#
✕ nick:pipcet[m]
#asahi · 2021-05-08 08:55
pipcet[m]ramfs
#asahi · 2021-05-08 08:54
pipcet[m]all Linux needs is a usb port, and we already have polling for that.
#asahi · 2021-05-08 08:54
pipcet[m]why woud I need the nvme driver?
#asahi · 2021-05-08 08:54
pipcet[m]I don't think coexistence is going to be much of a problem, actually.
#asahi · 2021-05-08 08:53
pipcet[m]which doesn't go through AIC
#asahi · 2021-05-08 08:53
pipcet[m]no, it needs FIQ
#asahi · 2021-05-08 08:53
pipcet[m]plenty of ways of doing that.
#asahi · 2021-05-08 08:52
pipcet[m]or use a linux version without the actual AIC driver.
#asahi · 2021-05-08 08:52
pipcet[m]or modify the driver to coexist peacefully
#asahi · 2021-05-08 08:52
pipcet[m]yes, I'm already doing that.
#asahi · 2021-05-08 08:49
pipcet[m]but that's post-HV fine tuning, really.
#asahi · 2021-05-08 08:48
pipcet[m]sorry, I don't understand. qemu would still use its own page tables to translate simulated virtual to simulated physical, which would be equal to actual physical.
#asahi · 2021-05-08 08:46
pipcet[m]marcan: I have a plan for that (start macos physically, teach qemu to identity-map /dev/mem for the relevant ranges)
#asahi · 2021-05-08 07:14
pipcet[m]and we could pass through everything but USB, I think.
#asahi · 2021-05-08 07:11
pipcet[m]dottedmag: https://worthdoingbadly.com/xnuqemu3/ apparently worked at some point
#asahi · 2021-05-07 21:41
pipcet[m]I'm sorry, I'll try to be more careful
#asahi · 2021-05-07 19:27
pipcet[m]just found the touchbar frame buffer, so this hack isn't totally worthless.
#asahi · 2021-05-07 19:11
pipcet[m]https://eclecticlight.co/2020/11/28/startup-modes-for-m1-macs/ says 11.2, no idea what's true here
#asahi · 2021-05-07 19:03
pipcet[m]"every article on how macs boot" says verbose mode was removed recently.
#asahi · 2021-05-07 19:03
pipcet[m]"in addition to", in fact. The nvram variable doesn't do anything either.
#asahi · 2021-05-07 18:46
pipcet[m]tba.cmdline = "-v cpus=1"
#asahi · 2021-05-07 18:03
pipcet[m]marcan: yes, the hv is going to be much better. no -v yet, no idea what I'm doing wrong. There's an allowed-boot-args property but setting it doesn't make a difference.
#asahi · 2021-05-07 16:56
pipcet[m]ooh, and I just realized we don't even need to patch the macos kernel to disable the WDT, we can do it from the stolen core.
#asahi · 2021-05-07 16:52
pipcet[m](but m1n1 can access macos RAM; macos doesn't know about m1n1 RAM)
#asahi · 2021-05-07 16:51
pipcet[m]it's not really one "under" the other. the two run in separate RAM areas on separate cores and are oblivious to each other. It's a bit surprising this all works.
#asahi · 2021-05-07 16:50
pipcet[m](raw code is available for now, and I'm not home right now so can't test this very second, but I'd be delighted to get this working on another machine, or to see how it fails to)
#asahi · 2021-05-07 16:49
pipcet[m]thanks! I'll try to put up instructions
#asahi · 2021-05-07 13:08
pipcet[m]well, "under". MacOS has an efficiency core, m1n1 has a performance core but less RAM.
#asahi · 2021-05-07 13:07
pipcet[m]sven: your USB driver works in m1n1 under macos, too! Thanks again, it's absolutely wonderful to be able to play with this hardware.
#asahi · 2021-05-07 09:10
pipcet[m]but m1n1 works.
#asahi · 2021-05-07 08:57
pipcet[m]unfortunately, the boot framebuffer is no longer used at that point.
#asahi · 2021-05-07 08:34
pipcet[m]okay, I can steal a CPU core from MacOS
#asahi · 2021-05-07 00:22
pipcet[m]good point. I was thinking maybe apple did something like rvbar = our_address &~ 0xffff, but maybe it's something else entirely. difficult to work out as long as "-v" doesn't work.
#asahi · 2021-05-06 17:23
pipcet[m]maybe "-v" counts as a custom bootarg and "-noprogress" doesn't, somehow.
#asahi · 2021-05-06 17:22
pipcet[m]I can pass bootargs, they have an effect (in the case of "-noprogress"), but if I pass "-v" instead of "-noprogress", nothing happens.
#asahi · 2021-05-06 17:21
pipcet[m]at least it did for my linux images; they were aligned to 128KB or something, not to 2MB as I'd requested.
#asahi · 2021-05-06 17:21
pipcet[m]marcan: I'm pretty sure iBoot2 looks at the alignment property of the macho sections it loads.
#asahi · 2021-05-06 16:10
pipcet[m]because I think what might be happening is that macos assumes it can use a physical address to write to the secondary CPUs' rvbar register, and that needs to be aligned to, I think, 64 KB
#asahi · 2021-05-06 16:09
pipcet[m]just segments?
#asahi · 2021-05-06 16:09
pipcet[m]you're not using macho sections for m1n1, are you?
#asahi · 2021-05-06 16:07
pipcet[m]just "-v"? No debug= or anything?
#asahi · 2021-05-06 15:43
pipcet[m]any way to tell macos to put anything but a progress bar there? -noprogress still works, the other options that used to control kernel output don't.
#asahi · 2021-05-06 15:40
pipcet[m]framebuffer would be ideal, yes.
#asahi · 2021-05-06 15:35
pipcet[m]of course if the hypervisor is done before then that would be much, much bett.r
#asahi · 2021-05-06 15:35
pipcet[m]I can't debug further unless someone tells me the magic trick to get macos kernel debug output without uart access :-)
#asahi · 2021-05-06 15:34
pipcet[m]had to do that to get it to boot, and also I'm still planning to steal CPU cores from macos.
#asahi · 2021-05-06 14:41
pipcet[m]load average at 52, it's not a happy system on a single efficiency core
#asahi · 2021-05-06 14:40
pipcet[m]currently trying to whittle down my various changes to what's actually necessary, will open a PR then :-)
#asahi · 2021-05-06 14:38
pipcet[m]and this time I'm on the right architecture, too!
#asahi · 2021-05-06 14:37
pipcet[m]single-core, but still

plus ›