#asahi-dev · 2026-03-30 10:19
jannaunoopwafel: yes. first step is to release m1n1 with GPU initialization#asahi-dev · 2026-03-27 14:22
jannauhmm, spi-hid has no subnodes#asahi-dev · 2026-03-27 14:18
jannaudisabling the keyboard node is probably enough#asahi-dev · 2026-03-26 17:26
jannaucy8aer: yes, pmp won't get enabled in the dt without m1n1 changes and everything works as before#asahi-dev · 2026-03-26 17:23
jannauwhy did you remove it? I'll test on m2*#asahi-dev · 2026-03-26 17:16
jannauwhy was pmp_report_gfx removed only from t600x?#asahi-dev · 2026-03-26 17:03
jannauprobably a similar path as pmgr reset, which existed but before supporting dptx#asahi-dev · 2026-03-26 17:02
jannauI fear it would also shutdown dcp further before pmp shutdown#asahi-dev · 2026-03-26 17:01
jannaudisp driver == linux driver?#asahi-dev · 2026-03-26 17:01
jannauI suppose the always-on for disp* take care of the boot framebuffer as well but their power-domains should be adjusted as well#asahi-dev · 2026-03-26 16:58
jannauI would prefer to wait rolling pmp out until the dt-bindings are reviewed upstream#asahi-dev · 2026-03-26 16:55
jannaua pity that we have to keep the disp* domains always on. I assume macos doesn't do that? not sure how that's related to the apple,min-state though. macos does the min-state PMGR states as well#asahi-dev · 2026-03-26 16:52
jannauchaos_princess: pmp seems to fine now on m1 max. I disabled the isp pmp_report domains for desktop devices#asahi-dev · 2026-03-26 13:21
jannaum1n1's adt takes care of that#asahi-dev · 2026-03-26 13:21
jannauin the case of m1 add 0x2000_00000#asahi-dev · 2026-03-26 13:21
jannauand remember that the regs are translated via /arm-io/ranges#asahi-dev · 2026-03-26 13:16
jannautryhard[m]: the rtkit protocol (used for communication with co-processors) and mmio reads/writes#asahi-dev · 2026-03-26 13:15
jannauthis would allow to turn PMP on on a per device basis#asahi-dev · 2026-03-26 13:13
jannauI pushed asahi-6.19.10-1 to github, it uses now APPLE_USE_PMP as define to control whether PMP/PMP_reporting is used. the commits to disable it got annoying with additional chnages#asahi-dev · 2026-03-26 13:10
jannauh13j pmp firmware asks for fast-die-ctrl-ce-map which seems to be only defined for m3+ devices#asahi-dev · 2026-03-24 10:07
jannausearch for udmabuf in https://gstreamer.freedesktop.org/releases/1.28/#asahi-dev · 2026-03-24 10:07
jannauthat might be a while, last year's 1.26 isn't packaged yet#asahi-dev · 2026-03-24 10:02
jannaugstreamer with udmabuf was afaik released, see https://mastodon.social/@rmader/114053192148587387#asahi-dev · 2026-03-24 09:58
jannaudoes it have dmabufs to begin with? not sure if it uses dma-heaps or udmabuf for software decoding so it would have dmabufs only for hardware decoding#asahi-dev · 2026-03-23 22:52
jannauah, yes, they don't have any cpufreq support#asahi-dev · 2026-03-23 22:35
jannauthe m3 airs don't have have them since those are spmi based. 14-inch m3 macbook pro and imacs have them#asahi-dev · 2026-03-23 14:08
jannaukettenis: thanks for checking. are the usb-c ports working? the main concern would be preventing new installs on laptops#asahi-dev · 2026-03-22 17:23
jannauI haven't tested yet#asahi-dev · 2026-03-22 17:19
jannauchaos_princess: do you want me to pull in my change to make DCAP optional?#asahi-dev · 2026-03-22 15:22
jannauchaos_princess: sorry, it was "might be a good time", fast forward 3-4 years, PMP driver #asahi-dev · 2026-03-22 15:15
jannaufor usb keyboards it's probably more involved but suspend is currently in a sad state for the native type-c ports#asahi-dev · 2026-03-22 15:13
jannauchaos_princess: "might be a good time to write a PMP driver" https://oftc.catirclogs.org/asahi-dev/2022-03-06#30702988#asahi-dev · 2026-03-22 15:11
jannauterr: afaik nobody has looked at it. if your lucky you just need to mark the right devices as wakeup source. I'd hope not that complicated for laptops keyboards#asahi-dev · 2026-03-22 14:29
jannaucould be related to macos not being able to wake dcp on m2 desktop devices after m1n1 used it#asahi-dev · 2026-03-22 14:25
jannauI can't remember what happens in macos#asahi-dev · 2026-03-22 14:23
jannauI suspect it might be crashing. rtkit doesn't comes up at all without min-state#asahi-dev · 2026-03-22 10:11
jannauI've tagged asahi-6.19.9-2, please use this instead of asahi-6.19.9-1#asahi-dev · 2026-03-22 09:37
jannauit was 6.19.7 based asahi-wip, I can push it somewhere#asahi-dev · 2026-03-22 09:33
jannaubut I wonder if it wouldn't make more sense if the power-domains notified pmp#asahi-dev · 2026-03-22 09:32
jannauI will remove the use of the pmp_report power-domains for now and tag another 6.19.9 kernel. In the worst case we will have at least one release with pmp-report support when the device tree starts using them#asahi-dev · 2026-03-22 09:29
jannauchaos_princess: no. integrated displays are always driven by dcp. on m1 and m1 pro/max/ultra the hdmi port is always driven by dcp#asahi-dev · 2026-03-21 21:23
jannauthat won't work, no m3 pro / t603x devicetrees in asahi-wip#asahi-dev · 2026-03-21 20:43
jannauyes, booting a 6.18 kernel with the 6.19 device tree#asahi-dev · 2026-03-21 20:29
jannaucan't really think a good alternative though#asahi-dev · 2026-03-21 20:29
jannaunot sure if that was such a good idea though. anyone tested how 6.18 kernels behave with the pmp_report power-domains? I expect it breaks the devices using them#asahi-dev · 2026-03-21 20:15
jannauI tagged and pushed asahi-6.19.9-1 (with disabled pmp)#asahi-dev · 2026-03-21 14:13
jannauI think that was only the copied name property but I realized that only after amking the changes#asahi-dev · 2026-03-21 14:12
jannauchaos_princess: pushed a few minor fixes to https://github.com/jannau/linux/commits/pmp-fixes-t8112/ feel free to squash or ignore. The main reason for dropping the tunables subnode owas that it broke `dtc -I fs -O dts -o - /proc/device-tree`#asahi-dev · 2026-03-21 13:44
jannauchaos_princess: does t60xx need the full "PMP_MMIO_SIZE: usize = 0x100000". t8112's pmp reg is 0x80000#asahi-dev · 2026-03-21 13:34
jannauinvis[m]: ps_ecpm and ps_pcpm (probably what refer to) should be dropped they are there mistakenly and I dropped them for the upstream submission