Asahi Linux Celebrates First Triangle On The Apple M1 With Fully Open-Source Driver

[luno]

> This is good work but the grump in me is fairly disappointed that all this work is being put into Apple's hardware of all things. I would rather a boycott ffrom the open-source community

problem is other arm vendors are only focused on Server architecture, Nvidia has Jetson but only dev boards and also focusing on Server Side. People are waiting for good arm laptop from long time. So if we can use Apple Laptop to use Linux why not.

> If you buy an M1 Macbook Air next year, you will need to wait *2* years for the community to catch up with its new changes from Apple. It isn't like the M1 is a stable platform, especially since it comes from a single vendor.

doesn't make sense I mean M1 Macbook not M2/M3, so in some years Linux will fully supported on M1

[dragon321]

I'm more interested in how soon the Pinebook makers sue them for $100 billion over totally frivolous nonsense like Apple used to do to everyone.

For what exactly Pinebook makers should sue Apple? For making ARM laptop?

I'm not big fan of Apple either but they sell capable ARM machines for customers and there are not many alternatives if you want to have ARM machine. Pinebook is fine but too slow to be capable of typical desktop use and it's hard to get because their production is not so big. Obviously Pinebook is not meant to be customer machine and PINE64 is honest about that (they clarify that Pinebook probably won't replace your actual machine and you shouldn't buy it for typical desktop usage). If Pinebook would be more available and with more powerful hardware then I wouldn't even think about getting anything else. But currently is not and other alternatives are not much more capable and/or more locked. Apple M1 machines are not only capable but also it seems that Linux will work fine on them in near future. So what hardware should I pick if I want to have capable ARM laptop that runs Linux?

I mean sure, I could get one of many x86 laptops but ARM has some nice advantages in terms of power usage.

[Developer12]

This is good work but the grump in me is fairly disappointed that all this work is being put into Apple's hardware of all things. I would rather a boycott from the open-source community

This took explicit effort from apple to open up the platform (apple engineered a custom boot flow on the M1 to allow 3rd party OSs, where as the iphone chip the M1 is derived from locks out absolutely everything). They've also fixed bugs for asahi which *only* affect linux. Unlike intel and AMD, there's also no management engine on this chip that has godmode access to the main CPU cores. There's no lingering SMM mode. Once boot hands off control to your software, you own the main cores completely.

There are dozens of smaller cores handling things like sound and display brightness, but they're isolated behind a mailbox interface or IOMMU, and for many of them the firmware is unsigned. If you don't think an equally large number of cores aren't secretly included in all x86 CPUs and chipsets, I have news for you. At least apple's are plainly visible.

Overall this machine is a *colossal* win for the open-source community. If you want apple to continue the trend, then the best way to communicate that is success in the market. Buy this machine heavily, and stop buying the next machine if it's less open. Carrot and stick.

Due to treadmilling, you are better off buying an M1 machine *now*. Then next year it might be usable once the hard work by the opn-source community catches up.

If you buy an M1 Macbook next year, you will need to wait *2* years for the community to catch up with its new changes from Apple. It isn't like the M1 is a stable platform, especially since it comes from a single vendor.

It's remarkably stable actually. The M1 is basically a small modification of their latest iphone chip, which was in turn a small modification of the chip the came before it and so on. It's far more work for apple *not* to make small incremental changes between chip generations.

They were able to start supporting the M1 pro and M1 max a year in advance of the launch because of clues in the architecture of where apple was going, and for the most part supporting those new chips just meant tweaking the device tree. Full support for them only took about a month to develop. The big cost is in the initial support, as was the case in the implementation of M1 support (first time supporting any apple CPU) and with this driver (first time supporting any apple GPU).

[Wooty]

I just started a YouTube channel focusing on performance of ARM Gaming paired with a PCIe GPU, and with Wine running as well as it does, I cannot wait for even a driver running at 50% the performance.

[andyprough]

For what exactly Pinebook makers should sue Apple? For making ARM laptop?

For what exactly did Apple sue Samsung for billions of dollars? Rounded corners? An app grid?

What's good for the goose is good for the gander.

[dragon321]

For what exactly did Apple sue Samsung for billions of dollars? Rounded corners? An app grid?

What's good for the goose is good for the gander.

For things that were patented by them. Why they were even able to get patents for such things is another story not related to Apple itself. Unless PINE64 get patent for ARM laptop then they are not really in position to sue Apple for anything. And even if they could then why would they? Pinebook is not Apple or most other laptops makers competitor. Pinebook has special case that doesn't really collide with Apple target.

And what about my other question? If not Apple then what I should buy if I want capable ARM laptop that runs usable (with things like GPU acceleration etc.) Linux? Sure, Apple M1 is not quite there yet but it seems that it will be in near future. But still if not Apple then what? There are not really many ARM laptops and most of them are locked. For example let's take Surface RT (not really a laptop but can be used as one with keyboard). Years after release and discontinuation is still not able to run usable desktop Linux due to locked UEFI. As I said I'm not big fan of Apple but at least their hardware is not blocked from using alternative OS.

[Vlad42]

Overall this machine is a *colossal* win for the open-source community. If you want apple to continue the trend, then the best way to communicate that is success in the market. Buy this machine heavily, and stop buying the next machine if it's less open. Carrot and stick.

I call bullshit. How the hell is it a "*colossal* win" when we still lack basic functionality 2 years later? At this rate, by the time we have mainlined graphics support for M1 laptops, they will already have been replaced by M2 variants (and the M1 variants will likely no longer be for sale). This does not even include the support for the NPUs, video encode/decode blocks, and any other accelerators. Meanwhile, we can buy pretty much any x86 based laptop, desktop, server, etc., and have it be fully functional on day 1 - at most we have to wait for the next kernel release. At best, one could argue it will be a colossal win for the developers that overcame Apples lack of documentation, but only after the M1 products are likely no longer being manufactured.

No, colossal wins are Nvidia finally trying to mainline an open source kernel driver. Just because Apple has not fucked us with closed source blobs that prevent hardware bring up and provided some minimal bug fixes on community made reverse engineered drivers, does not make it a colossal win. What Apple has done is nothing compared to Intel or AMD, both of whom could have just mainlined support for their server products and not bothered with consumer graphics or SOCs (the overwhelming majority of which are sold as Windows machines after all). Apple not screwing us with hardware bring up is bare minimum acceptable standard and them providing some fairly minimal Linux only bug fixes is a small win. A colossal win will be when they start mainlining drivers themselves in a timely manner and stop releasing products like keyboards that are so non-standard that it take years to make them functional.

Unlike intel and AMD, there's also no management engine on this chip that has godmode access to the main CPU cores. There's no lingering SMM mode. Once boot hands off control to your software, you own the main cores completely.

You are wrong when it comes to the security coprocessor/management engine/psp. It was part of Apple's T2 back when they were still on x86 and integrated it into the M1 SOC. In this respect, they are no better than Intel or AMD.

There are dozens of smaller cores handling things like sound and display brightness, but they're isolated behind a mailbox interface or IOMMU, and for many of them the firmware is unsigned. If you don't think an equally large number of cores aren't secretly included in all x86 CPUs and chipsets, I have news for you. At least apple's are plainly visible.

This is just conspiracy theorist appeasing nonsense. Intel and AMD have been very open about the fact that they have chipsets, SOCs, and IO chiplets providing USB, PCIe, etc. support for years now. What did you think they ran on, unicorns and pixie dust? Both x86 manufacturers have been open about using microcontrolers for this added functionality.

[Developer12]

At this rate, by the time we have mainlined graphics support for M1 laptops, they will already have been replaced by M2 variants (and the M1 variants will likely no longer be for sale). At best, one could argue it will be a colossal win for the developers that overcame Apples lack of documentation, but only after the M1 products are likely no longer being manufactured.

Maybe you should base your arguments on more than meaningless speculation.

This does not even include the support for the NPUs, video encode/decode blocks, and any other accelerators. Meanwhile, we can buy pretty much any x86 based laptop, desktop, server, etc., and have it be fully functional on day 1 - at most we have to wait for the next kernel release.

Sounds like you should go buy a windows laptop.

No, colossal wins are Nvidia finally trying to mainline an open source kernel driver.

Nvidia's open source driver will never be upstreamed. Ever. They don't intend to do it and the DRM developers don't either. This has been plainly stated since the initial announcement. Not only is it ineligible for upstreaming (it has no open userspace) but it's also completely wrong to fit into the kernel's DRM subsystem. Just two days ago we had an actual DRM maintainer stating that whatever does eventually land in the kernel will most likely not be based on it at all.

Just because Apple has not fucked us with closed source blobs that prevent hardware bring up and provided some minimal bug fixes on community made reverse engineered drivers, does not make it a colossal win. What Apple has done is nothing compared to Intel or AMD, both of whom could have just mainlined support for their server products and not bothered with consumer graphics or SOCs (the overwhelming majority of which are sold as Windows machines after all). Apple not screwing us with hardware bring up is bare minimum acceptable standard and them providing some fairly minimal Linux only bug fixes is a small win. A colossal win will be when they start mainlining drivers themselves in a timely manner and stop releasing products like keyboards that are so non-standard that it take years to make them functional.

Seems like you can't actually read.

To reiterate: This hardware doesn't have a lingering spyware SMM mode. It doesn't have an intel management engine or AMD platform security processor. (more on this in a moment) User-supplied code gets immediate and complete control of all the processor cores. There is not a single CPU made by AMD or intel since before the 386SL (the introduction of SMM) that meets those criteria.

The point isn't apple's support. Linux on their hardware isn't their concern, but they're helping anyway. The point is that the machine itself is pretty much everything you could ask for as far as user control and ownership. 100% control over what runs on the CPU with absolutely nothing spying on it is *exceptionally* rare. It can't even be found in ARM trustzone devices. The only other example is a POWER9 based system that costs thousands of dollars more.

You are wrong when it comes to the security coprocessor/management engine/psp. It was part of Apple's T2 back when they were still on x86 and integrated it into the M1 SOC. In this respect, they are no better than Intel or AMD.

Do you know anything at all about how the M1 is architected?

In fact, do you even know how the T2 is architected? The T2 is a completely separate chip and is completely incapable of spying on the generic intel CPU's activities. It's only function is to validate the hardware and the boot image, and to provide cryptographic services and a root of trust. Unlike the PSP/Intel ME, it has absolutely no access to the host RAM or any other aspect of the main cores' runtime operation. Ditto for the M1.

Intel and AMD have been very open about the fact that they have chipsets, SOCs, and IO chiplets providing USB, PCIe, etc. support for years now. What did you think they ran on, unicorns and pixie dust? Both x86 manufacturers have been open about using microcontrolers for this added functionality.

This is a complete lie that you have made up on the spot.

They don't mention firmware used to implement any of the functionality in any of the components you listed, in any of their documentation. Confidential or public. For example, we only recently learned through reverse engineering that all of intel's integrated GPUs feature an entire 486-based system with an OS just to perform shader scheduling and power management. Not only is it a closely guarded secret how they implement their hardware, but AMD or intel may even be unaware of firmware themselves when using eg. existing IP components from Synopsys DesignWare.

At any rate, the actual point (which you completely missed) is that much of this firmware is accessible and unsigned. If one wished, it could be replaced with open source software. Can't do that with any intel or AMD hardware.

You should educate yourself on the actual progress of the Asahi team, the way the M1 is actually put together and how it's protections work, how moden silicon is designed and fabricated, and what it's actually like to be a partner with AMD or Intel (or even to read their documentation).

But first you should learn to actually read and understand people's arguments. This hardware isn't a win because of apple's help. That's irrelevant to the question of the hardware itself. It's a win because of the unprecedented level of control and security that can be given to the user, in an everyday consumer computing device.

[schmidtbag]

Maybe it will be barely usable after a year of M2 released...

Well, M2 is likely an incremental improvement over M1, so it shouldn't take too much to get that caught up.

[jo-erlend]

Due to treadmilling, you are better off buying an M1 machine *now*. Then next year it might be usable once the hard work by the opn-source community catches up.

If you buy an M1 Macbook next year, you will need to wait *2* years for the community to catch up with its new changes from Apple. It isn't like the M1 is a stable platform, especially since it comes from a single vendor.

I think the idea was to buy second hand when the fans get the new one.