#
✕ nick:marcan
#asahi · 2021-08-08 18:36
marcansort of like linux-next
#asahi · 2021-08-08 18:36
marcanso people have something to pull and use
#asahi · 2021-08-08 18:36
marcanprobably just another repo with a thing that auto-rewrites a branch with a merge of "useful" stuff that is in development
#asahi · 2021-08-08 18:36
marcanonce I get the installer working I'll see if I can put together a workflow for our linux branch
#asahi · 2021-08-07 21:34
marcananyway, sleep ;)
#asahi · 2021-08-07 21:33
marcansure :p
#asahi · 2021-08-07 21:33
marcanso I'd need a separate error return for stop with error vs last payload, continue booting
#asahi · 2021-08-07 21:32
marcanthis is because the kernel header has a size that isn't the real file size, it's padded, so it's no good for finding the next thing, but it's good enough for loading the kernel
#asahi · 2021-08-07 21:32
marcanand that does return NULL because it means it can't find the next payload if there were one, but it *does* boot it
#asahi · 2021-08-07 21:32
marcanthe thing is, e.g. having the *kernel* uncompressed as the last payload *is* legal
#asahi · 2021-08-07 21:32
marcanyes, it's not a fatal error. maybe it should be :)
#asahi · 2021-08-07 21:31
marcanif you're using the hypervisor it will go to the secondary iface as vuart; if you're just chainloading it won't go anywhere because after the chainload is done your connection drops and presumably you aren't reconnecting quickly enough after that to get the spew before linux boots
#asahi · 2021-08-07 21:29
marcanyou *should* have gotten that
#asahi · 2021-08-07 21:29
marcan printf("Uncompressed cpio archives not supported\n");
#asahi · 2021-08-07 21:28
marcansince you can concatenate multiple things and it needs to be able to find the next one
#asahi · 2021-08-07 21:28
marcanso it requires compression
#asahi · 2021-08-07 21:28
marcanyeah, it needs to know the size and currently makes no attempt at walking a cpio to find the end
#asahi · 2021-08-07 21:26
marcanpeople actually want to run this stuff now and it's silly that everyone has to fight with one thing or another :-)
#asahi · 2021-08-07 21:25
marcanand see, this is why I really should spend some time building an installer and a kernel merge/something pipeline and simple docs that work for everyone
#asahi · 2021-08-07 21:24
marcan*at 12h
#asahi · 2021-08-07 21:24
marcanstarting 12h and continuing for 24h after that
#asahi · 2021-08-07 21:24
marcanso I guess overall 36h total effect after the shot
#asahi · 2021-08-07 21:24
marcanyeah, 12h after the shot (at night) I got a fever, 38.5 ish, hitting 39 at times. then it went down to 37.5 and fluctuated; the first half of the next day I was tired / sore. It fluctuated a bit, shot up to 38 again briefly in the afternoon, then seems to have just gone away. <37 now.
#asahi · 2021-08-07 21:22
marcanhah!
#asahi · 2021-08-07 21:22
marcanthat's interesting that you don't without it then :D
#asahi · 2021-08-07 21:20
marcanI think I've mostly recovered from the second shot
#asahi · 2021-08-07 21:20
marcangotta get some sleep :)
#asahi · 2021-08-07 21:20
marcanif you don't get a framebuffer that sounds like a kernel config problem
#asahi · 2021-08-07 21:16
marcanonce m1n1 hands off control USB is dead
#asahi · 2021-08-07 21:16
marcanthat looks normal
#asahi · 2021-08-07 21:11
marcanLB is intended to be a FAT32 EFI partition for u-boot/grub/kernel, LR would be reformatted as ext4 for / or whatever, but that's just one possible set-up
#asahi · 2021-08-07 21:11
marcanj_ey: LB/LR are just placeholders for linux partitions
#asahi · 2021-08-06 15:49
marcanand also if I can somehow determine if it's for icache/dcache or something else (maybe always ignore dcache ops and always turn icache invals into dcache flush/icache inval?) I can just disable all device cache management... and see what breaks
#asahi · 2021-08-06 15:48
marcanmight reveal some stuff
#asahi · 2021-08-06 15:48
marcanI think there's a trap bit for cache ops in the VM stuff, maybe I should turn that on and see how much cache management macos does
#asahi · 2021-08-06 15:46
marcanso far we have no evidence of *anything* being noncoherent on M1
#asahi · 2021-08-06 15:45
marcanyup
#asahi · 2021-08-06 15:43
marcanthought so
#asahi · 2021-08-06 15:43
marcan/home/marcan/asahi/git/linux/arch/arm64/include/asm/io.h:#define ioremap_wc(addr, size) __ioremap((addr), (size), __pgprot(PROT_NORMAL_NC))
#asahi · 2021-08-06 15:43
marcanyeah
#asahi · 2021-08-06 15:43
marcan(I forget how this ended up :))
#asahi · 2021-08-06 15:43
marcanand do we even do _wc? it probably falls back to uncached if it even works, which would be bad
#asahi · 2021-08-06 15:42
marcansince we know it's coherent
#asahi · 2021-08-06 15:42
marcanarnd: yeah but not here, here you just want flat out cached
#asahi · 2021-08-06 15:42
marcanheh, a bunch of arm drivers also use dma_alloc_wc I see
#asahi · 2021-08-06 15:40
marcanwhich ISTR being used to map framebuffers
#asahi · 2021-08-06 15:40
marcanI was thinking _wc
#asahi · 2021-08-06 15:38
marcanon M1 dma_alloc_coherent should be a normal memory allocation
#asahi · 2021-08-06 15:37
marcandon't model your memory management after x86 drm drivers; that is the ~one place weird stuff happens on x86 that has no place anywhere else
#asahi · 2021-08-06 15:36
marcandma_alloc_coherent is probably what you want

plus ›