#
…

2026-07-18

00:05maettu102 has joined #asahi-dev
00:14Mary has joined #asahi-dev
00:40voidah has quit [Ping timeout: 480 seconds]
00:47voidah has joined #asahi-dev
00:56voidah has quit [Ping timeout: 480 seconds]
01:03voidah has joined #asahi-dev
01:07hdbngr has joined #asahi-dev
01:16hdbngr has quit [Ping timeout: 480 seconds]
01:33r0ni has quit [Ping timeout: 480 seconds]
01:49hdbngr has joined #asahi-dev
01:51erincandescent_ has joined #asahi-dev
01:58hdbngr has quit [Ping timeout: 480 seconds]
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:12r0ni has joined #asahi-dev
02:27nela has quit [Ping timeout: 480 seconds]
02:30nela has joined #asahi-dev
02:33nnym has quit [Ping timeout: 480 seconds]
02:34hdbngr has joined #asahi-dev
02:42hdbngr has quit [Ping timeout: 480 seconds]
02:50hdbngr has joined #asahi-dev
02:59hdbngr has quit [Ping timeout: 480 seconds]
03:07hdbngr has joined #asahi-dev
03:15hdbngr has quit [Ping timeout: 480 seconds]
03:16pb17 has quit [Remote host closed the connection]
03:16pb17 has joined #asahi-dev
03:43nora has joined #asahi-dev
03:45nora_ has quit [Ping timeout: 480 seconds]
03:58yuyichao_ has joined #asahi-dev
03:59hdbngr has joined #asahi-dev
04:01marvin24 has quit [Ping timeout: 480 seconds]
04:07hdbngr has quit [Ping timeout: 480 seconds]
04:20hdbngr has joined #asahi-dev
04:28hdbngr has quit [Ping timeout: 480 seconds]
04:37hdbngr has joined #asahi-dev
04:46hdbngr has quit [Ping timeout: 480 seconds]
04:51hdbngr has joined #asahi-dev
04:56voidah has quit [Ping timeout: 480 seconds]
04:59nickchanyuka: tupai is base A18
04:59nickchanrefer to /System/Library/Frameworks/Kernel.framework/Versions/A/Headers/arm/cpuid.h in macos 27 sdk
04:59nickchanbase m5 is hidra
05:00hdbngr has quit [Ping timeout: 480 seconds]
05:00voidah has joined #asahi-dev
05:00yukaWhoops okay
05:01nickchanm5 pro and m5 max is sorta_s and sorta_c and there's something unifdef's immediately after based on the spacing
05:02nickchanthen it's a19 pro and a19 though the comments seems to be copy pasted wrong
05:02nickchanfollowed by even more unifdef ofc
05:03yukarhlvista[m] on t6050 there is cluster-type = M (marketing name: performance core) and cluster-type = P (marketing name: super core)
05:04yukaI guess I need to copy that file to my linux partition to save myself rebooting into macOS
05:05yukaoperating macOS is physical pain to me since it doesn't support my keyboard layout
05:06yuka(Well it can be installed but not for the login screen where I need to type a password at least once on every switch)
05:09nickchangiven that m5 pro and m5 max is both t6050
05:09rhlvista[m]cool on the wikipida for the m5 it says it have 4 super cores but on the adt the core name is efficiency
05:09nickchanand their device trees both have double the amount of cpu in m5 max
05:10nickchanexpect more multi die shenanigans in the mac studio soon
05:11voidah has quit [Ping timeout: 480 seconds]
05:14nickchanrhlvista[m]: that doesn't seem to be a problem
05:14nickchanthere are 10 cpu cores total in base m5
05:14nickchanand /cpus/cpu0 to /cpus/cpu5 in adt seems to be efficiency core
05:15nickchancpu6 - cpu9 is cluster-type = P (marketing name: super core)
05:15rosalyntg has quit [Read error: Connection reset by peer]
05:15rosalyntg has joined #asahi-dev
05:15voidah has joined #asahi-dev
05:16hdbngr has joined #asahi-dev
05:18rhlvista[m]ok that make sense
05:21rhlvista[m]and also for the 26.6bata 5 "msr((3, 5, 15, 1, 3), mrs((3, 5, 15, 1, 3)))" still crashes
05:21yukarhlvista[m] you adt should list 4 P-cores
05:21nickchanthis page is up to date afaik
05:22yukaIf this msr crashes that will also be the reason for the hypervisor crashing
05:22yukaAh yes sotra_d
05:22yukaWhat could that be /s
05:22nickchanappeared over a year ago inside some ane firmware
05:24rhlvista[m]what are some changes would need to be made to make the linux kernel to get the m5 framebuffer/aic controller to work
05:24yukaNot much
05:24nickchanremove access to that msr which would work as long as kvm is not enabled
05:25yukaBasically get some recent linux checkout
05:25hdbngr has quit [Ping timeout: 480 seconds]
05:25yukaWrite a device tree
05:25nickchanthough... i wonder if some *other* random register is unlocked instead
05:25nickchanwith how sloppy apple has been recently wouldn't be surprised
05:28rhlvista[m]i wrote my own device trees it was kida fun finding out about the address offset
05:28yukacan relate to that
05:28yukaCan you commit it with sign-off and push it somewhere?
05:28yukaThen I can include it in my branch
05:28rhlvista[m]im making a fork of it rn
05:32yukaYou probably already have the dapf_init_all commented out
05:33yukaAnd specify idle=nop in the cmdline
05:34yukaThat should get you booted to a framebuffer shell
05:34rhlvista[m]how would i specify idle=nop i just use the asahi config for building the kernel
05:35yukaYou load the kernel and initrd over the m1n1 proxy protocol using linux.py
05:35nickchanyou could hardcode it in the kconfig or use m1n1's linux.py
05:35rhlvista[m]oh you meant over the boot args
05:36yukaThe -b param of linux.py
05:37voidah has quit [Read error: Connection timed out]
05:49voidah has joined #asahi-dev
05:49rhlvista[m]compiling the kernel right now
05:49rhlvista[m]and also i checked the compatible regs for the m5
05:49yukaj615 should be j815?
05:50yukain the Makefile
05:50yukaoh you just moved a line
05:50rhlvista[m]yea made a mistake
05:51yukaif you run into problems booting into the framebuffer shell, try commenting out the pmgr stuff. I usually do the initial tests without pmgr (all devices remain in the state m1n1 left them in).
05:52yukaAnd a general note, you should use git cherry-pick to put together commits of other people to maintain attribution
05:52yukafeel free to dm me if you want to be guided through that, though there are also many tutorials online
05:52hdbngr has joined #asahi-dev
05:53yukanot only attribution, but also keeping each functional change in its own commit, so that I can cherry-pick your new changes back
05:56rhlvista[m]yea i probably should have done diffent commits with the git cherry-pick for those files. next time i will do that
06:00rhlvista[m]ok it is at the asahi logo didnt change at all
06:00rhlvista[m]b 'earlycon console=ttySAC0,1500000 console=tty0 debug idle=nop'
06:00yukatry console=tty0 instead of console=ttySAC0,1500000
06:01hdbngr has quit [Ping timeout: 480 seconds]
06:01yukaI see you have both, I don't remember which one is the primary console
06:02yukaas in, the last or the first specified
06:02yukabut yeah seems like something is going wrong
06:03voidah has quit [Ping timeout: 480 seconds]
06:03rhlvista[m]2 ports open with the m1n1 proxyclient ttyACM0 / ttyACM1
06:03rhlvista[m]im gonna try them both and see if stuff change
06:03yukaonce you run linux.py, the cdc-acm ports no longer work
06:04yukaat least if it actually started the kernel
06:04yukattyACM0 is for communication with m1n1 proxy, ttyACM1 is the vuart for communicating with the m1n1 hypervisor guest
06:05yukayour device tree is already wayyy too complicated and there are tons of things in there which will probably not work. I suggest you try commenting out most of it and only get aic and framebuffer to work
06:05hdbngr has joined #asahi-dev
06:05rhlvista[m]includeing the uart?
06:05yukathe s5l-uart: possibly
06:06yukayou don't have anything connected to it anyways
06:06yukaso it's just an unnecessary point of failure
06:06rhlvista[m]whats up with the timer node?
06:06yukathat is fine probably
06:07yukabut some nodes you should comment out: i2c*, fpwm0, pmgr*, pinctrl*
06:07yukaand the pmgr.dtsi include
06:09mattia013 has quit [Remote host closed the connection]
06:09rhlvista[m]affinities cpu_e0 ect
06:09rhlvista[m]do i comment those out too in the aic node?
06:09yukaaffinities you can comment out (but probably does not matter), but reg-names must stay
06:10nickchani don't think you need affinities unless unless one plans on using perf
06:10yukatake this an example of what is sufficient to boot to a framebuffer console
06:11yukaonce you have cleaned up the typos and added a sign-off to your commit, I can use it to give you a tree with (hopefully) working dockchannel-uart, which you can use to debug the early boot failures using a serial console from another pc
06:11mattia013 has joined #asahi-dev
06:11nickchana nit that device tree does violate the usual ordering guidelines
06:11nickchanat least timer should be after soc
06:12yukaare those documented somewhere?
06:12voidah has joined #asahi-dev
06:13yukaI copied from t6030, but I will fix it ;)
06:14hdbngr has quit [Ping timeout: 480 seconds]
06:14rhlvista[m]how do i find the vaules for the
06:14nickchanyou could just not include them now
06:15nickchani think people query it from macos sysctl
06:15yukaso cache-size = <0x600000>; for l2_cache_0 / e-cores and cache-size = <0x1000000>; for l2_cache_1 / 1-cores
06:16nickchanhw.perflevel1.l1icachesize, hw.perflevel1.l1dcachesize, hw.perflevel1.l2cachesize, hw.perflevel0.l1icachesize, hw.perflevel0.l1dcachesize, hw.perflevel0.l2cachesize
06:16nickchanfor pcluster and ecluster respectively
06:16yukathat's probably the more proper way if you can be bothered to reboot into macos
06:18nickchanand the most pedantic way of doing the t6050 dtsi is probably have 36 cpus but that doesn't matter until sorta_d or whatever comes out
06:19nickchanso at least i would not recommend that until there's more information on sotra_d
06:19nickchanerm, recommend upstream
06:20yukathe adt of j714s only has 18 cpus
06:20yukathe adt of j714c has 36 cpus
06:20yukaand I think the 36 cpus would be across two dies?
06:20yukaso it would be two dies of 18 cpus each
06:20yukaso t6050-dieX.dtsi with 18 cpus
06:20yukawhich is included twice
06:20yukabut yeah I decided not to do that for now
06:21voidah has quit [Ping timeout: 480 seconds]
06:22yukanickchan: in the kernel docs I only see rules for ordering of nodes _on a bus_
06:22yuka'/ {' is not a bus I think?
06:23nickchan"Nodes on any bus, thus using unit addresses"
06:24nickchanthe bus is relvant for the fact that it uses unit addresses for its children
06:24yukathat's one reading of it
06:25nickchanand if a node is not on a bus then it goes to 2
06:25nickchan"Nodes without unit addresses shall be ordered alpha-numerically by the node name."
06:25nickchanyou could also check the example
06:25yukarhlvista[m] ideally you would use git commit --amend and fixups/squashing of commits to create one single commit "add M5 device tree" which has all the changes and the sign-off
06:25rhlvista[m]ayo good news i saw the tux logo for a sec then it crashed
06:26yukayou might be the first person to see the tux logo on that device :)
06:26rhlvista[m]and sorry this is one of the first time of me useing git on the terminal i was prevulusy useing the vscode git section
06:28yukaYes no worries, there's lots to learn about git
06:28yukait's very valuable to learn though
06:28rhlvista[m]thanks im gonna do it again and see when it crashed give me a sec
06:28nickchanrhlvista[m]: you do want the reserved-memory node
06:28nickchankeep it as is filled by loader
06:29nickchanother things look either OK or pointless
06:29rhlvista[m]oh ok im gonna do that first and rerun the device tree
06:29hdbngr has joined #asahi-dev
06:29voidah has joined #asahi-dev
06:29nickchanyeah m1n1 really wants /reserved-memory
06:32rhlvista[m]so when the reserved memory is in the linux kernel/tux does not load
06:32rhlvista[m]without the reserved memory it shows the tux logos
06:33rosalyntg has quit [Ping timeout: 480 seconds]
06:33yukadid you also change the memory@ node at the same time?
06:34yukaand framebuffer status="disabled";
06:34yukasince these unrelated changes are in the same commit
06:35yukawell if framebuffer is disable, you're not going to see a tux logo
06:36rhlvista[m]oh then set it to "okay" then?
06:36yukait was "okay" before and you changed it to "disabled"
06:36nickchanm1n1 enables the framebuffer
06:36yukaaccording to your commit diff
06:36yukayes but linux needs to use it
06:36pb17 has quit [Ping timeout: 480 seconds]
06:37hdbngr has quit [Ping timeout: 480 seconds]
06:37nickchanyuka: m1n1 enables the framebuffer in dt_set_fb()
06:38nickchanand yes it should be disabled in the dtsi because it doesn't work before filled by loader
06:38voidah has quit [Ping timeout: 480 seconds]
06:38yukatrue, then nevermind that
06:39rhlvista[m]why would the reserved-memory break it then?
06:39yukaunknown, but what is your current reg = in the memory@ node?
06:39yukais it "reg = <0x1 0 0x2 0>;" or "reg = <0x100 0 0x2 0>;"?
06:39nickchanprobably some bad assumption in m1n1
06:40nickchanand the memory reg doesn't matter because m1n1 overwrites it
06:40rhlvista[m]one sec im gonna make some changes to the kboot.c see if that fixes it
06:42pb17 has joined #asahi-dev
06:43rhlvista[m]that fixed it
06:43rhlvista[m]but still crashes
06:46voidah has joined #asahi-dev
06:46rhlvista[m]crashes at starting system log ????: syslogd, klo??
06:47yukathat doesn't sound too bad
06:47nickchanmaybe it's something to do with wfi
06:48yukasyslogd sounds like far outside of kernel land
06:48nickchanwhatever the same crashes yuka has in m4 even with idle=nop
06:48yukathose are always in the kernel
06:48yukabut crashes is a very unspecific term
06:53rhlvista[m]if it helps when its running linux the laptops get hot very fast
06:53hdbngr has joined #asahi-dev
06:54voidah has quit [Ping timeout: 480 seconds]
06:59yukalet me write a device tree with dockchannel uart so that you can get proper debug output and paste it somewhere
07:00weirdtreething has joined #asahi-dev
07:00rhlvista[m]ok i will make the changes i made to the kboot push it and make the changes i made with the device tree and push that too
07:02hdbngr has quit [Ping timeout: 480 seconds]
07:03voidah has joined #asahi-dev
07:11voidah has quit [Ping timeout: 480 seconds]
07:18yukawith the feature/t8142 branch of my m1n1 repo, you should get a "Apple Debug USB" device when you connect to the dfu usb port
07:18yukayou can then run https://github.com/AsahiLinux/kisd and attach with picocom to /dev/m1n1
07:18rhlvista[m]you do realize the memory node is @800000000
07:18yuka@... is just a name
07:18yukait does not matter, I think
07:20voidah has joined #asahi-dev
07:26Yushi_ has joined #asahi-dev
07:29voidah has quit [Ping timeout: 480 seconds]
07:30Yushi_ is now known as Yushi
07:35rhlvista[m]ok the dts i think the next-level-cache is wrong because when i run it with it on it stays at the asahi logo without being ran the tux logos appear... (full message at <https://matrix.org/oftc/media/v1/media/download/AQmCT68FO5rZL0SsOEM50XNbuPjIga7UFjigy1j-9YgZL_kw8HsXJgjip5gqemlKMNx-HKS2mNhKBMziTJxWnuBCefvEAL6gAG1hdHJpeC5vcmcvWnpqdktqb1VhQ1VpT3lMVVB2aFVYZWtZ>)
07:36Carloschaos_princess: i'm logged into fedora over the serial console, is there a config for enabling the graphics/gui or should i enable it myself. you said get to kamoso so i'm assuming you want me to get onto the ui, but i don't even have network support on the target with that tree
07:37chaos_princessthat tree should be capable of having a ui
07:37voidah has joined #asahi-dev
07:37chaos_princessprobably bad .config
07:45voidah has quit [Ping timeout: 480 seconds]
07:53voidah has joined #asahi-dev
07:54hdbngr has joined #asahi-dev
08:01voidah has quit [Ping timeout: 480 seconds]
08:03mripard has quit [Remote host closed the connection]
08:03hdbngr has quit [Ping timeout: 480 seconds]
08:04mripard has joined #asahi-dev
08:07barusu has joined #asahi-dev
08:10voidah has joined #asahi-dev
08:12ravikant_ has joined #asahi-dev
08:13ravikant_ has quit []
08:18voidah has quit [Ping timeout: 480 seconds]
08:21voidah has joined #asahi-dev
08:29voidah has quit [Ping timeout: 480 seconds]
08:37voidah has joined #asahi-dev
08:39hdbngr has joined #asahi-dev
08:45voidah has quit [Ping timeout: 480 seconds]
08:47nickchanthe cache properties in dt is only used for sysfs so that makes zero sense whatsoever
08:47mira_ has joined #asahi-dev
08:48yukatrying to get mtp/keyboard to work on t6050:
08:48yukaapple-dart 294808000.iommu: DART is locked down until reboot: 0000000d
08:51rhlvista[m]the dt for the m5 works
08:53voidah has joined #asahi-dev
08:56yukanevermind, my dart-mtp regs were just wrong
09:01voidah has quit [Ping timeout: 480 seconds]
09:06hardy94 has quit [Quit: hardy94]
09:10voidah has joined #asahi-dev
09:18voidah has quit [Ping timeout: 480 seconds]
09:20svenwe can also deal with locked darts, it’s just very annoying
09:22voidah has joined #asahi-dev
09:23hdbngr_ has joined #asahi-dev
09:23yukano it's not locked
09:25yuka> apple-dart 294800000.iommu: DART [pagesize 4000, 16 streams, bypass support: 1, bypass forced: 0, AS 42 -> 42] initialized
09:26yukahowever:
09:27hdbngr has quit [Ping timeout: 480 seconds]
09:30voidah has quit [Ping timeout: 480 seconds]
09:34yukastill "Failed to wake up coprocessor"
09:35yukabuffer request for 0x5000 bytes at 0x0000000000000000 also does not sound right
09:36yukashouldn't that be somewhere in the sram region?
09:39voidah has joined #asahi-dev
09:48voidah has quit [Ping timeout: 480 seconds]
09:54barusu has quit [Remote host closed the connection]
09:56yukacould also be a firmware change in macOS 27
09:56yukaI had it working before on the Neo with macOS 26.x
09:56voidah has joined #asahi-dev
09:56yukabut now that also fails to boot on 27
09:56yukaso I'll try to get that working again
09:56yukaand then figure out what changed
10:01yukafound it!
10:01yuka> rtkit-helper 294600000.mtp: RTKit: syslog message: keyboard.c:63: Keyboard initialized
10:01yukaon t6050
10:03JamesCalligeros[m]What did they change -.-
10:04yukait was just a stupid device tree mistake again
10:04yukathere's no change
10:04yukaon the firmware side
10:04yuka(at least none I have encountered at this point)
10:04voidah has quit [Ping timeout: 480 seconds]
10:08fl0_id_afkoops. sry lol
10:09yukaSo keyboard works, yay
10:13voidah has joined #asahi-dev
10:16ddxtanx has quit [Remote host closed the connection]
10:17hdbngr_ has quit [Quit: leaving]
10:19barusu has joined #asahi-dev
10:31voidah has quit [Read error: Connection timed out]
10:40voidah has joined #asahi-dev
10:49voidah has quit [Ping timeout: 480 seconds]
10:56voidah has joined #asahi-dev
11:05voidah has quit [Ping timeout: 480 seconds]
11:14voidah has joined #asahi-dev
11:23voidah has quit [Ping timeout: 480 seconds]
11:24voidah has joined #asahi-dev
11:32voidah has quit [Ping timeout: 480 seconds]
11:40voidah has joined #asahi-dev
11:50voidah has quit [Ping timeout: 480 seconds]
11:58voidah has joined #asahi-dev
12:06voidah has quit [Ping timeout: 480 seconds]
12:15voidah has joined #asahi-dev
12:23voidah has quit [Ping timeout: 480 seconds]
12:25voidah has joined #asahi-dev
12:29coroner21 has quit [Remote host closed the connection]
12:34voidah has quit [Ping timeout: 480 seconds]
12:36coroner21 has joined #asahi-dev
12:43voidah has joined #asahi-dev
12:51voidah has quit [Ping timeout: 480 seconds]
12:52snek has quit [Remote host closed the connection]
12:52snek has joined #asahi-dev
12:52snek has quit [Remote host closed the connection]
12:52snek has joined #asahi-dev
12:57barusu has quit [Remote host closed the connection]
12:58barusu has joined #asahi-dev
12:59mini has joined #asahi-dev
12:59voidah has joined #asahi-dev
13:07voidah has quit [Ping timeout: 480 seconds]
13:12snek has quit [Remote host closed the connection]
13:12snek has joined #asahi-dev
13:15voidah has joined #asahi-dev
13:18snek has quit [Remote host closed the connection]
13:18snek has joined #asahi-dev
13:24voidah has quit [Ping timeout: 480 seconds]
13:26voidah has joined #asahi-dev
13:27snek has quit [Remote host closed the connection]
13:28snek has joined #asahi-dev
13:32snek has quit [Remote host closed the connection]
13:32snek has joined #asahi-dev
13:33chadmed has quit [Remote host closed the connection]
13:34snek has quit [Remote host closed the connection]
13:34snek has joined #asahi-dev
13:35voidah has quit [Ping timeout: 480 seconds]
13:38chadmed has joined #asahi-dev
13:38chadmed[ 69.579020] apple-dcp 38bc00000.dcp: swap failed! status 4
13:38chadmedwe inch closer and closer :P
13:42voidah has joined #asahi-dev
13:53voidah has quit [Ping timeout: 480 seconds]
14:00voidah has joined #asahi-dev
14:30chadmedkernel stuff is odd. struct sizes dont match by a significant amount and my eyes are glazing over, so problem for when i get back to brisbane perhaps
15:57voidah has quit [Ping timeout: 480 seconds]
16:05voidah has joined #asahi-dev
16:14voidah has quit [Ping timeout: 480 seconds]
16:23voidah has joined #asahi-dev
16:28junefish has joined #asahi-dev
16:32voidah has quit [Ping timeout: 480 seconds]
16:40voidah has joined #asahi-dev
16:50voidah has quit [Ping timeout: 480 seconds]
16:57voidah has joined #asahi-dev
17:06voidah has quit [Ping timeout: 480 seconds]
17:15voidah has joined #asahi-dev
17:23voidah has quit [Ping timeout: 480 seconds]
17:31voidah has joined #asahi-dev
17:39voidah has quit [Ping timeout: 480 seconds]
17:47voidah has joined #asahi-dev
17:56voidah has quit [Ping timeout: 480 seconds]
17:59nela has quit [Quit: Ping timeout (120 seconds)]
18:00fossdd has quit [Remote host closed the connection]
18:00nnym has joined #asahi-dev
18:01fossdd has joined #asahi-dev
18:04nela has joined #asahi-dev
18:05voidah has joined #asahi-dev
18:13voidah has quit [Ping timeout: 480 seconds]
18:21voidah has joined #asahi-dev
18:30voidah has quit [Ping timeout: 480 seconds]
18:38voidah has joined #asahi-dev
18:46voidah has quit [Ping timeout: 480 seconds]
18:48voidah has joined #asahi-dev
18:57voidah has quit [Ping timeout: 480 seconds]
19:06voidah has joined #asahi-dev
19:14voidah has quit [Ping timeout: 480 seconds]
19:17alxdrl[m] has joined #asahi-dev
19:23voidah has joined #asahi-dev
19:31voidah has quit [Ping timeout: 480 seconds]
19:39voidah has joined #asahi-dev
19:48voidah has quit [Ping timeout: 480 seconds]
19:49voidah has joined #asahi-dev
19:58voidah has quit [Ping timeout: 480 seconds]
20:06voidah has joined #asahi-dev
20:14voidah has quit [Ping timeout: 480 seconds]
20:15fossdd has quit [Remote host closed the connection]
20:16fossdd has joined #asahi-dev
20:22voidah has joined #asahi-dev
20:29hardy94 has joined #asahi-dev
20:30voidah has quit [Ping timeout: 480 seconds]
20:38voidah has joined #asahi-dev
20:46voidah has quit [Ping timeout: 480 seconds]
20:50voidah has joined #asahi-dev
20:55enverbalalic has joined #asahi-dev
20:57enverbalalici'm starting to thing that i'm cursed... tried yuka's dockchannel_irq experiment, interrupt really is different then ADT on t6041 (360 vs 816), but linux doesn't wanna use it, doesn't even create the ttyDC device, no earlycon. is there any m1n1 support needed for it if i do macvdmtool reboot debugusb - i can see m1n1's debugusb output, iBoot's output
20:57enverbalalici did disable pcie and dapf init in m1n1 (didn't figure it out for t6041), could that be breaking it ?
20:58yukadevice tree?
20:58yukakernel config?
20:59voidah has quit [Ping timeout: 480 seconds]
20:59yukavery likely to be one of those two
20:59enverbalalicmy own device tree, based off an existing one for m4 mini basically just changing the model, chassis-type, etc, including t6040.dtsi (with changed IRQ num) and enabling serial1
20:59yukacan you push it somewhere?
21:00enverbalalickernel config has CONFIG_SERIAL_APPLE_DOCKCHANNEL and CONFIG_SERIAL_APPLE_DOCKCHANNEL_EARLYCON, direct includes, not modules
21:00yukafwiw the irq does not matter for earlycon
21:00enverbalalict6040 dtsi from your dcuart-t8142 branch
21:01yukathat one is untested
21:01Yushi has quit [Read error: Connection reset by peer]
21:01yukabecause the t6040 I have remote access to is not hooked up for debugusb
21:01Yushi has joined #asahi-dev
21:02yukawhat cmdline / bootargs do you boot it with?
21:03enverbalalicwell i'm not sure what could even be breaking it at this point, regs in t6040.dtsi match what i'm seeing in the ADT, AIC matches what i'm seeing in the ADT, linux boots to shell, doesn't freeze
21:03yukaOh, I see an issue with the irq
21:03enverbalalic`console=tty0 loglevel=8 nokaslr idle=nop maxcpus=1 debug init=/init`, tried with `earlycon=dockchannel,mmio32,0x508828000 as well`
21:03yukamake that dockchannel,mmio32,0x50882c000
21:03yukathat needs to point to the data and not config reg
21:04enverbalalici tried all the regs already :(
21:06enverbalalichow are these regs figured out exactly ? looking at the ADT, 0x508828000 (config in dts) maps to dockchannel-uart[0], 0x50880c000 (irq in dts) maps to dockchannel-uart[1] and this 0x50882c000 (data)\ doesn't map to anything in the ADT
21:06voidah has joined #asahi-dev
21:07yukadata is config + 0x4000
21:07enverbalalicconfig is always [0] and irq is always [1] ?
21:07yukaand in the linux device tree, data always needs to be specified first
21:07yukado you have a working samsung serial?
21:07yukaor any other type of working console?
21:08enverbalalicno, that's why i'm stuck in this hell in the first place
21:08enverbalalici have a working framebuffer and i point an iphone camera in slow-mo mode at the screen when it boots to read dmesg
21:08yukayou can boot it without initrd
21:08yukathat reduces the amount of logs by a lot
21:08yukaand the relevant stuff happens before entering initrd
21:09yukathat could reveal some errors
21:09yukaotherwise, we could try to get your keyboard working
21:09yukathen you can use framebuffer + keyboard as console to debug the dockchannel-uart
21:09enverbalalicyeah but without initrd the first output on framebuffer is very late in boot, i don't see any early stuff
21:10enverbalalici have initrd to be able to do dmesg on init to dump the initial stuff
21:10enverbalalicsee, i had the same thought about keyboard a week ago, got nowhere with that either
21:10enverbalalicbased off your existing work for dchid on one of your branches
21:10yukaif you already have a custom initrd which runs dmesg on init, you could probably add some hacky thing to slow down the output deliberately?
21:11yukathinking about 'pv --rate-limit'
21:12yukaI just pushed my dcuart branch with a fix for the irq
21:12yukabut again, shouldn't matter for earlycon in theory
21:13enverbalalicchanges the serial1: dockchannel-uart address ?
21:13yukaonly the node name
21:13yukalet me try the keyboard thing
21:13yukasince I just got that working on t6050
21:14enverbalalici can send you the ADT for it if you need it
21:15voidah has quit [Ping timeout: 480 seconds]
21:16enverbalalicyeah node name does nothing
21:17enverbalalicthinking about it, when i tried integralchaos' dockchannel driver, that one did create ttyDC, and could output to my other device, but had the FIFO issue where just one message went through
21:17enverbalalicbut that one also had mailboxes and stuff in the dtb
21:17yukado you still have the dt for that?
21:22voidah has joined #asahi-dev
21:22enverbalalicmtp-hid is 99.9% wrong, i have no idea where those coproc reg addresses come from
21:24yukaI'm very confused how that could've worked for dockchannel uart
21:25yuka> <0x50 0x880c000 0x0 0x4000>
21:25enverbalalicthat didn't, that was me experimenting
21:25yukabut you said it could output?
21:25enverbalalic0x5 0x0880c000 did
21:25yukaand that's the same I have for my dockchannel-uart nodes
21:26enverbalalicit is the same, you just have reg-names in different order
21:26enverbalalicdata config irq vs irq config data
21:26yukayes and hopefully also the regs
21:26yukayes data needs to be first for earlycon
21:27enverbalalicintegralpilot earlycon didnt work either, just ttyDC and a straight echo to it
21:27enverbalalicdriver didn't even have earlycon IIRC
21:27yukathere was no earlycon implementation
21:27yukaexactly
21:27yukaearlycon over a mailbox driver would be a bit difficult
21:28yukaokay give me another 5 min
21:28yukathen I'll send you a new tree to try
21:28enverbalalicyeah, when i tried shoehorning earlycon into it i realized that...
21:31voidah has quit [Ping timeout: 480 seconds]
21:39voidah has joined #asahi-dev
21:46yukaHmm, I forgot there is no pmgr yet in this device tree
21:47crabbedhaloablut has quit []
21:47yukathis is going to take a while, possible not going to get it done today
21:47crabbedhaloablut has joined #asahi-dev
21:47yukasince this adds a lot of complexity, I also need to test it on the M4 Pro Mac Mini
21:47yuka(even if that can't test the actual keyboard part)
21:48enverbalalicwell, i lurk here daily and kb is the only hope to getting dc here, so whatever you need, tag me
21:49voidah has quit [Ping timeout: 480 seconds]
21:50enverbalalicyou want the m1n1 dump adt for 6041 or no ?
21:52voidah has joined #asahi-dev
21:53yukaI might ask for it in the future
21:53yukait's interesting mostly for tunables
21:54yukaso pcie, usb
21:54yukanot currently needed
21:54yukaeverything else I can pull out of ipsws
21:59yukam1n1 proxyclient does not work well over a high latency websocket proxy
22:00chadmed has quit [Quit: Konversation terminated!]
22:12yukaenverbalalic instead of disabling all of dapf_init_all, comment out only everything except mtp (and the trailing null) from the list of dapf instances in dapf.c
22:16yukaIt will likely not work, I need to check some more things tomorrow
22:26yukawhat is this madness
22:29WindowPain has quit [Read error: Connection reset by peer]
22:29enverbalalicthis dies before ever printing to the framebuffer
22:29enverbalalicjust the logo
22:31WindowPain has joined #asahi-dev
22:33yukacan you try if you get output over debugusb with earlycon=dockchannel,mmio32,0x50882c000 anyways?
22:34yukahow do you connect to debugusb?
22:35enverbalalici see output up until linux boots
22:35enverbalalicvectoring to next stage is last line
22:36yukaplease try this .config: https://x0.at/4H3s
22:36yukait should be rather quick to compile
22:37yukaotherwise I am out of ideas
22:38enverbalalicdoesn't work
22:39enverbalalicbut why is it dying before fb now
22:39yukabecause I added a lot of stuff which is probably broken
22:39yukayes, very smart, I know
22:39enverbalalicstuff's based on working stuff on t6050 ?
22:40yukapmgr mostly does weird things until all the right apple,always-on and status="disabled"; haev been added
22:42yukabut pmgr is kinda important to have for enabling things
22:43enverbalalic probably a stupid question, but if m1n1 has dockchannel working, shouldn't it leave all the stuff enabled for it work under linux as well ?
22:44yukadockchannel-uart, or dockchannel-mtp?
22:44yukadockchannel-uart does not need any additional power domains afaik
22:44yukadockchannel-mtp is not used/enabled by default by m1n1
22:44enverbalalicthat's probably the main thing that's bothering me here, in one of my experiments i basically copied the m1n1 dockchannel "driver" as a earlycon linux driver and tried it, linux would still not print anything
22:45yukathat's probably related to pagetables / memory mapping
22:46yukaearlycon is a bit special in that regard
22:46enverbalalictrue that's what i noticed in my experiment vs your earlycon driver, you map the regs as 0x04 i mapped them as 0x4004
22:46yukayes that won't work
22:46enverbalalicTX8 & TX_FREE
22:46yukaearlycon only maps 4K from the reg[0]
22:47yuka* up to
22:48yukathis is also why it's important that data is the first reg
22:49enverbalalicalso a stupid question, in the dtb, mapbase is the first reg, or the addr in the name ?
22:49yukamapbase gets populated from the first reg, no matter what is in reg-names
22:49yukadoes that answer your question?
22:50yukaI'm trying since half an hour to upload a kernel and initrd to the m4 max at flokli's home (80ms latency), and it's getting late...
22:51yuka*m4 pro
22:52enverbalalicup until tonight i thought that the `node-name: this-part@thisaddress {` actually mattered
22:52enverbalalicbut that's just a name
22:53yukalearning to (re-)write device trees is so useful
22:53snek has quit [Remote host closed the connection]
22:53yukaI finally got the i2s hat on my raspberry pi working on a mainline kernel
22:53snek has joined #asahi-dev
22:54yukawheeee
22:54yukathe m4 pro has booted
22:54yukainto a shell
22:54yukawith all cores
22:54yukaand pmgr
22:54yukalots of errors because m4 pro apparently has less memory channels and it's trying to enable the missing ones
22:55enverbalalicyeah... and what's even more insane, years ago in a previous company i worked on a HDMI CEC driver for a allwinner H2/H3, and got it working without knowing this about dtb's, all those nights thinking those addresses mattered
22:55yukapushed one more fix to the same branch
22:56yukaokay I'm heading off for tonight
22:56yukagood luck!
22:57enverbalalicthanks for all this, there's enough things here to keep me editing dtbs and building kernels for a good while
23:27enverbalalic has quit [Quit: enverbalalic]
23:53copticcurse has joined #asahi-dev
23:56nela has quit [Quit: Ping timeout (120 seconds)]
…