#
…

2026-07-25

00:04Mary has quit [Quit: .]
00:05Mary has joined #asahi-dev
00:05maettu102 has quit [Quit: brb]
00:05maettu102 has joined #asahi-dev
00:54unifiedgravity[m] has joined #asahi-dev
00:54unifiedgravity[m]Hi all, does anybody have the app/dmg installer for macOS 13.5.x for m1n1? I can only find ipsw which requires factory reset to install. I have an empty partition and want to triple boot macOS 26/13 and Asahi without wiping my machine please. All of the links to Apple's servers are now 403. Please, I mostly finished my PreSonus Quantum 2626 Linux driver - reverse engineered. I want to finish the USB4 bringup for Asahi please, but
00:54unifiedgravity[m]I need this OS installer first. Cheers :)
00:54unifiedgravity[m]Or, does anybody know if 13.7.8 works with m1n1, or what it'd take to make it work?
01:16chadmedthe next target is 14.8.5
01:16chadmedwe cannot possibly hope to properly support all point-releases
01:30unifiedgravity[m]That works for me, is it ready to use? Or do you have an ETA so I can decide whether I should just wipe my machine with an IPSW and get 13.5? Cheers
01:31unifiedgravity[m]I mean someone must have the installer for 13.5, no? Or do you all use the IPSW files with DFU restore?
01:31chadmedthe hypervisor works for 14.8.3 yes
01:33chadmedand yeah i mean, we have dev devices so its no issue to just dfu the things when we need to. the installassistant for 13.5 is on archive dot org
01:33unifiedgravity[m]Can somebody please update the docs https://asahilinux.org/docs/sw/m1n1-hypervisor/ to reflect 14.8.3 is now supported? It would have saved a lot of confusion
01:33hdbngr has joined #asahi-dev
01:37unifiedgravity[m]Also, I tried archive.org and could not find the file for 13.5 pkg install assistant. I'd still appreciate finding it if somebody can point me to it. Cheers :)
01:41aites[m] has joined #asahi-dev
01:41aites[m]<unifiedgravity[m]> "Can somebody please update the..." <- 😭😭😭😭 somebody else having the exact same issues as me
01:41aites[m]Yeah I think somewhere a few months ago was me struggling in the chat with the exact same thing
01:42hdbngr has quit [Ping timeout: 480 seconds]
01:42aites[m]I'm pretty sure this is the one I used. But also I think I was able to install it without this? It was kinda weird on my machine in the end I wasn't fully understanding how it worked in the end by the time I got the working 13.5 installation on my machine
01:43aites[m]I'm pretty sure the one you get from their online mirrors and/or mist also work
01:45SnipeX_[m] has joined #asahi-dev
01:53erincandescent_ has quit [Remote host closed the connection]
01:54erincandescent has joined #asahi-dev
01:59pb17 has quit [Ping timeout: 480 seconds]
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:01unifiedgravity[m]You can no longer get 13.5 installer from the Apple servers - 403. Frustratingly, none of the "repositories" ever downloaded them - only give links to Apple servers which Apple eventually retires. Therefore I also had no luck with Mist etc
02:01aites[m]Ah, right. In that case I must have gotten the 13.5 installer from archive.org then
02:02aites[m]It sounds like you did a lot of work trying to get this to work - Honestly, I feel ya, its an absolute pain to navigate I had to do this too
02:05pb17 has joined #asahi-dev
02:05hdbngr has joined #asahi-dev
02:10JamesCalligeros[m]archive.org/details/install-assistant_20240930
02:13hdbngr has quit [Ping timeout: 480 seconds]
02:42hdbngr has joined #asahi-dev
02:51hdbngr has quit [Ping timeout: 480 seconds]
03:17nora has joined #asahi-dev
03:19nora_ has quit [Ping timeout: 480 seconds]
03:57pb17 has quit [Ping timeout: 480 seconds]
04:08pb17 has joined #asahi-dev
05:29hdbngr has joined #asahi-dev
05:38hdbngr has quit [Ping timeout: 480 seconds]
06:16hdbngr has joined #asahi-dev
06:17rosalyntg has quit [Read error: Connection reset by peer]
06:17rosalyntg has joined #asahi-dev
06:25hdbngr has quit [Ping timeout: 480 seconds]
06:26ralim8 has quit [Quit: The Lounge - https://thelounge.chat]
06:26ralim8 has joined #asahi-dev
06:27roxfan has quit [Ping timeout: 480 seconds]
06:34hdbngr has joined #asahi-dev
06:42hdbngr has quit [Ping timeout: 480 seconds]
06:44nela has quit [Ping timeout: 480 seconds]
06:47nela has joined #asahi-dev
07:00weirdtreething has quit [Remote host closed the connection]
07:00weirdtreething has joined #asahi-dev
07:05hdbngr has joined #asahi-dev
07:18nela has quit [Ping timeout: 480 seconds]
07:21nela has joined #asahi-dev
07:39hdbngr has quit [Ping timeout: 480 seconds]
08:15hdbngr has joined #asahi-dev
08:19roxfan has joined #asahi-dev
08:23hdbngr has quit [Ping timeout: 480 seconds]
08:25hdbngr has joined #asahi-dev
08:34hdbngr has quit [Ping timeout: 480 seconds]
08:37nst has quit [Quit: WeeChat 4.7.2]
08:43nst has joined #asahi-dev
08:46nst has quit []
08:48nst has joined #asahi-dev
08:53TurboRigby[m] has joined #asahi-dev
08:54hdbngr has joined #asahi-dev
09:00yukaatcphy/hotplug for me on t6034, nice work chaos_princess!
09:00yuka* works for me
09:00chaos_princesswhat was your test device?
09:01yukaa yubikey and a usb hub
09:01yukaI can test a dp capable dock
09:02yukaseems to work
09:02chaos_princessi didn't expect it to, but neat
09:03yukaWell, obviously no display output but at least the usb and charging parts do work
09:03JamesCalligeros[m]…as in displayport works?
09:03yukaI don't have any dcp set up on this device ;)
09:03chaos_princessi expected dp to go full firework mode
09:04yukaI got this a few times:
09:04yukadrivers/phy/apple/atc.c:2328 at atcphy_mux_set+0x1154/0x1858
09:04yuka(WARNING)
09:05chaos_princessw/ dp or on usb3-only too?
09:05yukausb3-only
09:05chaos_princessdid your hub actually link up at 3.0 speeds?
09:06yukahow do I check that?
09:06chaos_princessis it under the 3.0 port in lsusb
09:06yukasomething superspeed in dmesg?
09:07chaos_princessi also plugged in a ssd enclosure and checked if the read speeds are what is expected for 3.0
09:07yukaand you mean specifically in the case where it produced the WARN?
09:07jannau`lsusb -t` displays the link speed
09:09yukaI have both a 2.0 and a 3.0 hub
09:09yukadoes that make sense?
09:09chaos_princessthat is what is supposed to be
09:09chaos_princessevery 3.0 hub is a 2.0 and 3.0 hubs under a trenchcoat
09:10yukaokay, so
09:10yukathe dock works at 3.0 speeds
09:10yukathe other hub not
09:10yukathe other hub is the one which produced the WARN
09:11svenfunnily, thunderbolt 3 can effectively tunnel usb2 and usb3 via pcie but usb4 then only tunnels usb3 and still passes through the D+/- lines with a hub :D
09:11yuka(though I do not get that WARN every time)
09:11JamesCalligeros[m]Someone ban the USB-IF from continuing to exist please
09:12svenalso, in cursed sprr hv news: i think my static pagetable rewrite code works \o/
09:12yukathat hub also does not work at 3.0 speeds on my t6000, so....
09:12chaos_princessoh, then i don't care :P
09:12hdbngr has quit [Ping timeout: 480 seconds]
09:12yukamaybe the cable is just broken
09:13chaos_princessdoes it link up at 3.0 speeds in macos?
09:13svencan't accurately represent every SPRR permissions but the ones xnu actually uses should be fine
09:14hdbngr has joined #asahi-dev
09:14chaos_princessstatic as in - just rewrite pagetables, do not hook tlbi/lazy fill yet?
09:15svenyeah, it's just python for now even :D
09:15sveni think i'll start with just marking everything read-only. it'll be slow but easier to get right I think
09:15yukaconfirmed the cable/plug of the other hub is worn out and does not do usb3 speeds on macos either unless I do cursed things
09:17yukaso yeah atcphy seems to work pretty much flawlessly
09:27TurboRigby[m]hey, M4 MacBook Air owner here. I've got AArch64 bare-metal/embedded background (built a small custom kernel on ARM before, comfortable with boot chains and low-level debugging). Is anyone actively working on M4 bring-up yet, or is it still fully untouched? Happy to help with initial m1n1/devicetree probing or tracing if there's groundwork that needs doing and nobody's on it.
09:27chaos_princessthe "sptm" thingy mentioned above is very m4+-relevant
09:29chaos_princesssprr/gxf/GLx - all of it is m4-relevant stuff
09:32yukarelevant for the hypervisor on M4+ anyways, which allows us to analyze interactions of macOS with hardware. The peripherals on M4 are luckily still somewhat similar.
09:34yukaTurboRigby[m] I have an M4 Mac mini myself, but if I need anything M4-specific tested on a laptop (keyboard etc.) I can ping you (if you keep hanging out here)
09:35yukait's definitely not untouched, lots of stuff has been happening, though slowly because of the lack of the hypervisor
09:38TurboRigby[m]yuka: yeah for sure, I'll keep hanging around here. happy to test whatever comes up on the Air
09:39nst_ has joined #asahi-dev
09:44nst has quit [Ping timeout: 480 seconds]
09:48TurboRigby[m]chaos_princess: is the XNU-shim/asahi_neo approach the current best lead, or is there official m1n1 work happening on this too?
09:49chaos_princessdon't go to that one, iirc that one is slop
09:49chaos_princesswork is happening under the actual AsahiLinux org
09:49chadmedasahi_neo is entirely slop
09:50TurboRigby[m]chaos_princess: ah good to know, thanks. is there a branch or issue in the AsahiLinux org I should track for the M4/SPTM work?
09:50chaos_princessnot really, just look at pull requests and see which ones reference t604x or t8132
09:51chaos_princesswe do not really do "feature" branches
10:00svenand for the sptm stuff probably just follow discussions in this channel
10:05yukaTurboRigby[m] fwiw asahi-wip should have a basic device tree for your device, just that ~0 of the hardware is included/supported (no nvme, no pcie, no usb, no wifi/bt, no keyboard/touchpad...), so no point in booting it unless you have something specific to test
10:07yukaand there is no support for M4 in the installer so it's a very manual process of installing a recent beta/RC into a separate partition, install m1n1 as custom kernel object via the 1TR recovery, and then pushing the kernel/initrd to the device via usb every time
10:08Tachyonray[m] has joined #asahi-dev
10:09Tachyonray[m]Is this process described in the docs somewhere?
10:09chaos_princessjust install macos and kmutil it
10:10chaos_princessisn't the page above missing the bputil -nckas step?
10:10yukapossible
10:21ddxtanx_ has quit [Remote host closed the connection]
10:44alessaba has quit [Ping timeout: 480 seconds]
10:44alessaba has joined #asahi-dev
10:44hdbngr has quit [Ping timeout: 480 seconds]
10:46hdbngr has joined #asahi-dev
10:52alessaba has quit [Ping timeout: 480 seconds]
10:53alessaba has joined #asahi-dev
10:59yuka> Bus 002 Device 002: ID 0bda:8156 CMI USB 10/100/1G/2.5G LAN
10:59yukaOn M4 Mac mini
10:59chaos_princessbus 002 is 3.0 link, right?
10:59chaos_princessso, t8122 phy just works(tm)?
11:00yukaonly thing is one tunable is named differently
11:00yukatunable_ATC0AXI2AF vs tunable_ATCAXI2AF
11:00yukathough that might also be a firmware version thing
11:01yukaunclear
11:01yuka/ iboot version
11:01yukaQuick question: where is the wdt re-enabled when booting linux? I couldn't find it
11:01yukabut jannau said something like this recently
11:02yukaor does the linux driver reenable it by itself
11:02yuka(I looked in m1n1 payload / kboot)
11:04chaos_princesslinux is supposed to re-arm it
11:07jannauhuh, it looks like I'm mistaken and it was just what I was expecting (to trigger reboots if the next payload doesn't start)
11:07jannauI think u-boot re-arms it as well
11:07yukaI think it changed again with some recent iboot/firmware update :(
11:08yukaat least I was sure I had tested a linux update of 10+ minutes
11:08yukaat some point
11:08yukabut now I can't get past 4 minutes
11:08yukabut doesn't even seem to be watchdog
11:09yukaif m1n1 doesn't re-arm it before booting linux
11:09Mrmaxmeier17 has quit [Quit: The Lounge - https://thelounge.chat]
11:10Mrmaxmeier17 has joined #asahi-dev
11:14jannausofus: are some fails / errors expected for vp9? t8103 passes 215/315 and none of the high bit depth ones
11:16jannauJVT-AVC_V1 passes 28/135 on t8103
11:21yukagetting >5min uptime sitting in m1n1, but once I load linux it reboots after 3-4mins (even without watchdog node and with pd_ignore_unused). weird
11:22yukawill try once more with everything removed from the dt except aic and serial
11:23sofusjannau: vp9 sounds right, iirc could not get macos to decode high bit/profile. i have 216/305, i wonder what that one test is
11:25sofus28/135 is very low, iirc its 77/135 on t6020
11:25Tachyonray[m]can I find this test suite somewhere?
11:26jannauthere is a single test with ERROR: vp90-2-02-size-lf-1920x1080.webm maybe it's that one
11:28sofusjannau: i think its a timing thing then, its one of the more "demanding" tests
11:36jannauframe processing timeout's in kernel log
11:37sofusdoes it say something like vp255 or vp3
11:38jannau"Vp: 3 (00)" and "H3 00 error"
11:39sofuswhat i mean by timing thing is, maybe we are sending instructions faster than it can process them
11:42sofusthe cm3 firmware has 5-6 different places where it checks this and "waits" (piodma transfer)
11:44hdbngr has quit [Remote host closed the connection]
11:44hdbngr has joined #asahi-dev
11:54hdbngr has quit [Ping timeout: 480 seconds]
11:58hdbngr has joined #asahi-dev
12:00aites[m]Heya! So I've been working on writing a new driver for controlling haptics on the trackpad for the MTP multitouch trackpad and I would love if somebody could take a look. I've spent way too long on this 😅 I had to dig through a lot of code and mailing list stuff to figure all this out
12:00aites[m]sofus: you might be interested
12:02chaos_princessdo you want an actual ml-style review?
12:02aites[m]What is meant by ml?
12:03aites[m]Ohh right.
12:03yukathere's still commented out code in there
12:03chaos_princessalso, run checkpatch.pl
12:06aites[m]chaos_princess: 😅 I'm a little anxious about a ml-style review I still have left some comments and codes strewn about, but sure
12:08chaos_princessstart by running checkpatch. also, it is unclear to me why it is a separate thing and not just a part of hid-magicmouse
12:08Mary has quit [Quit: .]
12:09chadmeduse --strict with checkpatch too. some maintainers will crucify you for not doing so
12:09Mary has joined #asahi-dev
12:11sofusaites[m]: amazing! in the events you send, do the rest of the "magic" numbers do anything?
12:17aites[m]chaos_princess: Dockchannel enumerates the actuator as a separate device that it probes and initializes independently from the trackpad itself and its own instantiation of hid_ll_driver which you can send a raw request using, independently of the raw_request of the trackpad itself
12:17aites[m]And yess I will clean up the code if the shape of the driver looks about right to you guys
12:20aites[m]sofus: Thank you!!! :D
12:20aites[m]I believe the other magic numbers do something, the clicks definitely feel different, but I'm not 100% sure what they do, but I think it might be that what it does might be non-standard compared to the rest of the haptic trackpads from other vendors so it might be difficult to finangle in controls for them in a way that's standardized if their feature was known
12:20aites[m]But if you wanted to take a look as well I'd be absolutely interested in what you find
12:20aites[m]chaos_princess: 🫡 this will be fixed
12:22chadmedmight be worth calling it taptic to match apple's name for it to make that less likely too
12:22chadmedwould be interesting if the interface is shared with the taptic engine on the idevices with it
12:27sofusaites[m]: does it not make sense to expose control to change the haptics without doing the host mode thing?
12:27hardy94 has joined #asahi-dev
12:31yukaThe M4 Mac mini is literally the only device I have where that atcphy tunable is called ATCAXI2AF. all newer devices use ATC0AXI2AF again
12:31yuka_think different_
12:32yukaI guess it should be handled like an alias
12:32yukado I add a generic aliases mechanism or do I just add this case in dt_append_atc_tunable?
12:35chaos_princesslet's not bother for now imo
12:35chaos_princessnot bother = just special case for now
12:36aites[m]sofus: What do you mean?
12:39sofusiirc it was possible to change the how the clicks feel and not use the host mode
12:40sofusso like you could change the downclick and the trackpad would still handle everything
12:41JamesCalligeros[m]Does libinput still not have all the bits to do this itself
12:41aites[m]JamesCalligeros[m]: It doesn't afaik
12:41aites[m]I think I'll send a PR to libinput too
12:41JamesCalligeros[m]There were concepts of plans to add it when i looked at this in 2022 :p
12:42jannaulinux itself has now support for haptic trackpads
12:44jannausee drivers/hid/hid-haptic.c
12:44aites[m]<aites[m]> "My design follows the interface..." <- Yeah this is in the patchseries I linked
12:45aites[m]sofus: Is this the `0x22` and `0x23` events?
12:45MikeM64 has quit [Ping timeout: 480 seconds]
12:50Mary has quit [Quit: .]
12:54svenhas anyone ever looked into what SPRR_UMPRR_EL1 does? xnu seems to access it afaict but i can't figure out what this changes
12:55Mary has joined #asahi-dev
12:58aites[m]sofus: Ohh yeah that makes sense, I could add that to the driver too I reckon, but I guess the problem is that we already have pre-established ioctls for configuring the haptic force feedback events that userspace can manually send but none for configuring the ones that happen autonomously
13:02sofusahh, makes sense. Dont bother doing that
13:13nicklvh has joined #asahi-dev
13:14hdbngr has quit [Ping timeout: 480 seconds]
13:17hdbngr has joined #asahi-dev
13:26svengod is this going to be cursed if i want to support gl0 which svcs (or any other exception) up to gl1 which i can't easily trap
13:26hdbngr has quit [Ping timeout: 480 seconds]
13:27chaos_princessdo we need to trap it?
13:31svenif i just replace the _GL1 exception regs with hvcs and then some exception happens while we're inside GL0 (which is actually cursed EL0 now) we'll get the exception state in the _EL1 registers and, uh, that's probably not gonna go great
13:32chaos_princessoh, the exception state regs differ
13:32chaos_princessok, makes sense now
13:32sveni think i can get away with swapping them around and keeping track in the hv, hence the "oh god, this is gonna be cursed"
13:33svenbut we'll see, i'm still busy figuring out how GXF actually works
13:39NickChan[m]sven: why do you need to replace registers? TIDCP should trap them
13:40svenbecause i only want to trap GXF/SPRR on M1 for now
13:43svenand xnu's code is effectively read-only anyway so i'm also fine with patching whatever
13:43sveni have to patch genter/gexit to some hvc anyway
14:00hdbngr has joined #asahi-dev
14:11nicklvh has quit [Remote host closed the connection]
14:11hdbngr has quit [Ping timeout: 480 seconds]
14:13hdbngr has joined #asahi-dev
14:26hdbngr has quit [Ping timeout: 480 seconds]
14:39hdbngr has joined #asahi-dev
14:50hdbngr has quit [Ping timeout: 480 seconds]
14:52hdbngr has joined #asahi-dev
14:53alessaba has quit [Ping timeout: 480 seconds]
14:53alessaba has joined #asahi-dev
14:58chadmed has quit [Quit: Konversation terminated!]
14:58chadmed has joined #asahi-dev
14:59svenanother cursed idea: since i control the page tables anyway i can just map my own code and force VBAR_GL1 there and just do whatever is required there before jumping to the real one
15:01hdbngr has quit [Ping timeout: 480 seconds]
15:07hdbngr has joined #asahi-dev
15:12hardy94 has quit [Remote host closed the connection]
15:16hdbngr has quit [Ping timeout: 480 seconds]
15:27hdbngr has joined #asahi-dev
15:36devadri has joined #asahi-dev
15:38yukahmm, wifi/bluetooth/ethernet show up in lspci, but the driver for ethernet does not probe: Unable to obtain 64 bit DMA for consistent allocations
15:38yukawere the gigabit ethernet cards in previous mac minis also brcm tg3?
15:44chaos_princessyep, tg3 on the m1 one here
15:45yukaand it's using the unmodified driver and fw for that, correct?
15:51jannausearch for dma_mask in tg3_init_one(), maybe try to 'apple,dma-range = <0x0 0x0 0x0 0xfc000000>;' to the ethernet pcie dart node
15:52yukaAha, this is downstream stuff
15:54jannauI don't think it'll help, different issue than I assumed\
16:00jannaunot sure why dma_set_coherent_mask() would fail. it should not with iommu. maybe print the error and follow the failure path up
16:01yukacould be my kernel config
16:01jannauthe wider address space shouldn't be an issue since the mac mini with m2 pro works with tg3 as well
16:02jannau16k page size, iommu / dart support?
16:02yukayes to all of them
16:02yuka(USB works in the same config)
16:07yukaDMA_ZONE maybe
16:07yuka(config option)
16:08yuka*ZONE_DMA
16:08yukabut no
16:13yukait goes to dma_supported() -> dma_direct_supported() when then returns 0 because the mask is < "phys_to_dma_unencrypted(dev, min_mask)"
16:13yukamask=7fffffff, phys_to_dma_unencrypted(dev, min_mask)=103e08fffff
16:14yukait's difficult to tell if any of the code path should already have returned earlier, since it checks for many ways that dma would be supported
16:17jannauI don't think it should call dma_direct_supported()
16:17yukamore like use_dma_iommu?
16:18jannauuse_dma_iommu() should return true. do you have CONFIG_IOMMU_DMA=y?
16:18yukaI will find out why it doesn't return true
16:19jannauI suspect 'iommu-map*' in the pcie devicetree node
16:22jannausofus: which decoder are you using on linux? JVT-AVC_V1 passes only 28/135 with GStreamer-H.264-V4L2SL on t6020 here
16:28sofusGStreamer-H.264-V4L2SL. I just tested it again
16:29sofuswait i have some local changes
16:31sofusthat did not change anything
16:36lion328 has quit [Ping timeout: 480 seconds]
16:36jannaunot sure if checked if the h264 samples are complete. the server was blocking me
16:37sofustry using FFmpeg-H.264 as the decoder? easy way to check
16:37jannauI'm testing on FAR so that will not work at all
16:39sofuswhat is FAR?
16:39svenyay, xnu survives my first emulated genter and then immediately dies at gexit :D
16:39jannauFedora-Asahi-Remix
16:40sofusjust ffmpeg (software decoder)
16:42daMaestro has joined #asahi-dev
16:43daMaestro has quit []
16:48yukaOkay I was missing APPLE_ADMAC but it's still failing (but I can now re-test all the device tree variations again)
16:50jannausofus: samples weren't complete. I'll retest
16:51jannauyuka: can you pastebin your t8132 pcie nodes?
16:52svengah, probably something wrong with the SP. and here i thought i understood how GXF works :/
16:53svenbut, progress \o/ now just gotta figure out why the SP seems wrong after that gexit: TTY> hv_sprr[0]: first genter #0 from 0xfffffe0012b801a4 -> 0xfffffe00135a5a1c
16:53svenTTY> hv_sprr[0]: SPRR_PERM_EL1 0x2020a5053020f0f0 -> 0x2020a5063020f0e0
16:53svenTTY> hv_sprr[0]: first gexit -> 0xfffffe0012b801a4 (ASPSR 0x2)
16:55hdbngr has quit [Ping timeout: 480 seconds]
17:00lion328 has joined #asahi-dev
17:01jannauyuka: I think iommu-map should be https://paste.debian.net/hidden/937709b0
17:02jannaunote that you're overriding it in t8132-j773g.dts
17:02yukaI will try what you pasted, but I'm following t8103-j274.dts
17:02yukawhich one was it
17:03yukat8103-j457
17:03yuka*t8112-j473
17:04yukaI swear one of the mac minis had the port1 disabled, and then overrides the iommu-map like this
17:05yukait was one of the Pro variants
17:05yukat6020-j474s: port01 is unused, remove the PCIe sdhci0 node from t600x-j375.dtsi and adjust the iommu-map.
17:06hdbngr has joined #asahi-dev
17:08lion328 has quit [Ping timeout: 480 seconds]
17:10svendlmalloc: internal error 🙃
17:12yukait works!
17:13yukathe regs of the darts were wrong
17:13yukathe iommu-map was correct
17:13jannauyuka: looks like it could be ok. the decompiled devicetree might be interesting to spot unexpected errors. `dtc -I fs -O dts -o - /proc/device-tree | less -I`
17:14jannaunevermind
17:14yukait works until I set the network interface to up %)
17:14yukathen SError
17:15jannauSError in the dart mmio space or somewhere else?
17:15yukapcie config space
17:15yukaI had one earlier too which I was able to work around
17:15yukapcie config + 0x200068
17:15yukaearlier:
17:16yukapcie config + 0x44
17:26yukathese SErrors are actually async though so I don't know yet which write triggered it
17:28lion328 has joined #asahi-dev
17:45hdbngr has quit [Ping timeout: 480 seconds]
17:45hdbngr has joined #asahi-dev
17:49lion328 has quit [Ping timeout: 480 seconds]
18:00nnym has joined #asahi-dev
18:01jannausofus: h265 decoding seems completely busted on t6001. ~35 tests so far and only erros, fails and timeouts
18:02jannauh264 is at 77/135 with all samples present
18:02svenmaaaaybe i should've written this in rust. i'm somehow corrupting dlmalloc's heap :/
18:02lion328 has joined #asahi-dev
18:08jannauh265 end result 0/147 on t6001, translation faults in kernel log
18:08nnym has quit [Quit: quitting out | tjmnkrajyej@gmail.com]
18:09beastd has quit [Ping timeout: 480 seconds]
18:11nnym has joined #asahi-dev
18:13bennofs30 has joined #asahi-dev
18:13jannaut6001 vp9 at the "expected" 215/305
18:14bennofs3 has quit [Ping timeout: 480 seconds]
18:14bennofs30 is now known as bennofs3
18:18jannaut8103 h264 with all samples present is 76/135
18:31amarioguy has quit [Quit: Textual IRC Client: www.textualapp.com]
18:38jannauI've pushed asahi-7.1.5-1 to github
18:40jannauwithout M3 aop / and usb. I'll look in the next days after those. I wanted to get a 7.1 release out quickly
18:49yukathank you!
19:20yukaBtw, the reboot after 3 minutes issue I had earlier was fixed my marking ps_ans as always-on
19:20yukanot caused the wdt thankfully
19:22hdbngr has quit [Ping timeout: 480 seconds]
19:25svenOH FOR FUCK'S SAKE. why do i keep running into this cursed stuff that's been in our tree for ages: https://github.com/AsahiLinux/m1n1/pull/640
19:26svennow i get EL0 trying to use SIMD instructions to update its pagetables 9/10 tries so there's something terribly wrong, but 1/10 times this happens: https://f.svpe.de/21e408e29be8f1a75614159eec87752bd839eca0adadea8c5d1a9e75bef10559_cursed_progress.jpg
19:27chaos_princesssimd is not _that_ insane? i think xnu allows using neon, right?
19:27sventhe more insane thing is EL0 updating pagetables
19:28sveni've been at this since like 8am and just realized i didn't even eat lunch. time to call it a day :D
19:29hdbngr has joined #asahi-dev
19:38hdbngr has quit [Ping timeout: 480 seconds]
19:39voidah has joined #asahi-dev
19:45hdbngr has joined #asahi-dev
19:48sofusjannau: thanks. on t6030 h265 is also broken, the others work fine
19:54hdbngr has quit [Ping timeout: 480 seconds]
19:54jmax[m](checks time in Europe. almost 10pm.. oof!)
19:58sofusdid anyone ever test h265 on t8112
20:01hdbngr has joined #asahi-dev
20:10hdbngr has quit [Ping timeout: 480 seconds]
20:11sofusim expecting it to be broken also, this might fix it
20:13chadmed has quit [Quit: Konversation terminated!]
20:13chadmed has joined #asahi-dev
20:13Tachyonray[m]It’s on my list. I have a t8112, but still have to figure out how to install Asahi on it
20:14jannausofus: t8112 results h264: 77 passed, vp9: 216 passed 216, h265: probably 0 passed with asahi-wip, test still running
20:14chaos_princessthose are just supported by normal installer?
20:14picnoir1 has joined #asahi-dev
20:15hdbngr has joined #asahi-dev
20:16Tachyonray[m]sorry, have a t8122
20:18chaos_princessjust wait a bit then
20:18chaos_princessit is almost in a release-able state
20:19jannausofus: t8112 h265: 142/147 passed. pushed to asahi-wip
20:19sofusjannau: thanks i expect the same on t600x
20:20sofuskinda weird m3 does not work then
20:20picnoir has quit [Ping timeout: 480 seconds]
20:20sofuschaos_princess: what about gpu, is that not a lot of work?
20:20chaos_princessrelease will not have gpu
20:20chaos_princesstarget is everything supported by m1/2 except gpu and dcp
20:23jannausofus: t6001 looks better but it's quite slow
20:25jvoisin` has joined #asahi-dev
20:27jannaut6001-h265: 140/147 passed
20:27sofusyeah i noticed that to, vp9 feels like the slowest one, took ~60 seconds on my t6020 is t6001 much slower?
20:28jannaut6001-h265 took 324 seconds
20:28svenstill very slow and very unstable because EL0 still wants to update its own pagetables somehow, but sometimes i get far enough for this https://f.svpe.de/6a1342f428a1724f2055ed3dddb6896aaab4603551f8d8802966e4ab170253c7_cursed2.jpg
20:29yukaThis is emulating SPRR on a device which would already have it?
20:29chaos_princess1. awesome. 2. lol
20:29svenyuka: yup
20:29yukanice...
20:30chaos_princesssven: just in case - are you sure that it is el0 and not gl0
20:30svenessentially i patch every genter/gexit and a whole bunch of msr/mrs to hvc and emulate everything
20:30svenchaos_princess: oh... nope
20:30chaos_princessmaybe its txm?
20:30sveni don't think this old mac os release has txm yet
20:31svenxnu's ttbr0 and ttbr1 get a guarded and a non-guarded shadow pagetable each which right now is fully rewritten whenever anything in what xnu believes the real pagetable to be is touched
20:31svenwhich probably explains why booting takes like 10 minutes
20:31svenanyway, now it's really enough for today :D
20:32sveni emulated it on a m1 because it's easy to compare what should happen with the real HW vs. what my emulation does and because i don't have a m4 yet ;)
20:33svenand because it's just PPL and not the entire SPTM/TXM mess
20:36jannausofus: t6001-vp9 takes 317s but feels faster because it's twice as many test cases
20:41sofusvp9 is +15 sec for me, kinda weird its is less for you.
20:41sofusthe adt describes some clocks, i gotta look into that at some point
20:43sofusi plan to take a lot at stability/speed once av1 is good enough
20:46sofusthe av1 tests are actually super cool, i have used the "standart" ones up until now
20:48jannaudecoding speed for a long high resolution sample looked okay but I did not compare it to videotoolbox
20:51sofusiirc i tested a sample where videotoolbox ~60fps and linux+gstreamer was ~30, but i also need to do a propper test
20:53jannaucharging_end is broken in asahi-7.1.5-1 (I think the upstream driver dropped it). there will be asahi-7.1.5-2 in any case for the t600x/t8112 hevc fix. I'll look at it tomorrow
20:59sofusshould a dfu restore also change the system fw? the boot picker sometimes has the new symbols
21:00rhlvista[m]If you do revive
21:00sofusi did restore
21:00yukam1n1 should tell you the system fw?
21:01yukain the boot log
21:01sofusyes its also new, 18000.x
21:01Yushi_ has joined #asahi-dev
21:03sofus(all os fw are reported as 10151.x)
21:05Yushi has quit [Ping timeout: 480 seconds]
21:05voidah has quit [Read error: No route to host]
21:05Yushi_ is now known as Yushi
21:08yukaHmm, I was able to get M3 nvme working with 25.5 fw both in m1n1 and linux
21:08yukaand M4 too
21:09yukaI do not like that this involves commenting out "nvme_set_queue_count"
21:09yukaseems to work though
21:14yukaOn A18 Pro too
21:16voidah has joined #asahi-dev
21:19yukathe trace I have covers only the mmio address space and no the queues
21:19yukaand nvme_set_queue_count is a command via the admin queue
21:31voidah has quit [Ping timeout: 480 seconds]
21:31yukaskipping APPLE_ANS_LINEAR_SQ, APPLE_ANS_UNKNOWN_CTRL & nvme_set_queue_count is FW 26.x thing (needed on M3 with 26.x too)
21:32yukasplitting up the offsets of nvme and nvmmu into two regs is an M4+ thing
21:32yukaand nvme + 0x1208 / 0x1200 (still got to find names for them) is also an M4+ thing
21:40voidah has joined #asahi-dev
21:41hdbngr has quit [Ping timeout: 480 seconds]
21:46hdbngr has joined #asahi-dev
21:47voidah has quit [Read error: No route to host]
21:59pb17 has quit [Remote host closed the connection]
21:59pb17 has joined #asahi-dev
22:18hdbngr has quit [Remote host closed the connection]
22:18hdbngr has joined #asahi-dev
22:33hdbngr has quit [Ping timeout: 480 seconds]
22:37yukat8122 atcphy also works fine on t8140 (it has compatible t8130)
22:37yukaexcept that it has no usb4 tunables
22:37yukabut I guess the ace firmware is then programmed in a way that it does not try to negotiate usb4
22:37yukaactually...
22:38yukais that signed?
22:38ddxtanx has joined #asahi-dev
22:39chaos_princessi think there was a hack
22:39yukathere is also a hub in between and currently it only gets detected (via dwc3 reset) once I pug something into the rear port, since that's the only one I have fully described in the device tree
22:40yukaI'm not sure what is the best way to model it in the dt (I do not see the hubs on any of the other models in the dts, but those also do not have such a cursed setup)
22:40chaos_princessthere is a hub on mac studio on the front ports
22:40nnym has quit [Ping timeout: 480 seconds]
22:40yukaand also m4 mac mini I think?
22:40yukathe m4 mac mini only has 3 hpms
22:41yukabut 5 usb typec ports
22:41chaos_princessi think that has a pcie usb controller
22:41yukaAhh possible
22:41chaos_princesslike the m1 mac mini for type-a ports and the 4-port imacs
22:41yukaHuh I do not see anything xhci in the j773g adt
22:41hdbngr has joined #asahi-dev
22:42yukaand I did not see any xhci controller in the lspci listing
22:42yukait's a hub
22:42yukausb-drd2 has usb[23]-hub-port[12]
22:43yukaokay but then I look at the mac studio
22:43chaos_princessfrom what i remember on some other device that has a hub - we just have a thingy that enables a power gpio and that's all
22:44yukat8140 does not need anything to enable the hub
22:44yukait's there, but there is no obvious signal when it's booted
22:44yukaAh right
22:44yukawhat happens if no atcphy is connected?
22:45yukato dwc3
22:45yukathat's not possible with the apple glue driver?
22:45yukasorry I meant no hpm remote-port
22:45chaos_princessso, like, you can drive it in a usb2 mode somewhat, as that is using the eusb repeater and not atcphy
22:50voidah has joined #asahi-dev
22:50hdbngr has quit [Ping timeout: 480 seconds]
22:51nela has quit [Ping timeout: 480 seconds]
22:52yukaOk, so the question is: does dwc3 just keep running with the usb2 hub enumerated?
22:52yukaor do we shut it down when no cable is connected to either of the ports
22:52yukaI think this logic is missing in the dwc3-apple glue driver
22:52chaos_princesswe shut it down
22:53yukathe glue driver mentions something about DWC3 being shut down when no cable is connected
22:53chaos_princessbecause of the whole needing a reset for it to work reliably
22:53yukathe problem is that currently there is no proper way to describe in the adt that a cable plug on _either_ of the ports should reset/re-enumerate dwc3
22:54chaos_princessso, we have one dwc and two hpms?
22:54yukabut only usb2 goes over the hub
22:55chaos_princessi think you will have to have a fake mux or something in front of hpms
22:55JamesCalligeros[m]We are reaching think different levels that should not be possible
22:56chaos_princessso for example if you plug in a device into port 0 and it establishes a 2.0 link, then plug in something into port 1, the device on port 1 does not get reset
22:56yukawell port 1 is usb2 only
22:56yukaso it never really needs a reset
22:56chaos_princessright, but the atcphy driver does not know that
22:57chaos_princessso needs a fake mux in front
22:57yukawhen I wired port 1 to the (only) dwc3, it kinda worked as long as I first plugged in a device into port 1 before plugging a second device into port 2
22:58chaos_princessthat would make sense, as atcphy will bring up the link including the 2.0 lanes
22:58yukasince there is only one port which supports usb3, there is only one atcphy
22:58voidah has quit [Ping timeout: 480 seconds]
22:59hdbngr has joined #asahi-dev
23:00chaos_princesswhat does macos do if you plug in a 3.0 device into port 0, plug in whatever on port 1, then unplug port 0
23:00yukaI'm pretty sure they spent a lot of time making this work
23:00yukado you mean if the port 1 device briefly disconnects?
23:01chaos_princessi mean the 3.0 one disconnects
23:01chaos_princessthat should cause an atcphy reset
23:01yukayes, and if they keep up the usb 2.0 part during that
23:01chaos_princessand bring down the link to the device on the 2.0-only port
23:01yukapretty sure they do some magic to prevent that
23:02chaos_princesswell, we kinda want to know the trick behind the smoke and mirrors :P
23:02yukaI can try to get a trace for exactly this scenario
23:03jannauI don't think they have to reset the usb-2.0 link because the hub is always connected
23:03yukaexactly
23:03chaos_princessbut atcphy reset resets the entire dwc3 core
23:03chaos_princessincluding the 2.0 parts
23:03yukathat's what the linux driver does
23:04jannaubut iirc the reason for the dwc3 reset is to reset the eusb2 repeater
23:05yukaso can we introduce some usb-2-always-on mode?
23:05jannauwhich isn't necessary with the always connected usb2 hub
23:06nnym has joined #asahi-dev
23:08hdbngr has quit [Ping timeout: 480 seconds]
23:15nnym has quit [Ping timeout: 480 seconds]
23:15nnym has joined #asahi-dev
23:15yukaproblem is... the hub is not fast enough to boot
23:16yukaso when I just init the dwc3 on atcphy probe, it doesn't detect the hub, and there is no indication of when it's available that I know of
23:16yukaiirc the hub is also not listed in macos until you plug something into either slot, which speaks for the hpm mux
23:16hdbngr has joined #asahi-dev
23:17yukaalthough that will of course also break things which are plugged into the second port at boot
23:17jvoisin` has quit []
23:17yukaso there needs to be some other indication of the hub becoming available
23:17yuka. o O ( sleep 5; reset_dwc3 )
23:19jvoisin has joined #asahi-dev
23:20chadmedive rebased my 14.8.3 dcp kernel branch on latest asahi-wip and added the surface caps stuff which should fix the apple-shaped hole in the framebuffer
23:21chadmedstill cant figure out how to get it to actually reinitialise the dfb or null a surface but those are the only two problems left for internal display use. everything else works the same as before including promotion
23:21chadmedstill havent looked at avservice-related stuff either
23:23chadmedactually its too cold to go outside today so i might look at hdmi :P
23:25hdbngr has quit [Ping timeout: 480 seconds]
23:33hdbngr has joined #asahi-dev
23:44hdbngr has quit [Ping timeout: 480 seconds]
23:51hdbngr has joined #asahi-dev
…