#asahi · 2021-12-01 19:07
marcan(openbsd can pick whatever version it claims to support the same stuff as :p)#asahi · 2021-12-01 19:07
marcanlinux version#asahi · 2021-12-01 19:06
marcandm-pre-reloc?#asahi · 2021-12-01 19:02
marcanand then it might be nice if we could get some feedback on the currently installed kernel version, e.g. dumping a file in <EFI PART>/asahi/kernel.ver or something. that could be used by our bootloader installer/updater to filter the firmware versions we offer, so users don't accidentally install firmware that's too new for their kernel#asahi · 2021-12-01 19:00
marcanbarring changes to distros, maybe we can have u-boot mark only the currently booted EFI part with the real GUID, so distros don't get confused by multiple EFI parts#asahi · 2021-12-01 19:00
marcan... just the default boot)#asahi · 2021-12-01 19:00
marcanthere are some subtleties to work out around multi-boot; my current thinking is we should recommend that users do multi-distro-boot by using completely separate OS containers from the machine's point of view, which lets you use the native boot picker, and means each distro gets its own EFI partition (which is important because we don't have a good way to do efivars, so no configurable EFI boot menu, ...#asahi · 2021-12-01 18:59
marcanfor the firmware, I also need to write a spec for that, but my current thinking is something like <EFI PART>/asahi/firmware.tar.gz; distros would be expected to untar that into /lib/firmware, and keep track of the mtime so they can do it again if it changes (e.g. the user ran a bootloader/firmware upgrade)#asahi · 2021-12-01 18:56
marcanI also want to offer distro images pre-configured in our installer so people can just dump those in in one shot; I need to get around to writing a spec for those...#asahi · 2021-12-01 18:55
marcanor you could preinstall a netinstall image or what have you#asahi · 2021-12-01 18:55
marcanonce that's installed it'll just be an EFI boot which should work with USB disks for distro installers and such#asahi · 2021-12-01 18:55
marcanwe provide m1n1+u-boot which will be required to boot the distro installers in the first place (which users will install using a script from macOS/1TR); I don't expect distros to attempt to roll their own from scratch replacement for that (trust me, you don't want to)#asahi · 2021-12-01 18:53
marcan(or whatever EFI bootloader the distro wants to use)#asahi · 2021-12-01 18:53
marcanso just like any other GRUB machine from a distro's point of view#asahi · 2021-12-01 18:53
marcanthe bootloader should be GRUB or whatever, u-boot provides EFI services#asahi · 2021-12-01 18:53
marcanmostly just the firmware thing#asahi · 2021-12-01 18:52
marcanwouter: I don't really expect distro installers to need any porting beyond adding some bits once we have them (e.g. some kind of bundled touchbard for the M1 13" MBP, or else you get no function keys...)#asahi · 2021-11-28 02:09
marcanhard UART is only really necessary if you're doing KVM development (where you can't run under the hypervisor), m1n1 bring-up itself, or some corner case relating to e.g. realtime or time-sensitive kernel code that might be affected by the hypervisor#asahi · 2021-11-28 02:07
marcanyou can run the kernel under the m1n1 hypervisor and it will virtualize the UART and expose it as a normal USB CDC-ACM device you can access with a normal USB cable from any machine#asahi · 2021-11-28 02:04
marcanthat said, most people shouldn't need serial comms for anything, given we have a hypervisor that can virtualize that over normal USB#asahi · 2021-11-28 02:03
marcanChristopherNugent[m]: it's proprietary; you need either another M1 Mac or custom hardware. it's not about the USB chipset.#asahi · 2021-11-27 12:21
marcanhonestly these SoCs make more sense than most#asahi · 2021-11-26 11:31
marcanI think the t6000 upstreaming branch has most of the immediately upstreamable stuff (and some not)#asahi · 2021-11-25 18:13
marcandottedmag: prores is a separate hardware block, not AVD#asahi · 2021-11-25 16:21
marcanthough apple calls VP9 some weirdo internal codename I forget all throughout their driver stack, because I guess it was secret or they're allergic to calling it VP9 or something?#asahi · 2021-11-25 16:20
marcanit supports AVC, HEVC, and VP9#asahi · 2021-11-25 16:20
marcanfirefox317: AVC is H264 and I don't think it supports Adaptive DNA Storage...#asahi · 2021-11-25 14:56
marcanI looked at the kext and it has the firmware embedded (a whole bunch actually, for different IP core versions)#asahi · 2021-11-25 14:56
marcanfirefox317: I was poking around ioreg to find out what codecs/levels it supports, and I found a firmware-size property and that set all the alarms ringing#asahi · 2021-11-25 14:52
marcanit's just "there happens to be an m3 inside avd"#asahi · 2021-11-25 14:52
marcanit's not an m3wrap#asahi · 2021-11-25 14:52
marcanor just custom#asahi · 2021-11-25 14:52
marcanit's just a random M3 core they threw into that IP block#asahi · 2021-11-25 14:51
marcanit's not an ASC style thing as far as I could tell#asahi · 2021-11-25 14:51
marcansven: I'm not even sure if AVD has the same mailbox though#asahi · 2021-11-25 13:58
marcan(people, please provide logs!)#asahi · 2021-11-25 13:57
marcanunfortunately there's no dmesg of the failed case#asahi · 2021-11-25 13:57
marcanhere, see cth451_desktop's last message#asahi · 2021-11-25 13:50
marcanmaz: btw, did you see that the PERST patches apparently break 10G NICs? my guess is it just takes longer to initialize after PERST...#asahi · 2021-11-25 13:49
marcan:D#asahi · 2021-11-25 13:45
marcanunfortunately being impatient doesn't make things go faster :(#asahi · 2021-11-25 13:40
marcanthat too :p#asahi · 2021-11-25 13:38
marcanthe important thing is MMIO will be inaccessible while suspended, so you need to make sure to always resume it before doing anything#asahi · 2021-11-25 13:37
marcanj_ey: look at the patch to add runtime-pm to the samsung driver for an example; also there's a good chance some other SPI driver has it too#asahi · 2021-11-25 13:37
marcanyes, suspend the device when it is idle. it's what tells pmgr to turn it off.#asahi · 2021-11-25 13:34
marcanj_ey: also, do you think you can implement runtime-pm in that spi driver? that'd be another good test case for the pmgr stuff, and it should be easy to add (and does not add a hard dependency in either direction)#asahi · 2021-11-25 13:33
marcanwe're not going for "it works" here, we're going for maintainability and correctness, and that means doing things right#asahi · 2021-11-25 13:33
marcan(those weird defs came from there)#asahi · 2021-11-25 13:33
marcanbecause they are mostly wrong :p#asahi · 2021-11-25 13:33
marcanmps: see this is why we don't send corellium drivers to the kernel ML