#asahi-dev · 2026-04-28 10:16
chaos_princessright, so wrong iova table#asahi-dev · 2026-04-28 10:16
sofusstartup#asahi-dev · 2026-04-28 10:16
chaos_princessand when does it crash? startup or suspend?#asahi-dev · 2026-04-28 10:15
chaos_princessremove from t6021.dtsi too, i want that field completely gone as it will cause it to send a zero address to IOVATableAck#asahi-dev · 2026-04-28 10:13
sofusstill crashes. Did you mean to remove them entirely so it uses t6021 pio-ranges?#asahi-dev · 2026-04-28 10:00
chaos_princesshttps://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t602x-die0.dtsi#L971 like for t600x https://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t600x-die0.dtsi#L745#asahi-dev · 2026-04-28 09:59
chaos_princessnot remotely what i expected. can you delete pio-ranges https://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t6020.dtsi#L30 entirely, delete pmp_report_gfx https://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t602x-die0.dtsi#L51 and only leave ps_gfx as the gpu power domain#asahi-dev · 2026-04-28 09:51
sofushere: https://pastebin.com/TmWZ9vuF#asahi-dev · 2026-04-28 09:34
chaos_princesscan you also break into the hv shell and run dart_pmp_tracer.dart.dump_all() once it settles#asahi-dev · 2026-04-28 09:31
sofushttps://gist.github.com/codemaster330/219b9ca640fceffecf5d5561a2cbdcc6#asahi-dev · 2026-04-28 09:27
chaos_princessyea, i guess i want the trace then#asahi-dev · 2026-04-28 09:27
chaos_princesswhat#asahi-dev · 2026-04-28 09:23
sofusdo you want the m1n1 traces? macos gives PMPMessage_IOVATableAck(TYPE=0x11, IOVA=0x0) which is why i assumed pio-ranges should be empty#asahi-dev · 2026-04-28 09:21
sofusit crashes on boot#asahi-dev · 2026-04-28 09:21
sofushttps://pastebin.com/tP5BHzuz#asahi-dev · 2026-04-28 09:12
sofusok will try#asahi-dev · 2026-04-28 09:11
chaos_princessand revert your pio change#asahi-dev · 2026-04-28 09:11
chaos_princesscan you do the changes jannau suggested, just like here https://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t600x-pmgr.dtsi#L429#asahi-dev · 2026-04-28 09:10
chaos_princesshttps://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t602x-pmgr.dtsi#L451#asahi-dev · 2026-04-28 09:10
chaos_princessright, we don't#asahi-dev · 2026-04-28 09:09
chaos_princessdo we not have those?#asahi-dev · 2026-04-28 09:09
jannauI think t602x pmp needs the same changes as t600x, always-on for the disp* pmp_report nodes and ps_pmp and ps_pms_sram#asahi-dev · 2026-04-28 09:08
chaos_princesscrashlog?#asahi-dev · 2026-04-28 09:08
sofusthen it crashes right after it starts#asahi-dev · 2026-04-28 09:07
jannaumaz: walking the hierarchy would be fine, the issue is that I can't see how to determine that parent doesn't support affinity before allocating the irq. I want to add MSI_FLAG_NO_AFFINITY so irq_set_affinity of the msi irq chip isn't set to msi_domain_set_affinity#asahi-dev · 2026-04-28 09:07
chaos_princesswhat happens if you don't remove them#asahi-dev · 2026-04-28 09:07
chaos_princesswell, it needs those#asahi-dev · 2026-04-28 09:07
sofushttps://pastebin.com/YXvxciZS#asahi-dev · 2026-04-28 09:04
chaos_princesspastebin the dt diff you have #asahi-dev · 2026-04-28 09:03
sofushttps://pastebin.com/gNDYEZnm#asahi-dev · 2026-04-28 09:03
sofushad to remove sio-ranges from the dt to get pmp to start#asahi-dev · 2026-04-28 09:02
chaos_princessoh, and what do you mean by crash? full machine freeze? just pmp crash?#asahi-dev · 2026-04-28 09:02
sofuspro, and yes. I have already run the tracer#asahi-dev · 2026-04-28 09:01
chaos_princessDo you already have the m1n1 development setup?#asahi-dev · 2026-04-28 09:01
chaos_princessand which specific t6020? Pro? Max?#asahi-dev · 2026-04-28 09:01
sofusyes#asahi-dev · 2026-04-28 09:00
chaos_princess*will you#asahi-dev · 2026-04-28 09:00
chaos_princesswere you be willing to run a m1n1 tracer#asahi-dev · 2026-04-28 09:00
chaos_princess> haven't validated it on all machines#asahi-dev · 2026-04-28 09:00
sofusHi, on t6020 pmp crashes when returning from suspend, is this a known issue?#asahi-dev · 2026-04-26 16:26
mazsomething that the "managed MSIs" framework allows on other designs.#asahi-dev · 2026-04-26 16:26
mazand that breaks any SW that uses per-CPU queues and using per-CPU data...#asahi-dev · 2026-04-26 16:25
mazwhat's even worse with the AICv2+ is that not only you can't set the affinity, but you can't really influence where the affinity lands at all.#asahi-dev · 2026-04-26 16:24
maz(where 1:N distribution is that the implementation picks an affinity based on a pre-defined population).#asahi-dev · 2026-04-26 16:23
mazthe only vaguely similar thing is GICv3 1:N, which Linux doesn't support for related reasons.#asahi-dev · 2026-04-26 16:23
mazjannau: no, walking the hierarchy is the only way. Linux is very ill-equipped for systems that can't set the affinity at the root controller level (Apple is the only one we have, I expect).#asahi-dev · 2026-04-26 10:58
jannauadding MSI_FLAG_NO_AFFINITY to struct msi_parent_ops.required_flags fixes that but I don't see a clean way to determine that the parent doesn't support set_affinity in apple_msi_init()#asahi-dev · 2026-04-26 10:57
jannaumaz: is there a way to check whether an irq_domain supports set_affinity without irq_data? Lina noted that the pcie msi irqs have writable smp_affinity but trying to set it results in -ENOSYS from irq_chip_set_affinity_parent#asahi-dev · 2026-04-26 09:22
chadmedNew progress report: https://asahilinux.org/2026/04/progress-report-7-0/#asahi-dev · 2026-04-25 15:44
freetolgaoh