#asahi-dev · 2026-01-19 19:38
jannausven: atc + cd321x trace is sufficient for the optical cable#asahi-dev · 2026-01-19 15:07
noopwafeljannau: ><, thanks!#asahi-dev · 2026-01-19 14:58
jannausee a8000096874b using "select" because other KMS driver use it#asahi-dev · 2026-01-19 14:54
jannaunoopwafel: I think I added that dependency (or select) in the last days#asahi-dev · 2026-01-19 14:25
noopwafelin kernel: DRM_APPLE should depend on BACKLIGHT_CLASS_DEVICE due to dcp_backlight? or new config depending on both?#asahi-dev · 2026-01-19 12:54
noopwafelthe hv console, for 'bt' etc - I realized I can feed macho to run_guest and then maybe it only needs the symbol base set, and then got distracted#asahi-dev · 2026-01-19 12:52
jnni won't have an answer, but i'm curious: which piece of code should consume the symbols?#asahi-dev · 2026-01-19 12:50
noopwafelI guess this is kind of a corner-case which no-one cares about for more than 5 minutes#asahi-dev · 2026-01-19 12:44
noopwafelis there any simple way to get symbols for m1n1 itself, when running m1n1 as guest?#asahi-dev · 2026-01-19 00:02
fl0_id___could also put up the whole log, but only if you think there might be sth interesting#asahi-dev · 2026-01-18 23:47
fl0_id___guest#asahi-dev · 2026-01-18 23:47
noopwafelyou're booting with linux.py or as a guest? (I guess the latter?)#asahi-dev · 2026-01-18 23:40
fl0_id___but smt in kernel is still enabled#asahi-dev · 2026-01-18 23:39
fl0_id___noopwafel, based on device tree and m1n1 output when loading kernel, I'm only using 1 core (cpu 4, first p core iirc)#asahi-dev · 2026-01-18 23:38
fl0_id___https://www.pastery.net/mkkeuj/#asahi-dev · 2026-01-18 23:37
noopwafelare you booting without smt? (have a link to backtrace?)#asahi-dev · 2026-01-18 23:35
fl0_id___right that makes sense. but as someone suggested memory layout, that could it be anything else there? on the last experiments, got a more specific backtrackt, where it crashed when printing sth.#asahi-dev · 2026-01-18 23:25
jannaudoesn't really explain any issues in linux though as the base m3 boots fine. m1n1 handles it automatically#asahi-dev · 2026-01-18 23:23
jannaufor the base m3 as well, I haven't noticed that before#asahi-dev · 2026-01-18 23:22
jannaufl0_id___: it's expected for m3 pro/max. dram starts at 0x10000000000 as noopwafel said#asahi-dev · 2026-01-18 23:15
noopwafel(on my phone so anything I say is more suspect than usual but maybe worth a try)#asahi-dev · 2026-01-18 23:13
noopwafelis that an m3 part? the dram base is likely at 0x10000000000 not 0x800000000#asahi-dev · 2026-01-18 22:49
fl0_id___s/resets/restarts/#asahi-dev · 2026-01-18 22:48
fl0_id___https://www.pastery.net/hbgbry/#asahi-dev · 2026-01-18 22:48
fl0_id___in the sense of confirming whether I have a memory problem - should memdump.py in experiments not just work? it crashes almost immediately and resets the proxy.#asahi-dev · 2026-01-18 22:01
Shizraise*#asahi-dev · 2026-01-18 22:00
noopwafelI did 'break "lol"' and it worked, and maybe it is difficult to be more elegant than this#asahi-dev · 2026-01-18 22:00
noopwafelis there an obvious way to break 'properly' into the shell from a python-side hook?#asahi-dev · 2026-01-18 20:37
GlanzmannI remember when we had to trace mmio in FAUmachine. It took us 6 context switches. We than used a JIT to replace the offending instruction into a HV call.#asahi-dev · 2026-01-18 20:36
svensync mode where every MMIO takes a round trip to Python is very slow indeed #asahi-dev · 2026-01-18 20:35
GlanzmannWow. Nice.#asahi-dev · 2026-01-18 20:34
svenit supports async mode where it just takes the exception, logs the access to a buffer and then erets to the guest and only bother with the entire pipeline to Python every now and then #asahi-dev · 2026-01-18 20:34
svenyeah#asahi-dev · 2026-01-18 20:29
Glanzmannsven: Can the HV trace mmio without adding to much latency so that usb does not break?#asahi-dev · 2026-01-18 20:24
GlanzmannI see, so in order to debug this you need to trace under hv, add some debug code to your driver and compare and hope to find the difference. Or sprinkle your working code and the broken one with mmio debug and compare. Not fun.#asahi-dev · 2026-01-18 20:22
svenprobably wrong order somewhere#asahi-dev · 2026-01-18 20:21
svensomething wrong with the atcphy mmio pokes#asahi-dev · 2026-01-18 20:21
Glanzmannsven: Do you think it is a bug in the code or do you run in some race condition?#asahi-dev · 2026-01-18 20:21
Glanzmannsven: And thank you for making the defconfig Glanzmann safe.#asahi-dev · 2026-01-18 20:18
sveni shouldn't have jinxed it by saying i was gonna submit that tbt rfc this weekend :D#asahi-dev · 2026-01-18 20:17
svensums up my feelings about this very well ;)#asahi-dev · 2026-01-18 20:17
Glanzmannsven: :-(#asahi-dev · 2026-01-18 20:05
svenand on the first plug with the old code nothing happens as well#asahi-dev · 2026-01-18 20:05
svenwith the new code nothing happens so I think there's no negotiation happening at all#asahi-dev · 2026-01-18 20:04
svenwhen i set the wrong cable info with the old code my dock starts blinking because it's likely trying and failing to negotiate the connection again and again#asahi-dev · 2026-01-18 20:04
svenyeah, definitely something wrong with atcphy#asahi-dev · 2026-01-18 18:45
sveni guess the current code just always run into what only happened on the first plug there#asahi-dev · 2026-01-18 18:44
svenokay, pretty sure it's some atcphy thing. the old branch the worked showed the same issue on the first plug event, and only the second one works#asahi-dev · 2026-01-18 17:18
fl0_id___a few days ago I think someone suggested that the memory layout might be different on m3 max. should this be visible in adts?#asahi-dev · 2026-01-18 16:34
jannauthe initial baudrate is 115200. if the proxyclent doesn't receive a reply it assumes it has already switched to 1.5 Mb