#asahi · 2021-11-21 16:27
marcanpresumably it is provisioned via that i2c link#asahi · 2021-11-21 16:27
marcansupposedly it talks IPv6/TLS using the onboard CPU and can turn the computer on and off#asahi · 2021-11-21 16:24
marcanhttps://support.apple.com/en-gb/guide/mdm/mdm580cf25bc/web#asahi · 2021-11-21 16:22
marcansilly overloaded terms#asahi · 2021-11-21 16:22
marcanoh sorry, "lights out management"#asahi · 2021-11-21 16:22
marcanthe second MAC would be for the BMC#asahi · 2021-11-21 16:22
marcanthis is server terminology#asahi · 2021-11-21 16:21
marcanlom = "lan on motherboard"#asahi · 2021-11-21 16:21
marcanfor the BMC#asahi · 2021-11-21 16:20
marcanso maybe indeed it does refuse to train at Gen1? if so try my patch#asahi · 2021-11-21 16:20
marcanhm, the datasheet says Gen2/Gen3#asahi · 2021-11-21 16:19
marcanthis has to do with the BMC type functionality that is supposedly supported on these macs maybe?#asahi · 2021-11-21 16:18
marcanaqc-a on i2c3#asahi · 2021-11-21 16:17
marcanthere is an i2c device associated with the 10ge#asahi · 2021-11-21 16:13
marcanyou can try this to fix the link speed problem, but I doubt that will make it work (it should still negotiate at Gen1): https://github.com/AsahiLinux/linux/commit/35ee181fdfdea4c5c87b2365c6f3de707ddef745#asahi · 2021-11-21 16:12
marcanah yes, that one#asahi · 2021-11-21 16:10
marcanISTR maz mentioned something about PERSTN?#asahi · 2021-11-21 16:09
marcanbut if the link isn't up then it isn't up#asahi · 2021-11-21 16:06
marcanjannau: pull/125 does nothing because Linux currently ignores the max speed prop, see -dev :-)#asahi · 2021-11-21 09:45
marcanso it's a bit weird that it wouldn't work at all#asahi · 2021-11-21 09:45
marcanOTOH, I would *expect* both sides to fall back to an older gen if things go wrong during link training#asahi · 2021-11-21 09:45
marcanit's entirely possible that Gen4 needs more initialization than we currently do#asahi · 2021-11-21 09:44
marcanso all links are 1x on t8103 and t600x (at least on current devices; not sure if they might be able to bond links into one root port, maybe? but not on current devices)#asahi · 2021-11-21 09:43
marcanyeah, definitely only 3 lanes on t8103#asahi · 2021-11-21 09:42
marcan(which also means the 16 bound in your patch isn't quite right :))#asahi · 2021-11-21 09:42
marcanI'm pretty sure t8103 only has 3 lanes total and t600x only has 4 lanes total#asahi · 2021-11-21 09:41
marcanmaximum-link-speed = 4 probably means PCIe Gen4#asahi · 2021-11-21 09:41
marcanI think you're confusing link speed with link width#asahi · 2021-11-21 08:54
marcanI guess 10ge would just about saturate a PCIe 3.0 link so maybe they do only need one lane#asahi · 2021-11-21 08:53
marcanthen again on the M1 Mini with 3 physical ports, how would they do that?#asahi · 2021-11-21 08:52
marcanI think all the other devices are 1x?#asahi · 2021-11-21 08:52
marcanI assume the 10ge variant has more than a 1x pcie link#asahi · 2021-11-21 08:52
marcanI'm wondering how link bonding works#asahi · 2021-11-21 08:52
marcando we have an ADT from that variant?#asahi · 2021-11-21 08:37
marcanit's not PCI#asahi · 2021-11-20 13:56
marcanthe way it does firmware is horrible#asahi · 2021-11-20 13:56
marcanI really need to get to that wifi patch#asahi · 2021-11-20 03:56
marcanj_ey: fwiw m1n1 should now propagate USB disabling from the HV to the dtb, so you no longer need to comment out the dwc3 to avoid it stealing it#asahi · 2021-11-20 00:28
marcans/repo/branch/#asahi · 2021-11-20 00:24
marcanRealityVoid: the old repo probably had AIC code that does not work as a VM guest (which needs to use different registers), hence the undefined instruction exception.#asahi · 2021-11-19 18:19
marcanyeah, it's easy enough to implement anyway#asahi · 2021-11-19 18:08
marcansince m1n1 can pass a kernel command line to u-boot (configured at installation time), we could use that to let users preconfigure u-boot a certain way at install time at least (or even via passing something through from boot-args if needed)#asahi · 2021-11-19 18:07
marcanyeah, I figured there's something for that#asahi · 2021-11-19 18:04
marcan(plus in theory we should eventually have support for setting boot-args from linux too)#asahi · 2021-11-19 18:03
marcanbut it seems convenient to be able to flip an existing install to proxy mode without throwing in a whole new m1n1#asahi · 2021-11-19 18:03
marcanit's up to people how they do this, of course#asahi · 2021-11-19 18:03
marcane.g. once we get SEP involved this might matter#asahi · 2021-11-19 18:02
marcanyes, but that's another 2.5G and a separate install, plus it makes it harder to test certain things because at that point it's a separate OS identity#asahi · 2021-11-19 17:58
marcanso people who want proxy mode would have to toggle it with bputil first (which requires auth), then set boot-args#asahi · 2021-11-19 17:57
marcanI wouldn't disable boot-args restriction by default since this is not password-gated and thus a backdoor