#
…

2026-07-04

00:01SternXD has quit [Remote host closed the connection]
00:04eConnah has quit [Quit: eConnah]
00:05maettu102 has quit [Quit: brb]
00:06maettu102 has joined #asahi-dev
00:08Mary has quit [Quit: .]
00:09Mary has joined #asahi-dev
00:09chadmed has joined #asahi-dev
00:16chadmedi think we can replace the "port + 6 + 1" magic number thingy in the mca syncgen enable sequence with #define CLUSTER_SYNCGEN_EN GENMASK(2, 0)
00:17chadmedi havent seen this get set to anything other than 0b111 yet
00:18chadmedand setting it to anything else breaks speakersafetyd/visense data anyway
01:04roundduckkira has joined #asahi-dev
01:23mischa85 has quit [Quit: Lost terminal]
01:53paulwalker_ has joined #asahi-dev
01:53paulwalker has quit [Read error: Connection reset by peer]
01:53paulwalker_ is now known as paulwalker
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:10paulwalker has quit [Ping timeout: 480 seconds]
02:24weirdtreething has joined #asahi-dev
03:39linuxgemini has joined #asahi-dev
03:59nora_ has joined #asahi-dev
04:00nora has quit [Ping timeout: 480 seconds]
04:04mattia013 has joined #asahi-dev
05:47valdo94[m] has joined #asahi-dev
06:08weirdtreething has quit [Remote host closed the connection]
06:08weirdtreething has joined #asahi-dev
06:30weirdtreething has quit [Remote host closed the connection]
06:30roundduckkira has quit [Quit: Konversation terminated!]
06:31weirdtreething has joined #asahi-dev
07:00weirdtreething has quit [Remote host closed the connection]
07:00weirdtreething has joined #asahi-dev
07:02StanfordAppliedCyber[m]i'm dealing with some strange agx problem right now, basically xnu is panicing because work isn't being done. i see work currently enqueued, but clearly the gpu isn't doing anything
07:03StanfordAppliedCyber[m]the strange part is that the firmware definitely hasn't printed any crash message
07:03StanfordAppliedCyber[m]i don't want to say it's for sure alive, because after all i can't get it to do anything, but i can't say if it's just hanging on something or dead and didn't print a crash for some reason
07:05StanfordAppliedCyber[m]i'm confident it's not my uat emulation, the only potentially concerning thing is the cache flush instructions i mentioned earlier (they're present in some uat endpoints, but these endpoints are never actually exercised) but i don't think that's the problem
07:05StanfordAppliedCyber[m]any suggestions on how to figure out what's going on here?
07:38starcatmeow has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
07:39Sobek[m](Isn't it getting late in your part ? Perhaps you're running out of computron, and should go sleep. The answer may be revealed to you in a Dream, and if not, a fresh batch on computron will help. (Credits to be given to Prof Dave Eckhardt from Carnegie Mellon))
07:42yukawhat is uat again?
07:43yukaah, some gpu thing
07:44StanfordAppliedCyber[m]uat is the gpu's iommu
07:44StanfordAppliedCyber[m]fwiw m1n1's uat implementation is pretty close to sptm
07:47StanfordAppliedCyber[m]<Sobek[m]> "(Isn't it getting late in your..." <- i only started working 3 hours ago and it's currently only 1am
07:47StanfordAppliedCyber[m]that's super early :)
08:03StanfordAppliedCyber[m]it appears that agx is asleep waiting for an interrupt
08:07LarstiQ has quit [Quit: ZNC - http://znc.in]
08:13LarstiQ has joined #asahi-dev
08:39picnoir has quit [Ping timeout: 480 seconds]
08:47picnoir has joined #asahi-dev
08:48ravikant_ has joined #asahi-dev
08:48ravikant_ has quit []
09:09yukaI really want to know what happens if you connect two macbook neos with a usb cable and try to airdrop something
09:09yukaI think it will not be able to use the cable connection
09:10yukaI tried this a while ago: Some Mac <-> USB-Hub <-USB-cable-> Macbook Neo
09:11yukaand nothing happened
09:14StanfordAppliedCyber[m]ive never gotten mac mini m4 to mac neo airdrop working
09:15StanfordAppliedCyber[m]and for some reason since i've done that test my apple id got fucked up so even regular airdrop doesn't work
09:15StanfordAppliedCyber[m]so i now just host a python http server on my mac mini and wget binaries from that onto the neo
09:45yukaI take suggestions (both on the contents, and on which category of the wiki this belongs in)
09:49snooze6214[m]yuka: we still don't know how to toggle the mux right?
09:49yukathat's why there's "???" in the picture
09:50snooze6214[m]Right.... Also, can kisd run on non-apple silicon hosts too?
09:50yukait just does normal usb bulk transfer
09:50snooze6214[m]Ah neat, thanks
09:54yukadebugusb is not quite as high-speed as cdc-acm / usb device mode
09:54yukabut it's generally more useful (you will be able to see linux boot logs over it in the future)
09:56yukaand we are now able to make m1n1 put the device into debugusb mode, so you don't need a central scrutinizer or other mac to send the usb-pd vendor defined message for that
09:56yukaI'll now try integralpilot's dockchannel uart driver again :)
09:57yukaand figure out why it's crashing
10:10mripard has quit [Quit: WeeChat 4.9.2]
11:11yukaThe only device I can test dockchannel-uart on right now is the Neo with 26.6b3, and I'm pretty sure it's getting the SError while reading DATA_TX_FREE
11:12yukaL2C_ERR_ADR is 0x2804c030812c004. 30812c000 is the base and +4 is DATA_TX_FREE
11:14yukaso probably the same problem as with the pmgr yesterday
11:14yukaand pc=arch_counter_get_cntpct was a red herring because whatever pmgr access it did before actually caused the serror
11:14yukaI keep forgetting which kinds of errors are delivered with delay
11:15yuka(the dockchannel-uart crash is now showing pc=raw_spin_unlock_irqrestore which is done just after the DATA_TX_FREE read)
11:35sventhat really sounds like they messed some chicken bits up or something similar
11:36svenwell, that or somehow power is turned off for random peripherals because something else is missing
11:36yukaaccesses to pmgr itself are experiencing this
11:36yukathis is on the boot core btw
11:37yukathe dockchannel uart works in principle as long as the blocks are bigger than the fifo size
11:38yukaonce I send a smaller batch, it crashes
11:39yukaeither direction fwiw
11:40yukawhoops
11:40yukathe full crashes btw
11:44yukathis was the same crash behavior of the dockchannel-uart driver as on enverbalalic's t6041
11:45yukajust that I was able to get logs for it
11:45yukaI don't know which iboot/firmware version they were using at that point
11:46sveni'm not even sure how we could file a feedback report for this tbh :/
11:46svenprobably needs a small binary that reproduces it
11:46eConnah has joined #asahi-dev
11:47yukafirst, I want to figure out if it happens on 26.5.1, and 26.6b1/26.6b2
11:47svenmakes sense
11:47yukabut again, don't have physical access to the machine at this point
11:47* svenreally needs to get his hands on a m4 or m5 as well
11:48svenanyway, first need to finish this psci-via-efi crap. or well, fix it after i broke something after cleaning up :/
11:50eConnah has quit []
12:35flokliMmh, did someone try to do a DFU restore /through/ the CS (its passthrough port)? it does briefly probe, but then the CS seems to make it do a hard reset and reboot again (into DFU mode), causing it to get to loop in there...
12:38chadmedif we want to take the ivsense stuff out of the speakersafetyd config files we're gonna have to do some pretty gross stuff in code or make some unpleasant changes to macaudio and the ucm conf files
12:39chadmedi think hiding the ugliness in speakersafetyd is the path of least resistance right now. the problem is that the desktops dont have sound-name-prefix
12:47yurix has quit [Read error: Connection reset by peer]
12:49yurix has joined #asahi-dev
12:50flokli(trying to make this remotely DFU'able, with a KVM and some remote button presser. cs-fw loops with https://gist.github.com/flokli/e3209cd9a7611cbf4122b104382885a2 )
12:53runxiyu_ has quit [Remote host closed the connection]
12:57first_thunder022 has joined #asahi-dev
12:58first_thunder02 has quit [Read error: Connection reset by peer]
12:58first_thunder022 is now known as first_thunder02
13:02runxiyu_ has joined #asahi-dev
13:06first_thunder02 has quit [Read error: Connection reset by peer]
13:08yukaintegralpilot: did the dockchannel driver interface change significantly to the downstream one? I'm wondering if I can cherry-pick the dockchannel-uart on top of asahi-wip for testing.
13:08first_thunder02 has joined #asahi-dev
13:10integralpilot has joined #asahi-dev
13:11integralpilotyuka: you cant sorry - it did change significantly because it was rewritten as a mailbox for upstream
13:11integralpilotif you cherry pick the commits that add the mailbox driver as well, then that mgith work
14:16MikeM64 has joined #asahi-dev
14:33integralpilot has quit [Quit: integralpilot]
14:50jeroenflvr has joined #asahi-dev
14:52jeroenflvr has quit []
15:18sven[ 0.026227] smp: Bringing up secondary CPUs ...
15:18sven[ 0.027371] CPU1: Booted secondary processor 0x0000000001 [0x611f0221]
15:18svenPSCI: CPU_ON: releasing CPU 1 (mpidr=0x1) to 0x818bf5510
15:18svenso that works. cpuidle doesn't quite like what i do yet though
15:24svenah, lol, because i don't define any idle states in the dt
15:48ansg191 has quit [Remote host closed the connection]
15:57ansg191 has joined #asahi-dev
15:59ansg191_ has joined #asahi-dev
15:59ansg191_ has quit [Remote host closed the connection]
16:06ConnorAlecks[m] has left #asahi-dev [User left]
16:07ConnorAlecks[m] has joined #asahi-dev
16:08yukathe new upstream-proposed dockchannel will be fun to coordinate with u-boot
16:09yukaI guess it uses a new compatible
16:10mattia013 has quit [Remote host closed the connection]
16:10yukaI guess u-boot and the device trees ship together
16:11yukabut once the device trees get updated, old downstream kernels won't be able to use the keyboard anymore
16:21sventhat’s what we get for keeping this downstream for so long :(
16:23yukaThis is new: apple-dockchannel 2a050c00.mailbox: error -ENXIO: IRQ index 0 not found
16:45yukaintegralpilot: do you see anything suspicious in these device tree nodes?
16:49yukahttps://x0.at/iG1K <- the dmesg of what dockchannel-hid does on my machine
16:49yuka(keyboard works)
16:49yukaThough I'm more interested in the tty right now
16:55yukaOr maybe it's some timing thing? This is in a big distro kernel with everything built as modules
17:06yukatypo in the device tree node
17:11yukadockchannel-uart crashes on m3
17:11yukaSError, but different stacktrace
17:11yukaI will try to fix it now
17:45mtimothy[m] has joined #asahi-dev
17:52mischa85 has joined #asahi-dev
17:56yukayesterday when dockchannel-uart was crashing on the Neo, I said 0x2804c030812c004 looked like DATA_TX_FREE
17:56yukabut that's bogus
17:56yukaif anything, it's TX8
18:22chaos_princesswebcam works on t6031, the changes are that the pmp message no longer exists, and the irq enable is now at mbox+8 instead of mbox+4
19:07coroner21 has joined #asahi-dev
19:52Lyude has quit [Read error: Connection reset by peer]
19:52Lyude has joined #asahi-dev
20:00yukaI think I might have found the culprit re dockchannel-uart
20:04yukathe different stack trace was just a timing thing of when the serror arrives
20:07coroner21 has quit [Remote host closed the connection]
20:07yukathat's it, it now works on the Neo too, and I have to check what's up with pmgr again
20:07yukathe issue was writeb_relaxed on the TX8
20:07yukawhen it should be writel_relaxed
20:07yukaI'll write a response on the list now to preserve the context
20:22oolong has joined #asahi-dev
20:22oolong has quit []
20:56Yushi has quit [Remote host closed the connection]
20:57Yushi has joined #asahi-dev
21:02Yushi_ has joined #asahi-dev
21:04coroner21 has joined #asahi-dev
21:05Yushi has quit [Ping timeout: 480 seconds]
21:05revadam has joined #asahi-dev
21:06Yushi_ is now known as Yushi
21:06revadam has quit []
21:14yukabooted t8140 with pmgr again
21:14yukawith 26.6b3 iboot
21:15yukathe difference is: we are no longer allowed to write the pmgr state of ps_disp0_cpu and ps_dispext0_cpu
21:15yukareproduced in m1n1
21:15yukathis produced the serrors
21:38nela has quit [Ping timeout: 480 seconds]
21:41nela has joined #asahi-dev
21:46leah has quit [Ping timeout: 480 seconds]
21:50chadmedchaos_princess: nice!
22:03StanfordAppliedCyber[m]is this a change vs 26.6b2?
22:04yuka26.6b?
22:05yukaas in, I do not know which of the betas introduced it
22:05yuka26.5.x was the last one I tested
22:05yukawhere writing ps_disp_cpu didn't serror yet
22:06StanfordAppliedCyber[m]ack yeah i just read the context
22:08yukaSo, when I implement earlycon for dockchannel uart, I would somehow obtain the data reg of the linked mbox node, and then write to it directly?
22:08JamesCalligeros[m]If you think this is a bug please file a radar or something
22:08yukaI don't think it is
22:09yukafrom what I could tell from playing with disp_cpu previously (at least without knowledge of dcp stuff), there is no point in touching it
22:11yukaI have now seen several times on the macbook pros when playing with disp power domains that the picture on the screen disappears (the lcd is in all-black/opaque state) but the backlight zones still retain the image that was previously shown. and re-enabling the power domains does not bring it back
22:12yukaalso from what I can tell runtime power management for at least the framebuffer device and as such ps_disp0_cpu is nonexisting or broken for these devices anyways
22:13yukawhen we start touching dcp and trace macOS, and we determine macOS tries to power off the ps_disp0_cpu in some case, we can still file a radar
22:13chaos_princessthe case is most likely a sleep cycle
22:13yukabut this is far far away
22:13StanfordAppliedCyber[m]if you give me a trace range i'd be happy to get you one as soon as i am able
22:14yukapmgr.get_reg(0)
22:14StanfordAppliedCyber[m]i can get traces right now of everything up to and including initdata for the agx
22:14StanfordAppliedCyber[m]and some transactions after
22:14StanfordAppliedCyber[m]ack, i'll get that sometime eventually
22:14yukaI don't think it makes sense right now
22:14yukabecause what we want to observe is sleep behavior or maybe closing the lid
22:15yukabut this means it should probably already be at the login screen at least
22:15yukaanyways I'm leaving for today
22:15StanfordAppliedCyber[m]yeah fair, not quite there yet, this accursed gpu
22:16yukathis may or may not be useful in its current state
22:16yukabut the commit messages are to some extent documentation of the state and some of the things I ran into
22:44enverbalalic has joined #asahi-dev
22:45enverbalalicyuka: which issue of uart does writel fix ? if it's the hanging after 1 message part, that's still present on 6041
23:02enverbalalic has quit [Quit: enverbalalic]
23:05croissant` has joined #asahi-dev
23:07leah has joined #asahi-dev
23:12croissant has quit [Ping timeout: 480 seconds]
23:15hdbngr_ has quit [Ping timeout: 480 seconds]
23:24hdbngr has joined #asahi-dev
23:32hdbngr has quit [Ping timeout: 480 seconds]
23:41hdbngr has joined #asahi-dev
23:45luma has quit [Ping timeout: 480 seconds]
23:49hdbngr has quit [Ping timeout: 480 seconds]
23:58hdbngr has joined #asahi-dev
…