#asahi-dev · 2026-01-23 12:41
yukahttps://cyberchaos.dev/yuka/linux/-/commit/864d804931e5d9864d9dea55312623bf29db9987#8d44d32e1d90e744a2166a88d974331d06c3fc15_0_22#asahi-dev · 2026-01-23 12:40
yukait is#asahi-dev · 2026-01-23 12:40
jannauplain earlycon might depend on /chosen/stdout-path being set correctly#asahi-dev · 2026-01-23 12:38
yukaat first I assumed earlycon just wasn't set up early enough for this#asahi-dev · 2026-01-23 12:38
yukajust earlycon doesn't work#asahi-dev · 2026-01-23 12:38
yukaYes!#asahi-dev · 2026-01-23 12:38
svenhuh, kinda surprised just earlycon isn’t enough #asahi-dev · 2026-01-23 12:37
yuka(without commenting out cpuhp_setup_state())#asahi-dev · 2026-01-23 12:37
yukais the info I'm getting now#asahi-dev · 2026-01-23 12:37
yuka> Internal error: Oops - Undefined instruction: 0000000002000000 [#1] SMP#asahi-dev · 2026-01-23 12:37
fl0_idok nice. I can try it only tonight but I'm excited now#asahi-dev · 2026-01-23 12:37
yukayes#asahi-dev · 2026-01-23 12:37
fl0_idis your address the early_uart_base?#asahi-dev · 2026-01-23 12:36
fl0_idyuka oh I should try that too...#asahi-dev · 2026-01-23 12:36
yukafl0_id you should also try earlycon=s5l,[addr]#asahi-dev · 2026-01-23 12:36
fl0_idnoopwafel not much, I have only suspicions. I added uart to my device tree, but that didn't seem to do anything. I rebased on asahi-wip, no observable change. somewhere along there I lost my symbols for whatever reason despite using the same config...#asahi-dev · 2026-01-23 12:36
noopwafel\o/#asahi-dev · 2026-01-23 12:36
yukaand now I'm actually getting useful output before the crash is happening!#asahi-dev · 2026-01-23 12:35
yukaadded earlycon=s5l,0x3ad200000 to my bootargs#asahi-dev · 2026-01-23 12:35
yukaOk I just found out how I could've skipped all the hacky debug_putc stuff#asahi-dev · 2026-01-23 12:33
noopwafeldid you get anywhere on working out what is broken, fl0_id?#asahi-dev · 2026-01-23 12:10
fl0_idat least after enabling early_uart_base and putc I have some console output from the kernel, so at least it works in theory#asahi-dev · 2026-01-23 12:07
fl0_idor a memory function that's used in lots of places#asahi-dev · 2026-01-23 12:07
fl0_id:/ feels almost like once when i put a hvc in bsearch or sth like that and then ofc it was like everywhere lol#asahi-dev · 2026-01-23 12:01
yukabut then placing my print after the __cpuhp_setup_state() call in cpu_setup_state(), it never happens#asahi-dev · 2026-01-23 12:01
yukaplacing my print in the last line (or any line really) of __cpuhp_setup_state(), the print happens multiple times which is confusing#asahi-dev · 2026-01-23 12:00
yukawhich calls __cpuhp_setup_state()#asahi-dev · 2026-01-23 12:00
yukacpuhp_setup_state() is an inlined thing in cpuhotplug.h#asahi-dev · 2026-01-23 11:59
yukamy printline debugging methods are definitely hitting limits#asahi-dev · 2026-01-23 11:59
yukaI'm so confused about cpuhp_setup_state() call in irq-apple-aic.c#asahi-dev · 2026-01-22 22:06
fl0_id__ok thx#asahi-dev · 2026-01-22 22:05
jannauboth. the hypervisor has a virtual uart at the same mmio address#asahi-dev · 2026-01-22 21:57
fl0_id__question... is early uart_base in m1n1 for the m1n1 proxy console/fb, or does the guest use this too? sry if this should be obvious#asahi-dev · 2026-01-22 09:40
sventhe PR looks like a good step towards m4 support to me then :)#asahi-dev · 2026-01-22 09:24
jannausven: indeed. I missed "new_base = u.base" and assumed new_base is where the binary is is loaded#asahi-dev · 2026-01-22 09:01
yukait "just works" (at least with chainload.py --raw)#asahi-dev · 2026-01-22 09:00
yukathis includes chainloading a second m1n1 and calling smp_start_secondaries()#asahi-dev · 2026-01-22 09:00
yukaI did not blindly remove or skip the rvbar write, meaning it never mismatched what was already there, or else it would've tried to write it and crashed#asahi-dev · 2026-01-22 08:59
svenhuh, i think it already tries to do that unless i'm reading the code wrong#asahi-dev · 2026-01-22 08:57
sveni guess we could also adjust chainload.py to just move the next stage to the same base address the previous one was #asahi-dev · 2026-01-22 08:55
yukawait, no#asahi-dev · 2026-01-22 08:54
yukajannau: chainload.py already starts the cores up with the stage-1 base#asahi-dev · 2026-01-22 08:48
sventrue, we could just put a "branch to current stage entry" stub there on chainload#asahi-dev · 2026-01-22 08:47
jannauI don't think we reuse the previous stage memory after jumping to the next stage so it should be enough to patch the old entry points#asahi-dev · 2026-01-22 08:43
svenif they’re locked by iboot we probably need to keep a page at rvbar „forever“ that catches them and forwards them to the currently running m1n1#asahi-dev · 2026-01-22 08:42
sventhe cores come up at least? Nice, last time someone tried they couldn’t even get them up#asahi-dev · 2026-01-22 08:40
jannauyuka: what do you mean by smp works in m1n1? With locked rvbar and your PR I would expect secondary startup to be broken after chainloading#asahi-dev · 2026-01-22 07:54
jannaubut yes if github pushes annoying non optional copilot features faster than they run out of money we should discuss switching#asahi-dev · 2026-01-22 07:43
NickChan[m]as to what that means it's up to anybody's guess#asahi-dev · 2026-01-22 07:42
NickChan[m]the option defaults to "select an option" by the way