ChanServ changed the topic of #asahi-dev to: Asahi Linux: porting Linux to Apple Silicon macs | Non-development talk: #asahi | General development | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Logs: https://alx.sh/l/asahi-dev
user982492 has joined #asahi-dev
Urriellu has quit [Ping timeout: 480 seconds]
Urriellu has joined #asahi-dev
Urriellu has quit [Ping timeout: 480 seconds]
skipwich has quit [Quit: DISCONNECT]
skipwich has joined #asahi-dev
rory_be has quit [Remote host closed the connection]
rory_be has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
Urriellu has joined #asahi-dev
rory_be has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
user982492 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
soonercane has quit [Quit: Bye]
soonercane has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
<marcan>
zzywysm: it's not pushed yet, no. it's merged locally.
krbtgt has quit [Ping timeout: 480 seconds]
enron8 has joined #asahi-dev
nafod has quit [Quit: Ping timeout (120 seconds)]
enron has quit [Quit: Ping timeout (120 seconds)]
nafod has joined #asahi-dev
enron8 is now known as enron
cylm has quit [Ping timeout: 480 seconds]
rory_be has joined #asahi-dev
gabuscus_ has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
gabuscus has quit [Ping timeout: 480 seconds]
nafod has quit [Remote host closed the connection]
<marcan>
sven, jannau & co: any objections to merging the 4K patches soon™?
<marcan>
I think 1 year in we've mostly covered the "making people fix their stuff on 16K" problem
<marcan>
and FEX actually works now for running real games to there's actually a valid use case for 4K (also waydroid I guess...?)
<marcan>
*so
<marcan>
some of the patches are... a bit questionable, but I don't think they're likely to regress 16K builds
<marcan>
this would be a linux-asahi-edge-4k package eventually (RIP my kernel build times...)
rory_be has joined #asahi-dev
<marcan>
in other kernel news, I'm going to do another 6.2 asahi tag with a couple trivialities, and then rebase on 6.3-rcX and start pushing through the asahi-soc queue
<sven>
im not sure what happens with thunderbolt/untrusted plus my old 4K patches
<sven>
I don’t remember if I disallow dma for those devices entirely
rory_be has quit [Ping timeout: 480 seconds]
<jannau>
the dcp size alignment patch is fine, I haven't looked at the iommu/dma changes in detail
<sven>
other than that im rather sure the patches work, they’re just a bit awkward due to the way the iommu api works
<marcan>
jannau: yeah, that one I already merged in, it's trivial
<marcan>
sven: if TB is broken on 4K builds, whatever, we can worry about that later. as long as it doesn't regress 16K I'm not worried.
<marcan>
this is mostly about giving people a way to use FEX (including the developers) on AS, since that stuff is starting to matter now
nyilas has joined #asahi-dev
rory_be has joined #asahi-dev
<sven>
the patches may expose adjacent pages on 4K for untrusted fwiw
<sven>
ill have to check again
<sven>
but it shouldn’t break anything on 16k builds
rory_be has quit [Ping timeout: 480 seconds]
<marcan>
can we add a kernel security taint on 4K?
<marcan>
I think TAINT_CRAP is the best fit here, heh
<jannau>
is there a matching TAINT? otherwise it's just 'add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);'
<marcan>
it's not exactly a firmware workaround...
<jannau>
nah, that's just copy-n-paste from dcp module unload code
<jannau>
just add the taint in dart_probe with 4k pages
rory_be has joined #asahi-dev
nyilas has quit [Remote host closed the connection]
leitao has joined #asahi-dev
leitao has quit []
nyilas has joined #asahi-dev
nsklaus has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
leitao has joined #asahi-dev
cuso4 has joined #asahi-dev
nsklaus has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rory_be has joined #asahi-dev
chadmed_ has joined #asahi-dev
chadmed has quit [Read error: Connection reset by peer]
<sven>
I should also get back to TB and clean that up now that $work has calmed down a bit
<marcan>
:)
<marcan>
tagged -asahi12, just that minor DCP change and some miscellanea from lina
<marcan>
also it's probably a good time to send off mailbox after a smoke test here, enough people have tested it in asahi now
* marcan
grabs the popcorn
<_jannau_>
for dart you have drop the t8112 patches but manual cleanup required after that
<_jannau_>
rebase for gpu is not trivial due to upstreamed rust changes, my first try didn't compile. I haven't tried to clean it up
leitao has joined #asahi-dev
rory_be has joined #asahi-dev
darkapex1 has quit []
darkapex has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
<kettenis>
sven: you don't use the subpage protection to avoid exposing more than 4k?
<marcan>
I think that isn't hooked up yet
<sven>
I wanted to do that but never got around because that also requires changes to the iommu api
<kettenis>
there are some gotcha's with exposing really small regions, but 4k regions should be fine
<sven>
I think I did a quick hack and the WiFi chip didn’t like just exposing very small (8 bytes or so) regions
<kettenis>
8-byte alignment is ok, smaller than that gets you into trouble with the onboard NIC
<kettenis>
(the NIC probably always does 64-bit aligned memory access, which is consistent with its mips heritage)
<marcan>
_jannau_: didn't run into anything significant for gpu, I just dropped all the upstreamed commits and replaced the arm64 series with what was on the ML + a conflict fix and it worked
<marcan>
should I merge the 13.2 DCP changes?
<_jannau_>
I think I have newer 13.2 dcp changes so please don't merge them from that branch
<marcan>
ok
<marcan>
smoke testing on t314 now
<marcan>
seems to work, I'll do a push
<marcan>
github seems to be hurting...
<marcan>
maybe this is a good time to move to fdo...
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
leitao has joined #asahi-dev
rory_be has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
rory_be has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
raveling has quit [Ping timeout: 480 seconds]
rory_be has joined #asahi-dev
<marcan>
ok, applied the pending soc stuff (I think)... now let's see about mailbox.
rory_be has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rory_be has joined #asahi-dev
salimterryli has quit [Remote host closed the connection]
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
salimterryli has joined #asahi-dev
rory_be has joined #asahi-dev
salimterryli has quit [Remote host closed the connection]
raveling has quit [Ping timeout: 480 seconds]
eiln has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
<eiln>
is there an easy way to get M2 device trees?
<eiln>
they're not in preboot and i dont want to install ventura :(
<ChaosPrincess>
download ipsw and grab from there
<jannau>
eiln: in the installer source: python src/urlcache.py $IPSW
<eiln>
thank you!
<jannau>
without arguments it will print the files. list files you want to extract as further arguments
<marcan>
sven, jannau: want to review the wording for the de-mailboxing before I send it? https://mrcn.st/p/ecCKPUy9
eiln has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
<_jannau_>
marcan: maybe replace "just isn't particularly high quality" with "made design decisions unfit for our use case"
<marcan>
done
<_jannau_>
you could add binary size comparisson. mailbox with only apple-mailbox vs. soc/apple-mailbox
<_jannau_>
might look more favorable than the source code size reduction
<marcan>
heh, sure
rory_be has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
pjakobsson_ has joined #asahi-dev
pjakobsson has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
<marcan>
_jannau_: meh, there's some alignment thing and .text doesn't change size (.data and .rodata do, but not by too much)
<marcan>
not worth spending more time on
<marcan>
I think something might be padding to 64K align
<j`ey_>
#define SEGMENT_ALIGN SZ_64K
<marcan>
eh, whatever :p
<marcan>
I'll send it
<_jannau_>
you could compare the size of the object files, not worth to spend much time on though
<ChaosPrincess>
maybe mention that gpu goes faster with your mailbox
raveling has quit [Ping timeout: 480 seconds]
rory_be has joined #asahi-dev
cuso4 has left #asahi-dev [#asahi-dev]
<marcan>
ChaosPrincess: it doesn't :p (not yet at least)
<marcan>
I could barely get a 1% improvement with regular benchmarks, mostly within margin of error
<marcan>
in sync mode it was actually worse, probably because the extra CPU busywork ended up helping with the scheduler or something
<marcan>
there's too much cpu/gpu interaction subtlety right now
<marcan>
but at least when we control the driver we can actually profile and optimize it without a pile of common code getting in the way
* marcan
hits b4 send
raveling has joined #asahi-dev
rory_be has quit [Ping timeout: 480 seconds]
<marcan>
(also first time I send a series with b4)
<marcan>
between this and my DIY patch tracker and maybe moving soc to fdo/gitlab, maybe I can make this whole kernel thing less painful enough to tolerate :p
<ChaosPrincess>
b4 is great, removes a lot of guessing 'how will i fuck up sending patches this time' :P
<j`ey_>
looks like it went through ok marcan
<marcan>
I like how it picked up my gpg key automatically, although it was a bit surprising (why is a dry-run send asking me to touch my yubikey?!)
<ChaosPrincess>
now, if people would actually respond to them in reasonable time, but thats not something you can solve with tech
<marcan>
ChaosPrincess: thankfully for this one we don't have to *actually* wait for Jassi, if he drags his feet on it we can just merge everything up until the removal ourselves :p
<j`ey_>
the only issue is the config change then
<marcan>
yeah, but that's a minor evil if it comes to that
<eiln>
that they're comparing a ulong to int ret in find_and_map_user_pages()
<eiln>
but i *really* dont want to join a @quicinc.com thread
nsklaus has joined #asahi-dev
fortich has joined #asahi-dev
<maz>
why not? we all interact with "competitors on the list.
rory_be has quit [Ping timeout: 480 seconds]
<eiln>
maz: ill send a patch if it gets merged.
raveling has quit [Ping timeout: 480 seconds]
<maz>
eiln: it'd be a lot easier to get it fixed before... :-/ doing it after the facts is a real pain.
rory_be has joined #asahi-dev
<eiln>
maz: i feel the opposite. the qaic patch is at v5(!) now, and i have no idea of its review status. only qualcomm would know if v5 gets scraped again.
raveling has joined #asahi-dev
<maz>
eiln: v5 is nothing. I have series that are up to v18 in my Inbox. getting things merged with blatant bugs usually end up in a disaster, specially with QC, as the crap immediately propagates in the Android trees...
<maz>
I'm sorry, but I really don't get the rationale.
leitao has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
<eiln>
maz: of course some unaffiliated normie would feel out of place in disturbing a massively corpo & proprietary operation (Cloud AI chip btw, WTF?)
<eiln>
this is less about email, but it sure doesn't help
<eiln>
like i said, i've been eyeing the thread, and once reviewed/staged/stable i'll send a patch
<ChaosPrincess>
i can bully some corporations on your behalf if you want it :P
raveling has quit [Ping timeout: 480 seconds]
<maz>
I still don't understand the corp vs unaffiliated statement. a bug is a bug, and the sooner it gets squashed, the better. anyway, I'll go back to my cave, I'm clearly missing some basics here.
<eiln>
e.g. intel vpu had to scrape userptrs in like v2; if qaic also does the patch becomes meaningless
<eiln>
maz: if a patch holds that much importance the focus should be on simplifying the process
raveling has joined #asahi-dev
<maz>
eiln: we'll have to agree to disagree here.
<marcan>
maz: there is a much bigger mental and practical barrier to email/patch interactions than most seasoned kernel devs acknowledge
<marcan>
I mean FFS I just had to make myself a boutique patch tracker for my own submissions/reviews because I just couldn't cope with email
<maz>
marcan: clearly, I'm not seasoned enough. and replying to an email to say "this is broken" isn't that hard.
<maz>
but hey, like I said, I don't understand most things.
<marcan>
it isn't, but replying to a merge request is even easier and has an edit button
<psykose>
there are plenty of email issues but i wouldn't say "i will send the bug report after it's merged and not before it's merged" is very related to that
<psykose>
and i refuse to use git-send-email on a good day :p
bcrumb has joined #asahi-dev
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
leitao has joined #asahi-dev
leitao has quit [Ping timeout: 480 seconds]
bcrumb has quit [Quit: WeeChat 3.8]
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
raveling has joined #asahi-dev
leitao has joined #asahi-dev
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
raveling has quit [Ping timeout: 480 seconds]
leitao has joined #asahi-dev
raveling has joined #asahi-dev
Dcow has joined #asahi-dev
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
leitao has joined #asahi-dev
alyssa has joined #asahi-dev
<alyssa>
my #1 bugbear with email patch review is that I'm apparently not allowed to resend patches after i fix things people report
<alyssa>
which means people are expected to review stale code
<alyssa>
because otherwise resending will annoy people by generaitng more email
<alyssa>
gitlab by contrast, the moment i fix something reported i repush the MR and reviewers always see the working code
<alyssa>
which means instead of waiting a week for every new version of the series, it can be minutes in between a reviewer sending in feedback and it getting fixed
Urriellu has joined #asahi-dev
<alyssa>
and instead of taking a month to get through 4 versions before merged, it can be (and often is) just a few days
raveling has quit [Ping timeout: 480 seconds]
<tobhe_>
is that a linux ml rule?
<alyssa>
etiquette/courtesy, but yes
<ChaosPrincess>
since we are talking ml, can someone explain to me what statuses on patchwork mean
<alyssa>
ChaosPrincess: no
<ChaosPrincess>
like, a patch is in status 'changes requested', except there were zero responses to series
<_jannau_>
my crystal ball answers: maybe unaddressed comments from a previous series but I'd vote for alyssa's answer
raveling has joined #asahi-dev
leitao has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
leitao has joined #asahi-dev
Urriellu has quit [Remote host closed the connection]
leitao has quit []
Dcow has quit [Quit: dcow]
salimterryli has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
Dcow has joined #asahi-dev
raveling has joined #asahi-dev
leitao has joined #asahi-dev
leitao has quit []
bcrumb has joined #asahi-dev
raveling has quit [Ping timeout: 480 seconds]
leitao has joined #asahi-dev
<marcan>
see my solution has emojis, that's how you know it's better than patchwork https://mrcn.st/t/fboard.png
raveling has joined #asahi-dev
<alyssa>
marcan: -1, needs lina sticker
<marcan>
lina sticker... where?
<alyssa>
idk
<alyssa>
everything is better with lina stickers
<alyssa>
:p
<marcan>
fair.
<marcan>
(tbf this is an off-the-shelf thing I just plugged a mailing list scraper into, I'm not planning on hacking up the UI :p)
<marcan>
it's... not exactly the most solid piece of software, it has some quirks, but it does the job
seeeath has joined #asahi-dev
<marcan>
I'm trying to use it both for general project/task management, and as a patch tracker
<marcan>
(and it's easy to deploy which is important, and has a somewhat-questionably-documented-but-functional API I could fire requests at from a python script running as a Postfix transport from a redirect and yes this is all very rube goldberg)