#asahi-dev · 2025-03-22 16:19
mischa85yeah it's frustrating. i'll see if *anything* in chickens.c works on t6041#asahi-dev · 2025-03-22 16:03
mischa85also freezes on the NEX powergating part: SYS_IMP_APL_HID13 / SYS_IMP_APL_HID14#asahi-dev · 2025-03-22 16:02
mischa85well i'm on t6041 duties again. fresh start. BRINGUP trying to see what's going on in cpu_init(). already freezes when we try to set SYS_IMP_APL_EHID4 / SYS_IMP_APL_HID4 which seemingly works on every other soc m1n1 has ever seen? is t6041 just an exception to the rule or is there something else going on? #asahi-dev · 2025-03-18 22:50
mischa85fl0_id thanks for your MR, i'll see if i missed some stuff and can try on t6041 too#asahi-dev · 2025-03-18 22:35
mischa85but it really seems to be correct#asahi-dev · 2025-03-18 22:35
mischa85*stole#asahi-dev · 2025-03-18 22:34
mischa85but i can't take credit for even that one because i still it from some other wip repo#asahi-dev · 2025-03-18 22:29
mischa85well there's at least one t6041 thing that can probably be added to m1n1: T6041 EARLY_UART_BASE: 0x429200000#asahi-dev · 2025-03-18 22:23
mischa85i'm almost 95% sure my chicken bits are wrong#asahi-dev · 2025-03-18 22:22
mischa85aight i'll try that then#asahi-dev · 2025-03-18 22:20
mischa85i have a custom bootpath, it does do mmu_init()#asahi-dev · 2025-03-18 22:19
mischa85at least it's not exploding#asahi-dev · 2025-03-18 22:18
mischa85sven: actually i did take over some mmu code, but i have no idea if it's correct#asahi-dev · 2025-03-18 22:17
mischa85apple seemed to be short on bits there so they went on to do some weird stuff there#asahi-dev · 2025-03-18 22:17
mischa85but the only problematic section for me was the pmgr section of course#asahi-dev · 2025-03-18 22:16
mischa85fl0_id: i actually had some fun in untangling the binary data adt.py spit out when i had it wrong#asahi-dev · 2025-03-18 22:13
mischa85noob question: the chicken bits are irrelevant in being able to/not being able to fire up the cores right?#asahi-dev · 2025-03-18 22:10
mischa85btw that repo is a mess but i will only start cleaning it up once i know at least a tiny little bit what is going on on m4#asahi-dev · 2025-03-18 22:07
mischa85i still think some of these are related to the power registers#asahi-dev · 2025-03-18 22:05
mischa85i did some more stuff to it, not only to get it to parse but also to index all the values i could find in it: https://github.com/mischa85/m1n1/commit/ad8c07411c9ac730615a141519652128a34a721a#diff-a41d8ed8cd219c858c64b5b44cc2f744f5199de7f65161b7e21a57aa3c18adf1#asahi-dev · 2025-03-16 19:30
mischa85upon closer inspection, those are also present in t6031 adt. misc-acg-offset = 0x98000, misc-cores-offset = 0x88000 are the same on m4/m3 but cluster-ctl-offset is 0x18000 on t6031 and 0x20000 on t6041#asahi-dev · 2025-03-16 17:22
mischa85thanks jannau#asahi-dev · 2025-03-16 17:22
mischa85i'm so glad i have a working shell for all this stuff#asahi-dev · 2025-03-16 17:21
mischa85so instead of feeding offset 0x20000 to smp_start_cpu i think i should be doing something else to that register#asahi-dev · 2025-03-16 17:19
mischa85the good news i mapped the whole register range by bruteforcing#asahi-dev · 2025-03-16 17:18
mischa85maybe this will be a hint to how pmgr is done on m4#asahi-dev · 2025-03-16 17:18
mischa85nope it just goes straight into firing the cores#asahi-dev · 2025-03-16 17:16
mischa85so my guess is that offset 0x20000 needs to be tickled in some way en then the cores are to be enabled from 0x88000 offset, same as m3#asahi-dev · 2025-03-16 17:15
mischa85or maybe i should be using my brain for a moment. the t6041 devicetree is full of mysteries but there are some new pairs in pmgr: misc-acg-offset = 0x98000, misc-cores-offset = 0x88000, cluster-ctl-offset = 0x20000#asahi-dev · 2025-03-16 16:10
mischa85okay i've exousted all the possibilities, cannot start the cpu. i have a list of addresses that result SError or Failed. only thing i could figure out is that offset 0x20000 fails, but the next call makes the system crash, fans go on (maybe safety mode?)#asahi-dev · 2025-03-16 10:41
mischa85but i gotta disable the bootcounter because it's going to be SError gallore and it boots into 1TR after a couple of "failed boots"#asahi-dev · 2025-03-16 10:39
mischa85not great, not terrible#asahi-dev · 2025-03-16 10:39
mischa85it's 1E0 options, 300#asahi-dev · 2025-03-16 10:39
mischa85only bad news is my math#asahi-dev · 2025-03-16 10:39
mischa85that's what i hoped for#asahi-dev · 2025-03-16 10:38
mischa85which would mean i have 0x1E = 30 options, considering the value of offset is always xxxxx1000#asahi-dev · 2025-03-16 10:37
mischa85correction: 0x502800000 is the first /arm-io/pmgr/reg, it has a size of 0x1E0000#asahi-dev · 2025-03-16 10:22
mischa85or am i chasing shadows?#asahi-dev · 2025-03-16 10:22
mischa85for the m1n1 devs: i'm trying to enable cpu0 like this: smp_start_cpu(0, 0, 0, 0, 0x210050000, 0x502800000 + cpu_offset) where i incremend cpu_offset by 0x1000 to find something that works. 0x210050000 (size: 0x9010) is the cpu-impl-reg of cpu0, and 0x502800000 is the start (end is 0x502818000) of dark-dcp/dapf-instance-0. all of this is of course without pmgr_init(). is this even possible like this #asahi-dev · 2025-03-15 23:22
mischa85it seems to be booting from cpu 4 from the sawtooth cores#asahi-dev · 2025-03-15 23:16
mischa85at least something turns on#asahi-dev · 2025-03-15 23:16
mischa85woosh#asahi-dev · 2025-03-15 23:16
mischa85okay start_offset 0x10000 makes the fans go vacuum#asahi-dev · 2025-03-15 17:09
mischa85or even better just create a function that allows me to set the cpu_start_off variable#asahi-dev · 2025-03-15 17:06
mischa85then i can brutefore it from shell#asahi-dev · 2025-03-15 17:06
mischa85i'll just alter smp_start_secundaries so can give it the start_off argument#asahi-dev · 2025-03-15 16:57
mischa85desperate times#asahi-dev · 2025-03-15 16:57
mischa85or can run on just the boot cpu maybe....#asahi-dev · 2025-03-15 16:52
mischa85i tried CPU_START_OFF_T6031 0x88000 but it's incorrect#asahi-dev · 2025-03-15 16:46
mischa85next destination hv