#
✕ nick:sven
#asahi-re · 2022-02-28 14:45
svenAUSPLL_CFG_DCO_EFUSE_SPARE.RODCO_BIASADJ_EFUSE, AUSPLL_CFG_FRACN_CAN.DLL_CAL_START_CAPCODE_EFUSE, AUSPLL_CFG_CLKOUT_DTC_VREG.DTC_VREG_ADJUST_EFUSE
#asahi-re · 2022-02-28 14:45
svenCIO3PLL_CFG_DCO_NCTRL.DCO_COARSEBIN_EFUSE0, CIO3PLL_CFG_DCO_NCTRL.DCO_COARSEBIN_EFUSE1, CIO3PLL_CFG_FRACN_CAN.DLL_CAL_START_CAPCODE_EFUSE, AUS_CMN_SHM_BLK_VREG_REG_0.VREG_TRIM, CIO3PLL_CFG_FRACN_CAN.DLL_CAL_START_CAPCODE_EFUSE, CIO3PLL_CFG_CLKOUT_DTC_VREG.DTC_VREG_ADJUST_EFUSE, AUSPLL_CFG_DCO_EFUSE_SPARE.RODCO_ENCAP_EFUSE,
#asahi-re · 2022-02-17 15:31
svenhrm, nope, both seem to be for older chips
#asahi-re · 2022-02-17 15:22
svenbut i think what they call interrupt workaround is separate from that
#asahi-re · 2022-02-17 15:21
svenDoWorkaround reads from a hardcoded physical address (lol) which seems to trigger an exception on the M1 though.
#asahi-re · 2022-02-17 15:20
svenlooks like the nvme drivers has a few suspicious functions like "IsInterruptWorkaroundRequired" and "DoWorkaround"
#asahi-re · 2022-02-12 11:52
svenfile a radar? :>
#asahi-re · 2022-02-12 11:52
sven:D
#asahi-re · 2022-02-11 09:11
svenjannau: afaict the 12.2.x kernel still requires com.apple.private.hypervisor.vmapple and/or com.apple.private.hypervisor to allow things like SPRR/GXF in EL1
#asahi-re · 2022-02-10 21:20
svenThat’s the thing that decides which features are allowed. Iirc it passes a “hv level” to the syscall which created the vm
#asahi-re · 2022-02-10 21:19
sventhe interesting part is inside hypervisor.framework fwiw
#asahi-re · 2022-02-10 21:16
svenBut it could also just be the virtual iboot thingy and gpu paravirtualization
#asahi-re · 2022-02-10 21:16
svenI’ll take a look at the kernel this weekend. It could be that it now allows GXF,SPRR and the apple PAC stuff
#asahi-re · 2022-02-10 21:13
sven(… and the gpu virtualization ofc)
#asahi-re · 2022-02-10 21:13
svenbut I should probably take a look again
#asahi-re · 2022-02-10 21:13
svenhm, so far I assumed that was just their virtual iboot thing
#asahi-re · 2022-02-10 21:07
svenso you can probably build a hackintosh and likely run that kernel inside qemu but you won’t get any graphics
#asahi-re · 2022-02-10 21:07
svenAnd as usual I keep forgetting the gpu…
#asahi-re · 2022-02-10 21:05
svenalyssa: ^—
#asahi-re · 2022-02-10 21:05
svenI wouldn’t be surprised if you could build a arm64 hackintosh with some patching of that kernel
#asahi-re · 2022-02-10 21:04
svenid be surprised if m1n1 runs in there
#asahi-re · 2022-02-10 21:04
svenWhen I looked into hvf a while ago you essentially got a regular armv8 without apple-internal entitlements
#asahi-re · 2022-02-10 19:10
svenit’s probably their kernel that can be booted on a “normal” arm machine without all their fancy extensions
#asahi-re · 2022-02-09 15:12
sven:D
#asahi-re · 2022-02-09 13:06
sven0x85000000000000 is "stop firmware for recovery inspection"
#asahi-re · 2022-02-09 13:05
sven0x84000000000000 can be "new mapping", "new unmapping" or "block for ring empty"
#asahi-re · 2022-02-09 13:02
sven0x83000000000010 is "kick firmware". 0x83000000000002 is "submit CL channel", 0x83000000000001 is "submit 3D channel", 0x83000000000000 is "submit TA channel". but it always ORs something else into bits 2 and 3
#asahi-re · 2022-02-09 12:58
svenalright, let's see
#asahi-re · 2022-02-09 12:58
svenphire: if you want to i can go look for 0x83000000000000 in the kext and see if there are interesting strings or function names where it's used
#asahi-re · 2022-02-07 08:10
sveni haven't looked at those things and i don't know if you want to but they at least contain more strings
#asahi-re · 2022-02-07 08:09
svenphire: fwiw, there's an old XNU release where Apple included LLVM bitcode for stuff that's running inside PPL. among other things it contains uat.o and uat_ppl_handoff.o
#asahi-re · 2022-02-05 13:04
sventhe actual DART, SART and NVMMU code is also in there
#asahi-re · 2022-02-05 13:04
sventhe only code allowed to mess with any pagetables
#asahi-re · 2022-02-05 13:04
svenpage protection layer, their kernel-inside-their-kernel running in GL2
#asahi-re · 2022-02-05 07:02
svenoh… duh. Ofc the actual pagetable code would be inside PPL in the actual kernel
#asahi-re · 2022-02-04 15:18
svenin the DART PTEs the lowest two bits are used for that
#asahi-re · 2022-02-04 15:17
svencould be kind of valid bit(s)
#asahi-re · 2022-02-04 14:04
svenhm... i think it does
#asahi-re · 2022-02-04 14:03
svendid you also check the second register range given in the adt?
#asahi-re · 2022-02-04 13:59
svenyeah, i assumed they were in the sgx node
#asahi-re · 2022-02-04 13:58
svenlet me check again, it's been a while
#asahi-re · 2022-02-04 13:58
sventhe gart node just loads this gart kext which is very very small and only calls back into the main kext
#asahi-re · 2022-02-04 13:58
sventhe registers are part of the agx node
#asahi-re · 2022-02-04 13:58
svenonly sort of
#asahi-re · 2022-02-04 13:57
svenstill a bit weird, but who knows
#asahi-re · 2022-02-04 13:56
svencould be
#asahi-re · 2022-02-04 13:56
svenif I were to speculate even more: AGXSecureSharedGartTableBacking maybe it's both actually? :D
#asahi-re · 2022-02-04 13:56
sventhere's also stuff like "kernel_gart_tables" @%s:%d which make it sounds it's the kernel itself
#asahi-re · 2022-02-04 13:55
svenhm, weird
#asahi-re · 2022-02-04 13:54
svenwhich strings make you think that the firmware is responsible for setting up GART?

plus ›