Strit@lemmy.linuxuserspace.show
on 25 Sep 2024 07:18
nextcollapse
There is currently work being done to get support for some snapdragon laptops into the kernel. I think 6.11 got preliminary support for a couple and patches for others are still waiting.
Iheartcheese@lemmy.world
on 25 Sep 2024 07:39
nextcollapse
They have not seen the light
Xiisadaddy@lemmygrad.ml
on 25 Sep 2024 07:57
nextcollapse
A lot of the people who would do the work simply don’t have access to the hardware to test stuff and so progress is slow. Would help a lot if qualcomm sent devs some hardware.
0x0@programming.dev
on 25 Sep 2024 08:24
nextcollapse
Aren’t they arm-based? What’s the hurdle?
Vitaly@feddit.uk
on 25 Sep 2024 08:53
nextcollapse
If the devs don’t have access to the hardware then it’s impossible to make drivers for a specific laptop
If it’s arm-based is the CPU so alien as to not being usable without a very generic kernel?
As for the hardware, is it so unique there are no drivers already?
Markaos@lemmy.one
on 25 Sep 2024 09:22
nextcollapse
I can’t speak for these specific laptops, but unlike x86, ARM generally doesn’t have a way for an OS to discover the available hardware, and most ARM platforms historically didn’t do anything to help. There is a standard for UEFI on ARM where the UEFI is supposed to tell the OS about the hardware, but as far as I know this is only a thing on ARM servers and these laptops might not support it.
Without any way of probing for hardware or getting the information from UEFI, Linux has to somehow be compiled with all the info about the hardware built-in. And the build will be model-specific (there’s a way to pass a file describing the hardware to Linux from the bootloader which enables a single kernel to be used on multiple models and have just a small part of the bootloader be model-specific, but somebody still needs to make that file and the manufacturers clearly don’t intend to do that).
woelkchen@lemmy.world
on 25 Sep 2024 14:35
nextcollapse
There’s more hardware in a notebook than just the CPU. It’s pointless without network and GPU drivers, for example. Also the ARM DeviceTree stuff is BS.
rotopenguin@infosec.pub
on 25 Sep 2024 19:31
collapse
ARM systems don’t have the whole ACPI thing to describe what hardware is where. Linux has to bodge together its view of the system with a devicetree instead. If you don’t know what device IP blocks are integrated into the SOC (and locked behind an NDA), good luck blindly guessing. You don’t even get EFI booting, you get shit like “the rpi gpu runs its own proprietary bootloader lol”.
Buffalox@lemmy.world
on 25 Sep 2024 11:06
collapse
Well there needs to be a working general boot-loader for one. Then the hardware needs to have drivers that work.
redxef@feddit.org
on 25 Sep 2024 09:50
nextcollapse
Tuxedo is working on one, they said it might be ready by christmas.
ShortN0te@lemmy.ml
on 25 Sep 2024 11:29
nextcollapse
Qualcomm did work together with Microsoft and the Vendors closely together before the launch to create those devices.
Linux device vendors probably did not get the same treatment. So give it time. Also, why not buy a windows laptop and put linux on it?
anon5621@lemmy.ml
on 25 Sep 2024 14:29
nextcollapse
I am personally not exicted about using arm on pc/laptop all because situation with them can repwat situation with phones where we have locked down devices without ability to unlock bootloader and hug problems with drivers as sequence. Also there no uefi with ACPI so each distribution should be custom build to exact laptop because of operating system should know about installed hardware in laptop/pc in DTB file,i would prefer stay on x86 long as i can and maybe risc v cpus gonna change this situation.
Exactly. All the hype and excitement over a locked down arm ecosystem with evaporating battery life advantages. No thank you. Development efforts are better served elsewhere. I would prefer the Linux community ignore it rather than support it over RISC-V.
data1701d@startrek.website
on 25 Sep 2024 18:03
nextcollapse
I wonder if the endgame for getting Linux running on these freakasauruses is not to create a custom UEFI firmware for each laptop that could abstract away the differences between each laptop with an ACPI API, rather that modifying the kernel itself.
It sounds daunting, but people have done it for the Raspberry Pi before. I don’t think it runs as actual firmware on the device - I think it’s just an ARM binary that could then execute and provide abstraction for a bootloader.
There are difficulties with that, obviously. For one, the Raspberry Pi is one hardware platform, and a Broadcom-based one at that. Still, I can’t imagine that you’d have to redo everything from scratch on every platform; it’d basically just be something like a device tree to define the ACPI info built into every firmware build variant. If this idea worked, people could just have an environment to install an operating system on that is almost like a normal UEFI PC but with ARM.
Truth be told though, I kind of wish Ampere would get more into the consumer space; I feel like they have the least insane configuration of almost any ARM device, being users of UEFI. I don’t know if they could viably scale down from their 192 core beasts, though. Now that BNL song is going through my head. “If I had a million dollars, I’d buy an Ampere workstation; a power-hungry ARM beast.”
rotopenguin@infosec.pub
on 25 Sep 2024 19:19
nextcollapse
Has Qualcomm ever been helpful?
SynopsisTantilize@lemm.ee
on 26 Sep 2024 11:47
collapse
I’m reading this on a bootloader locked S23U…while looking across the room I see my s10e,bootloader locked. And if I look in the distance my s7 is sitting there…locked…
LeFantome@programming.dev
on 25 Sep 2024 23:08
collapse
I thought there were two units working now ( maybe not everything ). One Lenovo and one Asus if I recall.
threaded - newest
There is currently work being done to get support for some snapdragon laptops into the kernel. I think 6.11 got preliminary support for a couple and patches for others are still waiting.
Do they have a list of devices that they work on?
They have not seen the light
A lot of the people who would do the work simply don’t have access to the hardware to test stuff and so progress is slow. Would help a lot if qualcomm sent devs some hardware.
Aren’t they arm-based? What’s the hurdle?
If the devs don’t have access to the hardware then it’s impossible to make drivers for a specific laptop
If it’s arm-based is the CPU so alien as to not being usable without a very generic kernel?
As for the hardware, is it so unique there are no drivers already?
I can’t speak for these specific laptops, but unlike x86, ARM generally doesn’t have a way for an OS to discover the available hardware, and most ARM platforms historically didn’t do anything to help. There is a standard for UEFI on ARM where the UEFI is supposed to tell the OS about the hardware, but as far as I know this is only a thing on ARM servers and these laptops might not support it.
Without any way of probing for hardware or getting the information from UEFI, Linux has to somehow be compiled with all the info about the hardware built-in. And the build will be model-specific (there’s a way to pass a file describing the hardware to Linux from the bootloader which enables a single kernel to be used on multiple models and have just a small part of the bootloader be model-specific, but somebody still needs to make that file and the manufacturers clearly don’t intend to do that).
Yep, an OS would need to be monolithic for a given device.
Something the computer world decided was a Bad Thing in about 1978.
Are we really going back to the dark ages?
Yes the drivers are all different afaik. You need a device tree, and hardware to debug what you wrote.
Just a kernel doesnt help much.
It would be cool if they provided at least a virtual machine to test on
Why a VM?
Because if they can’t provide them with real hardware at least a VM would be great
Like a VM emulating their exact hardware? Didnt know that was a thing
I’m not sure if it’s possible
There’s more hardware in a notebook than just the CPU. It’s pointless without network and GPU drivers, for example. Also the ARM DeviceTree stuff is BS.
ARM systems don’t have the whole ACPI thing to describe what hardware is where. Linux has to bodge together its view of the system with a devicetree instead. If you don’t know what device IP blocks are integrated into the SOC (and locked behind an NDA), good luck blindly guessing. You don’t even get EFI booting, you get shit like “the rpi gpu runs its own proprietary bootloader lol”.
Well there needs to be a working general boot-loader for one. Then the hardware needs to have drivers that work.
Tuxedo is working on one, they said it might be ready by christmas.
Qualcomm did work together with Microsoft and the Vendors closely together before the launch to create those devices.
Linux device vendors probably did not get the same treatment. So give it time. Also, why not buy a windows laptop and put linux on it?
I am personally not exicted about using arm on pc/laptop all because situation with them can repwat situation with phones where we have locked down devices without ability to unlock bootloader and hug problems with drivers as sequence. Also there no uefi with ACPI so each distribution should be custom build to exact laptop because of operating system should know about installed hardware in laptop/pc in DTB file,i would prefer stay on x86 long as i can and maybe risc v cpus gonna change this situation.
Exactly. All the hype and excitement over a locked down arm ecosystem with evaporating battery life advantages. No thank you. Development efforts are better served elsewhere. I would prefer the Linux community ignore it rather than support it over RISC-V.
I’m pretty sure some basic stuff is running on the Windows Dev Kit 2023 (no thanks to Qualcomm), which is very similar. See www.phoronix.com/news/Windows-Dev-Kit-2023-Linux
I wonder if the endgame for getting Linux running on these freakasauruses is not to create a custom UEFI firmware for each laptop that could abstract away the differences between each laptop with an ACPI API, rather that modifying the kernel itself.
It sounds daunting, but people have done it for the Raspberry Pi before. I don’t think it runs as actual firmware on the device - I think it’s just an ARM binary that could then execute and provide abstraction for a bootloader.
There are difficulties with that, obviously. For one, the Raspberry Pi is one hardware platform, and a Broadcom-based one at that. Still, I can’t imagine that you’d have to redo everything from scratch on every platform; it’d basically just be something like a device tree to define the ACPI info built into every firmware build variant. If this idea worked, people could just have an environment to install an operating system on that is almost like a normal UEFI PC but with ARM.
Truth be told though, I kind of wish Ampere would get more into the consumer space; I feel like they have the least insane configuration of almost any ARM device, being users of UEFI. I don’t know if they could viably scale down from their 192 core beasts, though. Now that BNL song is going through my head. “If I had a million dollars, I’d buy an Ampere workstation; a power-hungry ARM beast.”
Has Qualcomm ever been helpful?
I’m reading this on a bootloader locked S23U…while looking across the room I see my s10e,bootloader locked. And if I look in the distance my s7 is sitting there…locked…
I thought there were two units working now ( maybe not everything ). One Lenovo and one Asus if I recall.
EDIT:
ASUS Vivobook S15 & Lenovo Yoga Slim7x
www.phoronix.com/news/Linux-6.11-SoC-Platforms
6.11 is released now.