Linuxeden开源社区LINUX KERNELLinus Torvalds刚刚发布了Linux 6.2内核的稳定版，标志着2023年的第一个主要内核版本。
Linux 6.2接替Linux 6.1成为2022年的LTS内核，将至少维护到2026年底。预计Linux 6.2中的许多修正会被回传到Linux 6.1中，对于那些还没有被选中的长期支持内核。
Linux 6.2的功能非常多，有全面进展，而且像英特尔、AMD、谷歌、红帽和其他常见的问题有很多改善。Linux 6. 2促进了英特尔Arc图形（DG2/Alchemist）的稳定和开箱即用性，英特尔的按需驱动现在对第四代至强可扩展 “Sapphire Rapids “CPU有很好的支持，对NVIDIA GeForce RTX 30系列 “Ampere “加速图形有最初的支持，有Nouveau开源代码。苹果M1 Pro/M1 Max/M1 Ultra的支持已经上游化，鉴于Retbleed比启用IBRS的成本更低，呼叫深度跟踪已经合并，以帮助旧的英特尔Skylake时代的PC的性能，各种文件系统驱动程序的增强，安全改进，以及其他。
文末有Linux 6.2内核功能列表，以全面了解该内核版本的所有激动人心的变化。
Linux 6.2
由于Linux 6.2的开发周期是在圣诞节/新年假期前后开始的，而且许多开发人员/测试人员由于欢度新年而没有时间工作，所以多拖了一周。但是现在，Linux 6.2已经发布了，是时候开始同样令人兴奋的Linux 6.3开发周期了。
Linux 6.2将有望成为Ubuntu 23.04的默认内核，以及在4月下旬继任的v6.3内核推出之前在Fedora 38等中发现。从我们的持续测试来看，Linux 6.2总体上处于良好状态，特别是对于Skylake/Skylake衍生的内核来说，如果选择调用Depth Track，看起来会更好，而且在实验室里有稳定的Arc Graphics A380/A750/A770方便应用。
新的内核，像往常一样，可以从kernel.org下载，供那些推出自己的内核构建的人使用。
Linux 6.2 功能列表：
Processors / Architectures:
– AMD Zen 4 pipeline utilization data is now exposed for perf to help developers/administrators in profiling and finding performance bottlenecks with new Ryzen 7000 series and EPYC 9004 series processors.
– Ampere Altra’s SMPro co-processor has seens everal drivers upstreamed for Linux 6.2.
– Fixing the broken strcmp() implementation for Motorola 68000 series that turns out has always been broken in a subtle way for the m68k.
– A scalability improvement for large IBM Power systems.
– RISC-V support for persistent memory devices.
– Intel’s Linear Address Masking (LAM) feature had been proposed and shortly merged to Linux Git but then un-pulled it with Linus Torvalds being upset over some fundamental issues with the Intel LAM code. So it’s not in Linux 6.2 now and will need to be re-worked.
– The Intel IFS driver has been fixed up for this In-Field Scan feature for providing CPU silicon testing capabilities with upcoming Intel CPUs.
– Intel On Demand Driver is ready for use with more functionality implemented as well as now labeling it Intel On Demand rather than the former “Software Defined Silicon” naming. Intel On Demand / Software Defined Silicon is the controversial feature for licensed activation of certain CPU features in upcoming Xeon Scalable processors.
– Intel TDX guest attestation support has been merged as the latest Trust Domain Extensions (TDX) work.
– KVM prepares for exposing new Intel CPU instructions.
– A power savings tweak for Alder Lake N and Raptor Lake P processors.
– Intel SGX Async Exit Notification “AEX Notify” support for helping fend off some forms of SGX (Secure Guard Extensions) attacks.
– Various AArch64 improvements such as dynamic shadow call stack support.
– A new control for the split-lock detector due to a prior kernel change around split-lock detection/enforcement hurting the performance of some Steam Play games.
– Support for more Qualcomm Snapdragon SoCs as well as the Apple M1 Pro/Ultra/Max has now been mainlined. With the Apple Silicon enablement push is also the new CPUFreq driver being merged.
– AmpereOne mitigation for Spectre-BHB.
Graphics:
– Initial NVIDIA RTX 30 “Ampere” GPU acceleration within the Nouveau driver but the performance is still extremely poor.
– Energy sensor monitoring support for DG2/Alchemist graphics via the HWMON interfaces.
– Continued enablement around Meteor Lake graphics support.
– Intel DG2/Alchemist graphics are stable with no longer being hidden behind a module flag for enabling. This affects current Intel Arc Graphics, Flex Series, and other DG2-based Intel GPUs.
– Various other DRM graphics driver updates.
– FBDEV support for the “nomodeset” option.
– Raspberry Pi 4K @ 60Hz display support.
– Allwinner A100 and D1 display support within the Sun4i DRM driver.
– Tieing into the graphics DRM code is the new compute accelerator “accel” subsystem/framework.