#
#asahi-dev · 2026-01-10 15:11
fl0_id__sven, yeah, that's what I thought. I have the memory node, and ok that's what I thought that it gets filled in. Just kinda trying to brainstorm a bit what else to try
#asahi-dev · 2026-01-10 14:25
jannauremoving nonposted-mmio from mtp_dockchannel works for me on the 13-inch m2 mbp
#asahi-dev · 2026-01-10 14:13
integralpilotalso if you try it and doesnt work on M2 lmk, i have a theory why it might not but can confirm works on M3
#asahi-dev · 2026-01-10 14:12
integralpilotim going to log off now, if t8122 dt is sill wanted lmk i can send in morning.
#asahi-dev · 2026-01-10 14:11
integralpilotso that doesnt fix the need for nonposted-mmio
#asahi-dev · 2026-01-10 14:11
integralpilothavent seen that but did exact same thing in my refactoring for upstream
#asahi-dev · 2026-01-10 14:10
jannaufor dockchannel/mtp
#asahi-dev · 2026-01-10 14:10
jannau4398719b8c5fad might fix it if you haven't seen it
#asahi-dev · 2026-01-10 14:08
sveni have the same issue with the current tbt code fwiw, but haven't bothered to track down why yet
#asahi-dev · 2026-01-10 14:07
jannaumaybe parent is absent due to creating the plaform device manually
#asahi-dev · 2026-01-10 14:07
integralpilotmight have made a mistake in testing or modified other things inbetween
#asahi-dev · 2026-01-10 14:07
chaos_princessdo you have full dts? post it somewhere
#asahi-dev · 2026-01-10 14:07
integralpilotM3 baseline
#asahi-dev · 2026-01-10 14:07
integralpilothmmm idk
#asahi-dev · 2026-01-10 14:07
chaos_princesswhich machine btw?
#asahi-dev · 2026-01-10 14:06
jannauso that should still work
#asahi-dev · 2026-01-10 14:06
integralpilotit definetly needs it's own one and doesn't infer
#asahi-dev · 2026-01-10 14:05
jannauof_mmio_is_nonposted() recurses to the parents
#asahi-dev · 2026-01-10 14:05
integralpilotthat's what i'm trying to know too
#asahi-dev · 2026-01-10 14:05
chaos_princessmy question was more of a "shouldn't it infer from parent, why does it have it's own one"
#asahi-dev · 2026-01-10 14:05
integralpilotbut just tested again and it is definetly needed
#asahi-dev · 2026-01-10 14:04
integralpilotanyway, have to log off now but will do more research in the morning
#asahi-dev · 2026-01-10 14:03
integralpiloti think only marcan knows as they are the one who originally wrote dockchannel nodes with the (apparently needed) nonposted-mmio
#asahi-dev · 2026-01-10 14:03
integralpiloti was researching why now, to give an answer to chaos_princess
#asahi-dev · 2026-01-10 14:03
integralpilot(removed from the dockchannel node)
#asahi-dev · 2026-01-10 14:02
integralpilotit doesn't, i removed it and it breaks.
#asahi-dev · 2026-01-10 14:01
jannaushouldn't it infer nonposted-mmio from the parent soc node?
#asahi-dev · 2026-01-10 13:56
svenyou should just need a memory node filled by m1n1 with the actual memory size
#asahi-dev · 2026-01-10 13:56
sven(for the soc bus at least)
#asahi-dev · 2026-01-10 13:55
sven{address,size}-cells = 2 is correct, we have 64bit addresses on these systems
#asahi-dev · 2026-01-10 13:55
sventhe adt lies, don't trust it
#asahi-dev · 2026-01-10 13:54
fl0_id__(or at least not minimal ones/those i looked at)
#asahi-dev · 2026-01-10 13:54
fl0_id__any hints welcome... in the original adt there is like reserved_memory, but non of th dts rly have that
#asahi-dev · 2026-01-10 13:53
fl0_id__adjusted adsress-cells / size-cells for cpu, as my adt had these as 1 and 0 as opposed to what integral suggested. but didn't seem to change anything for memory. are there any parts of such a minimal dt that matter much for memory?
#asahi-dev · 2026-01-10 13:50
chaos_princessremind me, why does it have nonposted-mmio?
#asahi-dev · 2026-01-10 13:50
integralpilotthank you for your time!
#asahi-dev · 2026-01-10 13:49
svennice, i'll take a look tomorrow
#asahi-dev · 2026-01-10 13:49
integralpilothttps://github.com/IntegralPilot/linux-next/tree/apple-mtp-testing
#asahi-dev · 2026-01-10 13:49
integralpilotshould be live now at:
#asahi-dev · 2026-01-10 13:49
integralpilotthanks, i sorted it out!
#asahi-dev · 2026-01-10 13:47
svenyeah, you should be able to only push the changed objects which shouldn't be much if you fork torvalds/linux.git on github first
#asahi-dev · 2026-01-10 13:39
leiowouldn't a fork in github have most of the objects already there and not need to push it all?
#asahi-dev · 2026-01-10 13:38
chadmedgiven there are dt changes its probably worth an initial off-list review
#asahi-dev · 2026-01-10 13:36
integralpilotdoes anyone want to look at DockChannel/MTP stuff first in a branch or do i just send it? it's taking FOREVER on my slow WiFi for it to push the entire linux history to github so i can publish a branch
#asahi-dev · 2026-01-10 13:34
fl0_id__though rly don't know enough about memory stuff. I also modified the dt a bit and gonna try that next
#asahi-dev · 2026-01-10 13:34
fl0_id__ok, bisected it a bit further, it stops inside for_each_zone_zonelist(zone, z, zonelist, offset) {...
#asahi-dev · 2026-01-10 13:28
fl0_id__btw it's a hv problem if when dropping to hv shell it drops, but not before giving an indexError, right? such vector, target = self.vectors[idx], coming from handle_exception, handle_sync
#asahi-dev · 2026-01-10 13:25
fl0_id__also rly interesting how some stuff just gets optimized away, or maybe placed elsewhere... and then I end up with several hvc instructions after each other lol
#asahi-dev · 2026-01-10 11:43
chadmedsome things are really obvious of course, like the complaints about smc fourccs being mixed case "macros" or whatever
#asahi-dev · 2026-01-10 11:42
svenyup, this varies by subsystem. most are reasonable but there are a few exceptions ;)

plus ›