#
✕ nick:marcan
#asahi · 2022-01-02 11:21
marcanyes, similar mechanism to UEFI runtime services
#asahi · 2022-01-02 11:20
marcanso linux will never actually have this code itself
#asahi · 2022-01-02 11:20
marcanonce the PSCI stuff is implemented linux will call m1n1 (which will remain resident in RAM) to implement this, as well as full system sleep
#asahi · 2022-01-02 11:19
marcanyes
#asahi · 2022-01-02 11:18
marcanit's a completely different thing
#asahi · 2022-01-02 11:17
marcancpufreq just changes the clock speed
#asahi · 2022-01-02 11:17
marcancpufreq is not sleep
#asahi · 2022-01-02 11:17
marcanhttps://github.com/AsahiLinux/m1n1/blob/main/src/utils_asm.S#L155
#asahi · 2022-01-02 11:17
marcanno, that's for cpufreq
#asahi · 2022-01-02 11:16
marcanmaybe not even then, logically speaking
#asahi · 2022-01-02 11:16
marcanI don't think macOS attempts to actually power cores off at any time except when going into whole system sleep
#asahi · 2022-01-02 11:15
marcanm1n1 already implements this
#asahi · 2022-01-02 11:15
marcanthey lose GPR register state except pc, sp, which means you need a little push/pop deal around the sleep
#asahi · 2022-01-02 11:15
marcanwe actually don't know how to turn cores off, but we know how to put them into deep sleep, which should be the ~same
#asahi · 2022-01-02 03:43
marcanmaking that work acceptably well needs more power management stuff that isn't done yet (CPU deep sleep, whole system sleep, both of which depend on the PSCI stuff)
#asahi · 2022-01-02 03:42
marcanI would not recommend using Linux in a "portable" scenario without shutting down the full system every time you store it, yet
#asahi · 2021-12-31 15:39
marcan:)
#asahi · 2021-12-31 15:38
marcanregular dev should resume next week; I had one too many new year's related events and other IRQs this week...
#asahi · 2021-12-31 15:38
marcangoing to finish new year's dinner with a friend (eating between DJ switches is... a bit fragmented, heh) and then get some sleep :)
#asahi · 2021-12-31 15:37
marcan(this is why I have sent PRs to OBS :D)
#asahi · 2021-12-31 15:36
marcanjust signed off of an 8h shift doing central stream management for a round-the-timezone anime music new year DJ stream (yeah... the things I do :p)
#asahi · 2021-12-31 15:30
marcanhappy new year from japan :)
#asahi · 2021-12-30 07:58
marcanbut if it is, it won't help find the problem by itself because the pcie reset change was itself fixing a bug and resetting the card is supposed to work
#asahi · 2021-12-30 07:58
marcanit's almost certainly the pcie reset change as I said
#asahi · 2021-12-30 01:35
marcan(bbl, sorry, busy with some other stuff now)
#asahi · 2021-12-30 01:34
marcanyeah
#asahi · 2021-12-30 01:33
marcanyes
#asahi · 2021-12-30 01:33
marcanthat almost sounds like whatever the reset the driver thinks it's doing, it's not actually resetting the card, and when you actually reset it it takes quite a bit longer
#asahi · 2021-12-30 01:32
marcannot sure if we just need to wait longer or if there is something wrong causing it to take that long
#asahi · 2021-12-30 01:32
marcansee ee91cb570
#asahi · 2021-12-30 01:31
marcancth451_desktop: we actually reset the card properly now
#asahi · 2021-12-29 23:55
marcanit can't verify the seal, that would take forever; the whole point is it's a hash tree verified dynamically at runtime, so it has to be done by the kernel
#asahi · 2021-12-29 23:55
marcaniBoot does not verify the snapshot seal or any of that, it merely passes through the root signature to xnu
#asahi · 2021-12-29 23:54
marcanall those logs are from xnu
#asahi · 2021-12-29 23:54
marcanthat article is confused about iBoot/xnu
#asahi · 2021-12-29 04:26
marcanNamidairo: those aren't the user join messages, that was some silly script/bot (that wasn't even working properly, since it was greeting people on quit)
#asahi · 2021-12-28 12:03
marcanthis is distro-level stuff; I don't mind maintaining some trivial downstream patches like that just to save us headaches when people come here for support
#asahi · 2021-12-28 12:03
marcanbetter have some warnings in the actual code
#asahi · 2021-12-28 12:03
marcannobody reads wikis :p
#asahi · 2021-12-28 12:01
marcanI really need to start making a pile of "don't do that" patches for linux/u-boot/etc - something that gives you big "don't do that" warnings every time someone tries to do something that won't work, like load a dtb :p
#asahi · 2021-12-28 12:01
marcanand the framebuffer address can change depending on what monitor you have plugged in, or could until apple broke it on the mini entirely
#asahi · 2021-12-28 12:00
marcanand the memory base potentially
#asahi · 2021-12-28 12:00
marcanyes, kaslr changes the SEPFW address on every boot
#asahi · 2021-12-28 12:00
marcanthere are things that change every boot
#asahi · 2021-12-28 11:59
marcanstatic DTs only work for very simple machines; these machines aren't simple
#asahi · 2021-12-28 11:59
marcandevice trees are designed to be updated by bootloaders with runtime information
#asahi · 2021-12-28 11:59
marcanwe already make everything that can be static static
#asahi · 2021-12-28 11:59
marcanamw: no, the dtb has to be processed with *runtime* information from your machine
#asahi · 2021-12-28 10:54
marcanbypassing that will break a bunch of things
#asahi · 2021-12-28 10:54
marcanm1n1 needs to process the dtb before it can be used

plus ›