#asahi-dev · 2025-06-04 13:27
kettenisbut for some reason it is writable in my environment#asahi-dev · 2025-06-04 13:26
kettenishmm, I think the pre-defined stage1 version property string ends up in read-only memory#asahi-dev · 2025-06-04 09:22
chadmedactually that would just skip trying to set the prop#asahi-dev · 2025-06-04 09:21
chadmednickchan: instead of reverting can you try add the chosen_cnt = 0; line back?#asahi-dev · 2025-06-04 09:21
nickchanim not super sure the best way to properly fix it#asahi-dev · 2025-06-04 08:56
nickchanam going to revert it locally for now#asahi-dev · 2025-06-04 08:50
nickchansven: https://gist.github.com/asdfugil/70316615d46829202fa7f06d39416ef0#asahi-dev · 2025-06-04 08:36
nickchani will try to get more details first#asahi-dev · 2025-06-04 08:34
nickchanthe crash is very weird too it's a crash trying to access a variable's "="#asahi-dev · 2025-06-04 08:31
nickchan"payload: Fix m1n1 stage 1 version passing"#asahi-dev · 2025-06-04 08:31
nickchansince this commit trying to boot kernel on ipad with m1n1 would crash m1n1#asahi-dev · 2025-06-04 08:31
nickchan75329c51e280b23205a51547a7799369d4c5e23d is the first bad commit#asahi-dev · 2025-06-04 07:51
svenif the offset changed we can’t use aic2#asahi-dev · 2025-06-04 07:40
nickchanThen that doesn't break any existing device trees#asahi-dev · 2025-06-04 07:39
nickchani think it may be possible to keep the apple,aic2 compatible and have the aic2 values as default#asahi-dev · 2025-06-04 07:39
jannau"#asahi-dev · 2025-06-04 07:38
jannauthe only offsets that changed is AIC2_IRQ_CFG now specified in the ADT as "extint-baseaddress#asahi-dev · 2025-06-04 07:30
jannauI think m4 uses still "aic,3" and I'm not sure aic,3 differs that much from aic,2. not sure what changed beside moving offsets/strides to the ADT instead of hardcoding them#asahi-dev · 2025-06-04 07:30
sven*offsets #asahi-dev · 2025-06-04 07:29
svenI just don’t trust the ADT at all #asahi-dev · 2025-06-04 07:29
svenI’d really double check if any of the offers actually changed #asahi-dev · 2025-06-04 07:28
chadmedi feel like a massive idiot re that, because in the docs i wrote the part that was like "apple wont change aic2 for years it does everything they need" #asahi-dev · 2025-06-04 07:26
sveniirc robh also suggested that he would NAK aic3 because if the hw is changing once per generation anyway we may as well only use tXXXX-aic #asahi-dev · 2025-06-04 07:05
nickchanthere's some minor differences that don't line up but linux don't use them at all#asahi-dev · 2025-06-04 07:05
nickchanas far as linux is concerned two seems to line up pretty well so far though#asahi-dev · 2025-06-04 07:02
nickchantechnically speaking #asahi-dev · 2025-06-04 07:02
nickchanalso "aic" in ADT means the AIC block that handles IRQs but in Linux the "AIC" driver also handles FIQ even though technically speaker those are part of the CPU core complexes are not the AIC block#asahi-dev · 2025-06-04 07:00
nickchansaid#asahi-dev · 2025-06-04 07:00
nickchanso t6031 is definitely not compatible anyways follow what jannau set#asahi-dev · 2025-06-04 07:00
nickchant6000 is already incompatible#asahi-dev · 2025-06-04 07:00
nickchanfl0_id: the last soc to be compatible with "apple,aic" is t8103#asahi-dev · 2025-06-04 02:36
chadmed_fl0_id: sometimes those containers with multiple regs/sizes are for different parts of the peripheral. e.g. 0x4000-sized regs are probably the address of the asc mailbox associated with the peripheral. have a look at https://github.com/AsahiLinux/linux/blob/asahi-wip/arch/arm64/boot/dts/apple/t602x-die0.dtsi#L129#asahi-dev · 2025-06-03 14:00
fl0_idsth else - if there is not just one address and size for a register, is just the first one used for the dt, or is there a more complicated calculation e.g. how a set of five address / size pairs for a reg entry are transcribed?#asahi-dev · 2025-06-03 13:59
fl0_idmakes sense#asahi-dev · 2025-06-03 13:39
jannauiirc the offsets specified in the ADT differ from the aic2 values#asahi-dev · 2025-06-03 13:37
jannauI would read "aic,3" as "aic3"#asahi-dev · 2025-06-03 13:33
jannauI plan to send that (and the dt-bindings change) after 6.16-rc1\#asahi-dev · 2025-06-03 13:32
jannauDaniel and I followed Apple's v3 in m1n1 and linux https://github.com/jannau/linux/commit/3daf98a7eac1ba8ee972c2d9d381c1aba39cd7b2#asahi-dev · 2025-06-03 13:27
svenI know that v3 introduced some more properties for register offsets but they seems similar to v2#asahi-dev · 2025-06-03 13:27
sventhe adt node e.g. has a version property somewhere #asahi-dev · 2025-06-03 13:27
fl0_idok#asahi-dev · 2025-06-03 13:27
sventhey work differently there #asahi-dev · 2025-06-03 13:27
svenreally, don’t trust the ADT for stuff like compatibles #asahi-dev · 2025-06-03 13:26
svenit’s at least v2 for m3, maybe even v3 #asahi-dev · 2025-06-03 13:24
fl0_id(this is from the adt)#asahi-dev · 2025-06-03 13:23
fl0_idso I assume the s5l8960x is still there#asahi-dev · 2025-06-03 13:23
fl0_idnickchan are you sure that "apple,aic" cannot be compatible? while the adt has compatible = [aic,3] for the aic, the wdt has compatible = [wdt,t8101, wdt,s5l8960x]#asahi-dev · 2025-06-03 12:33
fl0_idjannau thx for explaining#asahi-dev · 2025-06-03 12:33
fl0_idnickchan thx, good to know that's what I was hoping. I have serial I think. not sure about wdt#asahi-dev · 2025-06-03 12:04
jannau#address-cells / #size-cells specify the layout of the "reg" property. address-cells = 1, #size-cells = 0 is perfectly valid in some places, for example for i2c devices