#asahi-dev · 2026-01-06 02:41
fl0_idis there another symbol to hook?#asahi-dev · 2026-01-06 02:41
fl0_idseems like stepping through works now. but the break point does not seem to work. I drop to shell with ctrl ^ c, set_sym_bp("start_kernel"), continue with ctrl^d, but this just runs to the "end".#asahi-dev · 2026-01-06 02:39
integralpilothmmm i've been doing some experiments with M3 GPU and the actual kernel DRM driver seems to be doing okay and might just need a little bit of modification unless i'm interpreting what's happening wrong (don't know too much about graphics)? i assume most of the work would then would be in mesa / the userspace stuff#asahi-dev · 2026-01-05 22:46
fl0_idthx#asahi-dev · 2026-01-05 22:42
integralpilot*then#asahi-dev · 2026-01-05 22:40
integralpilotfor BPs as soon as you see EL1 m1n1 come on, immediately press ^C before it starts Linux and you can set a breakpoint the#asahi-dev · 2026-01-05 22:00
fl0_id__integralpilot - as you were asking about backtrace before https://www.pastery.net/auaccd/#asahi-dev · 2026-01-05 21:04
fl0_id__dumb question maybe, what is the best way to add a breakpoint? I use run_guest_kernel.sh with a shell option. then I load System.map for symbols. however at this point, this fails because I self.ctx.cpu_id is not available yet. if I continue for a bit, then drop to a shell it works, but then the breakpoint didn't work. so clearly I am missing sth#asahi-dev · 2026-01-05 19:56
integralpilotthe hypervisor’s own shell is quite well featured though and has like everything you could need, thanks to the ppl who made it!#asahi-dev · 2026-01-05 19:54
integralpilotgdbserver didn’t work for me :(#asahi-dev · 2026-01-05 19:53
jannauhv has gdbserver support but it's also a while since I used that#asahi-dev · 2026-01-05 19:53
integralpilotstep has been working well for me, used it extensively when doing XNU bring up#asahi-dev · 2026-01-05 19:49
chaos_princessit somewhat worked recently while i was doing some macos experiments#asahi-dev · 2026-01-05 19:48
svenwell, 4.5 years. but it's rather simple and shouldn't really break#asahi-dev · 2026-01-05 19:47
fl0_id__haha#asahi-dev · 2026-01-05 19:47
sven*:D#asahi-dev · 2026-01-05 19:47
svenlooks like i added support for that 5 years ago and probably last tested it around that time as well :d#asahi-dev · 2026-01-05 19:47
fl0_id__thx#asahi-dev · 2026-01-05 19:46
svenset a breakpoint on kernel init, then just use step() from the hv shell#asahi-dev · 2026-01-05 19:46
fl0_id__sven: what's the best way to test this? I could drop to a HV shell.#asahi-dev · 2026-01-05 19:45
svenif single stepping is still working (unlike el1 calls *cough*) you can also single step through the kernel#asahi-dev · 2026-01-05 19:45
fl0_id__sven: thx#asahi-dev · 2026-01-05 19:45
fl0_id__ok. have to think about my setup then, as right now I use a thinkpad x230 as control device, which is fine, but it does not compile the kernel quickly...#asahi-dev · 2026-01-05 19:44
noopwafelscattered uart writes in the kernel are definitely your friend#asahi-dev · 2026-01-05 19:44
svenkernel, you're already past m1n1 at this point#asahi-dev · 2026-01-05 19:43
fl0_id__jannau: mmmh. you mean into the m1n1 part or the kernel itself?#asahi-dev · 2026-01-05 19:43
fl0_id__mmh seems to be about the same with earlycon. just asahi logo and a lot of pmgr "dangerous write", and mmu shutting down#asahi-dev · 2026-01-05 19:42
jannaumight be even worth sprinkling manual writes to the uart into linux early startup code if you don't where you're hanging. iirc marcan did that for the initial bringup#asahi-dev · 2026-01-05 19:40
fl0_id__sven: yeah, that was just b/c i was experimenting and wanted to remove all boot options#asahi-dev · 2026-01-05 19:39
svenyou're very blind otherwise#asahi-dev · 2026-01-05 19:39
svenstrongly recommend to always keep it enabled#asahi-dev · 2026-01-05 19:38
fl0_id__sven: not at the moment, because it didn't seem to make a difference#asahi-dev · 2026-01-05 19:38
fl0_id__both somewhat abbreviated#asahi-dev · 2026-01-05 19:38
fl0_id__and this is the usual m1n1 output on first port https://www.pastery.net/jcvrka/#asahi-dev · 2026-01-05 19:37
svenalso, do you have earlycon enabled?#asahi-dev · 2026-01-05 19:35
integralpilotif you press ^C and then type “bt”, what’s the kernel up to?#asahi-dev · 2026-01-05 19:30
fl0_id__this is the serial output from the second port https://www.pastery.net/tzbddr/#asahi-dev · 2026-01-05 19:29
fl0_id__integralpilot thx. yeah skipping cpufreq stuff got me further.#asahi-dev · 2026-01-05 19:16
integralpilotbut if it’s causing you issues there’s no harm in just skipping m1n1 cpufreq for now as it’s just early bring up so not important#asahi-dev · 2026-01-05 19:15
integralpilotI know t8122 does that not supported thing, got mixed up not sure about t6030 I’ll see#asahi-dev · 2026-01-05 19:14
integralpilotwait sorry I might be remembering wrong#asahi-dev · 2026-01-05 19:13
integralpilott6030 just goes “cpufreq: not supported on this chip” and m1n1 doesnt bother (though Linux has working cpufreq b/c of the device tree I made) so perhaps your issue is with m1n1 cpufreq somewhere and you should just skip it for now?#asahi-dev · 2026-01-05 18:54
fl0_id_looking at differences between t6030 and t6031 - is there a reason why in src/cpufreq, some switch statements only include t6031, not t6030 (but all other devices) or did integralpilot or others add this locally? if this was different between these, it would seem like a pretty big change. selectively disabling things did not help however haha#asahi-dev · 2026-01-05 17:13
fl0_id_(and good, then i don't have to check there)#asahi-dev · 2026-01-05 17:12
fl0_id_svenpeter[m]: probably, from src/cpufreq.c I just thought maybe it already jumped / started the kernel as it is after uncompressing the kernel. and m1n1(proxy) is presumably running on that one core, so what's the problem (for m1n1) ^^ I tried different values for the counter, but they don't seem to change anything. Would be just like me to have overlooked sth incredibly basic ... but not as far as i could find#asahi-dev · 2026-01-05 15:27
svenhttps://github.com/AsahiLinux/linux/tree/bits/080-wifi Is the branch with the wifi patches #asahi-dev · 2026-01-05 15:21
svenvery nice!#asahi-dev · 2026-01-05 15:02
ellyqwsven: i can help with that, sure :D#asahi-dev · 2026-01-05 14:55
jannau"right now" == "this cycle", still cleanup to do. driver is very chatty#asahi-dev · 2026-01-05 14:53
sveni like that "submit right now" part 😁