#
✕ nick:amarioguy
#asahi-re · 2022-10-27 12:39
amarioguymarcan: ah i see, that makes sense
#asahi-re · 2022-10-27 00:18
amarioguybut it may just be like that CYC_OVRD reg where 0 is on, 2 is off and 1 is.... just unused to my knowledge
#asahi-re · 2022-10-27 00:18
amarioguyprobably can gather a lot of the runtime data for values of 0x1 and 0x2 this weekend
#asahi-re · 2022-10-27 00:10
amarioguyit affects the heuristic that much is for sure
#asahi-re · 2022-10-27 00:10
amarioguyfwiw the register only supports setting bit 0 and bit 1 so i doubt you're gonna get much with only two bits
#asahi-re · 2022-10-27 00:08
amarioguymarcan: in other words it's time for scientific method testing :) (i'd realistically only need to test this on 2-4 cores or so and store the data on how many IRQs each core gets at what settings)
#asahi-re · 2022-10-26 17:18
amarioguy(probably unsurprising result here, but FIQs are totally unaffected by s3_4_c15_c10_4, makes sense as they're part of core complexes directly)
#asahi-re · 2022-10-26 16:39
amarioguyfrom my testing seems like BIT(0) or BIT(1) by themselves don't affect this (perhaps they're sort of a prioritization thing?)
#asahi-re · 2022-10-26 16:37
amarioguymarcan: update on that register - turns out *only* 0x3 will opt out the core
#asahi-re · 2022-10-26 16:37
amarioguyyes!! figured it out
#asahi-re · 2022-10-26 16:30
amarioguythink i almost have this figured out
#asahi-re · 2022-10-26 16:29
amarioguy...okay never mind lol was dumb
#asahi-re · 2022-10-26 16:21
amarioguyfurther testing with linux, with core 0 set to accept everything and cores 1-19 opted out of IRQs, doesn't affect anything
#asahi-re · 2022-10-26 14:55
amarioguy(also booting linux with the register changed seems to affect nothing (not sure if linux changes this register back or smth)
#asahi-re · 2022-10-26 14:54
amarioguyhmm been testing s3_4_c15_c10_4 (want to figure out what bits do what), doesn't seem to be doing anything (SYS_IMP_APL_CYC_OVRD does actually seem to disable irqs/fiqs) on t6002 (been poking it with m1n1, trying to raise IRQs to see what happens, but they seem to go off as they normally do)
#asahi-re · 2022-10-22 00:31
amarioguyhw:sep wiki page is updated with stuff for the record
#asahi-re · 2022-10-22 00:02
amarioguylooking forward to tackling this day over day, but think i did a lot for one day lol
#asahi-re · 2022-10-21 23:58
amarioguyokay lot of interesting findings about xART init (one big one is that the "tag" when fetching xART lockers seems to increase incrementally the more lockers it finds), i need to process this a bit more tomorrow but otherwise felt great doing the mmio tracing
#asahi-re · 2022-10-21 23:11
amarioguy(note: SKS uses the other kind of message, just found that out now
#asahi-re · 2022-10-21 22:12
amarioguyesp. since it seems only the "SEPMessage" is used in practice
#asahi-re · 2022-10-21 22:11
amarioguysven - btw why are there two different definitions of SEPMessage in the tracer? was that part of that "secondary" IPC you were talking about before
#asahi-re · 2022-10-21 22:05
amarioguy(this is explicitly printed in xnu bootlog btw)
#asahi-re · 2022-10-21 22:04
amarioguyponders why HDCP endpoint is turned on then off then on again
#asahi-re · 2022-10-21 22:04
amarioguyHW:SEP should be updated again by end of today
#asahi-re · 2022-10-21 22:03
amarioguy(not complaining or anything just a really interesting experience)
#asahi-re · 2022-10-21 22:02
amarioguyman the mmio tracing here is having me take a lot of disorganized notes lol
#asahi-re · 2022-10-21 17:06
amarioguyxnu single user mode is definitely helpful for this lol
#asahi-re · 2022-10-21 17:05
amarioguyis attempting to decode what the xART messages mean
#asahi-re · 2022-10-16 15:14
amarioguyah that makes sense
#asahi-re · 2022-10-16 15:06
amarioguy(though that is a bit lame :P, still w/e would be best for the project)
#asahi-re · 2022-10-16 15:01
amarioguy(though to be fair i can always just contribute to docs and have someone else write the driver for safety anyways)
#asahi-re · 2022-10-16 14:34
amarioguydon't want to taint myself or anything
#asahi-re · 2022-10-16 14:34
amarioguyi realize it may sound dumb but erring on the side of caution really
#asahi-re · 2022-10-16 14:33
amarioguyjust to clarify, does running strings count as "binary RE" in the context of this project
#asahi-re · 2022-10-16 14:33
amarioguyalright i'll verify that when i can
#asahi-re · 2022-10-16 14:31
amarioguy(that was an endpoint name that popped up in the tracer)
#asahi-re · 2022-10-16 14:30
amarioguyah gotcha, "xarm" isn't the endpoint for that?
#asahi-re · 2022-10-16 14:28
amarioguybut probably will stick to release tracing ngl if it's not that different it's probably just better
#asahi-re · 2022-10-16 14:27
amarioguyso idk if you want to run it yourself or smth
#asahi-re · 2022-10-16 14:27
amarioguyrequirement*
#asahi-re · 2022-10-16 14:27
amarioguythough idk if strings runs afoul of the "binary RE" clean-room rq
#asahi-re · 2022-10-16 14:25
amarioguystill i can run strings on both dev/release applesepkeystore to see if there's a major difference?
#asahi-re · 2022-10-16 14:25
amarioguyoh i see - probably just gonna stick to release
#asahi-re · 2022-10-16 14:24
amarioguyat minimum the sep key store driver might be more verbose?
#asahi-re · 2022-10-16 14:23
amarioguybtw i've had the idea of MMIO tracing using that dev kernelcache that apple accidentally sent out in the 12.3.1 IPSW
#asahi-re · 2022-10-16 14:22
amarioguyyea i should give that a look over again
#asahi-re · 2022-10-16 14:20
amarioguyah that makes sense
#asahi-re · 2022-10-16 14:19
amarioguyor is that just a random UUID generated per req
#asahi-re · 2022-10-16 14:19
amarioguysven: thanks! for UUID is that the UUID of the request being sent over the ASC mailbox?
#asahi-re · 2022-10-16 13:59
amarioguysven: maybe want to drop that gigalocker stuff?

plus ›