marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | General project discussion | GitHub: | Wiki: | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-offtopic | Keep things on topic | Logs:
phiologe has joined #asahi
PhilippvK has quit [Ping timeout: 480 seconds]
bgb_ has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
marvin24_ has joined #asahi
marvin24 has quit [Ping timeout: 480 seconds]
artemist has quit [Ping timeout: 480 seconds]
TypoKign[m] has joined #asahi
artemist has joined #asahi
s-urabe[m] has joined #asahi
VinDuv has joined #asahi
hir0pro has quit [Ping timeout: 480 seconds]
VinDuv has quit [Quit: Leaving.]
hir0pro has joined #asahi
pugguu has joined #asahi
pugguu has quit [Ping timeout: 480 seconds]
beej has quit [Quit: Connection closed for inactivity]
bps has joined #asahi
choozy has joined #asahi
pcm720 has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
pcm720 has joined #asahi
choozy has quit [Remote host closed the connection]
<brentr123[m]> What happened to the stream?
<marcan> brentr123[m]: starting now :)
<brentr123[m]> Sorry I’m so impatient, I’m kind of addicted to them 😂
<j_ey> is part of the thing that got merged the 'pmgr' and 'acpie' in the mmio output?
<_jannau_> marcan: looks like the buffer overflow I saw on mac os
<_jannau_> marcan: try disabling the mmio addr lookup
<sven> marcan: uh, does it always run on the same core?
bradfier has joined #asahi
<_jannau_> lookup_all is not used
branon has joined #asahi
cadubentzen has quit [Quit: ZNC 1.7.5+deb4 -]
leah2 has quit [Quit: trotz alledem!]
hir0pro has quit [Ping timeout: 480 seconds]
<sven> simd in pmgr :D
<_jannau_> SIMD not disabled via compile flags and auto vectorization?
<sven> for MMIO reads/writes?
<sven> or maybe it's not MMIO but some shared memory region for one of the subprocessors?
jeh has joined #asahi
jeh is now known as Guest141
Guest141 has quit []
jehh has joined #asahi
jehh has quit []
jehh has joined #asahi
skali has joined #asahi
_whitelogger has joined #asahi
<kettenis> marcan: the pcie initialization code that's now in m1n1 might make macos unhappy
<kettenis> I had u-boot hang if I repeated some of the initialization steps that m1n1 now does
<kettenis> so I removed that code from u-boot
hir0pro has joined #asahi
<kettenis> it may make sense to only do the pcie_init later when booting a "kernel"?
<kettenis> good enough I'd say ;)
hir0pro has quit []
jkc has joined #asahi
jkc has left #asahi [#asahi]
jkc has joined #asahi
choozy has joined #asahi
<jn> nice work on booting to login!
<thunfisch> nice work man!
<thunfisch> that was fun to watch :)
<j_ey> :O
<thunfisch> aw yeah, rainbow apple <3
<jn> very nice :)
skali has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<kettenis> you're lucky that the GPU uses main memory instead f VRAM out on the PCI bus
pcm720 has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
pcm720 has joined #asahi
arekm has joined #asahi
arekm has quit []
arekm has joined #asahi
pcm720 has quit []
pcm720 has joined #asahi
arekm has quit []
arekm has joined #asahi
arekm has quit []
arekm has joined #asahi
arekm has quit []
arekm has joined #asahi
<jix> marcan: the map_hw call still has reg.addr instead of addr (you missed undoing that one)
<jix> marcan: (also I think the map_sw at the top should be a map_hw, at least you said that while writing it)
Andalu30 has joined #asahi
<jannau> pmp[17] overlaps with sgx
<ar> marcan: sysctl hw.ncpu
VinDuv has joined #asahi
<jix> marcan: missing space at the start of the string appended to the end I think
Izumoo has quit [Ping timeout: 480 seconds]
<ar> hm. "networksetup -setairportpower en0 off" should disable wifi, but i'm not sure if it's gonna be permanent
<ar> also, sysctl machdep.cpu.brand_string shows "Apple M1" here
<kettenis> marcan: the snapshot stuff might be because the kernel panic'ed
<ar> oh, yeah, and after 3 kernel panics on shutdown it *might* want to go into a "reinstall macos" screen on next boot
<ar> marcan: ^
<ah-[m]> interesting, it runs on one of the efficiency cores?
<ar> (it happened to me on my work mbp13 m1)
<ar> marcan: command shift 3
jehh has quit [Quit: Konversation terminated!]
<jannau> just kick wifi off from the ADT
bps has quit [Ping timeout: 480 seconds]
Andalu30 has quit [Ping timeout: 480 seconds]
Andalu30 has joined #asahi
<Fanfwe> marcan: I didn't actually mean to push you to implement the m1racles mitigation right away :-P
<marcan> too bad :p
<Fanfwe> At least you now have a good story to tell: Revealing the vuln, and the next day publishing software that allows to mitigate the problem
<marcan> :D
<Fanfwe> (as long as you don't need all of your M1 cores)
pugguu has joined #asahi
<pugguu> Hello stupid question but will i be able to run asahi linix on a 256gb m1 mac mini
<Emantor> pugguu: pretty sure you can, setting it up at the moment.
<Emantor> I'll update you when the second mac os installation is finished and m1n1 is running.
<j_ey> pugguu: that's probably the one marcan has
<Emantor> Note that this is only interesting if you want to do driver development/hardware investigation/testing for Asahi Linux. Not really useful for general purpose computing yet.
<pugguu> I am waiting for the boot selection menu so that i can use m1n1 but still use macos welst waiting for general purpose stiff
<pugguu> Stuff*
<j_ey> pugguu: that already works
<pugguu> Wait so i dont need a secondary computer
<Emantor> You'll need a second computer to run the shell/m1n1 proxy.
<Emantor> One could probably hack HID support into m1n1 and use the framebuffer…
<jn> Emantor: you'd still need a machine to run the python side of m1n1
<jn> not a beefy machine, and not a mac (with the recent USB-ACM driver), but still something
<jn> when doing something interactive with m1n1
<Emantor> marcan: src/hv_wdt.c:5:10: fatal error: cpu_dbg_regs.h: No such file or directory
<Emantor> bfe8c94c adds the include but not the file itself :-)
hir0pro has joined #asahi
h_ro[m] has joined #asahi
hir0pro has quit [Quit: Konversation terminated!]
hir0pro has joined #asahi
hir0pro has quit []
hir0pro has joined #asahi
Izumoo has joined #asahi
pugguu has quit [Read error: Connection reset by peer]
pugguu has joined #asahi
pcm720 has quit [Ping timeout: 480 seconds]
hir0pro has quit [Ping timeout: 480 seconds]
<marcan> oh that's a derp, I never used that
<marcan> just remove the include
<marcan> fixed
choozy has quit [Ping timeout: 480 seconds]
<Emantor> Ye, I found that too. Unfortunately it now refuses to boot m1n1 with a failure to boot and reverts to macOS recovery D:
<marcan> odd :/
Andalu30 has quit [Ping timeout: 480 seconds]
Andalu30 has joined #asahi
choozy has joined #asahi
pugguu has quit [Ping timeout: 480 seconds]
<Emantor> No, I am just too tired. curl <host>:8000/m1n1.macho > m1n1.macho will happily put the 404 into m1n1.macho, and kmutil does not even check whether the file your setting up for boot is a valid macho. Turns out a 404 error can't boot.
<sven> lol
* hell__ can't avoid reading "macho" in Spanish (which means "male")
<Emantor> python3 -m http.server seems to serve nice 404 error pages, I don't remember python2 -m SimpleHTTPServer doing that.
<emilazy> it's a deliberate pun, macho is a word in english too
<hell__> oh, lmao
<jannau> would it make sense to make a CI build m1n1.macho available?
<Emantor> I am wrong, python2 also serves an error page…
bps has joined #asahi
<jannau> a little annoying that 90% of the build time is spent installing the cross compiler
<sven> sounds good to me. in the worst case it provides a sanity check that only code which actually compiles is merged. in the best case it makes someone's life easier :)
<Emantor> 17 seconds to install the cross compiler sounds fine to me.
<Emantor> If you want to speed this up you can probably create a container which includes the cross-compiler and have github spin this up instead.
<jn> hell__: "Macho" has also found its way into the german language, meaning "someone who is aggressively, stereotypically male"
<j_ey> jn: same in english
<jannau> building m1n1 just takes 3 second. let's hope azure gets Ampere Altra nodes soon
<VinDuv> I think I remember a story about some software whose auto-updater replaced itself with a 404 page, and the resulting binary would actually cause bad things when executed
pcm720 has joined #asahi
<hell__> jn: yeah, it can also mean this in Spanish
<jn> interesting
<jn> (as a somewhat useless side-note, i pronounce the file format differently, as makh-o)
<emilazy> I believe makh-o is the correct pronunciation, since it's Mach-O after all
<emilazy> but it obscures the pun :p
Andalu30 has quit [Quit: Konversation terminated!]
VinDuv has quit [Quit: Leaving.]
<TypoKign[m]> Psalm 74:9
<yrlf> I pronounce it as in "Mach One" (as in the speed of sound), which probably is the intended association
<balrog> it's Mach Object
<balrog> but maybe?
<balrog> and yeah, emilazy I agree, I believe makh-o is correct, since it comes from the Mach kernel, and object file, back from the NeXT days
Ferluci[m] has joined #asahi
pinskia has joined #asahi
fff has joined #asahi
<fff> hello all
<fff> what do we know about the SEP?
<fff> particularly the interface it exposes
<jn> the first step would be to document the interface :)
<jn> not sure how much documentation exists at the moment
<fff> well i am curious if in fact, one day, i will be able to use it from linux to e.g. hold ssh keys
<fff> i guess that's still an open question then
<sven> that's the goal at least
<sven> it should be possible since you can use the SEP from mac os for ssh keys
<fff> good
<fff> and the current state is that there's some memory-mapped region presumably for communication with the SEP and not much more known that that?
<fff> *than that
<sven> pretty much. i've seen some slides about the mailbox protocol though
<sven> but that's the easy part
<fff> where can i find those slides? :)
<sven> don't remember, sorry :-)
<fff> that's okay
<sven> but iirc i googled for secure enclave mailbox or something like that
<sven> but the registers looked very simple. it's probably obvious how to send messages from looking at a single mmiotrace
pcm720 has quit [Ping timeout: 480 seconds]
<fff> these would be the slides probably
<fff> although i might prefer reading mmiotraces to reading slides
<ah-[m]> what's the hard part?
<sven> figuring out the protocol itself and then figuring out how to nicely integrate that into the whole gpg/ssh/etc. ecosystem
<sven> and yeah, those look like the slides
fff has quit [Quit: Page closed]
<rkjnsn[m]> Just out of curiosity, I see that the goals for Asahi include not just hardware functionality but also polish.
<rkjnsn[m]> Is, e.g., taking advantage of the P3, 500nit display for HDR playback in scope as part of that polish?
<mjg59> rkjnsn[m]: The state of HDR in Linux overall isn't great at the moment
<mjg59> Other than the extent that there's GPU support required, I think it's probably out of scope
choozy has quit [Quit: - Chat comfortably. Anywhere.]
kettenis_ has joined #asahi
kettenis has quit [Ping timeout: 480 seconds]
pcm720_ has joined #asahi
<yrlf> I guess that if they find out how the needed GPU support for HDR works, they'll add it to the driver, but toolkit and application support for HDR is very "not there" yet on linux so I don't expect that to get further soon
<emilazy> colour management on linux is very bad :(
<emilazy> wayland may improve it at some point in the future
<emilazy> the only thing you can really do is assume every application is in sRGB unless you have good reason to believe otherwise
<yrlf> there is a work in progress protocol specification for color management for wayland
<yrlf> though that specification is still missing details for HDR, mostly color space stuff until now
<ar> last time i tried wayland (sway and gnome3 around february), i hit a lot more serious bugs than "colorspace management is wrong"
<ar> like, with multiple monitors connected, menus in some apps ended up waaaay off
<emilazy> colourspace management being wrong can be pretty bad
<emilazy> it means that linux gui is inevitably super oversaturated and weird looking on many displays
<emilazy> wayland's multi-display story is ... in theory better than X's by quite a lot
<emilazy> but Xwayland probably has issues with it I imagine
<ar> also, does wayland still have 24bit colorspace hardcoded?
<ar> i vaguely remember it being the case that 8b/ch/pixel was hardcoded in the protocol around 2 years ago