Enterprises that care about using a kernel for many years typically put their own resources into providing the level of maintenance they require (or getting it from a distro that maintains their own LTS kernels), rather than depending on upstream to keep the patches coming for that branch. But if the upstream kernel support timelines become more closely aligned with what those downstream users want, they may shift to tracking upstream LTS branches.
Are we seeing Android phones upgrade their kernels yet? This Samsung S22 is still on 5.10. I thought that part of the idea for Android GKI was that phones would start getting kernel upgrades. But I'm not sure if that's actually happening.
I wish there was more pressure for this. Especially as Android Virtualization Framework starts really arriving & being useful, having a more modern kernel could be a very nice help, could offer neat new capabilities.
Okay, but 6.1 is still from December 2022. Like... it's an improvement, but as my desktop sits at 6.19 and 7.0 is impending, I have to question why they lag so much.
OP was talking about that they now have and pursue the intention of upgrading the kernel during the lifetime of the device.
Instead of device launching with LTS kernel, which is supported for many years upstream, and always using it, instead LTS kernels are supported for 2 years (or extended like here), and the devices keep moving on to the next lts branch during their lifetime (usually not immediately, but after the regressions fixed for next branch, tested well before that in avf VMS etc)
It could be, but are vendors actually upgrading kernels along with firmware updates? In my experience it's more like, ship 5+ year old kernel and then forget it forever.
So long as they keep up with patches that can be fine, but newer kernels also have useful feature improvements. If nothing else, performance tends to improve over time.
In practice upgrading kernel can easily cause performance regressions and cause multiple other issues (reduced battery life) so there's a lot of risk for zero reward for an OEM to do that.
After all, they're on the hook for not breaking users already working devices and don't get anything by risking lawsuits and recalls.
I'll grant that changes leave the possibility of regressions, but that's true for minor patches too, so you already need a lab set up to catch those regressions, and if you've got a lab set up to catch regressions and engineers who can fix them, then you might as well take the bigger upgrades too.
A couple times a year I get the joy of reading Kernel Newbies release notes for new kernels. And just being so delighted at all the amazing improvements happening. So many won't affect me or won't be big changes. But often there are amazing new capabilities and options too that do entice. Performance wins keep landing. Compatibility with other devices expands. Improvement is ongoing & continual. https://kernelnewbies.org/Linux_6.17
It just takes my breath away, is existentially scary, to hear folks be ok with being totally stuck on place. With devices that use open source but which are so fundamentally dead, are tuned out from that amazing growing goodness, are shovelware devices thrown over the wall that never improves while time marches forward.
This would be totally unacceptable for anyone on a computer. But somehow on consumer devices, it's just expected and accepted that everything is just stuck where it is, that it's ok to be so so so much worse than what everyone else is doing. It sounds like such a miserable existence, and I think it's just gobsmacking that such apathy & pass giving gets a break.
Will some things break sometimes? Honestly I think that's way overblown a concern, but yes, some, a tiny bit of things will break. Especially at first. I tend to think the risk is generally quite small. And often with engineering, the way to deal with your hard parts is to keep doing them. If anything, I expect that the risk of staying where we are is huge: we have thousands or perhaps millions of different kernel trees out there, bespoke special magical trees, for various devices, forked from magical point in time vendor BSPs, with special magical changes that we have to keep perpetuating, while integrating important fixes. This all sounds ridiculously unstable and risky, and inordinately costly to maintain. It seems reckless and dangerous to stick to this absurd course, to do everything so badly, at such human cost. Getting the fricking drivers upstreamed, maybe getting rid of this ridiculous anti-support GKI layer that apparently does no good and only makes it easier to be bad at updating & negligence & caring: that would reduce risk. That would decrease society from having to test these thousands or millions of kernels, and let us create a known predictable focus for our energies and tests, rather than this madcap batshit infinite vendor diversity that GKI has only sort of tamed.
This is such a shit situation to be in, and Android brings shame to computing, and if it's going to be so bad at Linux, it ethically doesn't deserve to have Linux. It is breaking the pact of what Linux can and should mean, and betraying consumers, by letting itself be a product that rots into obsolescence like this. This is a techno-spiritual sin, and it is a mortal sin.
Here: 20 years of desktops and laptops basically installing the latest kernel asap on Gentoo then Arch. I did break stuff, especially Gentoo, a lot... but out of those I maybe got hit once by a kernel regression?
I don't think people realize how long it takes for the kernel to eventually catch up with your hardware.
My one year old framework laptop motherboard STILL don't have properly implemented usb-c PD apis in the kernel today. Imagine if I took a 5y old kernel?
That is news to the millions of linux users who upgrade their kernels regularly, and suffer zero consequences.
It's cowardice and FUD, in my view, to clutch to such old versions. It's just bad practice and bad engineering, and a crock of scary tales to make other people (& the people doing this) think their bad engineering & absurd self-injuring time-wasting practices are good, actually.
A lot of devices need to change their expectations around what qualification means, for a lot of systems & devices where requalification is such a pain that a kernel upgrade is a daunting task.
GKI is only stable within the point release. It means that 5.10 LTS Linux can be safely updated to the latest versions 5.10 LTS Linux. The regular LTS branch has no compatibility guarantees that drivers for one release will be compatible with the next release on that branch.
No 6.1? That's disappointing. Also I am surprised the previous decision wasn't reverted sooner. Linux foundation surely has enough resources to upkeep LTS kernels for longer.
We neither need, nor benefit from this precis, which is longer than the headline but contains no additional information and insight. On hn people are encouraged to read tfa for themselves.
If you have showdead on, that user's comment history is rather full of this sort of thing. Seemingly restarted half a year ago, but with similar conduct in the pre-LLM era as well.
Actually, as the article falls into that "ad begging" category and requires time-consuming disabling of tracking, I can understand why someone posts a summary.
> the Linux kernel is catching up with its users’ wants when it comes to longevity.
> Kernel end-of-life dates mean very little for users, even at the enterprise level.
So... no one cares about longevity? Or they do? I'm confused.
reply