Better power management, security, and scheduling in Linux kernel 6.13

Linux kernel 6.13 is here, but don't get too excited. It's not a biggie and, given the timing, probably won't appear in many familiar distros.

Head penguin herder Linus Torvalds announced Linux version 6.13 this week, and he's also chosen five lucky winners of guitar pedals. The KernelNewbies site has a full rundown of what's new, which, if we were a newbie, we feel would look pretty intimidating.

Although there are important changes in this release, they are either things that are only significant to those using certain specific models of CPU or other hardware, or they're steps in large-scale changes that will have a noticeable impact in future, but don't right now.

So, for instance, there's improved power management on certain models of AMD CPU and GPU. Specifically, there's better management of AMD's fancy 3D V-cache. There's support for Intel "Panther Lake" CPUs and Xe3 GPUs. There's better Apple device support too, covering some MacBooks, plus support for certain Apple SoC devices that were used in some older iPads and iPhones. However, this does not mean that you'll soon be able to install a kernel 6.13-based Linux distro on your old iDevice. It's just a step toward that maybe happening one day. Also in Arm land, there's improved graphics support for the Raspberry Pi's VideoCore GPU, support for running Linux inside an Arm64 Confidential Compute Architecture (CCA) realm, and more.

One terabyte SD cards are old news now. Users can already buy SD cards with capacities exceeding 2 TB. The Secure Digital Ultra Capacity (SDUC) standard defines card sizes up to 128 TB. As of this kernel, Linux supports them.

There's a new pre-emption model for the kernel's scheduler, referred to as Lazy Pre-emption, which aims to improve scheduler efficiency. Until recently, Linux had three pre-emption modes. At the end of last year, it gained another with the integration of the long-standing PREEMPT_RT patches in last November's kernel 6.12. Linux Weekly News explained the four modes in 2023 over two articles: part 1 and part 2. (Be warned, it took LWN two articles, so this is complicated.) Now, there's another model, and in October there was another LWN explainer. The plan is that the new mode will eventually enable the other non-realtime modes to be simplified down to just two.

The Ext4, Btrfs, F2FS, and XFS file systems all got tweaked. Notably, 6.13 can do atomic writes on Ext4 and XFS, which means it's now possible to write blocks of data larger than a single sector in a single operation. This will be handy for some databases, notably 2024 "Database of the Year" PostgreSQL.

Regular watchers of the rolling übernerd soap opera that is the kernel development community will have seen this coming, but there aren't any improvements to bcachefs, a file system designed for better performance and flexibility, since developer Kent Overstreet got his wrists slapped in November. Undaunted, he's sent in a large patch to refine bcachefs in the forthcoming kernel 6.14. This, sadly, means another on-disk format change.

We noted that this coincided with the announcement from Norwegian tech blogger Sesse Gunderson that he is migrating away from bcachefs. This comes a year after his now-deleted post on "Life with bcachefs," excerpted here. By April, he was having problems but finding ways around them. It looks like one of the more visible experiments in the Linux blogosphere is over. The future of bcachefs remains uncertain. ®