Google Finally Shifting To "Upstream First" Linux Kernel Approach For Android Features

[phoronix]

Phoronix: Google Finally Shifting To "Upstream First" Linux Kernel Approach For Android Features

Google's Android had been notorious for all of its downstream patches carried by the mobile operating system as well as various vendor/device kernel trees while in recent years more of that code has been upstreamed. Google has also been shifting to the Android Generic Kernel Image (GKI) as the basis for all their product kernels to further reduce the fragmentation. Looking ahead, Google is now talking of an "upstream first" approach for pushing new kernel features into mainline Linux before deploying them on Android...

https://www.phoronix.com/scan.php?page=news_item&px=Android-Linux-Upstream-First

[andrei_me]

Awesome plan, but I fear they won't be able to execute it 100% as planned, specially having to follow upstream release cadence and code review 🤔

[sverris]

Why not switching to mainline Linux kernel entirely?

[Old Grouch]

Hmm. Could this be an easy PR win for Google, right before they pivot to using the Zircon kernel from Fuchsia OS with Starnix instead of Linux as the Android kernel.

https://fuchsia.dev/fuchsia-src/contribute/governance/rfcs/0082_starnix

Isn't it easier to build the ABI shim/compatibility layer based on the mainline kernel rather than one that has been heavily patched?

Cynical, moi?

[uid313]

What about system-on-a-chip (SoC) device drivers?
Things like Qualcomm Snadragon, Samsung Exynos, MediaTek Helio, Google Whitechapel, etc?

[caligula]

I wonder how this would help at all. The SoC and phone vendors don't want to provide more than 0 to 3 years of support (starting from the product launch day). Any Android device will have millions of lines of code for proprietary drivers and firmware.

[birdie]

It's all about saving money - it has nothing to do with wanting to mainline or support upstream. Maintaining out of tree patches is bloody expensive due to Linux kernel developers' stable API nonsense mantra.

It's kinda beneficial for everyone in the end.

[sinepgib]

Hmm. Could this be an easy PR win for Google, right before they pivot to using the Zircon kernel from Fuchsia OS with Starnix instead of Linux as the Android kernel.

Bold to assume anyone but some nerds like us care about what's the running kernel... For PR to work you need someone interested.

Isn't it easier to build the ABI shim/compatibility layer based on the mainline kernel rather than one that has been heavily patched?

Besides, no, because most of the kernel changes aren't really at the ABI level and applications can't see them.

[phoenk]

Why not switching to mainline Linux kernel entirely?

Because the mainline kernel doesn't support the devices and features they need to support. That's what this effort is trying to fix: they are upstreaming what they can, and modularizing what they can't, to close the gap. No one maintains a kernel just because they can, and Google is trying to make money, after all.

[bemerk]

I'd love to see this approach being extended to AOSP in general. Lots of teams, companies, oems and individual people have great Forks of android with nice features and changes. If we had a Linux kernel style development with AOSP, with area maintainers and merge requests, android would be amazing