#
#asahi · 2025-04-13 09:52
PaulFertserWell, recently I had issues simply booting an x86 laptop (some asus expertbook). Debian 6.1 booted fine but lacked Xe driver for GPU. Debian 6.12 hanged hard after initing USB host interfaces, no panic no error no nothing, and I have no clue if SysRq worked because I couldn't find SysRq. Same about Debian 6.13. But somehow Fedora 6.12 booted just fine. With 6.1 the ax211 firmware was older and
#asahi · 2025-04-13 09:46
chaos_princessmight be for kernel dts, but if it is provided by platform firmware, then who knows
#asahi · 2025-04-13 09:46
fl0_idchaos_princess mmh ok makes sense. thx for clarifying
#asahi · 2025-04-13 09:45
PaulFertserI thought Rob and friends care a lot about not breaking compatibility where it matters.
#asahi · 2025-04-13 09:45
chaos_princessexcept the dts is not forward compatible apparently, as i had repeatedly found out :P
#asahi · 2025-04-13 09:44
PaulFertserDTs are forward-compatible. So the vendor releases a new shiny server, prepares DT which works with at least some Linux version actual at the time, ships this DT with platform firmware. Newer kernel shouldn't have a problem using this older DT. And when some important changes are needed a newer DT can be used instead.
#asahi · 2025-04-13 09:43
svenyeah, i prefer bytecode i could RE if i wanted to over some invisible mystery blob running in EL3 or whatever
#asahi · 2025-04-13 09:42
PaulFertserchaos_princess: bytecode is probably good because you can kind of RE it easier and do you own thing if needed.
#asahi · 2025-04-13 09:42
chaos_princessright, so kernel. And at this point you no longer have forward compatibility, as "random iso you found on a usb stick" will not have the dts for your shiny new arm server
#asahi · 2025-04-13 09:41
PaulFertserWell, normally it should be the platform firmware but in fact usually Linux+BSD devs are much better than the vendor so you probably have to count on them keeping it sane and up-to-date and use what's on kernel.org rather than provided by the vendor.
#asahi · 2025-04-13 09:41
chaos_princess(both choices have downsides, and there are no other choices)
#asahi · 2025-04-13 09:40
chaos_princessno, what loads it, kernel or platform firmware?
#asahi · 2025-04-13 09:40
PaulFertserThe schematics for the board.
#asahi · 2025-04-13 09:39
chaos_princessPaulFertser: where is your dts coming from
#asahi · 2025-04-13 09:39
PaulFertserOn typical Arm you have DT describing hardware, standard PSCI calls to handle power management. You just grab aarch64 image and boot it on any system, feels OK in my experience.
#asahi · 2025-04-13 09:38
chaos_princessPaulFertser: hot(?) take: bytecode snippet good. what shouldn't exist is smm calls
#asahi · 2025-04-13 09:38
chaos_princessfl0_id: not the way device trees are used rn. device trees are tied to kernel version, so what you need is a "prototype device tree" that is provided by the bootloader, and gets transformed into "real device tree" that is used by the kernel. And acpi could be alright as a "proto device tree"
#asahi · 2025-04-13 09:36
PaulFertserfl0_id: device trees do not provide those bytecode snippets. And that's a great thing to not have to use them.
#asahi · 2025-04-13 09:36
chaos_princessPaulFertser: oh, i know, but it is still way better than the situation on arm/riscv
#asahi · 2025-04-13 09:35
PaulFertserchaos_princess: only that talking about laptops you know how many custom acpi-specific (also wmi) code there is in Linux with probably a hundred of model-specific quirks. And how Linux pretends to be a certain windows version when interpreting the obscure bytecode, and then on some models or UEFI versions it needs to pretend being a different windows version etc etc. Pure filth.
#asahi · 2025-04-13 09:34
fl0_idafaik the device trees used on many arm devices could also work very well, at least if they would come with docs and would contain all necessary info
#asahi · 2025-04-13 09:30
PaulFertserAnd Asahi devs do RE and implement APIs of the (typically) most recent stable version available so that it works on the widest range of devices possible (Apple provides newer firmware for older devices but not older firmware for newer devices).
#asahi · 2025-04-13 09:30
chaos_princessPaulFertser: imo, while i do not like some engineering choices in acpi, it is still an insanely great advantage of x86 platform. being able to just flash a random iso and it mostly just working is great
#asahi · 2025-04-13 09:28
PaulFertserI guess for someone familiar with WiFi this analogy might work: where you're used to have SoftMAC-like solutions the Apple hardware offers you FullMAC, and they change the API to those "FullMAC" devices any moment they feel like because they control both sides of the API.
#asahi · 2025-04-13 09:27
chaos_princessi do not remember what is in that spec, but "arm requires X" is kind of a meme, their platform requirements tend to be 80% exceptions by volume
#asahi · 2025-04-13 09:26
PaulFertserAnd I'm not sure it's just a great advantage.
#asahi · 2025-04-13 09:26
PaulFertserArm requires ACPI for their server platforms now :/
#asahi · 2025-04-13 09:25
chaos_princessthey have uefi, they pretend that they have acpi and so on
#asahi · 2025-04-13 09:25
chaos_princessright, but the qualcomm thing that WoA runs on is much closer to "standard x86 platform but arm" than arm macs
#asahi · 2025-04-13 09:23
PaulFertserAnd of course those "CISC" CPUs like their microcode updates.
#asahi · 2025-04-13 09:23
monthebiffstill a tease
#asahi · 2025-04-13 09:23
chaos_princessmonthebiff: (the actual answer is that i haven't looked at it since the last time i gave an "update")
#asahi · 2025-04-13 09:22
PaulFertserTF-A via PSCI on common Arm-based systems.
#asahi · 2025-04-13 09:22
PaulFertserIBM PC mental model isn't much useful for any modern hardware, including amd64. The days of VESA VBE, APM and A20 gating are long gone, things are done very differently now. But you still depend on things outside the OS kernel even after passing control to it: ACPI tables for detecting on-board hardware and ACPI bytecode for runtime power-related actions on typical x86 platforms, and calls into
#asahi · 2025-04-13 09:21
monthebiffbut thats such a tease though :P
#asahi · 2025-04-13 09:21
chaos_princessdone when its done and not a second sooner :P
#asahi · 2025-04-13 09:17
monthebiffis there any news/ rumblings on power efficiency/ battery life on the m1pro chips? I had the sad experience of wondering why my laptop wasn't turning on after someone unplugged it to move it :P
#asahi · 2025-04-13 09:12
monthebiffooooo https://www.reddit.com/r/AsahiGaming/comments/1jao6nz/does_world_of_warcraft_with_under_asahi_via_fex/ somebody else is looking for it as well ... :P
#asahi · 2025-04-13 08:58
PaulFertserfeatures are likely to require firmware versions that would be current by the time those features are developed.
#asahi · 2025-04-13 08:58
PaulFertserByronJackson: on a desktop you also get firmware updates, e.g. for wifi adapters (quite often), for USB printers, for GPUs etc etc. It's just that apple hardware is special and you have plenty of additional CPUs on board that you need to talk to for doing useful things. If the current Asahi feature set covers all your usecase then you won't be forced to upgrade firmware most probably. But new
#asahi · 2025-04-13 08:50
chaos_princessif the way this platform does things looks confusing, it is because the mental model should not be "a pc with an arm cpu", but "an oversized iphone with macos and unlocked bootloader"
#asahi · 2025-04-13 08:43
chadmedwe have mechanisms to deal with this sanely but you have to keep macos around
#asahi · 2025-04-13 08:42
chadmedyeah but no point sooking about it. we dont control the platform, thats just how it is
#asahi · 2025-04-13 08:42
ByronJacksonI guess? If the firmware works well enough in the first place then constant abi changes are just a nuisance.
#asahi · 2025-04-13 08:40
chadmed or runtime services or whatever, the firmware on each coprocessor is tightly integrated with that device's function. a lot of our linux drivers are basically just writing a pointer to a firmware command structure to a mailbox and hitting a doorbell
#asahi · 2025-04-13 08:39
chadmedthere is no "bios
#asahi · 2025-04-13 08:39
chadmedByronJackson: it's a philosophical difference and also partly due to the demarcation between "firmware" and "operating system" being different on this platform
#asahi · 2025-04-13 08:36
fl0_idif nothing else, to have a consistent firmware across supported devices afaik
#asahi · 2025-04-13 08:36
fl0_idByronJackson afaik it changes usually at least once per major version update
#asahi · 2025-04-13 08:35
ByronJacksonI'm not sure I understand why the firmware should change so often. The mac replaces my eleven-year-old desktop, which recived exactly one firmware update.

plus ›