#
#asahi-dev · 2026-05-28 15:01
yukayes
#asahi-dev · 2026-05-28 15:01
chaos_princessdid you manage to get macos to fully boot? i wasn't able to, it kept crashing
#asahi-dev · 2026-05-28 15:00
chaos_princessyep
#asahi-dev · 2026-05-28 15:00
yukaregs apcie[5-7] are not accessed at all
#asahi-dev · 2026-05-28 14:59
chaos_princessfuses no longer seem to be a thing past t602x
#asahi-dev · 2026-05-28 14:59
chaos_princesswell, not fuses, the other new calibrations
#asahi-dev · 2026-05-28 14:58
chaos_princessso, i strongly suspect fuses are only applied on sleep exit as, iboot already does it for us on cold boot
#asahi-dev · 2026-05-28 14:57
yukaas in mac os booted in the hv after a bit of fiddling and a few more tries
#asahi-dev · 2026-05-28 14:57
yukaThis thing pretty much worked: https://gist.github.com/IntegralPilot/85c355124fe54e670e6d5586c7a07826
#asahi-dev · 2026-05-28 14:55
yukaI traced pcie on M3 Max (the 14 core one, T6034 if it matters), but I saw none of the fuses stuff
#asahi-dev · 2026-05-28 14:48
chaos_princessobligatory check with macos, attach dmesg, probably grab the edid of the device you were trying to use.
#asahi-dev · 2026-05-28 14:42
sklugHey everyone. I am at Embedded Recipes in Nice and wanted to do a presentation using a M2 MBP with HDMI out. Unfortunately the HDMI detection failed and we ony got garbage on the output. After reboot it kind of worked, but I still saw rendering issues. Is there any debug data I could collect that might help in debugging the issue afterwards?
#asahi-dev · 2026-05-28 12:31
yukabut good enough
#asahi-dev · 2026-05-28 12:30
yukajust the shell is slightly broken then
#asahi-dev · 2026-05-28 12:30
yukaAh, this also works: 2>&1 | tee -i /path/to/log
#asahi-dev · 2026-05-28 12:24
yukanice, thanks
#asahi-dev · 2026-05-28 12:23
jannauthere's `-l` to log to files, note that not everything uses log. there are things only printed to the terminal
#asahi-dev · 2026-05-28 12:18
yukabut having traces go to another iodev would be nice
#asahi-dev · 2026-05-28 12:18
yukaI guess I'll need to switch over to something more fancy for this
#asahi-dev · 2026-05-28 12:17
yukaI use foot
#asahi-dev · 2026-05-28 12:16
chaos_princessboth Konsole and stock macos terminal support that
#asahi-dev · 2026-05-28 12:16
yukaAh well
#asahi-dev · 2026-05-28 12:15
chaos_princessi just set infinite scrollback
#asahi-dev · 2026-05-28 12:15
yukaand my scrollback buffer is too small to copy it afterwards
#asahi-dev · 2026-05-28 12:14
yukasince I have to interact with the shell to start it properly, and also piping the output somewhere doesn't seem to work
#asahi-dev · 2026-05-28 12:14
yukaHow do you all record/save your hv traces?
#asahi-dev · 2026-05-28 10:56
yukaBut if I managed to enter device mode in Mac OS, I would now try to focus the regs of the _repeater_ (I2C addr 0x21), and not the hub.
#asahi-dev · 2026-05-28 10:56
yukaFor some reason I can't properly dump the i2c regs using the m1n1 code. Maybe because it doesn't implement the correct i2c reading mode (I know too little about i2c...)
#asahi-dev · 2026-05-28 10:55
yukait is possible that apple's custom variant allows controlling it over i2c
#asahi-dev · 2026-05-28 10:55
yukathe device tree doesn't mention anything of a gpio
#asahi-dev · 2026-05-28 10:54
yukaOn TIUSB2E221 (which is a newer variant of TIUSB2E22 with i2c) the 2x2 crosspoint mux is controlled via a gpio
#asahi-dev · 2026-05-28 10:53
yukabut all variants of ti 2e22 usb repeaters I can find do have a 2x2 crosspoint mux
#asahi-dev · 2026-05-28 10:53
yukabut not the same, since the i2c register map seems to be completely different
#asahi-dev · 2026-05-28 10:52
yukaThe device tree says the usb-repeater is a "ticd2e22", ti = texas instruments, cd = custom design?, it's probably related to this one: https://www.ti.com/lit/ds/symlink/tusb2e221.pdf
#asahi-dev · 2026-05-28 10:51
yukaWe know the usb-hub is a variant of the Via VL122 (https://www.via-labs.com/product_show.php?id=126). While it is not super obvious, it seems unlikely to me now that the upstream/downstream ports of this hub can be reversed.
#asahi-dev · 2026-05-28 10:50
yukaa friend noted that the usb-repeater is likely responsible for bypassing the usb-hub
#asahi-dev · 2026-05-28 10:50
yukaRe macbook neo usb device mode
#asahi-dev · 2026-05-28 09:09
sofusleio: ha that actually makes sense! I have used verisilicon as reference (and copied a lot), i will look into rockchip also, thanks!
#asahi-dev · 2026-05-28 09:09
leio(I guess old codec meant like H.264 in that?)
#asahi-dev · 2026-05-28 09:08
leiothese were*; gave up*
#asahi-dev · 2026-05-28 09:07
leio on these chip concurrently to Apple giving up on them, don't think it was a coincidence, and since we know by now that the GPU is somewhat of a fork of Imagination design, its a possiblilty that the codecs are a fork of their design too; at least for the old codec, and how to compute auxiliary / scratch buffer sizes, that might be key
#asahi-dev · 2026-05-28 09:07
leiosofus: I've been told that if you look at eiln's stuff and compare against driver/media/platform/verisilicon or drivers/media/platform/rockchip decoder drivers, you'll find some strangely similar parameters and patterns, same kind of address and offsets, so you might find some inspiration from there. TI released some open source codec drivers for now abandonned Imagination CODEC; these where hybrid, you could do firmware access, or cpu access; Imagination gave on
#asahi-dev · 2026-05-28 09:07
sofusI will need to look into a few more things before i have a real answer
#asahi-dev · 2026-05-28 09:05
chaos_princessnot like the command stream is big, but still
#asahi-dev · 2026-05-28 09:05
chaos_princesswhy not use the (hopefully much faster) ap cores to generate the command stream?
#asahi-dev · 2026-05-28 09:04
sofusI would rather do it on the cm3, but i need to look more into the other codecs before i know if its possible
#asahi-dev · 2026-05-28 09:04
chaos_princessdrm/accel is actually kind of a perfect fit for this kind of thing - you have map/unmap memory, and "run a command buffer" api. Should be all you'd want
#asahi-dev · 2026-05-28 09:03
JamesCalligeros[m]Getting the m3 to do stuff isnt a bad idea either but i would be wary of NIHing a firmware interface if it’s not necessRy
#asahi-dev · 2026-05-28 09:03
JamesCalligeros[m]If you want to delegate responsibilities to userspace youd be better off skipping v4l2 and implementing a drm/accel device and using vk video
#asahi-dev · 2026-05-28 09:01
chaos_princessright, so, from my understanding v4l2 has all the parts in kernel, with no userspace part that can generate the instruction stream

plus ›