#
…

2026-09-25

00:05distantorigin has quit [Ping timeout: 480 seconds]
00:34pb17 has quit [Ping timeout: 480 seconds]
00:34pb17 has joined #asahi-dev
00:38jn has joined #asahi-dev
00:42jnn has quit [Ping timeout: 480 seconds]
01:37bennofs3 has quit [Ping timeout: 480 seconds]
01:39bennofs3 has joined #asahi-dev
01:45distantorigin has joined #asahi-dev
01:46distantorigin has quit [Remote host closed the connection]
01:46distantorigin has joined #asahi-dev
01:48distantorigin has quit []
01:49distantorigin 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]
03:02Guest368 has quit [Remote host closed the connection]
03:53nora has joined #asahi-dev
03:55nora_ has quit [Ping timeout: 480 seconds]
04:15saiprasad_ has joined #asahi-dev
04:28saiprasad_ has quit [Remote host closed the connection]
04:28saiprasad_ has joined #asahi-dev
04:46bennofs3 has quit [Quit: Ping timeout (120 seconds)]
04:46bennofs3 has joined #asahi-dev
05:34integralpilot has quit [Quit: integralpilot]
05:52integralpilot has joined #asahi-dev
05:52integralpilotbtw! when i was running dchid under lockdep to see if there were issues i found smth else in a different part
05:53integralpilotthe gpio driver was giving these lockdep warnings, but its actually the default of regmap, it was allocating memory while holding the GPIO raw spinlock
05:54integralpilotit looks like the lock actually isnt needed during creation/destruction (where the issue was) cause its not available yet or ppl must have stopped by then
05:55integralpilotand not during cache reinit either b/c the API asks caller to prevent concurrent access
05:55integralpilotso just removing it fixes all the lockdep warnings
05:55integralpilotas a seperate patch would it be OK to send https://pastebin.com/9b2fKwzV upstream or have i misunderstood smth here
05:58integralpilot has quit [Quit: Lost terminal]
06:03HexDroidUser has joined #asahi-dev
06:05integralpilot has joined #asahi-dev
06:53lucy_ has joined #asahi-dev
06:55lucy_ has quit []
06:57HexDroidUser has quit [Ping timeout: 480 seconds]
07:01hardy94 has joined #asahi-dev
07:03N_Tempus_04 has joined #asahi-dev
07:38Peiran has quit [Read error: Connection reset by peer]
07:38Peiran has joined #asahi-dev
07:38pb17 has quit [Ping timeout: 480 seconds]
07:49N_Tempus_04 has quit [Ping timeout: 480 seconds]
07:53pb17 has joined #asahi-dev
07:55MichaelReeves[m] has joined #asahi-dev
08:07myriaWait, all the M5 chips are T6050? Wikipedia lied to me (crazy, right?). Should I change https://github.com/AsahiLinux/docs/pull/278 to reflect this?
08:10yukaM5 base is T8142
08:11yukaM5 Pro/Max/Ultra are probably all T6050 but it's a bit unclear at this point
08:12chaos_princessjannau: as you asked, reminder about the r4l thing
08:12svenalright, now that the watchdog thing is applied i can pick up the m4 series this evening
08:13svenchaos_princess: i think he got the day wrong, he wanted to speak to the r4l guys on wednesday. but given that he disappeared then for a few hours i think that's what he did at kernel recipes :D
08:16yukasven: you mean A18 Pro?
08:16svener, yes
08:16yukanice nice
08:16sven"fake m4" :P
08:16yukaI'm looking into the pinctrl 'protected' pins
08:17yukaon A18 Pro
08:17yukatrying to determine if it really needs a new device tree property but I think yes
08:18yuka(background: on the MacBook Neo you get SErrors trying to read the REG_GPIOx of a specific pinctrl_ap pin
08:18yuka* of a specific pinctrl_ap gpio
08:18yukawhich happens to be listed in the adt under "no-restore-gpio-pins"
08:18chaos_princessno-restore-gpio-pins exists on previous socs too, no?
08:18yukabut we ignored it
08:19yukaOn this device it seems to actually matter
08:19yukaso my solution for that is adding .readable_reg to the regmap_config, but the question is where to source the list of gpios not to read
08:20sveneither device tree property or table based on compatible in the driver
08:20yukaI think device tree property is nicer
08:20svenit's also describing hardware so should be fine i think
08:20chaos_princessi would be against a compatible, gpio numbers sounds like a thing that change based on what device the soc is in, not just the soc itself
08:20yukaOr I also considered if it would be possible to get a list of all gpios the which _are actually used_ and then ignoring all others
08:21yukabut doesn't seem easy to do that
08:21chaos_princesshow would that work with opening /sys/class/gpio or whatever that is called
08:21yukayeah probably doesn't work at all, so device tree property it is
08:25integralpilotshould i CC the asahi list for the regmap fix or not really bc its not directly asahi related
08:26jannauI got the day wrong but I remembered. unfortunately Miguel was sick so I did not ask him
08:29aissen has quit [Quit: brb]
08:32svenintegralpilot: no need to cc it, and no idea if that fix is correct. i've seen the lockdep splat before and just ignored it :D
08:34integralpilotthanks! i guess only one way to find out if its right, will send soon
08:34integralpilotim pretty sure it is but i might be misunderstanding
08:35integralpilotbusy night tonight cause there's 3 patchsets to send out, then on my other laptop im constantly reloading the collegeboard website cause scores are coming out tonight too
08:35aissen has joined #asahi-dev
08:37saiprasad__ has joined #asahi-dev
08:38saiprasad_ has quit [Remote host closed the connection]
08:38integralpilotcollegeboard (us)
08:38integralpilotim just keeping my options open so took it earlier this month
08:39integralpilotthankfully atar not coming out until next year!
08:46N_Tempus_04 has joined #asahi-dev
09:12svenin other news, made some progress on the m4 hv on the train. sptm checks a lot of crap now on m4:
09:12sven# [cpu4] SPTM panic at 0xfffffe001900d620 (SPTM-rx+0x5d620 = 0xfffffff0270fd620): format=0xfffffe0018f17e34 args=0xfffffe001903fe80
09:12sven# [cpu4] ctXr_check_all_regions: CTRR-A per-core register not locked by iBoot (expected: 8000000000000001, got: 0)
09:15yuka(the progress)
09:49midou has quit [Ping timeout: 480 seconds]
10:16integralpilotthank you jannau for sending nvme :)
10:18svenwill have to do that properly but i like this progress
10:23JamesCalligeros[m]oh neat it started turning on coprocessors nice
10:24svenyeah, it's also done a full trip through all the guarded levels at this point
10:24integralpilotlooks like issue happens when talking to SMC?
10:24integralpilotso cool it's gotten this far!
10:24svenso that's the issue now: panic(cpu 4 caller 0xfffffe003d541ea8): Kernel data abort. at pc 0xfffffe003d390098, lr 0xfffffe001854446c (saved state: 0xfffffe4eaf65b240)
10:24svencould be guarded emulation going wrong
10:26svenor maybe some of those SPTM checks actually checked something important
10:26sveni didn't really like being loaded at ram_base + >2GiB
10:27svenmaybe i just need to steal the memory for the shadow pagetables from the end of RAM
10:28ohtwoone has quit [Quit: Cya]
10:43ohtwoone has joined #asahi-dev
10:44integralpilotOK so regmap is sent; dchid almost ready just double checking; am writing wifi cover letter; SAT not yet out
10:45integralpilot(i did end up CC-ing asahi list because i wanted a lore.kernel.org/asahi/... link for the spreadsheet)
10:52svenurgh.. i think this might be irq delivery going wrong when i'm inside an emulated guarded level :/
10:53yukaI have the pinctrl protected-gpios thingy working
10:54yukathough it gives me a warning when compiling the device tree
10:54yuka"apple,protected-gpios = <134>;"
10:54yuka"Warning (gpios_property): /soc/pinctrl@31a000000:apple,protected-gpios: cell 0 is not a phandle reference"
10:54svenhow did you add that property to the dt-binding? it should be a uint32 array or something like that
10:55yukaI think the reason is just that the property contains the string "gpio"
10:55yukaif I rename it to protected-pins, the warning disappears
10:55yukafwiw this is not even dtbs_check, just normal make dtbs
10:55yukaSo I guess it's protected_pins then? or would that name be inaccurate
10:57nimprod3l has joined #asahi-dev
10:58integralpilotARGHH sashiko caught something right about regmap
10:58integralpilotcause its only 1 tiny patch how long do i have to wait before v2
11:02myriaUhhhh, idk? I'd at least give it a day or two
11:02svenyuka: works as well i think
11:03svenintegralpilot: i like to wait at least 24h-48h because the maintainer might comment on something else especially if it's a tiny patch that can be reviewed quickly, but i've also sent a v2 after an hour before and wasn't screamed at so *shrug*
11:08integralpilotthank you, makes sense :)
11:09jannauyuka: "apple,protected-pins" (no underscore) is fine and and matches "apple,npins". I could bikeshed over "apple,unused-pins", "…,skipped-pins", "…,externally-used-pins"
11:10jannaudid you check if any other pinctrl dt-binding has something similar
11:16yukajannau: it's a chicken and egg problem if I don't know what other dt-bindings would call this
11:16yukaI grepped for .readable_reg and searched for implementations which read them from the device tree and couldn't find any
11:17yukajust looking through pinctrl dt-bindings randomly, I haven't found anything comparable
11:17JamesCalligeros[m]it's probably fine then. if there is a different way maintainers would prefer they can tell you
11:18svenquick grep for reserved shows something like gpio-reserved-ranges
11:18sventhough i dunno what that does
11:22jannaulooking through the other dt-bindings and not seeing something that might be similar is enough
11:23jannauanother possible solution would be to add a property to Documentation/devicetree/bindings/pinctrl/pincfg-node.yaml
11:29yukaI assume this problem is somewhat unique to Apple though
11:30yukausually performing an MMIO _read_ in the middle of the gpios mmio space would not _crash_
11:30chaos_princessi remember some other platform having "trusted" gpios, but there they mediate stuff via el3
11:33jannaufrom Documentation/devicetree/bindings/gpio/gpio.txt "gpio-reserved-ranges" seems to the right property
11:34SigneRGK[m] has joined #asahi-dev
11:35JamesCalligeros[m]oh my god no way that was the issue lmfaoooo
11:36SigneRGK[m] is now known as Spriggy[m]
11:38sofusbut like p010 did not look completly wrong so i kinda wonder if avd could also use the 'padded' formats for interchange
11:39JamesCalligeros[m]if it supports it, maybe
11:39sofusand my mystery format is actually just p030
11:39JamesCalligeros[m]but if the issue was just BOMDAS then whatever
11:42sofusnot only no, avd's interchange uses the new formats i added
11:43svenyuka: you ran into some pmgr mmio write serror on m4 before, didn't you?
11:43sveni think i see the same now with that hypervisor
11:43JamesCalligeros[m]sofus: yeah thats fine, dcp supports them
11:43yukasven: yes
11:43JamesCalligeros[m]i was under the impression for a long time that the p*** fourccs were for fully-planar formats and kept trying to get that to not crash :P
11:44svendo you have a list of pmgr register that are off-limits? or is there some way to identify them from the adt?
11:44yukasven: it had a pattern. ps_disp*_cpu was definitely affected
11:45sofusJamesCalligeros[m]: lol, im not actually sure its nv15/20/30 but the math is mathing. the 'p' could be for packed
11:45yukaI think that was it
11:45midou has joined #asahi-dev
11:49sofusavd v2 is gonna have quite a lot of changes, not sure if i should also include the multi-core stuff
11:51jannauI think P___ are semi-planar high-bit-depth (Y plane and interleaved CbCr plane). I wouldn't read too much into it and just look up how the format is defined
11:56sofusthe definition did not make me a lot smarter. But nv15 gives the correct bpp
12:02N_Tempus_04 has quit [Ping timeout: 480 seconds]
12:05nimprod3l has quit [Quit: Leaving]
12:06jannauit packs 4 consecutive 10-bit values in 5 bytes
12:15integralpilot2 down, 1 to go
12:15yukayippie
12:15yukaI'll test it shortly
12:17JamesCalligeros[m]ill finish v2 of audio tomorrow hopefully, just need to sprinkle some locks and semaphores around the driver now
12:19yukaAh wow apple,protected-pins clashes with the -pins$ patternProperties used for pinmux nodes
12:23integralpilotphew dchid v4 sashiko dropped and its all hypothetical/not right so at least we're starting off with no known bugs until the maintainers see it
12:26integralpiloteh regmap maintainer replied and i might have misunderstood smth and v2 needs to a lot bigger
12:26integralpilotanyway will look into that after brcm4388 sent, its almost ready
12:28myriaWoohoo! Go integralpilot
12:33aissen has quit [Remote host closed the connection]
12:33aissen_ has joined #asahi-dev
12:33sofusjannau: i might have confused myself. Is that nv15 or p420 you are talking about? 4 values in 5 bytes is nv15
12:34jannausofus: I did not look at your tree and replied to "But nv15 gives the correct bpp"
12:36jannauyuka: I think "gpio-reserved-ranges" is at least close enough that you would have to argue why you can't use it
12:37sofusi confused myself then. nv15 gives the correct bpp for the interchange buffer calculations, i dont know if avd can output nv15 withouth interchange
12:37sofusi was unsure if NV15 translates to dcp format P420
12:38saiprasad__ has quit [Remote host closed the connection]
12:38saiprasad__ has joined #asahi-dev
12:41chadmeddoes it output correctly
12:42sofusi dont know how to check. I could only test with interchange
12:46jannauno, 'p420' packs 3 values into 4 bytes with 2bits of padding (or 6 into 8 bytes for CbCr) if we're speaking about "PixelFormat = 1882468912;"
12:47jannaufor nv15 I'd expect YCbCr format with a plane with "BitsPerBlock = 40"
12:50sofusthanks! That makes sense, now i also get how to read it. Thats p030 then
12:56sofuswe cant use 'drm_format_info_bpp' then...
12:58chadmedthats what i thought
12:58chadmed(that it would be p030)
13:03chadmedfrom what i can see the format should describe iomfb's 420p
13:05aissen_ has quit []
13:05aissen has joined #asahi-dev
13:07midou has quit [Ping timeout: 480 seconds]
13:09svenokay, wtf, not even sure these serrors are related to pmgr anymore
13:14yukasven: does L2C_ERR_STS point to a pmgr reg?
13:19midou has joined #asahi-dev
13:22nano has quit [Quit: User left the chat]
13:24N_Tempus_04 has joined #asahi-dev
13:29integralpilotomg no brcm4388 has merge conflicts with wireless-next, it already moved since i prepared it monday
13:29integralpilotlucky i didnt send it yet!
13:29integralpilotbut now ive got to fix them and re-test before i can send
13:29integralpiloti was just about to and thought "hmm... should probably check" so lucky
13:35integralpilot has quit [Quit: integralpilot]
13:41integralpilot has joined #asahi-dev
13:43integralpilotok tried it and everything passes!! so nervous but sending in 3... 2... 1...
13:43integralpiloti pressed send
13:45integralpilotok it's 11:45 PM so i'm going to bed now cause it's been such a big day omg, goodnight everyone and thx for the help :)
13:46integralpilot3 of 3 sent, can't believe it
13:47svenyuka: it's somewhere in PMGR but it's not power-domain
13:50yukamaybe cpu / cluster related?
13:50chaos_princess^, can you catch the value being written?
13:51svenyeah, i'm going to try to just shadow that area now
14:10myriaIs there any PRs I could test? I have an M3 Max I could use for dcp tests?
14:14integralpilot has quit [Quit: integralpilot]
14:14yukaI tested dcp recently on M3 Max and it didn't work. Probably my next step would be comparing against a macOS trace.
14:14chaos_princessthe annoying bit is that there are no obvious differences, and the differences that there are do not seem to matter
14:19yukajannau: gpio-reserved-ranges seems perfect, I just need to figure out how to interleave regmap and gpiochip init so that the regmap init does not crash before gpio-reserved-ranges has been parsed
14:29jannauyuka: you might need to parse it in the apple pinctrl driver as well
14:32svenyay, it dies in launchd now. probably userspace sprr stuff
14:32svenguess m4 needs more than aprr_jit=0
14:33chaos_princesswhat was the pmgr-adjacent issue?
14:33sveni'm just shadowing the page now and it boots :D
14:33svenstill gotta figure out what exactly goes on there
14:35svenwell, pages. cpu-impl-reg points to them i think
14:37saiprasad_ has joined #asahi-dev
14:38chaos_princessi have a suspicion that the name of at least one function in backtrace will contain "lockdown"
14:38saiprasad__ has quit [Remote host closed the connection]
14:38svencan't easily see the backtrace, but it's an undefined instruction EL0 panic
14:38svenbut yeah... wouldn't surprise me
14:39svenyes, it's a userspace panic
14:39svenprobably a writing to that SPRR sysreg
14:39chaos_princessi meant the pmgr-adjacent panic, not launchd one
14:39sventhat was an serror
14:49nirgo has joined #asahi-dev
14:49nirgo has quit []
14:50bennofs3 has quit [Quit: Ping timeout (120 seconds)]
14:50bennofs3 has joined #asahi-dev
15:02hardy94 has quit [Remote host closed the connection]
16:02svenblergh.. now PMP doesn't want to init i think and that makes xnu panic after a timeout
16:03lucy_ has joined #asahi-dev
16:06lynndotpy72 has quit [Quit: bye bye]
16:08lynndotpy72 has joined #asahi-dev
16:09crabbedhaloablut has quit []
16:11crabbedhaloablut has joined #asahi-dev
16:27N_Tempus_04 has quit [Ping timeout: 480 seconds]
16:43rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
16:43Guest825 has quit [Quit: Bridge terminating on SIGTERM]
16:44rhysmdnz has joined #asahi-dev
16:45Guest971 has joined #asahi-dev
16:46sefidel has quit [Remote host closed the connection]
16:46lucy__ has joined #asahi-dev
16:50lucy_ has quit [Ping timeout: 480 seconds]
16:51sefidel has joined #asahi-dev
16:52Calandracas__ has joined #asahi-dev
16:59Calandracas_ has quit [Ping timeout: 480 seconds]
17:07lucy__ has quit [Quit: Konversation terminated!]
17:07lucy__ has joined #asahi-dev
17:10svenunrelated, but either my hdmi capture card is broken or i somehow dont get any display output in m1n1 already
17:11lucy__ is now known as hello0092
17:14jannaum1n1 would be expected as the iboot interface appears to be incomptible on m4 and/or macOS 26
17:15svenah, that probably explains it. actual xnu might just die too early to initialize the display
17:17jannaualso possible that whatever m1n1 does breaks macos. try uncommenting #define NO_DISPLAY in config.h
17:18sveni already have that in both the boot object and the chainloaded obj
17:18svenright now it looks like all cores somehow end up in their wfi loop and never leave just after launchd starts
17:19jannauthen macOS either dies too early or it is otherwise broken under HV
17:20sven~all impdef sysregs right now are shadowed, could be that something in there is supposed to trigger an interrupt that just never happens
17:37remexre has joined #asahi-dev
17:56sven…. Like the timers :D
18:03aditya has joined #asahi-dev
18:15bennofs3 has quit [Ping timeout: 480 seconds]
18:18bennofs3 has joined #asahi-dev
18:20croissant has quit [Quit: Leaving]
18:28croissant has joined #asahi-dev
18:32svenand i'm back to a corrupted stack pointer somehow :(
19:06bennofs3 has quit [Ping timeout: 480 seconds]
19:10bennofs3 has joined #asahi-dev
19:10svenwell.. either that or the cores end up in what's effectively an endless 1: wfi; b 1b loop
19:37pb17 has quit [Ping timeout: 480 seconds]
19:43averne has quit [Ping timeout: 480 seconds]
20:05N_Tempus_04 has joined #asahi-dev
20:12aditya has quit [Quit: Connection closed for inactivity]
20:16pb17 has joined #asahi-dev
20:37N_Tempus_04 has quit [Ping timeout: 480 seconds]
20:42alessaba has quit [Ping timeout: 480 seconds]
20:57cy8aer has quit [Remote host closed the connection]
21:08cy8aer has joined #asahi-dev
21:13bennofs3 has quit [Ping timeout: 480 seconds]
21:19bennofs3 has joined #asahi-dev
21:24svenand now I’m at some PAC failure I think
21:24sventime to call it a day though
21:32hello0092 has quit [Remote host closed the connection]
21:32hello0092 has joined #asahi-dev
21:36jules[m]1234 has joined #asahi-dev
21:36msimonelli2 has quit [Ping timeout: 480 seconds]
22:05pb17 has quit [Ping timeout: 480 seconds]
22:19msimonelli22 has joined #asahi-dev
22:29pb17 has joined #asahi-dev
23:26hello0092 has quit [Remote host closed the connection]
23:30hello0092 has joined #asahi-dev
…