How Google's Android Maintains A Stable Linux Kernel ABI

[rastersoft]

God, I thought rabid open source fans will start pouring tons of shat right after my comment but it just happened so fast it took me by surprise, lol. I will now gladly leave this thread 'cause it's already been hijacked by bigotry and absence of any rationale.

https://www.youtube.com/watch?v=1t3cBTb3xPc

[loganj]

birdie i have to agree with you. i have a nvidia gpu and i know the problem of drivers not working with newer kernel. and of course opensource fanatics will tell that nvidia its at fault for not helping with opensource driver. its a must for drivers to be opensource otherwise we ran into problems with kernels. so kernel is not at fault here. just the driver

[geearf]

They intend to write their own kernel? Where did you see that?

He's probably talking about Fuchsia and Zircon.

[ssokolow]

They intend to write their own kernel? Where did you see that? They modify the upstream Linux kernel. That is not the same thing as writing a new one. It's not even close. I'm sorry if that runs counter to your "Linux sucks" narrative.

I believe it's a reference to the hypothesizing that Fuchsia is intended to be a successor to Android, taking advantage of how guidelines-compliant Android apps stick to ABIs that have no dependence on the Linux kernel.

[skeevy420]

tg--
What good is it to have stable ABIs just to allow derived code to live outside of the project, just to allow external - possibly license-infringing - code to influence the kernel itself, by forcing it to maintain code that wouldn't otherwise be maintained?

Seeing as how the current model is "oh, that project uses this part of the kernel so let's wall it off or remove that part" means that possible license infringing code is already influencing the kernel itself. A stable API means that an external project has to work within the limitations of the API versus use what's available until it isn't and needs to be fixed.

Do you honestly think that the kernel devs would have tweaked the GPL_EXPORTS if it weren't for ZoL? A stable API would have had what's supposed to be used by what set in stone years ago with revisions to the API set to go into effect with new LTS kernels. Like, 4.14 would have introduced something new or changed something old and all the kernels from there to the very last 4.18 release would be expected to respect that without a single tweak until 4.19.

If a stable API was actually the case, the ZoL people would have known what is good to use, what is safe, what is set in stone, what will be removed soon, etc. Extend that to an entire OS and Google making Fuchsia makes a lot of sense. A stable API would also mean that the kernel devs can say "if you don't hook into our kernel using these methods then you're in violation" whereas the current method is a cat and mouse free-for-all that ZoL and Nvidia users know all too well.

[tg--]

Which kinda invalidates the claim that Linux is the most popular kernel in the world 'cause Google doesn't use the mainline kernel. They have a very special fork with tons of patches which they maintain on their own. That also explains why they're busy developing Zircon - they want their own fast and efficient kernel which features stable ABIs/APIs out of the box, unlike the Linux kernel where stable APIs are nonsense.

I'd venture to say that kernel developers' position in this regard is nonsense but ardent open source fanboys will find the most bizarre arguments to defend this point of view even though many are using ZFS and other out-of-kernel modules whose developers consonantly struggle to maintain compatibility.

By that logic Linux is almost irrelevant, since very few people actually use mainline kernels.
Basically all distributions ship their own Linux forks, in all cases with their own patchset that's not in mainline.
Even completely community based distros maintain their own kernel patchsets, like Debian and even Gentoo (where you build your own kernel) - so it shouldn't come as a surprise that the largest, and most divergent, Linux distribution (Android) uses its own patchset.
Since the Linux development model is to have thousands of forks, I fail to see the issue.

It is also perfectly valid to use your own fork, since Linux isn't "one thing". Linus Torvald's fork is by no means the only way to use Linux. There even is another fork that's far more widely used than Linus' version and generally considered "mainline", which is GregKH's "stable"-series, and also a number of LTS-series by various other developers.

Of course it is also valid to make the argument, that stable internal APIs/ABIs are nonsense.
Linux' model is to pull all code into a single place, which makes sense for a project that's GPL licensed and demands derived code to be made available under the same terms anyway.
What good is it to have stable ABIs just to allow derived code to live outside of the project, just to allow external - possibly license-infringing - code to influence the kernel itself, by forcing it to maintain code that wouldn't otherwise be maintained?

Having no stable APIs isn't just because they don't want them, it's to actively push people to honor the license and upstream code.

Of course it is a very different situation for Windows' and iOS' kernel, which are closed source software. Third party projects actually HAVE to use a different license and can't possibly be upstreamed (with small exceptions in the FreeBSD portion of iOS' drivers). That development model also demands stable ABIs since a much larger portion of external code is used.

[DanL]

A major software developer which has struggled with the Linux kernel so much, they intend to write their own and I'm writing crap?

They intend to write their own kernel? Where did you see that? They modify the upstream Linux kernel. That is not the same thing as writing a new one. It's not even close. I'm sorry if that runs counter to your "Linux sucks" narrative.

Hundreds of Phoronix users mourn each time ZFS, NVIDIA and other out-of-kernel modules break when Linux developers start hiding kernel calls behind GPL, or remove or change them drastically and I'm writing crap?

Yes, crap. There are other factors at play there beyond stable interfaces. I too, wish Linux devs could be a bit more accommodating towards out of tree modules, but that doesn't mean I want them to move to a stable ABI.

God, I thought rabid open source fans will start pouring tons of shat right after my comment but it just happened so fast it took me by surprise, lol.

Hey, if you come on to a Linux news site to repeatedly proclaim that Linux sucks, people will call you out on your antagonistic BS. How about that?! Of course, you already knew that. That's how you get your lulz.

I will now gladly leave this thread

One could only hope. It would be even better if you left the site.

[sa666666]

I'd venture to say that kernel developers' position in this regard is nonsense but ardent open source fanboys will find the most bizarre arguments to defend this point of view even though many are using ZFS and other out-of-kernel modules whose developers consonantly struggle to maintain compatibility.

Your trolls are obvious, and always the same. Your first paragraph is about some big, vague conspiracy or veiled comment on Linux ('damning with faint praise'). Then your second paragraph is basically flamebait, instigating an argument before anyone has even said anything. Basically setting up any response to be adversarial; aka, starting a fight.

God, I thought rabid open source fans will start pouring tons of shat right after my comment but it just happened so fast it took me by surprise, lol. I will now gladly leave this thread 'cause it's already been hijacked by bigotry and absence of any rationale.

Yep, and you got exactly what you wanted. You start off in an adversarial fashion, get your rocks off by the responses, then get to claim to be an innocent victim and leave the thread. Look, we're all on to you and your antics here. If you're so interested in leaving, next time why don't you consider not even posting, and save everyone else the trouble. Oh, that's right, then you won't get your jollies from trolling.

Looks like another one for the global killfile soon. I really wish blocklists were working; your spewage is the perfect place for it.

[birdie]

A major software developer which has struggled with the Linux kernel so much, they intend to write their own and I'm writing crap? Hundreds of Phoronix users mourn each time ZFS, NVIDIA and other out-of-kernel modules break when Linux developers start hiding kernel calls behind GPL, or remove or change them drastically and I'm writing crap? Linux desktop is nowhere to be seen and I'm writing crap? Native ports of AAA games basically don't exist and I'm writing crap? Looooool.

God, I thought rabid open source fans will start pouring tons of shat right after my comment but it just happened so fast it took me by surprise, lol. I will now gladly leave this thread 'cause it's already been hijacked by bigotry and absence of any rationale.

Now please discuss how Google are idiots and how they should force thousands of hardware vendors to open source their device drivers. As if Google doesn't already have enough sway. Oh, wait, those vendors depend their survival and profits on Google yet they refuse to open source their drivers. The world must be incompatible with open source ideals.

[DanL]

Which kinda invalidates the claim that Linux is the most popular kernel in the world 'cause Google doesn't use the mainline kernel.

Semantics. Google certainly does use the Linux kernel (especially LTS ones). Read the slides.

I'd venture to say that kernel developers' position in this regard is nonsense but ardent open source fanboys will find the most bizarre arguments to defend this point of view

I'd venture to say that at least 90% of the crap you write here is nonsense. Just because Google wants a stable ABI in their ecosystem does not mean the approach is necessary/desirable upstream or for other Linux users. The great thing about Linux is that both approaches can exist, especially with LTS kernels.