#
…

2026-06-28

00:05a4acg has quit [Ping timeout: 480 seconds]
00:06hdbngr has joined #asahi-dev
00:14hdbngr has quit [Ping timeout: 480 seconds]
00:15gargantua has quit []
00:19chadmed has quit [Read error: Connection reset by peer]
00:22hdbngr has joined #asahi-dev
00:23chadmed has joined #asahi-dev
00:30hdbngr has quit [Ping timeout: 480 seconds]
00:33chadmedsofus: i think maybe the reg containers in the t8112 dtsi are incorrectly sized. the adt reckons we have 0x1404000 at 0x268000000 but we're only mapping 0x1048000. whatever is in what we're calling sram is code.
00:38hdbngr has joined #asahi-dev
00:43chadmed has quit [Quit: Konversation terminated!]
00:46hdbngr has quit [Ping timeout: 480 seconds]
00:54hdbngr has joined #asahi-dev
00:54gargantua has joined #asahi-dev
01:00marvin24 has quit [Ping timeout: 480 seconds]
01:09hdbngr has quit [Ping timeout: 480 seconds]
01:17hdbngr has joined #asahi-dev
01:26hdbngr has quit [Ping timeout: 480 seconds]
01:35hdbngr has joined #asahi-dev
01:44hdbngr has quit [Ping timeout: 480 seconds]
01:51hdbngr has joined #asahi-dev
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:00hdbngr has quit [Ping timeout: 480 seconds]
02:07hdbngr has joined #asahi-dev
02:07gargantu1 has joined #asahi-dev
02:12gargantua has quit [Ping timeout: 480 seconds]
02:17hdbngr has quit [Ping timeout: 480 seconds]
02:19Ariadne has quit [Read error: Connection reset by peer]
02:20tobhe_ has joined #asahi-dev
02:25hdbngr has joined #asahi-dev
02:27tobhe has quit [Ping timeout: 480 seconds]
02:28Ariadne has joined #asahi-dev
02:34hdbngr has quit [Ping timeout: 480 seconds]
02:42hdbngr has joined #asahi-dev
02:51hdbngr has quit [Ping timeout: 480 seconds]
02:58hdbngr has joined #asahi-dev
03:05ddxtanx has joined #asahi-dev
03:07hdbngr has quit [Ping timeout: 480 seconds]
03:09ddxtanx has quit [Remote host closed the connection]
03:11nora_ has joined #asahi-dev
03:12ralim8 has quit [Quit: The Lounge - https://thelounge.chat]
03:12ralim8 has joined #asahi-dev
03:13nora has quit [Ping timeout: 480 seconds]
03:13ddxtanx has joined #asahi-dev
03:15hdbngr has joined #asahi-dev
03:17chadmed has joined #asahi-dev
03:20cow123 has quit [Read error: No route to host]
03:22cow123 has joined #asahi-dev
03:23hdbngr has quit [Ping timeout: 480 seconds]
03:25pb17 has quit [Ping timeout: 480 seconds]
03:29pb17 has joined #asahi-dev
03:32hdbngr has joined #asahi-dev
03:36chadmed has quit [Remote host closed the connection]
03:42hdbngr has quit [Ping timeout: 480 seconds]
03:44marvin24 has joined #asahi-dev
03:45hardy94 has quit [Quit: hardy94]
03:50hdbngr has joined #asahi-dev
03:59hdbngr has quit [Ping timeout: 480 seconds]
04:08chadmed has joined #asahi-dev
04:08hdbngr has joined #asahi-dev
04:21a4acg has joined #asahi-dev
04:23a4acg has quit [Remote host closed the connection]
04:23a4acg has joined #asahi-dev
04:24a4acg has quit [Remote host closed the connection]
04:25hdbngr has quit [Ping timeout: 480 seconds]
04:32hdbngr has joined #asahi-dev
04:41hdbngr has quit [Ping timeout: 480 seconds]
04:50hdbngr has joined #asahi-dev
04:58hdbngr has quit [Ping timeout: 480 seconds]
05:06hdbngr has joined #asahi-dev
05:37hdbngr has quit [Ping timeout: 480 seconds]
05:43hdbngr has joined #asahi-dev
05:46sofus has joined #asahi-dev
05:50sofuschadmed: the sizes look correct, it works on my t6020
05:52JamesCalligeros[m]I tried writing obvious values to various places in ram and scanning code and sram for it but it was in neither
05:52JamesCalligeros[m]No big deal, not important
05:52sofusif you share your code i would be happy to take a look
05:52integralpilot has joined #asahi-dev
05:52integralpilot has quit [Remote host closed the connection]
05:53hdbngr has quit [Ping timeout: 480 seconds]
05:53JamesCalligeros[m]Its just the syslog thing with the pointer to the buffer at various places in memory. I just had a thought though and will check something when i get home
06:01hdbngr has joined #asahi-dev
06:10hdbngr has quit [Ping timeout: 480 seconds]
06:11yukaHuh, I would've expected some PMPTracer output during operation when I turn off/on the wifi / display / ...
06:11yukabut it's all been during boot
06:14gargantu1 has quit [Ping timeout: 480 seconds]
06:17hdbngr has joined #asahi-dev
06:25hdbngr has quit [Ping timeout: 480 seconds]
06:31ralim8 has quit [Quit: The Lounge - https://thelounge.chat]
06:32yukaAn idea regarding the M4 "kernel write to read-only memory":
06:33yukaamcc-ctrr-a and amcc-ctrr-b seem to be enabled, write-disabled, and locked
06:33hdbngr has joined #asahi-dev
06:33yukaI can't quite parse the lower-limit and upper-limit yet
06:34chaos_princesswifi does not connect to pmp. And you should get constant gpu on/off spam. if you are expecting stuff over rtkit - that is not how it works, all power state communication is via DEV_STATUS_TGT_WR and such
06:34nickchan-there's also c and d on m4
06:34nickchan- has quit [Quit: ZNC 1.9.1+deb2+b3 - https://znc.in]
06:34nickchan has joined #asahi-dev
06:34yukachaos_princess: so the PMPTracer traces the iokit part only, and if I want to see the power state communication I add another trace_device()?
06:35yukanickchan: yes but c and d are not enabled
06:35yuka(and not locked)
06:35chaos_princessyes, you figure out which pmgr mmio is the correct one, and put PMPRegTracer on it
06:36yukaAhh makes sense
06:36nickchanif you are talking about the amcc part and not cpu register part write to protected region should come up as either a SError or amcc interrupt
06:37nickchanand not some sync exception
06:37chaos_princess"pmgr-debug=0xffffffff pmgr-debug-devices=0xffffffff pmgr-debug-always-on=0xffffffff pmgr-debug-energy-model=0xffffffff pmgr-debug-restore-counter=1 pmgr-trace=1 -disable_mtx_chk" is what you will need for that. And make sure you have i_can_has_debugger enabled
06:37yukaI mean I can also trace all pmgr and then see where the pmp shaped writes end up
06:37yukaiirc I already had one trace containing those writes
06:38yukanickchan: ah fair, was just an idea but then probably not the cause
06:38chaos_princessor you can do that, read the xnu log, and just look at what pmgr logs. it dumps its mmio maps with names for each region if you pass that stuff to it
06:39yuka(the ranges I end up from plain read32() on amcc {lower,upper}-limit-reg are 0x1-0xa6 and 0x0-0x35d, which seems quite odd to me)
06:40yukaI guess the addresess should be 64 bit and are spread over multiple 32-bit mmio reads?
06:41yukano, the upper address reg is lower address reg + 4
06:41yukaso it's 32 bit
06:42hdbngr has quit [Ping timeout: 480 seconds]
06:44yukaSo, uh
06:46hdbngr has joined #asahi-dev
06:48yukait wouldn't make sense anyways if it was amcc it would affect all cores equally
06:48yukathis whole "kernel write to read-only memory" feels like some mechanism for locking down address space has been left enabled
06:52chadmed has quit [Quit: Konversation terminated!]
06:57mattia013 has joined #asahi-dev
06:59hdbngr has quit [Ping timeout: 480 seconds]
06:59yukachaos_princess: from what i_can_has_debugger is a patch to the kernel?
06:59yuka* from what I understand
07:00yukado you usually apply it on your kernelcache before uploading it, or is there some function to let m1n1 patch it?
07:00chaos_princessyou just pass -d to run_guest.py
07:00chaos_princessand you do not want it on by default. Causes extra panics
07:04yukaOk yeah, it didn't boot fully with that but I probably got the output I need
07:05hdbngr has joined #asahi-dev
07:09yukaWhich name am I looking for in the Register Map?
07:09yukathere is none with PMP
07:10chaos_princesspastebin the map
07:12chaos_princessPMS_PTS_PTD iirc
07:16hdbngr has quit [Ping timeout: 480 seconds]
07:17pb17 has quit [Ping timeout: 480 seconds]
07:23JamesCalligeros[m]sofus: so unsurprisingly the issue was that I am stupid. had not reserved a section in the linker script and the pointer to the buffer was pointing at garbage
07:23hdbngr has joined #asahi-dev
07:32hdbngr has quit [Ping timeout: 480 seconds]
07:32sofuswhelp, thats a fun bug :p
07:33sofusi have piodma mostly working
07:35sofusbut something is wrong with interrupts, i think i have enabled mbox1 empty but i cant figure out where
07:36yukaHmm, I added the PMS_PTS_PTD reg to PMPRegTracer, and now it crashes "PMP Failed to Come Online within 30 seconds"
07:36yukaI do get lots of spam from PMPRegTracer
07:36yukaup to that point
07:36hdbngr has joined #asahi-dev
07:37LarstiQtaking more time under hypervisor, or missing some bit to come online?
07:38yukaI assume it's just too slow with the tracing, but I don't know for sure?
07:40yukachanged the tracer to async mode
07:40yukathat boots
07:42pb17 has joined #asahi-dev
07:42yukaI see things
07:44yukaI see both named and unnamed reads/writes in the PMPRegTracer
07:45hdbngr has quit [Ping timeout: 480 seconds]
07:45hdbngr has joined #asahi-dev
07:45yukaDEV_STATUS_TGT_RD / DEV_STATUS_TGT_UNK / DEV_STATUS_TGT_WR / PMP_STATUS / PMP_STATUS_UNK / DEV_STATUS_ACT / DEV_STATUS_ACT_UNK
07:50chaos_princesssounds about right
07:50yukaoffsets seem to be the same as t602x
07:52yukais there any problem with making PMPRegTracer async? If not I would submit that as a PR
07:53chaos_princessshouldn't be, its sync since that worked on t6000
07:55hdbngr has quit [Ping timeout: 480 seconds]
07:58hdbngr has joined #asahi-dev
08:02yukashould I also add the dev.flags.notify_pmp -> dev.id1 != 0 which you mentioned the other time?
08:03chaos_princessactually, maybe just both? since the second catches irrelevant stuff too
08:04yukaboth as in "notify_pmp || id1 != 0"?
08:04yuka"&&" makes little sense to me unless there is some device which has notify_pmp && id == 0
08:05chaos_princessas in id1 != 0 for filter, and print notify_pmp next to address
08:05yukaThe PMPRegTracer reg 41 was for t6000?
08:05yukaI would add a table
08:07chaos_princessyes. the tracer has addresses for t600x, t602x and t8112
08:07hdbngr has quit [Ping timeout: 480 seconds]
08:07chaos_princessactually, i am going to look at the adt in a bit, the names must be there somewhere
08:08yuka"has addresses"? for me it only has a hardcoded "start, len = u.adt['/arm-io/pmgr'].get_reg(50)"
08:08yukaor is the 41 correct for all these devices?
08:08yuka(50 is the one for t6034)
08:15chaos_princessdid i never push it
08:15hdbngr has joined #asahi-dev
08:15yukaOopsie
08:16yukaI think on t6020 it would be reg 45
08:18chaos_princessand no, the names are in kext
08:18yukano to what?
08:19chaos_princess>> the names must be there somewhere
08:24yukaOkay, we should definitely untangle that somehow, and have one list of which pmgr reg (or directly the pmp regs base address), and then use base address + tgt_write offset there
08:24hdbngr has quit [Ping timeout: 480 seconds]
08:24yukaso that support for a new device can be added in both tracer and the pmp_init experiment in one step
08:27hdbngr has joined #asahi-dev
08:29yukalots of pmp stuff is basically duplicated 3x
08:35hdbngr has quit [Ping timeout: 480 seconds]
08:36hdbngr has joined #asahi-dev
08:37yukaI see 3 PMPMessage (proxyclient/hv/trace_pmp.py, proxyclient/experiments/pmp_init.py, proxyclient/m1n1/fw/pmp.py), but they are not the same
08:37yukabut trace_pmp.py and pmp_init.py are the same
08:37integralpilot has joined #asahi-dev
08:38sofus has quit [Ping timeout: 480 seconds]
08:42chaos_princessjack mic works too, right?
08:44integralpilotsorry i just read "jack" and skipped over that and sent too soon!
08:44integralpilothaven't tested
08:44hdbngr has quit [Ping timeout: 480 seconds]
08:46integralpilotwill test mic now and report back
08:46integralpilot has quit [Quit: integralpilot]
08:47hdbngr has joined #asahi-dev
08:56hdbngr has quit [Ping timeout: 480 seconds]
09:05hdbngr has joined #asahi-dev
09:13yukachaos_princess: trying to make pmp_init.py a bit more generic, how can I determine (dynamically) if a device has IOVATable or not?
09:15yukahow did you obtain the bootargs? iop-pmp-nub dram_capacity does not exist in my adt
09:15chaos_princessalso only on t600x
09:16yukaonly BDID, DVID otherwise?
09:17hdbngr has quit [Ping timeout: 480 seconds]
09:21yukafrom what I can tell, macOS passes a lot more, but we can ignore that?
09:22chaos_princessdoes it _change_ a lot more?
09:22chaos_princessthe boot args are patched by the driver, not generated
09:23chaos_princesson t600x, it only changes those 3
09:25yukano you're right
09:25yukaonly those two
09:26hdbngr has joined #asahi-dev
09:30yukais it possible that PMPMessage in fw/pmp is for pmpv1?
09:31chaos_princesseverything not in mine is for v1
09:32gargantua has joined #asahi-dev
09:34integralpilot has joined #asahi-dev
09:35integralpilotchaos_princess: yes! jack mic works (tested with parecord)
09:35hdbngr has quit [Ping timeout: 480 seconds]
09:39integralpilothmm okay so audio patches pushed, and i'll send dc HID on-list tmrw night AEST unless there's any further feeback besides jannau's and sven's which i've addressed. anything else i can do (if i have time) that might be useful?
09:41integralpiloti was thinking 1. get AOP working on m3 and then 2. help upstream AOP (just iio parts for angle/ALS as audio is a bit more complex) unless there is anything better?
09:41chaos_princessaop works on m3, i already have a private branch
09:41integralpilotoh perfect then, cool!
09:41chaos_princessthe issue is that something about admac has changed, and it no longer works
09:42chaos_princessupstreaming aop would be cool, but hard, due to rust
09:43integralpilotoh yeah, true
09:43hdbngr has joined #asahi-dev
09:44integralpilotmy other idea was upstream WiFi, looks like our current patches ontop of mainline for that are mostly firmware loading changes to suit how Apple does it. the current way interferes a bit with the normal patch for standard cards but i do have an idea
09:44integralpiloton how to make it nicer for upstream
09:44chaos_princesswifi would be great
09:44svenyeah, WiFi would be useful
09:44integralpilotperfect, will have a look tmrw then! thanks :)
09:45integralpilot has quit [Quit: integralpilot]
09:50yukathere's also the pcie enable thingy
09:50yukarequired for wifi
09:51jannauintegralpilot: I haven't looked much beyond the mailbox changes. I'm heat throtled until tomorrow
09:52hdbngr has quit [Ping timeout: 480 seconds]
09:52hdbngr has joined #asahi-dev
09:54yukabig question: is the PERST gpio thingy upstream
09:54yukaanswer: no
09:55yuka"Allow pwrctrl framework to control PERST# GPIO if available"
09:56integralpilot has joined #asahi-dev
09:56jannauupstreaming of drivers in bits/250-aop (more like rust-drivers-except-agx) would be appreciated. It's blocked on the rtkit rust bindings which might not play well with rust device lifetime changes currently in progress. so it might not be the best target
09:58lederstr- has quit [Ping timeout: 480 seconds]
10:00lederstrumpf has joined #asahi-dev
10:01chaos_princessthere is a secondary problem in that aop contains common code between multiple drivers, which i now just stick into the kernel crate. And the actual solution is currently mid-bikeshed on ml
10:02hdbngr has quit [Ping timeout: 480 seconds]
10:03chaos_princessand sep should be easy imo, not like it does much anyway
10:03jannauyuka: our downstream pcie-poweron changes were soft-NAKed, it should be reimplemented with PCI_PWRCTRL_GENERIC (v7.1) / https://lore.kernel.org/linux-pci/20260220-topic-sm8650-ayaneo-pocket-s2-base-v5-3-1ad79caa1efa@linaro.org/
10:04hdbngr has joined #asahi-dev
10:06yukaAh, but good to know there is now a way forward!
10:07jannaure common code: I thought that would be taken care of via rust kbuild changes (afaik planned for v7.3). nova-{core,drm} has it's own way. not sure if that still going forward
10:08chaos_princessright, but 7.3 is in the future, so aop is blocked on that for now
10:13jannauwith v7.3 the device lifetimes and io projection are hopefully stable so that makes it a good future target as well
10:21hdbngr has quit [Remote host closed the connection]
10:21hdbngr has joined #asahi-dev
10:25yukafl0_id / chaos_princess: could one of you send me the binary version of a t6031 adt, since the one extracted from ipsw is missing pmp stuff
10:25yuka(fl0_id sent me a text version of it a long time ago)
10:25integralpilot has quit [Remote host closed the connection]
10:25chaos_princesswhat's missing?
10:26yukaOh actually they just changed the names
10:26yukathe extracted one is from a newer mac os version
10:26yukaKeyError "Child node 'pmp' not found"
10:26yukabecause it's pmp0
10:27chaos_princesspmp { - from DeviceTree.j514cap.im4p
10:27yukaYes, depends on which macOS version you got it from
10:27yukain 26.5, it's called pmp0
10:27yukaalso j514cap
10:28yukathe pmp bits list for t6030 (j514sap) looks very empty though
10:28chaos_princessright, mine are from 14.8.3
10:29yukacan you quickly run pmp_assist.py on j514{m,s,c}ap and tell me how long the output is?
10:30yukaoutput for j514m is 61 lines, so ~55 devices, which is about what I expect, but j514s (t6030) is only 30 lines
10:30hdbngr has quit [Ping timeout: 480 seconds]
10:31chaos_princessi don't have binary adts on this device, what is missing?
10:31chaos_princessis it just various dcs things?
10:32yukaAFNC*_IOA
10:32chaos_princessthat is memory/fabric/whatever things, looks normal
10:32pb17 has quit [Ping timeout: 480 seconds]
10:32chaos_princessafnc is fabric
10:33yukaPRORES*
10:33chaos_princessare those two completely missing, or is there just less of them
10:33yukacompletely missing
10:34chaos_princessis there anything "new"? Like APR or sth
10:34pb17 has joined #asahi-dev
10:37chaos_princessanyway, missing afncs is fine, atc/prores is weird, but maybe fine too?
10:37hdbngr has joined #asahi-dev
10:38yukaignoring the numbers:
10:39yukamissing: AFNC_IOA AMCC ATC_COMMON DCS DISPEXT_FE DISP_FE ISP-SOC-VMAX ISP-SOC-VNOM PRORES PRORES-SOC-VMAX PRORES-SOC-VNOM SCODEC-SOC-VMAX SCODEC-SOC-VNOM VENC-FAST
10:39yukanew: DISPEXT_SYS DISP_SYS SCODEC_STREAM SDEC-FAB-VMAX SDEC-FAB-VNOM VENC-SOC-VNOM
10:40yukasorry, I should really just paste it for you
10:40chaos_princessprobably fine
10:40chaos_princessi wouldn't worry about it
10:41yukaok i'll proceed with what I have
10:41yukabut t6031 also needs new offsets
10:41yuka*t6030 also needs new offsets
10:41chaos_princessiirc prores is not even a dedicated bit, right? it shares its with something
10:42yukaah whoops I pasted the files I used for diffing
10:43yukanot the full files
10:44yukaj514s/t6030 (26.5 adt): https://x0.at/Ofem.txt
10:44chadmed has joined #asahi-dev
10:45chaos_princessthe only ones that concern me is PCPM{0,1} in t6031. Need to check when macos turns those on
10:45yukabecause they have notify_pmp?
10:45chaos_princessalso, at least some of afncs might be always-on, and if you do not do it, it dies on sleep exit
10:46yukasleep does not work anyways (the display never turns on after waking up) :D
10:46yukabut yeah
10:46hdbngr has quit [Ping timeout: 480 seconds]
10:46chadmedjannau: when youre feeling up to it and get around to rebasing on 7.1, please also drop ASoC: apple: mca: Move clock shutdown to be shutdown
10:46chaos_princessoh, thats a worse kind of "dies", you can't even ssh into the machine as you kinda turned off half the chip fabric
10:46chadmedit is not required and the wrong thing to do anyway
10:51hdbngr has joined #asahi-dev
10:59integralpilot has joined #asahi-dev
10:59hdbngr has quit [Ping timeout: 480 seconds]
10:59integralpilotjannau: hope you get some good rest and it cools down soon, no worries!
11:00chadmedill happily swap right now, it is unpleasantly cold
11:02integralpilot has quit []
11:03svencold i can deal with, heat without AC i can't :/
11:04hdbngr has joined #asahi-dev
11:04chadmedive got the heater on but brisbane houses are notoriously poorly insulated and it is often colder inside than outside in winter
11:05LarstiQhow cold is it now?
11:06chadmed15 degrees outside but it feels colder than that
11:06chaos_princessi'd trade in a heartbeat :P
11:07chadmedbut that said, it is unseasonably warm due to the coming el nino. at this time of night at this time of year it should be closer to 8-11 degrees
11:07LarstiQah, that's a nice temperature we had a week ago, but even here it's now 22
11:12hdbngr has quit [Ping timeout: 480 seconds]
11:13yukachaos_princess: PCPM0/PCPM1 also have notify_pmp on t602x
11:18nickchanio-pgtable-dart's pagetable could work on A10
11:18ddxtanx has quit [Read error: Connection reset by peer]
11:18ddxtanx has joined #asahi-dev
11:18nickchanfor the DART1 variant, read protection is support from A10 onwards and there's no subpage protection but allowing *more* memory is generally fine
11:20nickchanFor some reason, the pagetable driver supported 4K paging even though none of that is in M1
11:20nickchannot complaining that the 4K paging worked perfectly ofc
11:20hdbngr has joined #asahi-dev
11:26yukaon t6021, pmp report entries pmp_report_afnc{4,5}_ioa exist and are always-on, but the other afnc ones not?
11:27yuka(the other afnc report entries do not exist)
11:29hdbngr has quit [Ping timeout: 480 seconds]
11:31chaos_princessthose are the only ones macos turns on
11:36yukawould that be triggered in some specific scenario? mine never turns on any of the afnc ones
11:37yuka...so I can leave them out?
11:37chaos_princessif it never turns them on, just leave them be
11:37chaos_princesson t6021, those are turned on shortly after kernel starts and never turned off
11:39hdbngr has joined #asahi-dev
11:44yukaok, possible that similar happens on t6031 but not t6034
11:44yukasomeone should test it to be sure
11:47hdbngr has quit [Ping timeout: 480 seconds]
11:51chadmedi think we can get kernel-side audio fully upstream by the end of the year barring any major issues with the machine driver
11:51chadmedtheres not really much left. ~all the codec stuff is merged, and hopefully this will be the last of the MCA patches (new variants notwithstanding)
11:52yukajannau: pmp breaks some assumptions around t603[124] grouping
11:52yukat603[14] have notify_pmp on the ans, but t6032 not
11:53svenchadmed: i'd immediately trade the 32 deg that i have inside and 37 degrees outside here for your 15 degrees :D
11:54yukaso t6031-nvme.dtsi needs some special conditions
11:54chaos_princessdoes it actually set the bit?
11:54yukaI don't have an M3 ultra to trace it :)
11:54chaos_princessi mean on t6034
11:54yukaYes it does
11:55yukaas a reminder: M3 ultra has the nvme on die 1
11:55chaos_princess.... why, apple
11:55yuka_think different_
11:55chadmedthink different
11:55chaos_princessalso, we have no idea if pmp even works on any ultras
11:55yukawhy wouldn't it?
11:55chaos_princessand if any changes are needed
11:56yukaah you mean our code
11:56yukaguess someone needs to test and potentially trace it eventually :-)
11:56yukawriting device tree nodes with lots of guessing and assumptions here
11:56hdbngr has joined #asahi-dev
11:57yukaI'm trying to cover all of M3, and fixups for everything non-t6034 can come later
11:58chadmeddo we need to buy an m3 ultra studio
11:58chaos_princessit is not even tested on m1/2 ultras
12:01yukacomplete support is cool and everything but maybe there's more important things than fully supporting m3 ultra
12:01jannauoverriding ans' power-domain from t6032.dtsi isn't much of an issue, it will need anyway to add all the devices from the second die. although the sram is probably used on both dies
12:01Ami[m]yuka: Sorry for interrupting again - but is M3 Ultra a different design to M3 Pro and similar? If so, why??? 😭
12:02chaos_princessit is 2 m3max glued together
12:02Ami[m]Ah I see I see
12:02chaos_princessbut m1/m2ultra had nvme on first die
12:02chaos_princessand m3pro and m3max are materially different, while for m1/2, the pro was a chopped max
12:03yukajannau: the thing is, j575dap re-uses the same ANS pmp report entry
12:03jannauI have access to a m3 ultra but it's low priotrity since it's macos 15 and later
12:03Ami[m]chaos_princess: Does it really have that much an affect on the software end? Sorry that seems naïve to ask
12:04chaos_princessthe adt is completely different, as for example everything starts at 0x21{zeroes} instead of 0x20{zeroes}
12:04jannauit doesn't because USE_PMP is not defined. we can sort this out when we decide to use PMP
12:04Ami[m]chaos_princess: Oh I see I see, thanks for explaining :)
12:04yukaAh, true. then it will have to override the power domain of ans back to the normal pmgr one
12:04yukaI can just do that
12:05hdbngr has quit [Ping timeout: 480 seconds]
12:06gargantua has quit [Ping timeout: 480 seconds]
12:07hardy94 has joined #asahi-dev
12:08renoo has joined #asahi-dev
12:09renoo has quit [Remote host closed the connection]
12:13hdbngr has joined #asahi-dev
12:14yukahuh, the dispext[0-3] report entries are always on in macOS
12:20sofus has joined #asahi-dev
12:34sofus has quit [Ping timeout: 480 seconds]
12:37yukaAh, during boot
12:37yukaafter boot they are turned off
12:54sofus has joined #asahi-dev
13:16yukafirst try to boot t6034 with pmp, wish me luck
13:16yukaAnd it's broken lol
13:17yukauboot....
13:17yukadoesn't know how to enable power for the nvme
13:18yukaso I guess I put the ane in always-on for now
13:18yukadon't have a good idea to teach uboot to use ps_ans while keeping pmp_report_ans for linux
13:19yukaI get why it's done this way, but the device tree structure for pmp does not seem quite right. As marcan said in some streams, device trees should describe the hardware and not how the software behaves
13:20LarstiQwhere firmware is included with hardware or software? I imagine the border is sometimes hard to draw
13:21yukafirmware is hardware in this case
13:21hdbngr has quit [Ping timeout: 480 seconds]
13:21yukaI guess this could be fixed in uboot by providing a no-op pmp driver just to bind to the compatible and pass through the dependency?
13:22yukasimilar to atcphy
13:25yukaAh shit, that doesn't even work
13:25yukaI think m1n1 would have to provide a safe starting state
13:25chaos_princessfor what, pmp?
13:25chaos_princessit already does
13:25chaos_princessjust don't turn it on
13:25yukaso the kernel turns it on later?
13:26yukaI wonder why my uboot isn't working then
13:26chaos_princessit actually takes a while to boot
13:26chaos_princessthe idea is that you send the reports no matter if it finished booting or not, and it will catch up
13:27yukaOkay, then the no-op pmp driver in uboot would work
13:30yukaapple_pmp 290700000.pmp: RTKit: syslog message: io.cpp:41: PMP started
13:30chaos_princessnow do a sleep cycle
13:30hdbngr has joined #asahi-dev
13:31chaos_princessand probably compare total system power with it on and off
13:31yukathe system power is drastically lower
13:31yukaat least what I see in upower -b
13:31yukalike almost unrealistic
13:31chaos_princess's the before/after
13:32yukabefore is ~7W
13:33yukawhoops, I still had a charger plugged in
13:33yukathat's why it was unrealistic
13:33chaos_princessdo total system power, not charge/discharge rate
13:33chaos_princessyou can read it from smc hwmon
13:33yukaOk I need to get a new 'before' reading then
13:33yukanever paid attention to that
13:34chaos_princessok, and whats the after?
13:34Sobek[m](I took the initiative of answering the Github question asking about MB Neo, basically saying, "MacBook Neo will probably, hopefully, be supported, on day. I is not harder than other M5 devices, but not easy either".
13:34chaos_princessit is about as hard as m4, and easier than m5
13:35Sobek[m]Ah, wait... Oh yes, it's a M4 Arch, not M5, I'll go and fix that.
13:36chaos_princessm5 uses exclaves more extensively, and has new radio chips
13:36chaos_princessalso new gpu, on which zero work has been done (m3 has some)
13:37Sobek[m]Anyway, I hope dealing with this sort of simple question answering is helpful to you guys ?
13:37chaos_princessas long as you don't post misinfo - sure, why not
13:38Sobek[m](I'll try to double check what I say, and not post stuff I'm unsure of).
13:38yukaIn the TTY with most system services stopped, *with* pmp: *4.17W*
13:39yukaI'll now reboot and perform the same test without pmp
13:39yukascreen on
13:39yuka(I don't know how to turn the screen off on this device)
13:39yukaas in the power domain?
13:39chaos_princessthough not sure if that turns off backlight
13:41amarioguy has quit [Ping timeout: 480 seconds]
13:42yukaah I'll test suspend before I reboot
13:42yukait does wake up
13:42yukascreen is off
13:42chaos_princessyou can now test with screen off :P
13:43yuka(though that won't disable the disp pmp report entry)
13:43yuka(anyways)
13:43chaos_princessbacklight itself is what is power-hungry
13:44yukaI think the backlight might still be on
13:44yukatotal system power is about the same
13:44chadmedisnt there a gpio for the backlight power?
13:44yukaalso I'm now using wifi to check the power
13:44chaos_princessthere is, but we don't know which gpio is it on m3max
13:45yukais it usually an smc gpio?
13:45yukaor which kind of gpio
13:48chaos_princesson j316 the gpio is connected to pmic, so that is probably via smc
13:54yukaokay, now same setup, in TTY, some background services stopped, but *without pmp*: 5.48W
13:54yukaI would say success
13:55chaos_princess1-ish W sounds about right
14:00averne has joined #asahi-dev
14:07yukaI'm looking into a dummy driver for uboot now
14:10yukaworks!
14:10yukashould I send that directly upstream?
14:13chaos_princessto u-boot? sure.
14:13chaos_princessit is not ever getting a "real" pmp driver anyway
14:51hdbngr has quit [Ping timeout: 480 seconds]
14:52hdbngr has joined #asahi-dev
14:54ddxtanx has quit [Ping timeout: 480 seconds]
15:00hdbngr has quit [Ping timeout: 480 seconds]
15:04hdbngr has joined #asahi-dev
15:04yukathe uboot dummy driver was too easy, it turns out when I tested it I still had the hack enabled
15:07ddxtanx has joined #asahi-dev
15:12ddxtanx has quit [Quit: Konversation terminated!]
15:12ddxtanx has joined #asahi-dev
15:12ddxtanx has quit []
15:12ddxtanx has joined #asahi-dev
15:15hdbngr has quit [Remote host closed the connection]
15:15hdbngr has joined #asahi-dev
15:18ddxtanx has quit [Quit: Konversation terminated!]
15:18ddxtanx has joined #asahi-dev
15:18ddxtanx has quit []
15:18ddxtanx has joined #asahi-dev
15:18enverbalalic has joined #asahi-dev
15:30hdbngr has quit [Ping timeout: 480 seconds]
15:31sofus has quit [Ping timeout: 480 seconds]
15:33ddxtanx has quit [Quit: Konversation terminated!]
15:34ddxtanx has joined #asahi-dev
15:34ddxtanx has quit []
15:34ddxtanx has joined #asahi-dev
15:39hdbngr has joined #asahi-dev
15:48hdbngr has quit [Ping timeout: 480 seconds]
15:48hdbngr has joined #asahi-dev
15:49enverbalalic has quit [Remote host closed the connection]
15:50enverbalalic has joined #asahi-dev
15:50nickchani could make use the apple,dma-range property in the downstream dart driver
15:50nickchani probably should write a binding for it and send it upstream
15:56hdbngr has quit [Ping timeout: 480 seconds]
16:03enverbalalic has quit [Remote host closed the connection]
16:03enverbalalic has joined #asahi-dev
16:04hdbngr has joined #asahi-dev
16:06lederstrumpf has quit [Remote host closed the connection]
16:13hdbngr has quit [Ping timeout: 480 seconds]
16:14ddxtanx has quit [Quit: Konversation terminated!]
16:15yukau-boot mailing list...
16:15yuka> Post by non-member to a members-only list
16:15yukado they actually check their moderation backlog or do I have to re-send it?
16:15hdbngr has joined #asahi-dev
16:15ddxtanx has joined #asahi-dev
16:21lederstrumpf has joined #asahi-dev
16:23hdbngr has quit [Ping timeout: 480 seconds]
16:27sofus has joined #asahi-dev
16:27Myria[m] has left #asahi-dev [User left]
16:32hdbngr has joined #asahi-dev
16:38yukachaos_princess: t600x and t602x have GFX notify_pmp, but only t8112 has it wired up in our dts
16:38yukadid you trace it and determine the adt lies?
16:38chaos_princesswe can't reinit the gpu, so we just don't let pmp power it down
16:38yukaOh, huh
16:39yukahow does that work on t8112 then
16:39yukajust nobody tested it?
16:39chaos_princesspmp does not work there for unrelated reasons
16:39yukaAh oki
16:39yukathen I'll not include pmp_report_gfx on M3*
16:40yuka(wait, shouldn't it be an always-on reporting node then, instead of no node?)
16:40chaos_princessno node also works
16:40chaos_princessi think it just never connects to it or something
16:40yukabecause it starts out as enabled and we don't touch it
16:40yukapresumably
16:40chaos_princessand in fact always-on breaks something
16:44Guest12079 has quit [Quit: Bridge terminating on SIGTERM]
16:44rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
16:47hdbngr has quit [Ping timeout: 480 seconds]
16:50ddxtanx has quit [Quit: Konversation terminated!]
16:50Jamie has joined #asahi-dev
16:50rhysmdnz has joined #asahi-dev
16:50ddxtanx has joined #asahi-dev
16:50Jamie is now known as Guest12229
16:55ddxtanx has quit [Quit: Konversation terminated!]
16:55hdbngr has joined #asahi-dev
16:56ddxtanx has joined #asahi-dev
16:56yukathe disp/dispext power domain dependencies were inverted on t6031
16:57yukait's fe -> cpu and not cpu -> fe
16:58yukaI see two paths for this:
16:58yuka1. Make pmp_report_disp* depend on ps_disp*_fe
16:59yukaeven though this is not actually what we want to signal (the id1 != 0 is still on disp_cpu not disp_fe)
17:00yuka2. replace the dependency of ps_disp*_fe on ps_disp*_cpu to pmp_report_disp*
17:00yukawhich would mean having ifdef APPLE_USE_PMP in the -pmgr.dtsi
17:00chaos_princesspmp stuff should be after all pmgr stuff
17:00yuka"after" as in run-time order?
17:01chaos_princesspowerup = pmgr, then pmp, then driver. powerdown in reverse
17:02yukait still sounds wrong to have pmp report of DISPEXT_SYS technically report the state of dispext_fe
17:02yukabut I guess we treat them all the same
17:03yukawe never power on dispext_sys without also powering dispext_fe
17:03yukaso it's fine
17:03yukaand also on these machines the whole pmgr disp* are always-on
17:04yukaso I'll try to not worry about this too much
17:04yukauntil we get DCP, I guess
17:04hdbngr has quit [Ping timeout: 480 seconds]
17:04yukalike proper DCP, not the iboot framebuffer
17:06enverbalalic has quit [Remote host closed the connection]
17:06ddxtanx has quit [Quit: Konversation terminated!]
17:06ddxtanx has joined #asahi-dev
17:06enverbalalic has joined #asahi-dev
17:11hdbngr has joined #asahi-dev
17:16ddxtanx has quit [Quit: Konversation terminated!]
17:16ddxtanx has joined #asahi-dev
17:18beastd has quit [Ping timeout: 480 seconds]
17:20hdbngr has quit [Ping timeout: 480 seconds]
17:23beastd has joined #asahi-dev
17:27lederstrumpf has quit [Ping timeout: 480 seconds]
17:28hdbngr has joined #asahi-dev
17:34enverbalalic has quit [Ping timeout: 480 seconds]
17:36ddxtanx has quit [Quit: Konversation terminated!]
17:37ddxtanx has joined #asahi-dev
17:37hdbngr has quit [Ping timeout: 480 seconds]
17:42yukaok, my PMP work is done for now
17:44sofus has quit [Quit: Lost terminal]
17:45hdbngr has joined #asahi-dev
17:46pb17 has quit [Ping timeout: 480 seconds]
17:50sofus has joined #asahi-dev
17:52ddxtanx has quit [Quit: Konversation terminated!]
17:52ddxtanx has joined #asahi-dev
17:54sofus has quit []
17:54hdbngr has quit [Ping timeout: 480 seconds]
17:58hdbngr has joined #asahi-dev
18:00truita_ has joined #asahi-dev
18:02enverbalalic has joined #asahi-dev
18:04truita has quit [Ping timeout: 480 seconds]
18:07hdbngr has quit [Ping timeout: 480 seconds]
18:07ddxtanx has quit [Quit: Konversation terminated!]
18:08ddxtanx has joined #asahi-dev
18:08ddxtanx has quit []
18:08truita_ is now known as truita
18:08ddxtanx has joined #asahi-dev
18:12hdbngr has joined #asahi-dev
18:13ddxtanx has quit [Quit: Konversation terminated!]
18:13ddxtanx has joined #asahi-dev
18:14ddxtanx has quit []
18:14ddxtanx has joined #asahi-dev
18:14ddxtanx has quit []
18:14pb17 has joined #asahi-dev
18:14ddxtanx has joined #asahi-dev
18:17apc8 has joined #asahi-dev
18:17apc is now known as Guest12231
18:18Guest12231 has quit [Read error: Connection reset by peer]
18:18apc8 is now known as apc
18:24hdbngr has quit [Ping timeout: 480 seconds]
18:24yuyichao has quit [Quit: Konversation terminated!]
18:24yuyichao has joined #asahi-dev
18:29ddxtanx has quit [Quit: Konversation terminated!]
18:30hdbngr has joined #asahi-dev
18:30ddxtanx has joined #asahi-dev
18:31yukau-boot people are asking about upstreaming the device tree bindings
18:31yukawhich is fair
18:32enverbalalic has quit [Remote host closed the connection]
18:33enverbalalic has joined #asahi-dev
18:40elvishjerricco2 has joined #asahi-dev
18:41hdbngr has quit [Ping timeout: 480 seconds]
18:45elvishjerricco has quit [Ping timeout: 480 seconds]
18:45elvishjerricco2 is now known as elvishjerricco
18:45hdbngr has joined #asahi-dev
18:45ddxtanx has quit [Quit: Konversation terminated!]
18:50ddxtanx has joined #asahi-dev
18:51mattia013 has quit [Quit: mattia013]
18:51mattia013_ has joined #asahi-dev
18:53mattia013_ is now known as mattia013
18:55ddxtanx has quit [Quit: Konversation terminated!]
18:55ddxtanx has joined #asahi-dev
19:09mattia013 has quit [Remote host closed the connection]
19:11mattia013 has joined #asahi-dev
19:11yuyichao has quit [Quit: Konversation terminated!]
19:11yuyichao has joined #asahi-dev
19:11yuyichao has quit []
19:12yuyichao has joined #asahi-dev
19:16hdbngr has quit [Ping timeout: 480 seconds]
19:20ddxtanx has quit [Quit: Konversation terminated!]
19:21ddxtanx has joined #asahi-dev
19:25hdbngr has joined #asahi-dev
19:31ddxtanx has quit [Quit: Konversation terminated!]
19:31ddxtanx has joined #asahi-dev
19:31ddxtanx has quit []
19:32ddxtanx has joined #asahi-dev
19:42ddxtanx has quit [Quit: Konversation terminated!]
19:43ddxtanx has joined #asahi-dev
19:52hdbngr has quit [Ping timeout: 480 seconds]
19:53mattia013 has quit [Remote host closed the connection]
19:58ddxtanx has quit [Quit: Konversation terminated!]
19:58ddxtanx has joined #asahi-dev
19:58ddxtanx has quit []
19:58ddxtanx has joined #asahi-dev
20:01hdbngr has joined #asahi-dev
20:04ddxtanx has quit [Quit: Konversation terminated!]
20:04ddxtanx has joined #asahi-dev
20:05yuyichao has quit [Quit: Konversation terminated!]
20:05yuyichao has joined #asahi-dev
20:11hdbngr has quit [Ping timeout: 480 seconds]
20:14ddxtanx has quit [Quit: Konversation terminated!]
20:15ddxtanx has joined #asahi-dev
20:15ddxtanx has quit []
20:19hdbngr has joined #asahi-dev
20:22jannauI've rebased asahi-wip onto v7.1.2, not tagged and I've ignored PRs for now
20:23jannauchadmed: I didn't drop the mca shutdown commit since I did not want to deal with the conflicts
20:24jannauaudio tested on j504/j514s
20:28sofus has joined #asahi-dev
20:30ddxtanx has joined #asahi-dev
20:30ddxtanx has quit []
20:31ddxtanx has joined #asahi-dev
20:32hdbngr has quit [Ping timeout: 480 seconds]
20:39hdbngr has joined #asahi-dev
20:41ddxtanx has quit [Quit: Konversation terminated!]
20:41ddxtanx has joined #asahi-dev
20:41ddxtanx has quit []
20:41enverbalalicyuka: i tested the dockchannel tty driver, after i finally figured out that i was loading the wrong dtb... console=ttydc0 doesn't output anything, but manually echoing to /dev/ttydc0 works with a caveat - it only works once, after that the entire machine freezes. no crashdump displayed on screen, and i have no other serial to figure out what's wrong
20:42ddxtanx has joined #asahi-dev
20:44yukaintegralpilot
20:45yukawhich direction worked once?
20:45yukaah you said echo
20:49yukaI guess I still need to test that direction too, to get a stacktrace
20:49hdbngr has quit [Ping timeout: 480 seconds]
20:52enverbalalicinteresting, if i echo a huge message where it just cuts out and doesn't output it all - it doesn't kill the machine then
20:52enverbalalicbut it will still kill the output pipe, second echo still doesn't do anything
20:53enverbalalichuge message = >2048 chars
20:54jannau2048 is the fifo size
20:57sofus has quit [Quit: Lost terminal]
20:58hdbngr has joined #asahi-dev
21:07ddxtanx has quit [Quit: Konversation terminated!]
21:07hdbngr has quit [Ping timeout: 480 seconds]
21:08ddxtanx has joined #asahi-dev
21:17hdbngr has joined #asahi-dev
21:20kettenisyuka: changing things because the bindings change just ends up being painful
21:21yukayeah absolutely fair
21:21yukaI'm actually not too happy with the way the pmp is abstracted in the device tree, but also can't see it getting better anytime soon
21:25hdbngr has quit [Ping timeout: 480 seconds]
21:26JamesCalligeros[m]It’s not too late to change of course
21:33hdbngr has joined #asahi-dev
21:38enverbalalic has quit [Quit: enverbalalic]
21:42hdbngr has quit [Ping timeout: 480 seconds]
21:42lederstrumpf has joined #asahi-dev
21:47lederstr- has joined #asahi-dev
21:50hdbngr has joined #asahi-dev
21:50lederstrumpf has quit [Ping timeout: 480 seconds]
21:57lederstr- has quit [Ping timeout: 480 seconds]
21:57jamie__ has joined #asahi-dev
21:59jamie_ has quit [Read error: Connection reset by peer]
22:03lederstrumpf has joined #asahi-dev
22:13hdbngr has quit [Ping timeout: 480 seconds]
22:18lederstrumpf has quit [Ping timeout: 480 seconds]
22:20hdbngr has joined #asahi-dev
22:29hdbngr has quit [Ping timeout: 480 seconds]
22:35hdbngr has joined #asahi-dev
22:36enverbalalic has joined #asahi-dev
22:46hdbngr has quit [Ping timeout: 480 seconds]
22:55hdbngr has joined #asahi-dev
23:04hdbngr has quit [Ping timeout: 480 seconds]
23:12hdbngr has joined #asahi-dev
23:21hdbngr has quit [Ping timeout: 480 seconds]
23:29hdbngr has joined #asahi-dev
23:30pb17 has quit [Ping timeout: 480 seconds]
23:32pb17 has joined #asahi-dev
23:38hdbngr has quit [Ping timeout: 480 seconds]
23:48hdbngr has joined #asahi-dev
23:50enverbalalic has quit [Ping timeout: 480 seconds]
23:56hdbngr has quit [Ping timeout: 480 seconds]
23:57hdbngr has joined #asahi-dev
…