#
#asahi-re · 2026-01-11 21:30
amarioguyhey, read dev backlogs, i saw that there might've been interest in figuring out how gl2 lockout works?
#asahi-re · 2025-12-28 11:32
nickchanstill no good same-EL xnu tracing...
#asahi-re · 2025-12-28 11:30
nickchanuses m1n1's hv load/store emulator of course :)
#asahi-re · 2025-12-28 11:29
nickchannot anything related to specific peripherals thogh, i think
#asahi-re · 2025-12-28 11:28
nickchan+ the sizes of the mmio region is used for patchfinding and that info is also available from adt
#asahi-re · 2025-12-28 11:26
nickchanmostly some structs related to generic arm stuff
#asahi-re · 2025-12-28 11:25
nickchanhttps://github.com/HoolockLinux/iboot_tracer
#asahi-re · 2025-12-28 11:25
nickchani have released my a11 iboot tracer (note this thing itself is filled with binary RE cause it needs to force link code into iboot)
#asahi-re · 2025-12-27 12:40
chadmedlikewise the "interchange" format fourccs used by macos simply do not exist anywhere in the firmware image
#asahi-re · 2025-12-27 12:40
chadmedre dcp firmware i can see where verify_surfaces is checking if the format is tiled/twiddled but cannot find where it's getting that from. its not in the surface, swap request or call struct the swap lives in afaict
#asahi-re · 2025-12-21 07:32
chadmedtwilight: read the documentation, set up m1n1, pick something you want to RE and have fun
#asahi-re · 2025-12-21 06:35
twilightlooking for joining the reverse engineering, where could i get started?
#asahi-re · 2025-12-21 06:35
twilightsup
#asahi-re · 2025-12-17 11:56
chadmedhttps://asahilinux.org/2025/12/progress-report-6-18/
#asahi-re · 2025-12-14 03:59
nickchanso i presume when bits are set in AGTCNTRDIR_EL{1,2,12} your reg access gets redirected to those 24 MHz custom registers instead of the standard 1 GHz ones
#asahi-re · 2025-12-14 03:58
nickchanso they kind of have to keep the 24 MHz counter going
#asahi-re · 2025-12-14 03:58
nickchanbut... apple exposed those regs to EL0 and maybe their vmapple guests has assumptions about timers as well
#asahi-re · 2025-12-14 03:57
nickchanarmv8.6 does require 1GHz timer
#asahi-re · 2025-12-14 03:57
chadmedoh right
#asahi-re · 2025-12-14 03:57
nickchanchadmed: <https://github.com/apple-oss-distributions/xnu/blob/43a90889846e00bfb5cf1d255cdc0a701a1e05a4/pexpert/pexpert/arm64/H15.h#L49> probably just means 24 MHz timer
#asahi-re · 2025-12-14 03:55
nickchanmore clearly: git diff xnu-11215.1.10 xnu-11417.121.6 -- osfmk/arm64/machine_routines.c
#asahi-re · 2025-12-14 03:55
chadmed"the apple timebase"
#asahi-re · 2025-12-14 03:54
chadmedthe way that comment reads, theyre doing something nonstandard with the timer so idk if theyre just shadowing the arch registers
#asahi-re · 2025-12-14 03:53
nickchanAGTCNTV_TVAL_EL0 = S3_1_C15_C15_4, AGTCNTVCT_EL0 = S3_4_C15_C11_7, AGTCNTVCTSS_EL0 = S3_4_C15_C10_6, AGTCNTV_CTL_EL0 = S3_1_C15_C0_5, AGTCNTKCTL_EL1 = S3_4_C15_C9_6
#asahi-re · 2025-12-14 03:52
nickchanhttps://github.com/apple-oss-distributions/xnu/blob/43a90889846e00bfb5cf1d255cdc0a701a1e05a4/osfmk/arm64/machine_routines.c#L2324 a newer xnu version changed secret names into encodings
#asahi-re · 2025-12-14 03:52
nickchanfound encodings
#asahi-re · 2025-12-14 03:43
nickchan<https://github.com/apple-oss-distributions/xnu/blob/8d741a5de7ff4191bf97d57b9f54c2f6d4a15585/osfmk/arm64/machine_routines.c#L2333> and most of those regs are probably 1:1 mapping to arm ones so yeah go find them
#asahi-re · 2025-12-14 03:43
nickchanAGTCNTVCTSS_EL0, ACNTVCT_EL0, AGTCNTP_CTL_EL0 exists at least
#asahi-re · 2025-12-14 03:41
nickchanthere's more than AGTCNTRDIR_EL{1,2,12}
#asahi-re · 2025-12-14 03:40
nickchanthis is real
#asahi-re · 2025-12-14 03:40
nickchan<https://github.com/apple-oss-distributions/xnu/blob/8d741a5de7ff4191bf97d57b9f54c2f6d4a15585/osfmk/arm64/machine_routines.c#L2579>
#asahi-re · 2025-12-14 03:39
nickchanif so the register probably stands for "Apple Generic Timer Counter Redirection" or something like that
#asahi-re · 2025-12-14 03:39
nickchanOkay I recommend checking if there's a whole class of custom sysreg that behave similarly to the ones stated in the Arm docs
#asahi-re · 2025-12-14 03:36
nickchanhttps://developer.arm.com/documentation/ddi0488/c/ch09s03s01 AArch64 Generic Timer register summary
#asahi-re · 2025-12-14 03:34
nickchanhttps://developer.apple.com/documentation/ios-ipados-release-notes/ios-ipados-18-release-notes apple only expose 1 GHz to apps compiled with new enough SDKs... but guest & host kernel probably would appericate it if the counter didn't change in their EL
#asahi-re · 2025-12-14 03:31
nickchanApple...GT...Counter...RDIR
#asahi-re · 2025-12-14 03:31
nickchananyways back to the naming A can be taken as "Apple" and "CNT" can be taken as counter
#asahi-re · 2025-12-14 03:29
nickchanjannau: maybe check cntfrq_el0 from EL0?
#asahi-re · 2025-12-13 16:16
jannauWhen bit(1) is set CNTFRQ_EL0 is 24MHz instead of the scaled 1GHz. no idea what bit(0) does
#asahi-re · 2025-12-13 16:14
jannauany good guesses what AGTCNTRDIR stands for? Apple GT Counter RDIR
#asahi-re · 2025-12-09 12:41
integralpilotyeah, i found it out earlier today. but still thank you!
#asahi-re · 2025-12-09 12:34
nickchanbut maybe that's not relvant now given the messages in #asahi-dev
#asahi-re · 2025-12-09 12:31
nickchanintegralpilot: it's a 16 byte struct so the return value would be spread across x0 and x1
#asahi-re · 2025-12-09 03:25
integralpilotthe issue happens when loading __DATA from the cache into new memory
#asahi-re · 2025-12-09 03:00
integralpiloti'm finding it now
#asahi-re · 2025-12-09 03:00
integralpilotso we can probably just skip it, but i want to know the actual reason initWithPrelinkedInfoDict is failing because the same thing (likely smth with memory) might cause more issues down the line
#asahi-re · 2025-12-09 02:59
integralpilotlooks like it handles blu-ray
#asahi-re · 2025-12-09 02:58
integralpiloti just checked and the problematic kext being loaded on the problematic run is com.apple.iokit.IOBDStorageFamily
#asahi-re · 2025-12-09 02:57
integralpilotoh thanks nickchan, ill look into that
#asahi-re · 2025-12-08 13:26
nickchan(i do not know the exact version used here)

plus ›