#
…

2026-09-26

00:00hello0092 has quit [Quit: Konversation terminated!]
00:03Mary has quit [Quit: .]
00:03Mary has joined #asahi-dev
00:05maettu102 has quit [Quit: brb]
00:06maettu102 has joined #asahi-dev
00:09hello0092 has joined #asahi-dev
00:20hello0092 has quit [Quit: Konversation terminated!]
00:43jules[m]123 has joined #asahi-dev
01:34saiprasad_ has quit [Remote host closed the connection]
01:35saiprasad_ has joined #asahi-dev
01:35XeR has quit [Remote host closed the connection]
01:49XeR has joined #asahi-dev
01:49XeR has quit [Remote host closed the connection]
01:50XeR has joined #asahi-dev
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:46tobhe_ has joined #asahi-dev
02:53tobhe has quit [Ping timeout: 480 seconds]
02:58ohtwoone has quit [Quit: Cya]
02:59ohtwoone has joined #asahi-dev
03:06kzzzl has quit [Remote host closed the connection]
03:09kzzzl has joined #asahi-dev
03:12ohtwoone has quit [Quit: Cya]
03:12ohtwoone has joined #asahi-dev
03:52nora_ has joined #asahi-dev
03:54nora has quit [Ping timeout: 480 seconds]
04:10pb17 has quit [Ping timeout: 480 seconds]
04:30pb17 has joined #asahi-dev
04:37saiprasad__ has joined #asahi-dev
04:38saiprasad_ has quit [Remote host closed the connection]
05:50hello0092 has joined #asahi-dev
06:37svenso let’s where that pac issue comes from now
06:38sobek[m]Best luck !
06:38sobek[m](I wonder if something like the pacman transient execution attack might help).
06:39svenmight be because SPTM and xnu both share el1 now and I’m not tracking that state correctly
06:40svenno, this environment is much too cursed for any normal test suite at this point
07:00weirdtreething has quit [Remote host closed the connection]
07:00weirdtreething has joined #asahi-dev
07:15N_Tempus_04 has joined #asahi-dev
07:26saiprasad__ has quit [Remote host closed the connection]
07:31saiprasad__ has joined #asahi-dev
07:47jannaum5 max has chip-id 0x6050. would be interesting what m5 ultra uses. unfortunately ioreg doesn't print chip-id
07:48jannauI guess we can use t6050.dtsi and t6050s.dtsi for M5 Pro and t6050c.dtsi for M5 Max
07:49jannauM5 Ultra can use then either t6050d.dtsi or t6052.dtsi
07:54nickchanjannau: ioreg does print chip-id
07:54nickchanioreg -p IODeviceTree -c0 -l | grep '"chip-id" = '
08:01jannauit does, I guess I'll check later in the Apple store
08:03svenpicking up that m6 or are you extending your collection of Macs even further? ;)
08:03JamesCalligeros[m]do they let you run ioreg on the demo boxes :P
08:04chaos_princessioreg does not need sudo, and they don't monitor them super closely
08:05nickchanplus ioreg is preinstalled software you are merely trying out the Mac experience
08:05myria has quit [Remote host closed the connection]
08:05JamesCalligeros[m]"man this terminal can fit so much devicetree in it wow"
08:06myria has joined #asahi-dev
08:18jannauyes, picking the m6 mini up. not sure what apple store people should have against looking at ioreg output. not that would contain any secret information
08:19jannauexcept maybe for serial numbers but I'm not sure what I can do with the serial numbers of a random store unit
08:27saiprasad_ has joined #asahi-dev
08:28saiprasad__ has quit [Remote host closed the connection]
08:35svenyuup, it was a pac state issue with the sctrl enable bits between sptm and xnu
08:36svennext up: looks like irq delivery is sometimes broken :/
08:39svenit gets fairly far into userspace now but i still dont see anything on my display :(
08:39svennot even the progress bar which is weird
08:41svenit doesnt survive long enough to bring up tb unfortunately
08:41svenso dunno if its just the usb capture card that broke down
08:49CaptainIRS3164 has quit []
08:49CaptainIRS3164 has joined #asahi-dev
09:09JamesCalligeros[m]..not even the rainbow apple logo?
09:15svenalso no m1n1 framebuffer
09:16sveni could try connecting it to my tv but that's in another room
09:17sveni think the irq issue is that an irq that arrived while the cpu is somewhere in SPTM messes something up
09:27chaos_princessare interrupts supposed to kick you out of gl?
09:27sveni'm not sure
09:27sventrying to figure that out on my mini now, but i thought they just get delivered to gl
09:28sven*m1 mini
09:28sventhis worked on sptm for m1 so maybe something has changed on m4. or maybe i was just lucky there
09:48svenstill nothing on the display though
09:50chaos_princessthats pretty far into macos boot
09:51svenyup, which is why i'm confused that i see nothing on the display
09:51svenusually it dies much much earlier, i added more logging and that made it survive longer
10:20devdrago has quit [Quit: WeeChat 4.8.1]
10:20devdrago has joined #asahi-dev
10:36chadmedlets go macaudio v2 works on every class of device
10:41svennice!!
10:45svenit looks like some curse has stricken again. now secondaries sometimes just don’t come up :/
10:45WindowPain has quit [Ping timeout: 480 seconds]
10:46chadmedwhat the hell have they done :/
10:47WindowPain has joined #asahi-dev
10:49svenNot sure if this is an Apple issue or yet another cursed bug in our secondary startup code
10:55N_Tempus_04 has quit [Ping timeout: 480 seconds]
11:17bennofs3 has quit [Ping timeout: 480 seconds]
11:17bennofs3 has joined #asahi-dev
11:19jannaujoke is on me, none of the two apple stores had a m5 ultra on display. 27km cycling, wouldn't have been as bad had I thought to avoid the area around the brandenburg gate
11:19WindowPain8 has joined #asahi-dev
11:20jannaualready full of runners for berlin marathon adjacent events
11:24WindowPain has quit [Ping timeout: 480 seconds]
11:24WindowPain8 is now known as WindowPain
11:26mmediouni[m]as far as I heard a lot of M5 Ultra buyers are waiting for the 512GB model...
11:35jannaumac mini 2.5 gbps ethernet is Realtek 8125
11:38mmediouni[m]Is 10GbE also using realtek or did they stick with Marvell AQC
11:42jannauthe template devicetree still has aqc references so I suppose they kept it. Their ILO appears to be a integrated aqc solution
11:50jannauhow does the min macos storage in the installer work? Do we need to adjust it for macOS 27.0? the m6 uses almost 60GB after its initial update
11:52mmediouni[m]between the newly staged OS and the space needed for Apple Intelligence (tm) quite a bit of space is needed for updates
11:53jannauerr, jumps between 58 and 63 GB, not sure what it does? maybe refreshes apple intelligence model data in the background?
11:53chaos_princessshould it not delete the models if you have it off and are getting close to end of disk space?
11:54chaos_princessand yes, they are re-downloaded, even if you delete the asset from recovery and have siri off
11:54mmediouni[m]> The Storage tab of the System Settings app says that the M1 MacBook Air and MacBook Neo needed around 14GB of space for Apple Intelligence after a fresh install. On the M3 MacBook Air, that number was 22.42GB.
11:54mmediouni[m]~22.5GB for the machines with 16GB of RAM or more on M3 and later
11:55jannau23.5G of apple intelligence data. Where do I charge Apple for that? multiply by 2-3 installtions per device
11:56svenok, so I think any interrupt while the cpu is in Gl ends up in GL als well and SPTM then forwards them to xnu because it doesn’t want to handle them and I think that’s what breaks
11:56chadmedsend the invoice directly to john apple
11:56chadmed"AI has made everything 500% more expensive so we're cutting back on our ram and storage also you have to dedicate most of whats left to AI"
11:57chaos_princesssven: so, would it be a stupid move to just mask interrupts when in gl
11:57mmediouni[m]that's what I did for a cursed project
11:57mmediouni[m](implementing GXF on top of FGT using non-private isa VMs...)
11:57sventhat sounds like a good test 😁
11:57svenlet’s see what happens
11:59mmediouni[m]jannau: I was told that if you disable all Apple Intelligence toggles through Screen Time and then delete from recovery it doesn't download them again
11:59chadmedit seems unreasonable to expect users to do all this
11:59mmediouni[m]they want Apple Intelligence to be always-on for everyone yes
11:59svenbut which user wouldn’t want AI? that’s just a bunch of weirdos obviously!
12:00mmediouni[m]the only disable toggle is in screen time...
12:00chadmedthen again if i were a PM at apple i would say it seems unreasonable for the OS to eat 60 GB of storage so that you can make siri into a dommy mummy but whatever what do i know my company isnt worth 6 trillion dollars
12:01chaos_princessmmediouni[m]: failed for me, i have classic siri, siri and system language desynced, and siri blocked via mdm, and it still re-downloaded it
12:03mmediouni[m]it could have been even bigger, at least Image Playground got moved to "cloud"/PCC only
12:03svenseems stable with irqs in gl disabled
12:03svenor maybe this boot was lucky again
12:03svenstill nothing on the display though :(
12:03chadmedmmediouni[m]: people are complaining about that because its no longer device-local, cant win :P
12:04* svenreally wants that picture with „Apple M4 Pro on cursed m1n1 hv“
12:07JamesCalligeros[m]Try tracing dcp?
12:07jannauanything dcp related in the macos log? macos may timeout external display setup if the HV is too slow, seen with dcp tracing specifically
12:07JamesCalligeros[m]All the callbacks will be wrong/unknown of course but see if it at least tries to do something
12:10svengonna grab lunch since I’ve been at this since 7am or so and then try that with hdmi attached again
12:10sventb did come up but it looks like the dptx capability status thingy timed out
12:40chadmedsome of the macaudio patches were rather superfluous as it turns out. not sure why we needed to do the flakey wildcard matching thing when strstr exists
12:41chadmedhopefully a few of the other things mark has nacked previously also turn out to be unnecessary :P
12:42kevans91 has quit [Ping timeout: 480 seconds]
12:44nickchandelete model and set the simmutable flag from recovery
12:50N_Tempus_04 has joined #asahi-dev
13:09jannaulol, the m6 mac mini froze during resizeContainer. I initially assumed it wasn't making any more progress at 15%
13:20sofusjust sendt v2 :) Its actually ~3 seconds faster on all fluster tests
13:24hello00921 has joined #asahi-dev
13:28hello0092 has quit [Ping timeout: 480 seconds]
13:35hello00921 has quit [Remote host closed the connection]
13:39hello00921 has joined #asahi-dev
13:39mattia013 has quit [Quit: mattia013]
13:39mattia013 has joined #asahi-dev
13:55saiprasad_ has quit [Quit: Leaving]
14:20AlLyu has joined #asahi-dev
14:24AlLyu has quit [Remote host closed the connection]
14:25jannauthank FSM, the M6 has again "parent_addr = 0x200000000"
14:37nickchanso what is fsm again
14:38jannauflying spaghetti monster. I almost written a 3 letter word
14:40alanlyu has joined #asahi-dev
14:41alanlyu has quit []
14:42alanlyu has joined #asahi-dev
14:42alanlyu has quit []
14:42alanlyu has joined #asahi-dev
14:42alanlyu has quit []
14:42alanlyu has joined #asahi-dev
14:43alanlyu has quit []
14:43alanlyu has joined #asahi-dev
14:47alanlyu has quit [Remote host closed the connection]
14:49sven*sigh* so I think there’s another weird bug in our smp code, I can occasionally hit it even without the hv on this m4 pro
14:51alanlyu has joined #asahi-dev
14:52alanlyu has quit [Remote host closed the connection]
14:52alanlyu has joined #asahi-dev
14:52alanlyu has quit [Remote host closed the connection]
14:56truita has quit [Remote host closed the connection]
14:58truita has joined #asahi-dev
15:01hardy94 has joined #asahi-dev
15:14truita has quit [Remote host closed the connection]
15:15truita has joined #asahi-dev
15:17jcbm has quit [Remote host closed the connection]
15:18jcbm has joined #asahi-dev
15:24jannauyuka: was there a known offset for the macos wireshark captured address and the actual base?
15:25jannaufor debugusb
15:25yukalook at the kisd code?
15:25jannauI already forget everything
15:25yukaAh wait
15:26jannauI see 0x34980000 in wireshark but kisd guesses 0x348000000 after I added bcd 5.00
15:26yukaLike... there were several possible offsets
15:26jannauit seems to work
15:26jannau0x0 works
15:26yukawhat you actually want to capture with wireshark is one of the keypress messages
15:29hardy94 has quit [Remote host closed the connection]
15:29jannauah, I thought the first large package on connect was enough
15:33jannauno m1n1 output on M6, dockchannel uart stops with "End of iBootStage2 serial output." and samsung uart is completely silent
15:35jannaum1n1 doesn't appear to be crashing but a watchdog appears to be active. after a few minutes it restarts to recovery
15:40aditya has joined #asahi-dev
15:44pb17 has quit [Ping timeout: 480 seconds]
15:49pb17 has joined #asahi-dev
15:54jannauuart is my fault. I looked at the wrong adt and ended up with the wrong address
15:56jannauuart works
16:00svenalright, so dcpext seems to be broken but the capture card also doesn't show anything on my m1 mini
16:00sveni guess that thing just broke :(
16:04jannauthe M6 boot core has both MIDR_CORE_TYPE_P and MIDR_CORE_TYPE_M set. MIDR is 0x611f0731
16:08nickchandespite how it's named in m1n1 MIDR_CORE_TYPE_P and MIDR_CORE_TYPE_M should be checked against MPIDR
16:09nickchanand is actually how m1n1 check if a core is ecore
16:13jannauI looked at the code and missed it
16:14nano has joined #asahi-dev
16:16jannaucomes up on core 6, S core
16:16jannauor P or whatever
16:17N_Tempus_04 has quit [Ping timeout: 480 seconds]
16:22jannaum1n1 dockchannel uart works now as well
16:35jannauunkown, I need to fix pmgr first
16:37chaos_princessdw, just got too excited
16:37jannaudart,gen3 is different. SErrors when I try to handle it as t8110 dart. At least the first 4 param regs look mostly the same
16:38svenhm... nope, no output on my tv screen either
16:39svenit's weird that neither hdmi nor dp-via-tbt nor dp-altmode works
16:44jannauI suppose that is all using dptx. dcp does dptx on M1 mini's HDMI port
16:48svenhuh, i don't see a dcp in the adt
16:48svenjust 4x dcpext
16:49sventhere's dcp0 = /arm-io/dcp0 in aliases but i don't see that node
16:55svenoh... is this because i'm stealing that one atc from the adt for the hv and something changed on m4?
16:55jannauwhy does it have 4 dcpexts if it only supports 3 simultaneous displays?
16:56jannauapple already omits dcp0 from the m2 ultra as they have enough dcpexts there
16:56svenit doesn't :D
16:56sventhere's just an alias for dcpext3 that points to an invalid node
16:57sventhe atc i stole is still in dfp-endpoints of the crossbar, let's see if removing it helps
16:57sventhough i don't see why that would make hdmi fail
16:58jannauI suspect dcp0 is a less powerful display controller only good enough for the 24-inch imac display and not supporting 6k resolution
16:59sventhat's under the hv, i had to drop the atc from that dfp-endpoints
16:59svenbut my usual password doesn't work
17:00jannauwould make sense if that makes dcp0 more power-efficient than dcpext for notebooks
17:01svennow it's up to 5 minutes :(
17:01jannauwhy would the password not work? subtly breaking sep?
17:01svenno, i probably used something unusual for some dumb reason and forgot
17:02sventhis is clearly PEBKAC
17:02svensep is up just fine from what i can tell
17:04svenyeah, that password doesn't work in recoveryos either
17:22jannau"fixing" pmgr is trivial: replace "/arm-io/pmgr" with "/arm-io/pmgr-child". why apple?!
17:24jannausmp_start_secondaries() works
17:26chaos_princessSo, we are now in a stupid situation where it works on m6, but not m5pro/max?
17:26nickchanit probably doesn't work on m5ultra either though that's not confirmed
17:27chaos_princessyea, all the t605x
17:47tobhe has joined #asahi-dev
17:49aditya has quit [Quit: Connection closed for inactivity]
17:52tobhe_ has quit [Ping timeout: 480 seconds]
17:56hardy94 has joined #asahi-dev
18:04svenwell fuck... recovery ofc installed another macos version and this still relies on binary patches to SPTM to remove some checks
18:04user890104_ has joined #asahi-dev
18:08chaos_princesssequential install trick is annoying but should work, right?
18:09svenyeah, i guess i'll try that tomorrow
18:11user890104 has quit [Ping timeout: 480 seconds]
18:14svenlol, old SPTM boots with the slightly newer kernelcache it seems :D
18:15svenwell.. i should've have celebrated already panic(cpu 9 caller 0xfffffe003a12ae60): "ApplePMGR: void ApplePMGR::_waitForXNUClusterPowerGatingThreadCall():11868 " "Deadline: 0xf796445f, AbsTime: 0xf7b91b9b XNU ACC cluster rail-gating callback failed to respond in 30 seconds" @ApplePMGR.cpp:11868
18:17sven*shouldn't
18:19svenhttps://f.svpe.de/b2947e94340fd9094450e37430424e7ee38c1a37a56cfc906154056d8b69162c_yay.jpeg well.. very unstable, i see those PMGR panics and just got another ACC one
18:19svenbut that's good enough for today
18:19chaos_princessis this w/ mismatched sptm and xnu?
18:19chaos_princessstill, hella impressive progress
18:20sveni'll do the double reinstall tomorrow
18:20chaos_princesswhat was your previous?
18:20sventhis is probably useable for tracing random hardware but it still dies very often
18:20svenone RC earlier
18:20jmax[m]woohoo! very cool! the cursed m1n1 on m4 screenshot seals the deal!
18:21chaos_princessso, another 26.7, just different build? i wouldn't expect them to change sptm abi that often
18:35jannauM6 P and M cores have the same CPU part number
18:40jannauidentical MIDR
19:29amarioguy has joined #asahi-dev
19:30amarioguyjannau: what is "parent_addr" here in this case?
19:35hardy94 has quit [Quit: hardy94]
19:38Lyude has quit [Quit: Bouncer restarting]
19:39jannauamarioguy: parent_addr for MMIO? "bus_addr = 0x0000000000000000 parent_addr = 0x0000000200000000 size = 0x0000000400000000"
19:39Lyude has joined #asahi-dev
19:39amarioguyjannau: ah IO base address, figured it was that but wanted to be sure
20:06hardy94 has joined #asahi-dev
20:07commandoline has quit [Quit: Bye!]
20:09hardy94 has quit []
20:10commandoline has joined #asahi-dev
20:49chaos_princess has quit [Quit: chaos_princess]
20:49chaos_princess has joined #asahi-dev
20:56amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
21:25amarioguy has joined #asahi-dev
21:34bennofs3 has quit [Ping timeout: 480 seconds]
21:38bennofs3 has joined #asahi-dev
22:43zkrx has quit []
22:44hello00921 has quit [Quit: Konversation terminated!]
22:45zkrx has joined #asahi-dev
23:02amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
23:03jules has joined #asahi-dev
23:08jules is now known as julesmeow
23:10julesmeow is now known as jules
23:10jules is now known as julesmeow
23:11julesmeow is now known as mrrpmeow
23:45distantorigin has quit [Ping timeout: 480 seconds]
23:51amarioguy has joined #asahi-dev
…