marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | "Does XXX work yet?": https://alx.sh/fs | 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
ten157237743 has joined #asahi
Z750 has quit [Quit: Ping timeout (120 seconds)]
Z750 has joined #asahi
Soni has quit [Ping timeout: 480 seconds]
Hibyehello_ has joined #asahi
Hibyehello has quit [Ping timeout: 480 seconds]
ptudor_ has joined #asahi
ptudor has quit [Ping timeout: 480 seconds]
Soni has joined #asahi
amarioguy has quit [Remote host closed the connection]
vitalino has quit [Remote host closed the connection]
ptudor_ has quit [Read error: No route to host]
ptudor has joined #asahi
alyssa has joined #asahi
<alyssa>
22:43 <ChaosPrincess> you can reduce macos to like 70gb and leave it be
<alyssa>
truly progressive, only 70gb for the base OS, wow! (-*
bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
baozich has quit [Ping timeout: 480 seconds]
baozich has joined #asahi
baozich has quit [Ping timeout: 480 seconds]
unicordion has quit [Ping timeout: 480 seconds]
baozich has joined #asahi
bisko has joined #asahi
nyilas has joined #asahi
guillaume_g has joined #asahi
giskard has joined #asahi
chadmed_ has joined #asahi
chadmed has quit [Read error: Connection reset by peer]
baozich has quit [Ping timeout: 480 seconds]
<as400>
alyssa - amazing ! :)
baozich has joined #asahi
bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<raph_ael>
70g...
bisko has joined #asahi
<ChaosPrincess>
the updater is not great and needs a lot of free space to not fail. the os itself isnt that big.
<ChaosPrincess>
and y'all kinda making me feel like im lucille from arrested development asking how much a banana could cost :P
<chadmed_>
its one operating system michael, how much space could it take up
<chadmed_>
3 petabytes?
the_CLA has joined #asahi
the_CLA_2 has quit [Ping timeout: 480 seconds]
baozich has quit [Ping timeout: 480 seconds]
roxfan has joined #asahi
baozich has joined #asahi
roxfan2 has quit [Ping timeout: 480 seconds]
<raph_ael>
my first hdd was 2.5Go :)
<mps>
heh, my first hdd was 10MB
<mps>
and before that I used a lot 360KB disketes
<as400>
mps: wow, my first hdd was 40MB :) And it was big, fat and heavy
<mps>
sorry for more OT, I still have armv7 chromebook with 16GB emmc which I used as my workstation for long time and I would still use it if I didn't bought m1
zanfix has joined #asahi
<zanfix>
Hi! question: is it safe right now (today) to update my ALARM KDE installation on Macbook air M1 with pacman -Syu? I was reading about the various problems in the last week...
bcrumb has joined #asahi
baozich has quit [Ping timeout: 480 seconds]
baozich has joined #asahi
baozich has quit [Ping timeout: 480 seconds]
baozich has joined #asahi
bcrumb has quit [Quit: WeeChat 3.8]
jmsunseri_ has joined #asahi
vitalino has joined #asahi
<jmsunseri_>
pardon me if this is a dumb question but i see a lot of M1 mac mini on ebay selling for nothing because they are locked. if i wanted to remove macos and replace with linux would i be able to do it with one of these machines? or would the lock still prevent me from doing so?
<ChaosPrincess>
the lock will prevent you from doing so
<ChaosPrincess>
those are bricks, only useful for parts
<jmsunseri_>
so sad
jmsunseri has joined #asahi
jmsunseri_ has quit [Quit: Page closed]
baozich has quit [Ping timeout: 480 seconds]
jmsunseri has quit [Quit: jmsunseri]
vitalino has quit [Remote host closed the connection]
baozich has joined #asahi
hightower4 has quit [Ping timeout: 480 seconds]
baozich has quit [Remote host closed the connection]
baozich has joined #asahi
zanfix has quit [Quit: Konversation terminated!]
baozich has quit [Ping timeout: 480 seconds]
zanfix has joined #asahi
<zanfix>
Did I ask a dumb question? My system i stable at the moment, I'm asking just not to screw up my installation...
<marcan>
zanfix: wait until the next update if you want to be certain (real soon now)
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
leitao has joined #asahi
alyssa has quit [Quit: leaving]
unicordion has quit [Quit: Konversation terminated!]
emberdex has joined #asahi
<emberdex>
Hi, are PRs being accepted to the GitHub repos at this time? I have a small PR up on the asahi scripts repo and didn't even stop to think about whether they were even being accepted
unicordion has joined #asahi
<emberdex>
No worries if not, I can close it out otherwise.
unicordion has quit []
<ChaosPrincess>
in my experience - they are, but you need to annoy important people enough for it to happen :P
<emberdex>
it's really not important, so i won't bother going that far - thanks ChaosPrincess
Moprius has joined #asahi
Brainium has joined #asahi
emberdex has left #asahi [#asahi]
unicordion has joined #asahi
gladiac is now known as Guest8347
gladiac has joined #asahi
Guest8347 has quit [Ping timeout: 480 seconds]
possiblemeatball has quit [Remote host closed the connection]
Moprius has quit [Quit: bye]
qwe has joined #asahi
giskard has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
chadmed_ has quit [Ping timeout: 480 seconds]
alxdrl has joined #asahi
___nick___ has joined #asahi
chadmed_ has joined #asahi
___nick___ has quit []
___nick___ has joined #asahi
___nick___ has quit []
___nick___ has joined #asahi
<qwe>
Hi folks. I'm doing some small research for a class on cryptography, revolving around the question "how needed is firmware security and authentication on modern x86 computing devices?"
<qwe>
It occurs to me that, since Apple does so many things differently and has so many firmware blobs for e.g. RTKit on the GPU firmware, perhaps it would be helpful to ask the following to aid my research:
<qwe>
err, that last message didn't send correctly
<qwe>
It occurs to me that, since Apple does so many things differently and has so many firmware blobs for e.g. RTKit on the GPU firmware, perhaps it would be helpful to ask the following to aid my research:
<qwe>
How many pieces of firmware blobs must be uploaded to the M-series in order to make various aspects of the device functional, *and* are signed/have a tamper-proof mechanism?
<qwe>
I've been stalking the Asahi Linux project since its inception and I'm not personally aware of any documentation of this, but if there is such docs I'd be happy to go dig in to them
<ChaosPrincess>
if loaded by is 'secureRom', iboot1 or iboot2 - means its signed and tamperproof
leitao has quit [Ping timeout: 480 seconds]
<qwe>
so for instance one could modify and upload their own Touch Bar firmware, since the OS loads it?
<ChaosPrincess>
yes, you can
<ChaosPrincess>
i even techincally did it, by accident :P
<ChaosPrincess>
as in - i screwed up the upload code and uploaded parts of it wrong
<marcan>
the Touch Bar firmware is signed just like everything else
<marcan>
it's just not verified *by the Touch Bar*
<marcan>
(it's verified by the recovery process / the OS)
<ChaosPrincess>
its only verified by the macos
bcrumb has joined #asahi
<ChaosPrincess>
not by tamper-proof iboot or on-chip things
<marcan>
right, which only means it can be modified the same way the main OS can be modified (which still has a bunch of process to make it secure)
<marcan>
(which is why the Asahi Linux installer exists and does what it does)
<ChaosPrincess>
we are technically not nearly as tamperproof as macos unless the user does a bunch of extra work
<marcan>
put another way: you have to distinguish between firmware that can be tampered with *persistently* and which is stored in off-SoC storage, and firmware that can be tampered with *transiently* but which is ultimately loaded by the OS, and therefore trustworthy if the OS itself is.
<marcan>
ChaosPrincess: no, what I'm saying is the installer is in charge of following through apple's process so you can install it at all
<qwe>
my learning goals are sorta tangential to that point, if I'm not mistaken; my understanding is that currently, Secure Boot on x86 inherently allows you to tamper with firmware blobs that can be uploaded and are blindly trusted
<marcan>
which is a controlled process for this and other reasons
<marcan>
I'm talking about boot policies, not the firmware extraction
<qwe>
my extremely simple thought in my head is "tell the camera firmware not to turn the camera light on when the camera is on"
<ChaosPrincess>
that one is impossible
<ChaosPrincess>
there is a non-reprogrammable fpga (oxymoronic, but still is) in charge of that
<qwe>
or perhaps have a rogue Thunderbolt controller allow DMA in any case, which I believe is exactly what Thunderspy accomplishes
<ChaosPrincess>
thunderspy is impossible if you do not misconfigure the darts
<ChaosPrincess>
dart = iommu in x86 speak
<marcan>
the thunderbolt controllers are internal and the firmware (which doesn't even have the access to do that) is signed anyway
<qwe>
ChaosPrincess: Is that particularly in the case of Macs, or are all cameras in laptops this way?
<marcan>
so no
<marcan>
Macs in particular, because Apple cares about all this quite a bit more than the x86 world
<ChaosPrincess>
qwe: arm macs, other laptops do whatever they want
<ChaosPrincess>
x86 secureboot is a lot less secure than what apple made here
<qwe>
marcan: Right, which is why I feel that this is an appropriate conversation to have with you all
<qwe>
Because Apple is doing this stuff right
<qwe>
And I want to assert that x86 is... not
<marcan>
tbh I don't know much about x86 secureboot so it's hard for me to provide very specific context, but that wiki page that was linked is a pretty good overview of how it all works in practice :)
<ChaosPrincess>
x86 is not inherently less secure
<ChaosPrincess>
like, think xbox one and newer
<marcan>
as an architecture no, but actual x86 implementations mostly are
<ChaosPrincess>
that
<qwe>
that's true, isn't it? I don't have a deep insight into the Xbox security platform, but they don't mess around
<marcan>
MS is basically the only vendor that can somewhat get somewhere close to where Apple is on that front
<marcan>
and to do that they had to get AMD to make custom silicon for them to their specification and actually designed part of it
<marcan>
(Sony tried and failed a bunch of times, even though they were working with the same AMD)
<qwe>
As I recall, MS did the same thing with IBM; I'm saying this to frame it in my head
<ChaosPrincess>
arguably sony failed not due to earlyboot level exploits
<marcan>
(want to know why? MS hired the Xbox hackers and Apple hired the iPhone hackers. Sony did not hire the Playstation hackers.)
<ChaosPrincess>
ps4 is usually hacked via webkit
<qwe>
Does Xbox Series * use transient memory encryption?
<ChaosPrincess>
yes
<ChaosPrincess>
google AMD SME/TSME
<qwe>
err not transient
<qwe>
what is it
<ChaosPrincess>
transparent?
<marcan>
ChaosPrincess: sony failed due to earlyboot stuff lol, we owned the PS4 via PCIe injection due to bad IOMMU perms
<marcan>
webkit came later
<marcan>
and lower level security stuff of PS4/5 has been compromised too
<ChaosPrincess>
did y'all get persistence?
<marcan>
with a modchip sure, that's how that worked
<qwe>
might be transparent
<qwe>
IDR
<ChaosPrincess>
eh
<marcan>
not that it was ever productionized, but in principle it would work
<qwe>
w/ever
<ChaosPrincess>
i mean, hats off, but thats not exactly ps3 and lower level fails
<marcan>
well PS3 was *very* special
<marcan>
like it is now the academic reference for how to fuck that up special
<marcan>
hard to screw up *that* bad twice
<marcan>
but anyway, there's nonpublic stuff affecting lower levels on newer consoles too
<ChaosPrincess>
i might disagree, i poked at ps5's bd jvm, holy shit the shitshow that is there is something extra
<marcan>
ChaosPrincess: "we have your private keys now" is hard to top
<ChaosPrincess>
like, im newbie af at those and even i could own userland there
<qwe>
there was something public released about that, no?
<qwe>
w/r/t bd jvm
<marcan>
like "your private keys to burned-in bootloaders"
<ChaosPrincess>
yea, i seen that talk
<marcan>
seriously though we keep getting cited on how to screw up ECDSA, it's nuts
<marcan>
actual academic creds there
<sven>
lol, i wrote some other research papers during to my PhD and that ps3 stuff is still cited more often than those :D
<marcan>
(meanwhile the PS4 PCIe anecdote is a favorite of the hardware hacking world, I just got quoted on hackaday again :p)
<qwe>
marcan: coming back to secure platforms: how is it the case that Thunderspy is not possible if you don't screw up your IOMMUs?
<ChaosPrincess>
you cant write to arbitrary locations in system memory over tb
<ChaosPrincess>
iommu will say no
<qwe>
wait, am I confusing something?
<sven>
yeah, you have to specifically allow memory regions that can be accessed inside the iommu
<marcan>
there are no Intel thunderbolt controllers on AS machines
<qwe>
Or was that on the CPU side?
<ChaosPrincess>
iommu is between tb controller and soc memory
<marcan>
it's all Apple-designed, and the firmware is all signed
<sven>
iommu sits between the thunderbolt pcie controller and all of memory
<ChaosPrincess>
(logically, physically its all on same soc)
<qwe>
oh, you're referring to AS again
<qwe>
right, of course
<marcan>
qwe: everything is inside one chip on AS, and the DART is only controlled by the OS, and the TB firmware is signed anyway so you can't just compromise it like you can on intel
<qwe>
I'm asking x86 questions in an AS channel :)
<ChaosPrincess>
yea :P
<sven>
even if you can override the thunderbolt firmware (which might actually be possible since it seems to be inside some weird MMIO mapped thingy) you can’t bypass the iommu
<marcan>
but yes, even on x86 the IOMMU should still apply if you do it right
<ChaosPrincess>
even on x86 you can also iommu the thunderspy away
<ChaosPrincess>
just dont screw up your drivers
<ChaosPrincess>
and have an iommu in the first place
<sven>
and usb4 requires a properly configured iommu as far as I understand the spec
<qwe>
do you all have any other obvious examples of how an insecure firmware on a device could be used as a drive-by attack on a computing device?
<qwe>
Of course this is an obvious question, I just want something to point at
<ChaosPrincess>
actually, maybe AS is an example, macos had a bug in DCP firmware, and rather lax iommu config
<ChaosPrincess>
so some malware used it to pivot
<qwe>
I vaguely remember this
<qwe>
There wasn't a blog post on it, was there?
<qwe>
and marcan deleted his Twitter so anything he wrote on it is gone now
<marcan>
can't work around poorly written drivers :p
<marcan>
(our DCP driver got that right)
<marcan>
DCP had a lovely callback to ask the OS to give it access to random memory, and the instant I saw that I knew we were checking against a whitelist.
<marcan>
turns out... Apple had no whitelist.
<qwe>
thanks for the link, and the story. Project Zero is fairly reliable, I think I'll have no trouble including it into my presentation
<qwe>
also, insert meme about turing-completeness here
<marcan>
there's also that talk Lina gave on her AGX exploit that can pivot from a shader to full system compromise
<marcan>
it's in Japanese though :p
<qwe>
ah... explains why I hadn't heard of that one :/
<qwe>
that sounds incredible, though, and I'd love to watch it if there are subtitles available
<marcan>
I wonder if the autosubs on that one are any good...
<marcan>
she said she'd do it in english some day though
<qwe>
she's not in here, is she? I'd love to give a +1 to "here's an interested student!" if possible
<marcan>
not on this channel, no
<ChaosPrincess>
yea, getting an en version would be nice, i know only like 3 words in jp :P
<qwe>
I do regret not following along with the GPU streams the last couple years
<qwe>
although they were primarily at 3AM or smth around here...
delsol has quit [Remote host closed the connection]
<marcan>
speaking of, it's 4AM here and I really should sleep :p
delsol has joined #asahi
<qwe>
is it an inherent prerequisite to being good at hacking that you keep a horrifying sleep schedule?
<sven>
it’s surprisingly correlated :(
<qwe>
might explain why I've still yet to do anything interesting :P
alxdrl has quit [Quit: Konversation terminated!]
<qwe>
going to bed at midnight somewhat limits your coding time
<ChaosPrincess>
oh, so 4 am is the benchmark, and here me thinking that i can achieve something with only 2am :P
<marcan>
I *usually* try to go to sleep before 4AM.
<handlerug>
might be a bell curve. I go to sleep at 6 AM yet I'm not even close to a hacker
<marcan>
blog/update release days tend to be some of the worst :p
<marcan>
I would say these days I average aiming for 2 and ending up at 3...
<psykose>
i think going to sleep at 6am is just depression and not much of a hacking thing
bcrumb has quit [Quit: WeeChat 3.8]
<sven>
the hacking thing at least for me usually was „just one more attempt and it’s totally gonna work” and suddenly it’s 6am
<handlerug>
that is how it usually ends up
delsol has quit [Ping timeout: 480 seconds]
Cyrinux9 has quit []
Cyrinux9 has joined #asahi
<mkurz>
Question: Until now I was using xorg but will be switching to wayland now since it is recommend now according to todays release blog post. However, will all the KDE "System settings" be preserved or do I need to reconfigure everything?
<ChaosPrincess>
you will lose some settings, like dpi configs and such
<mkurz>
ChaosPrincess: But settings not related to display/graphics should stay the same?
<ChaosPrincess>
iirc they did
<mkurz>
ChaosPrincess: OK I will see, thanks!
mkurz has quit [Quit: Leaving]
mkurz has joined #asahi
mkurz has quit []
<Cyrinux9>
hello! thx for the last update, im on a m2 air, should I be able with the last update to get OpenGL version: 3.2 in supertuxkart ? a friend update on a m1 pro, he got OpenGL version: 3.2, and its stuck on opengl 2.1 on mine :)
jacksonchen666 has joined #asahi
Techcable_ has joined #asahi
Techcable has quit [Read error: Connection reset by peer]
<Cyrinux9>
ha, sorry for the noise, it was a bad combo for the m1 pro, after update and reboot, opengl 2.1 like my m2 air.
brad_ has joined #asahi
unicordion has quit [Ping timeout: 480 seconds]
mkurz has joined #asahi
<mkurz>
I just upgraded my Macbook Pro M1 Pro to the latest edge linux/mesa, switched to Wayland, set Scala to 150% and disabled "Force Font DPI" but I have the feeling everything is quite small and the fonts the fonts are not correctly rendered...
<mkurz>
I logged out and logge din to xorg again and then again back to wayladn
<mkurz>
ok I will log out now and back in
<mkurz>
see you in a minute
mkurz has quit [Quit: Leaving]
gabuscus has joined #asahi
jacksonchen666 has quit [Remote host closed the connection]
rvalue has joined #asahi
<i509vcb>
in the case of xwayland you'll have to restart the app to get the new scale
mkurz has joined #asahi
ioldoortileotm^ has joined #asahi
<mkurz>
jannau: it's better now. However 150% was to small for my eyes, so I ended up with with 200%
<mkurz>
however the cursor is very big now...
jacksonchen666 has joined #asahi
<jannau>
mkurz: logging out a single time after changing the scale is enough
<mkurz>
any idea how to downscale the cursor only?
<mkurz>
in the kde settings it's size is 24
<mkurz>
which is the smallest
<jannau>
you're probably avoiding some fractional scaling issues with 200%
<mkurz>
so I would need to set it to 12 probalby
<mkurz>
jannau: It was ok with 150% as well after log out/login, however it just was to small for me
<qwe>
is that bug where GTK3 apps display the cursor as enormous with KDE fractional scaling still present?
<mkurz>
200 is much better for me
<jannau>
no idea about the cursor
<mkurz>
jannau: thanks anyway
mkurz has quit [Quit: Leaving]
<ayke_>
The cursor is buggy with Firefox and I think other GTK3 programs
<ayke_>
This is one of the main reasons I'm still on X11
<ayke_>
It should be fine in Qt apps
<ayke_>
IIRC you can also pick a different cursor theme to work around the bug.
<opticron>
my main reason for being on X11 is synergy/barrier support
mkurz has joined #asahi
<mkurz>
Is WebGL2 working in Chromium? In chrome://gpu/ it says "WebGL2: Hardware accelerated" but e.g. https://summer-afternoon.vlucendo.com/ shows "Seems like WebGL2 is not supported by your browser"
<bluetail3>
opticron please explain what you mean with synergy/barrier support
<mkurz>
Or do I need to start Chromium with some special flags?
<opticron>
bluetail3, synergy and barrier are applications that allow you to share a keyboard/mouse across any number of machines with monitors
<bluetail3>
opticron have another reason too. On my particular system wayland crashes all the time on sway, but I need such manager, cause I want rules to put windows in place. I tried this extensively and it just didnt workout on anything but i3/sway.
<bluetail3>
theres apparently quite a bit of bugs with wlroots/sway that are still not patched out
<bluetail3>
so my 6950xt just does amdgpu-reset, so does my nvidia gt930 that I could somehow run on sway..
<bluetail3>
but takes a bit of patience to happen
<opticron>
Cyrinux9, that's the first I've heard of that, so no
<opticron>
I'll take a look, thanks
<bluetail3>
btw, sway does not leave a coredump in my case
mkurz has quit [Remote host closed the connection]
<Cyrinux9>
opticron tell me if this works well then ;)
chadmed_ has quit [Ping timeout: 480 seconds]
mkurz has joined #asahi
___nick___ has quit [Remote host closed the connection]
___nick___ has joined #asahi
jacksonchen666 has quit [Remote host closed the connection]
mkurz has quit [Quit: Leaving]
<Retr0id>
what do people use as a hypervisor, on asahi as host? Normally virtualbox is my go-to, but I don't see it in the repos, doesn't seem like aarch64 support for it is very mature either (although I'm attempting to build from source rn, will see how that goes...)
jacksonchen666 has joined #asahi
<j`ey_>
Retr0id: qemu
<j`ey_>
maybe with libvirt or something as a frontend
<Retr0id>
I did give qemu a quick try, but it seemed like there was no support for gui frontends compiled in (no gtk, no sdl, and it defaulted to vnc)
<Retr0id>
will give gnome-boxes a try
<j`ey_>
oh right, yeah, idk about the qemu from the repos
<Retr0id>
welp, the gnome-boxes package seems to be broken: libsoup2 symbols detected. Using libsoup2 and libsoup3 in the same process is not supported.
<psykose>
ah yes that ole thing
<Retr0id>
virt-manager looks like it could work tho
ptudor has quit [Ping timeout: 480 seconds]
ptudor has joined #asahi
<Retr0id>
hm it wants uefi firmware, that's gonna be fun (or maybe I can do direct kernel boot, might be a pain tho)
<j`ey_>
Retr0id: you can grab that from the qemu build
<j`ey_>
cant remember the exact name, but it ends in .fd :D
<Retr0id>
do you know what package that's from?
<j`ey_>
it's edk2 firmware
<Retr0id>
ah there's a virt-firmware package
<j`ey_>
I grabbed it from debian, but that might have it too
<j`ey_>
QEMU_EFI.fd, or maybe AAVMF.fd or something