Linux Kernel 6.12 Has Been Released

Slashdot unixbhaskar writes: Linus has released a fresh Linux kernel for public consumption. Please give it a try and report any glitches to the maintainers for improvement. Also, please do not forget to express your appreciation to those tireless folks who did all the hard work for you.
The blog OMG Ubuntu calls it "one of the most biggest kernel releases for a while," joking that it's a "really real-time kernel." The headline feature in Linux 6.12 is mainline support for PREEMPT_RT. This patch set dramatically improves the performance of real-time applications by making kernel processes pre-emptible — effectively enabled proper real-time computing... Meanwhile, Linus Torvalds himself contributes a new method for user-space address masking designed to claw back some of the performance lost due to Spectre-v1 mitigations.

You might have heard that kernel devs have been working to add QR error codes to Linux's kernel panic BSOD screen (as a waterfall of error text is often cut off and not easily copied for ad-hoc debugging). Well, Linux 6.12 adds support for those during Direct Rendering Manager panics...

A slew of new RISC-V CPU ISA extensions are supported in Linux 6.12; hybrid CPU scaling in the Intel P-State driver lands ahead of upcoming Intel Core Ultra 2000 chips; and AMD P-State driver improves AMD Boost and AMD Preferred Core features.
More coverage from the blog 9to5Linux highlights a new scheduler called sched_ext, Clang support (including LTO) for nolibc, support for NVIDIA's virtual command queue implementation for SMMUv3, and "an updated cpuidle tool that now displays the residency value of cpuidle states for a clearer and more detailed view of idle state information when using cpuidle-info." Linux kernel 6.12 also introduces SWIG bindings for libcpupower to make it easier for developers to write scripts that use and extend the functionality of libcpupower, support for translating normalized error addresses reported by an AMD memory controller into system physical addresses using a UEFI mechanism called platform runtime mechanism (PRM), as well as simplified loading of microcode patches on AMD Zen and newer CPUs by using the family, model, and stepping encoded in the patch revision number...

Moreover, Linux 6.12 adds support for running as a protected guest on Android as well as perf and support for a bunch of new interconnect PMUs. It also adds the final conversions to the new Intel VFM CPU model matching macros, rewrites the PCM buffer allocation handling and locking optimizations, and improves the USB audio driver...

"Please give it a try..."

[Valgrus Thunderaxe]

"...and report any glitches to the maintainers for improvement."

If you want to get publicly flamed, just try that.

PREEMPT_RT

[VaccinesCauseAdults]

Why not call it PREEMPT_REALTIME? What’s with the 1970s C abbreviations?

Re:

[fahrbot-bot]

Why not call it PREEMPT_REALTIME?

That one's used to cut-off Bill Maher [wikipedia.org]. :-)

Re:

[VaccinesCauseAdults]

Okay I’ll pay that.

Re:

[backslashdot]

Memory is expensive .00000000001 cents per byte these days. That quickly adds up when you have billions of devices.

Re:

[VaccinesCauseAdults]

It is a kernel build-time configuration flag for enabling or disabling code paths via conditional compilation (preprocessor macro). It doesn’t ship on billions of devices.

Re:

[bn-7bc]

Is is really this cheap when uou neeed/want a TB+ in a single server?

Linux Preempting

[TheRealMindChild]

Robert Love created and released realtime patches way back in the 2.4 kernel days. There was nothing wrong with them, either.

https://www.kernel.org/pub/lin... [kernel.org]

Re:

[NoWayNoShapeNoForm]

And his patches were not brought into the Linux kernel "mainline"?

I guess rml got no love from the larger Linux development world, eh?

Intel/AMD whatever, how is Apple Silicon improved?

[drnb]

Enough about legacy CPUs. What does this update mean for ARM and in particular Apple Silicon?

Re:

[ArchieBunker]

Sir this is slashdot, you’re not supposed to like Apple.

Re:

[DrMrLordX]

Apple needs to play nice and open up their hardware to alternative operating systems. Asahi makes do as best it can, supporting only M1 and M2.

Re:

[vbdasc]

Apple doesn't need that. They don't give a rat's azz about alternative operating systems, and never did. What they really need is to keep their users locked in the pen, walled garden or whatever. And they excel in that art.

Linux can't support something whose owner doesn't want it supported by Linux. At best, you get hackers who reverse-engineer the closed parts and create something like Asahi Linux, but it's not a sustainable solution and can't be officially supported.

Re: Intel/AMD whatever, how is Apple Silicon impro

[vbdasc]

Linux is an open collaborative project. You are welcome to participate in improving its ARM support, and in particular, to petition Apple to provide open specifications of their hardware, so it could be properly supported by open-source OSes.

Re:

[bn-7bc]

Not to mention RISC-V, ARM is not exactly new, or open for that matter

Re:

[test321]

What does this update mean for ARM

According to the merge summary from LWN:
* "The Arm "permission overlay extension" feature is now supported, making memory protection keys available on that architecture." https://lwn.net/Articles/82655... [lwn.net]
* "The Loongarch, 64-bit Arm, PowerPC, and s390 architectures have all gained support for the vDSO implementation of the getrandom() system call." https://lwn.net/Articles/98318... [lwn.net]

Classism!

[Gravis Zero]

Also, please do not forget to express your appreciation to those tireless folks who did all the hard work for you.

I will show my appreciation by purchasing tires for the tireless programmers that have worked on making Linux possible but I have to say, it seems really rude to point out that some kernel programmers can't afford tires to wear. BTW, when did wearing tires become en vogue? Are they like, bicycle tires or should I be walking around with a Goodyear around my waist?