#
#asahi-re · 2025-12-08 13:23
nickchani wonder if the same version even works in hypervisor for m1/m2
#asahi-re · 2025-12-08 13:12
nickchanhttps://github.com/apple-oss-distributions/xnu/blob/f6217f891ac0bb64f3d375211650a4c1ff8ca1ea/osfmk/vm/vm_kern_xnu.h#L68 there should be something in kmr_return if the pointer is null
#asahi-re · 2025-12-08 13:11
nickchan integralpilot: kmem_alloc_guard_internal returns a struct with result code and pointer
#asahi-re · 2025-12-08 12:26
integralpilotfurther debugging: it happens on the 233rd run of initWithPrelinkedInfoDict and occurs due to _kmem_alloc_guard_internal seemingly returning null
#asahi-re · 2025-12-08 11:05
integralpilotso the actual underlying problem here is the exception which i'm collecting more info on rn
#asahi-re · 2025-12-08 11:04
integralpilotthe actual problem isn't the panic - __ZN6OSKext25initWithPrelinkedInfoDictEP12OSDictionaryb7kc_kind causes an exception when processing a kext (i'm trying to find which one) and then the exception handler does something it's too early for then causing the panic
#asahi-re · 2025-12-08 09:56
nickchanfrom your description it seems like it starts printing out logs already
#asahi-re · 2025-12-08 09:55
nickchanintegralpilot: btw what's the panic string when initializing iokit
#asahi-re · 2025-11-24 21:22
nnymAfter redownloading the KDKs and extracting them with ipsw (github.com/blacktop/ipsw), I can find `ps-regs` in each of them.
#asahi-re · 2025-11-24 19:38
nnymGood to know.
#asahi-re · 2025-11-24 19:38
chaos_princessand yes, Payload is compressed
#asahi-re · 2025-11-24 19:37
chaos_princessmacos upgrades screw up the installed kdks somehow, making them useless. you will need to reinstall
#asahi-re · 2025-11-24 19:36
nnymOh. No. It's probably because I didn't install the 13.0 beta 1 and 13.1 KDKs; I only `7z x`d them on my Linux computer until I got their `Payload`s. I'm guessing that they're compressed then. Size checks out. I did install the other KDKs that I mentioned though. Does upgrading macOS do something to the KDKs?
#asahi-re · 2025-11-24 19:27
chaos_princesshave you upgraded macos since you installed the kdk
#asahi-re · 2025-11-24 19:27
chaos_princessyou are doing something wrong
#asahi-re · 2025-11-24 19:27
chaos_princess`KDK_13.5_22G74.kdk/System/Library/Extensions/ApplePMGR.kext/Contents/MacOS/ApplePMGR`
#asahi-re · 2025-11-24 19:24
nnymI ran `rg -a ps-regs` on macOS 13.0 beta 1, 13.1, 15.0, 15.6, and 26.0 KDKs and didn't get any hits so it seems to be unused for a long time. Is it used in an earlier version?
#asahi-re · 2025-11-14 13:38
nickchanhttps://github.com/torvalds/linux/blob/e9a6fb0bcdd7609be6969112f3fbfcce3b1d4a7c/drivers/nvme/host/apple.c#L1111
#asahi-re · 2025-11-14 13:37
nickchangraphine27: merely never enable them in the first place
#asahi-re · 2025-11-13 08:18
Administrator8
#asahi-re · 2025-11-13 08:15
graphine27do you know the sequence to disable the quircks?
#asahi-re · 2025-11-13 01:24
nickchan(compatible according to sven anyways)
#asahi-re · 2025-11-13 01:23
nickchanhad also tried to disable quirks on t6000 but that makes rtkit crash
#asahi-re · 2025-11-13 01:22
nickchanafter disable quirks the t8103 nvme behaves in a compatible way to t8015 nvme
#asahi-re · 2025-11-13 01:22
nickchanas for external pcie over tbt no idea
#asahi-re · 2025-11-13 01:22
nickchangraphine27: i think quirks can only be disabled in t8103
#asahi-re · 2025-11-12 21:46
graphine27I will try to trace what T2 does, if m1n1 can work correctly when booted via pongo I should be able to fully trace it
#asahi-re · 2025-11-12 21:45
chaos_princessi'd put it as "not impossible" at least on m1. maybe blindly poke the mmio areas using t2 as reference and see what happens?
#asahi-re · 2025-11-12 21:44
graphine27It would be an option yes, but I was interested if the T2 way would work for bare metal functionality
#asahi-re · 2025-11-12 21:41
chaos_princessw.r.t your thing, can you synthesize a fake pcie bus by trapping it with the hypverisor? in theory, you only need to trap config registers, and then expose the real nvme mmio as BARs
#asahi-re · 2025-11-12 21:40
graphine27Yes. Although I suspect T2 could have done it over thunderbolt, like when using the target disk mode you could see the ssd name
#asahi-re · 2025-11-12 21:39
chaos_princessthen yea, ignore everything i said before
#asahi-re · 2025-11-12 21:38
chaos_princessoh, so it is not about exposing it via thunderbolt to external host, but to internal sw
#asahi-re · 2025-11-12 21:38
graphine27Being able to use NVMe in the standard way would make things much easier
#asahi-re · 2025-11-12 21:37
graphine27Trying to enable the hardware on a different platform where the driver framework is not as flexible as linux
#asahi-re · 2025-11-12 21:37
chaos_princessis there a specific reason why you want it?
#asahi-re · 2025-11-12 21:36
chaos_princessi may be misremembering, but i think you could on m1, but no longer can on m2+ or m1pro+
#asahi-re · 2025-11-12 21:36
graphine27I'm guessing the quircks also can't be disabled on M1
#asahi-re · 2025-11-12 21:35
graphine27In the first NVMe patch I read this about T2: disable some quirks (the NVMMU and the weird submission array) and then exposes it over a PCIe interface.
#asahi-re · 2025-11-12 21:34
chaos_princessdifferent silicon, so doesn't make much sense to leave it there. If you accept "custom drivers on host" - then should be possible, can even expose it over usb3, since m1 is capable of role switching on usb.
#asahi-re · 2025-11-12 21:31
graphine27I was hoping this would be possible sice T2 can do it, but chances are pretty low.
#asahi-re · 2025-11-12 21:30
chaos_princessis m1 capable of being a pcie endpoint? devices are usually either root or endpoint, i haven't seen many that can flip to either
#asahi-re · 2025-11-12 21:30
graphine27From what I've seen CreatePCIeUPDARTMapping is one of the functions used and it's present in M1 kexts as well. But some other things are missing from dt like mapper-pup-ans2 or apcieup-fn-nub-ans2
#asahi-re · 2025-11-12 21:28
graphine27nickchan: do you know if it would be possible for M1 chips to make the ANS controller available over PCIe like T2?
#asahi-re · 2025-11-01 06:15
nickchans/tend of/tend to/
#asahi-re · 2025-11-01 06:15
nickchanthe ones that get development builds are the ones that 3rd party kexts tend of interface with directly like iokit families
#asahi-re · 2025-11-01 06:14
nickchanmost of the "interesting" kexts for the purpose of this channel only has release builds
#asahi-re · 2025-11-01 06:13
nickchaninternally all the kexts are development builds too but in the KDK only core XNU and a few kexts are development
#asahi-re · 2025-11-01 06:13
nickchani am pretty sure the development xnu builds *are* the same as what apple uses internally, but
#asahi-re · 2025-10-28 21:26
nicolas17not the kexts

plus ›