey>
> Box86 is useless on 64-bit only systems
<dhewg>
yeah, i know, but still, some ppl run it successfully
<never_released>
yeah, on 32-bit systems
<never_released>
on 64-bit box86 won't get you anywhere
<never_released>
it's a nice approach though, with thunking to use native libraries for tons of stuff
<dhewg>
if you install a 32bit userland it'll work there too
<dhewg>
but yeah, not that clean and nice
<never_released>
dhewg: M1 does *not* have ARMv7 support in hardware
<never_released>
good luck
<klaus>
asahi linux won't be able to run arm32 bit softs ? like on x64 users have ability to run x86 code through multilib ?
<never_released>
klaus: the hardware doesn't support ARMv7 code exec
<never_released>
it's gone
<klaus>
never_released hmm
<dhewg>
oh hehe
<dhewg>
ok, i've seen reports of other aarch64 users running that, but if that's the case that onviously won't help on m1
<never_released>
yeah those added aarch32 libs
<dhewg>
yeah
<dhewg>
still interesting, some are even using that with an amd gpu and using box86 with gallium nine
<never_released>
yep, and you can't mechanically build a 32 <-> 64 thunking layer
<never_released>
so the approach there is a dead end sadly
<dhewg>
yeah m1 yeah
<never_released>
not even just the M1
<never_released>
the Cortex-A78 successor, Matterhorn is the last Cortex-A big core with ARMv7 code exec
<never_released>
Makalu, the gen after that, can no longer run AArch32 code
<klaus>
i understand it's very early to talk about all that, but it's interesting so start planning early too so i'm voicing concern about it..the ability to run x86_64 from inside asahi linux would be a must have .. just like macos does it's rosetta2.. they made it quite nicely, wine for example runs fast..
<j`ey>
klaus: it's a separate topic really
<dhewg>
yeah, that's a long term dead end indeed
<klaus>
i understand it's a separate topic
<klaus>
you're right
<klaus>
just as a future user i'm thinking about what i'll probably need :)
<dhewg>
for wine there was also another translation project, "hangover" iirc?
<never_released>
dhewg: Hangover depends on an external JIT
<never_released>
which can be Qemu or FEX today
<never_released>
it doesn't include its own x86->arm JIT
ch has joined #asahi
<dhewg>
haven't really followed that as i don't have a use case myself
<dhewg>
but that works with fex today? nice
nspaced has quit [Ping timeout: 260 seconds]
<klaus>
fex looks good
<klaus>
thanks for telling
baryluk has quit [Ping timeout: 265 seconds]
irl25519 has joined #asahi
nspaced has joined #asahi
baryluk has joined #asahi
odmir has joined #asahi
raster has joined #asahi
amw has quit [Ping timeout: 240 seconds]
irl25519 has quit [Quit: irl25519]
amw has joined #asahi
odmir has quit [Ping timeout: 240 seconds]
acelogic has quit [Ping timeout: 260 seconds]
TheJollyRoger has quit [Ping timeout: 268 seconds]
<marcan>
you can get a serial console and use m1n1 proxyclient with just another M1 device and a normal type C cable
<j`ey>
:O
<j`ey>
marcan: it *has* to be M1 to M1?
<marcan>
yes
<j`ey>
how come?
<sven>
nice :)
<marcan>
j`ey: because only the M1s have this feature as far as I know
<marcan>
the VDM stuff may or may not work on some Intel macs, but I'm pretty sure the serial console won't
irl25519 has quit [Quit: irl25519]
isacc has joined #asahi
stemnic has quit [Ping timeout: 264 seconds]
mndza has quit []
stemnic has joined #asahi
mndza has joined #asahi
<bastilian>
marcan: nice! Does the target M1 need to be in DFU already or is there any other requirement to start a console, or is it basically just "plug and play"?
<marcan>
no requirement
<marcan>
you can literally hard reboot it if it's in macos
<bastilian>
Sweet! Awesome! Thank you!
<marcan>
obviously to actually *get* anything on serial you need to have installed m1n1; iBoot and macOS won't speak serial by default (but if you reduce security and tweak the nvram you can get macOS debug spew on serial)
<marcan>
the remote reboot stuff is super handy for a fast test cycle
<marcan>
just one command chain to reboot, chainload a new m1n1 and/or kernel, etc
<marcan>
no need to hold down the power button for ages
<marcan>
there is also a command to go into DFU mode
<marcan>
it's the same as that `areboot` thing I've been doing on my streams with the arduino/USB-PD rig
<marcan>
(the regular reboot that is, I wasn't using DFU)
irl25519 has joined #asahi
<bastilian>
Cool! I was just waiting for this tool to start testing and playing around, without having to build a USB-PD cable!
<marcan>
I figured enough people would have two of them for this to be useful :)
<bastilian>
Truth be told, I did get a second M1, because I hoped this would work.
<marcan>
ha :D
<marcan>
glad I could deliver then :)
<Emantor>
I like the disclaimer.
<marcan>
I am barely a macos user (as of late) and *definitely* not a macos programmer
<marcan>
(I suspect that will have to change pretty soon...)
raster has quit [Quit: Gettin' stinky!]
irl25519 has quit [Quit: irl25519]
<davidrysk[m]>
marcan: is there any point testing this on a T2 Mac?
<roxfan>
those are probably easier to get right now
<Shiz>
"
<Shiz>
I have no idea what I'm doing with IOKit and CoreFoundation -marcan"
<bastilian>
david.rysk: i think the T2/Intel machines do not have the serial component.
<davidrysk[m]>
bastilian: how does the serial component display itself in the ioreg tree?
<bastilian>
david.rysk: On the M1s I found it by just looking for uart or "serial-console" the same or a similar entry isn't there on T2/Intel machines.
zkrx has quit [Ping timeout: 246 seconds]
<davidrysk[m]>
okay, that was user error. It sees the device, it sees the M1 Mac, but unlocking fails.
<bastilian>
You mean you can establish a connection from an Intel -> M1?
<davidrysk[m]>
it's in APP mode
<davidrysk[m]>
when the Intel and M1 are connected, it sees the M1 in APP mode\
<davidrysk[m]>
but it does not appear that changing modes works
<davidrysk[m]>
the reset command (Gaid) does work though :)