#asahi-re · 2025-10-16 07:57
jannauthey also broke adt parsing with having a "reg" property in /arm-io/dptx[01]-aux/dptx[01]-aux-core without #{address,size}-cells in its parent#asahi-re · 2025-10-16 07:51
svenpossibly, or they got better at preventing leaks#asahi-re · 2025-10-16 07:31
chadmedhave they just given up naming their cpu cores then?#asahi-re · 2025-10-16 07:14
jannaustill "aic,3", "t8110-dart", "t6020-pcie", "bcm4387" and of course everest/sawtooth compatible cpu cores#asahi-re · 2025-10-16 06:12
jannauthere's aop and aop2 firmware#asahi-re · 2025-10-16 02:11
nicolas17there's now exclaves on mac#asahi-re · 2025-10-16 01:32
nicolas17wait Apple screwed up and didn't upload the fcs/wkms keys so you won't be able to mount the fs -.- can look at kernelcache and boot components though#asahi-re · 2025-10-16 00:39
nicolas17MacBook Pro 14" M5 (J704AP, Mac17,2): https://updates.cdn-apple.com/2025FallFCS/fullrestores/093-49300/3AC98479-C75E-46E8-8AF4-1DDB5EAE564D/UniversalMac_26.0.1_25A8364_Restore.ipsw#asahi-re · 2025-10-10 10:58
nnymSounds good. I'll check it out.#asahi-re · 2025-10-10 10:56
chaos_princessit is mostly all there, template has some keys zeroed out, or specified as references to per-system info. all nodes are there iirc almost all keys also are#asahi-re · 2025-10-10 10:54
nnymI think that there's too much information for me to capture with my phone. What kind of information is missing from the IPSW?#asahi-re · 2025-10-10 09:28
nnymThanks. I want to see the whole devicetree so I'll use the boot argument.#asahi-re · 2025-10-10 09:06
chadmedyes the IPSWs for all the macos versions will have "default" ADTs in them for each machine, you just need to know what jxxx model your machine is#asahi-re · 2025-10-10 09:06
chadmedi interpreted that "only" as a signifier of simplicity rather than "is there no other possible way" #asahi-re · 2025-10-10 08:56
chaos_princessdepending on what specific ip block you are poking at, the template adt from the ipsw can be good enough#asahi-re · 2025-10-10 08:53
chadmedsee rust/src/adt.rs for how to parse it#asahi-re · 2025-10-10 08:53
chadmedyes one of the iboot boot args is a pointer to the head of the device tree#asahi-re · 2025-10-10 07:58
nnymDo I understand correctly that I can get the complete devicetree only from the boot arguments?#asahi-re · 2025-10-07 13:42
nnymOr the other way around. Stacks confuse me.#asahi-re · 2025-10-07 13:01
nnymBut I think that that should be simple because that means that the next frame pointer is bigger instead of being smaller.#asahi-re · 2025-10-07 12:58
nnymDoes preventing a fault require any more than verifying that the frame pointer and the PC point to the right sections before dereferencing them? And if getting stuck otherwise means following a bad frame pointer that causes a cycle, I don't check that.#asahi-re · 2025-10-07 11:55
svenprinting a stack trace is a bit tricky because that can easily cause an additional fault or get stuck otherwise#asahi-re · 2025-10-07 11:53
nnymm1n1 didn't seem to be able to print a stack trace so I added that to `print_regs`. Would it be desirable upstream?#asahi-re · 2025-10-07 08:25
nnymLooks like the value of `_text_size` is actually the value at the address `. - _text_start` so I have to take its adddress but that just gives me a meaningless virtual address.#asahi-re · 2025-10-07 07:59
nnymI declared `extern size_t _text_size` matching the symbol in `m1n1-raw.ld` and when I printed it I got 0x6162203a43495045 which is bigger than the entire linear address space. `_text_size` is just `. - _text_start` immediately after the text section.#asahi-re · 2025-10-06 07:36
nnymIt works but I had to subtract 4 or pass `-i` to get the right file; no big deal. I can also get the column if I `aarch64-linux-gnu-objdump -dW` but that's more involved. Anywway, this makes things much easier; thanks.#asahi-re · 2025-10-06 07:23
nnymNope.#asahi-re · 2025-10-06 07:23
chaos_princesstried addr2line?#asahi-re · 2025-10-06 07:22
nnymwithout `--strip-debug` but the resulting flat binary file didn't boot at all so I couldn't print the source location.#asahi-re · 2025-10-06 07:22
nnymI need a good method of finding the piece of source code associated with an instruction. Currently I have 5-second delay after an unhandled exception so I find the IP and then disassemble and search `m1n1-raw.elf` for that address and try to figure out which statement or expression produced it; it takes a while. I also tried `aarch64-linux-gnu-objcopy -O binary build/m1n1.elf`#asahi-re · 2025-10-06 00:38
nnymIt cleared my kernelcaches so I copied the last one from my other computer over, installed it, and rebooted; and now m1n1 behaves as before.#asahi-re · 2025-10-06 00:30
nnymI ran `bputil -nkcas` and `csrutil disable` again and now `/dev/disk3s2` is automatically mounted at `/Volumes/Preboot`.#asahi-re · 2025-10-06 00:24
nnymI think that I've tested all of them now and they all ran for less than a second but I know that m1n1 was running for longer—around a minute perhaps—apparently calling `print_regs` over and over until it stopped for a while, the screen became blurry, and the computer rebooted.#asahi-re · 2025-10-05 23:59
nnymI've gone through almost all of my kernelcaches—including the first—and every one of those that was built without `BRINGUP` exits almost immediately. I think that something changed outside my m1n1 and I think that it started after I ran m1n1 with `BRINGUP`. Is there a connection?#asahi-re · 2025-10-05 23:13
nnymI was editing a script in Vim and recoveryOS crashed. Is this normal?#asahi-re · 2025-10-05 22:56
nnymThanks.#asahi-re · 2025-10-05 22:56
nnymOh yeah. They're all there.#asahi-re · 2025-10-05 22:53
nnymOh that's what you mean. `/System/Volumes/Preboot` is a tmpfs which is not `/dev/disk3s2` ("Preboot"). Let me see.#asahi-re · 2025-10-05 22:50
chaos_princesshave you confirmed via diskutil or something that /s/v/preboot has what you expect mounted#asahi-re · 2025-10-05 22:43
nnymThe UUIDs match my volume groups but there's only the official kernel in each.#asahi-re · 2025-10-05 22:41
nnym/System/Volumes/Preboot/54B91C04-0D42-448A-806D-E2251A3F1DF6/boot/FC163BEB29178A53F9CC278EFA5992045C407B8A203446FF6CCA5B4F57D3A6D19070A1D84CF967EAEE8FC737B4034FB6/System/Library/Caches/com.apple.kernelcaches/kernelcache#asahi-re · 2025-10-05 22:41
nnym/System/Volumes/Preboot/EF3C096E-D78F-49E3-AD0F-26A11FF83EE6/boot/FC163BEB29178A53F9CC278EFA5992045C407B8A203446FF6CCA5B4F57D3A6D19070A1D84CF967EAEE8FC737B4034FB6/System/Library/Caches/com.apple.kernelcaches/kernelcache#asahi-re · 2025-10-05 22:41
nnymOnly `kernelcache`. I powered off and booted recoveryOS but I still see only `kernelcache`.#asahi-re · 2025-10-05 22:36
nnymI ran `kmutil configure-boot` with the XNU kernelcache and then m1n1 again and now I'm in recoveryOS but the directories in `/System/Volumes/Preboot` don't match my volume groups anymore. Also, I ran `find / -name 'kernelcache*'` (or `kernelcache.*`) earlier and it didn't find my kernelcache. I'll take a look here before rebooting.#asahi-re · 2025-10-05 22:26
chaos_princesscheck if it mounted the thing you expect to /s/v/preboot, the kernelcache should be there and named sth like kernelcache.custom.<hash>#asahi-re · 2025-10-05 22:24
nnymYeah.#asahi-re · 2025-10-05 22:23
chaos_princessyou are using the right recovery volume, correct?#asahi-re · 2025-10-05 22:22
nnymAlso, I wanted to look at the current kernelcache in recoveryOS but there seems to be only the official XNU kernelcache. (There's only the single directory whose name is the hash of some `apticket.der` in `/System/Volumes/Preboot/<long hash>/boot/<the hash>`.)#asahi-re · 2025-10-05 22:22
chaos_princessfresh builds are about a mb, goes up to 4 on this machine with dts+u-boot#asahi-re · 2025-10-05 22:21
nnymI see. Perhaps I misremembered.