#
✕ nick:marcan
#asahi-dev · 2022-10-16 11:59
marcanI'm thinking tomorrow I'm going to rewrite cpufreq one more time and then do another kernel rebase/update dance
#asahi-dev · 2022-10-16 11:59
marcanyeah, we should
#asahi-dev · 2022-10-16 11:21
marcanplan is to switch to PW once all the speaker EQ/safety/etc magic is in
#asahi-dev · 2022-10-15 04:36
marcanthe asahi-dev firmware extraction flow & scripts support ESP-as-/boot, but also a lot of the firmware stuff is getting another complete redesign soon
#asahi-dev · 2022-10-15 04:36
marcanI intend to push for fixing that soon, given grub's recent issues
#asahi-dev · 2022-10-14 12:49
marcan(and given the safety implications involved, I think I'm more comfortable if I get to do that part myself)
#asahi-dev · 2022-10-14 12:49
marcanpovik: fwiw I'm just waiting on VISENSE to get to the speaker safety userspace magic :>
#asahi-dev · 2022-10-12 00:07
marcan__TEXT is probably mapped via DAPF?
#asahi-dev · 2022-10-11 15:25
marcanif piodma represents hardware that exists, even if the driver doesn't talk to it, I think it's fine having it as a separate node/subnode in the DT
#asahi-dev · 2022-10-11 15:02
marcanwe could probably just always use the bools and pass the node names to the function at the caller? dcpext would always have X, false, false and NULLs for the latter 2 nodes.
#asahi-dev · 2022-10-11 15:01
marcanwe can always refactor later
#asahi-dev · 2022-10-11 15:01
marcan_jannau_: I'm fine with what we have now, the bools seem preferable to repeat lookups and inits
#asahi-dev · 2022-10-11 12:53
marcanthen confirmed with gpiola
#asahi-dev · 2022-10-11 12:53
marcanIIRC I just regmon'd the gpio stuff and saw which bits flipped when I used the SPI interface
#asahi-dev · 2022-10-11 12:52
marcanyeah it's broken up
#asahi-dev · 2022-10-11 12:52
marcanI think that was guesswork and testing with the logic analyzer?
#asahi-dev · 2022-10-11 12:52
marcanthe gpio mappings?
#asahi-dev · 2022-10-11 12:52
marcanoh that wasn't even stripped, 1.0MB stripped
#asahi-dev · 2022-10-11 12:51
marcanif you're ending up with 35MB binaries, that's not rust, that's that there's 35MB of stuff in there
#asahi-dev · 2022-10-11 12:50
marcanI have a Rust binary for a thing I wrote in 2016 here (first time I tried it! didn't touch it again until m1n1 :p) which does use a bunch of crates (mio, net2, log, some more), and it's 1.6MB
#asahi-dev · 2022-10-11 12:48
marcanand wouldn't really apply to something simple like an nvram util if you're careful about it
#asahi-dev · 2022-10-11 12:48
marcanthat's a different ecosystem problem that has little to do with the language itself
#asahi-dev · 2022-10-11 12:47
marcanChaosPrincess: Pulling in cargo dependency hell doesn't mean "rust binaries are huge", it means you have cargo dependency hell ;)
#asahi-dev · 2022-10-11 12:38
marcanyes
#asahi-dev · 2022-10-11 12:20
marcanso the Rust runtime needed to support hello, world is ~33% the size of the underlying glibc runtime needed to support it. I don't think a 33% overhead over C qualifies as "huge binaries".
#asahi-dev · 2022-10-11 12:19
marcanhello.c compiles to ~nothing, but if you statically link it, it's 768kB (and if you statically link the Rust one, it grows by about that much)
#asahi-dev · 2022-10-11 12:16
marcanpeople regularly throw `lvm` into initramfses and that's like 2.6MB statically linked.
#asahi-dev · 2022-10-11 12:16
marcanChaosPrincess: "huge" is largely FUD from people building in debug mode. with statically linked runtime in release mode, hello.rs is 256kB
#asahi-dev · 2022-10-11 11:25
marcan:D
#asahi-dev · 2022-10-11 11:24
marcangoing back to nvram: FWIW, nvram userspace tooling should be written in C (or Rust if you want to be adventurous); we don't want to rely on Python for this because that's the kind of thing we/someone might want to use from an initramfs for something, so it has to be a small self-contained tool
#asahi-dev · 2022-10-11 11:20
marcanbut that aside, also turn off clpc because it spams due to time warping (boot with clpc=0)
#asahi-dev · 2022-10-11 11:20
marcanyou asked it to log every guest exception so it's doing just that ;)
#asahi-dev · 2022-10-11 11:19
marcandon't use -e unless you're trying to do that
#asahi-dev · 2022-10-11 11:19
marcanI only used that way, way early on to debug really early XNU startup issues
#asahi-dev · 2022-10-11 11:18
marcanamarioguy: why are you enabling syscall hooking? that will indeed log every single page fault in the guest (and also probably break things after a while)
#asahi-dev · 2022-10-11 11:17
marcanlast I checked it's a perfect match except that apple invented some funny escaping mechanism to store binary values
#asahi-dev · 2022-10-11 11:16
marcanhttps://stuff.mit.edu/afs/sipb/contrib/doc/specs/protocol/chrp/chrp_hrpa.pdf PDF page 165 and onwards
#asahi-dev · 2022-10-11 11:15
marcanbut yeah, that should be handled in userspace IMO, just expose the raw mtd device (with the proper partition descriptions in the DT)
#asahi-dev · 2022-10-11 11:14
marcanChaosPrincess: You didn't have to figure out how nvram works, it's straight out of CHRP (except for the funny binary escaping business) :)
#asahi-dev · 2022-10-07 02:09
marcanit's been like that for a while in ventura betas
#asahi-dev · 2022-10-07 02:08
marcanjannau: if you have -v. what changed is that -v is no longer filtered out of boot-args.
#asahi-dev · 2022-10-05 09:52
marcanpovik: no, that's rmk ;)
#asahi-dev · 2022-10-05 09:52
marcanyeah, that makes sense :)
#asahi-dev · 2022-10-05 09:45
marcanarnd: ack
#asahi-dev · 2022-10-05 09:43
marcan6.0 is out, so it's probably too late for me to send a PR to arnd now, so I'll just rebase on top for 6.2 and there won't be a merge conflict at that point
#asahi-dev · 2022-10-05 08:44
marcanI should use this chance to get a bunch of submissions going. cpufreq needs another rethink for the latest bindings idea and then it should hopefully be mergeable.
#asahi-dev · 2022-10-05 08:43
marcanarnd: let's punt the PR for the next release. the timing has been kind of bad this time around, with the Akademy talk the same weekend of release and that bindings patch sneaking in via another tree, and all the other stuff going on...
#asahi-dev · 2022-10-01 23:51
marcanI also don't see anything about firmware updates in that repo
#asahi-dev · 2022-10-01 23:49
marcanhttps://gitlab.manjaro.org/manjaro-arm/packages/apple-silicon/linux-apple-silicon/-/blob/master/PKGBUILD
#asahi-dev · 2022-10-01 23:49
marcanapparently they are taking my latest (non prod) tag and putting the rc6-rc7 diff on top? that's... something

plus ›