#asahi-dev · 2026-01-23 13:47
mazare you booting at EL2?#asahi-dev · 2026-01-23 13:47
mazSYS_IMP_APL_VM_TMR_FIQ_ENA_EL2 seems to be the guilty one.#asahi-dev · 2026-01-23 13:46
mazyuka: clearly, the AIC on this machine does not behave as the previous generations.#asahi-dev · 2026-01-23 13:06
mazyuka: if you can dump this UNDEF exception with a full trace somewhere, it shouldn't take long to narrow it down.#asahi-dev · 2026-01-10 11:35
mazbut I doubt anyone will actually care.#asahi-dev · 2026-01-10 11:35
mazdo it at the end in one go. that's not a functional change, and better grouped together.#asahi-dev · 2026-01-05 13:41
mazfrom what you describe, the NHI and IOMMU are sharing a power rail (or clocks), gated by the PHY. describing that in DT is going to be fun. not.#asahi-dev · 2026-01-05 13:39
mazlook at the google pixel stuff (not sure it is merged yet). they definitely had some weird integration with one iommu per sswrap.#asahi-dev · 2026-01-05 13:37
mazI wonder if other SoCs that use an iommu per subsystem have similar constructs. I definitely remember something along these lines at $JOB-1...#asahi-dev · 2026-01-05 13:34
mazI guess that's an insurance policy. no external traffic before the IOMMU can filter things. but yeah, pretty annoying.#asahi-dev · 2026-01-05 13:32
mazwell, you need to start somewhere...#asahi-dev · 2026-01-05 13:32
mazsven: whooho!#asahi-dev · 2026-01-05 11:20
mazif there's a Co-Dev tag, you must have a SoB.#asahi-dev · 2026-01-05 11:15
mazcheckpatch is not meant to be an absolute oracle. it is mostly advisory.#asahi-dev · 2026-01-04 17:44
mazShiz: the other trick is to look for cables that do DP. they mandate the SBU pins to be wired, making them pretty useful for serial.#asahi-dev · 2026-01-04 10:52
mazat this point, the onus is on the x86 folks to do the work.#asahi-dev · 2025-12-24 09:27
mazfl0_id: no worries, I don't think anyone needs this urgently, if at all. when you get a chance to test it, let me know.#asahi-dev · 2025-12-23 11:36
mazfl0_id: I've just pushed a branch updated to SDK 2.2.0. on a sample of one, it seems to work fine.#asahi-dev · 2025-12-22 16:21
mazfl0_id: yup, got it. in my case, works means compiles, boots, and performs reliably. from past experience, it hasn't always been smooth sailing.#asahi-dev · 2025-12-22 16:18
mazOK. I'll see if 2.0.0+ works with the current code. If it does, I'll pull that in.#asahi-dev · 2025-12-22 15:27
mazmaybe I should just bump the SDK version.#asahi-dev · 2025-12-22 15:26
mazor is that in the bowels of pico-sdk?#asahi-dev · 2025-12-22 15:25
mazfl0_id: I'll take a patch if you can be bothered to write one.#asahi-dev · 2025-12-16 14:17
mazthe only case where CNTPOFF applies is when HCR.TGE is set. and Apple stuff doesn't implement CNTPOFF at all.#asahi-dev · 2025-12-16 14:09
mazreading from EL2, neither views have any offset applied to it.#asahi-dev · 2025-12-16 14:08
mazconfused. if the block has its own offset, how is that better by reading the physical counter on the CPU?#asahi-dev · 2025-12-16 13:59
maza read of CNTVCT_EL0 should results in the same thing, and is used all over the shop already.#asahi-dev · 2025-12-09 13:05
mazgood stuff!#asahi-dev · 2025-11-29 11:31
mazcould well be. I wouldn't be surprised if they had started with some embryonic SME hacking way before M4.#asahi-dev · 2025-11-29 11:29
mazdo you have a SMSTART before that?#asahi-dev · 2025-11-29 11:27
mazwell, the registers are.#asahi-dev · 2025-11-29 11:27
mazthey are the same instructions.#asahi-dev · 2025-11-29 11:27
mazat best, SME on M4+.#asahi-dev · 2025-11-28 09:17
mazintegralpilot: if your cable is display-port capable, you should be good.#asahi-dev · 2025-11-16 16:14
mazI'd humbly suggest you start learning about the basics of architecture before embarking on this sort of project.#asahi-dev · 2025-11-09 10:58
mazjannau: glad I could be of help.#asahi-dev · 2025-11-09 10:54
mazbeastd: I only use USB for mass storage, and don't have any HID plugged to it (this machine is used in completely headless configuration, with only a serial console).#asahi-dev · 2025-11-09 10:26
mazyup. it makes my life so much better, as I can finally test my own code without having to forward port complicated patch series on top of whatever -rc I'm working at the moment.#asahi-dev · 2025-11-09 10:16
mazI was pleasantly surprised to notice that I can boot 6.18 without any additional patches on my M2 Pro mini. Some stuff missing, of course, but the basics are there, finally. thanks to everyone who helped on this!#asahi-dev · 2025-10-28 09:21
mazah, now you're talking! :)#asahi-dev · 2025-10-28 09:20
mazsven: don't say that in public -- you know how it goes... :-/#asahi-dev · 2025-10-04 12:42
mazwe had A57 drops early 2011, and APM was already showing XGene on FPGAs by then.#asahi-dev · 2025-10-04 12:39
mazRTL emulation.#asahi-dev · 2025-10-04 12:36
mazApple's A7, not the contemporary 32bit core.#asahi-dev · 2025-10-04 12:35
mazI was doing ARMv8 porting long before A7 was a thing.#asahi-dev · 2025-10-04 12:34
mazI don't think you get to decide.#asahi-dev · 2025-10-04 11:45
mazs/our/out/#asahi-dev · 2025-10-04 11:44
mazchadmed: there were 6 different implementations going on at the very same time, and while Apple got the first CPU our in a device, they were not the first implementation.#asahi-dev · 2025-09-26 14:29
mazwell, you know it's way too warm to be there! :)