What Didn't Make The Cut For The Linux 3.8 Kernel

While there's a lot of features that are new to the Linux 3.8 kernel as covered in The Feature Overview For The Linux 3.8 Kernel, there's also several promising new features and functionality that didn't make the cut for this next kernel release.

- As talked about last weekend, VIA kernel mode-setting still hasn't been proposed for merging into the mainline Linux kernel. The VIA KMS support is now two years old, but hopefully we'll see the support called for pulling in 2013, for those few unfortunate souls still using VIA hardware.

- Reiser4 hasn't seen any new call for mainlining... While it's been talked about as a possibility in prior years, it doesn't look like it will ever happen unless some company is interested in investing heavily in the Btrfs/EXT4-alternative and is willing to maintain and support the code-base originally started by Hans Reiser. At least though Reiser4 is still being maintained out-of-tree and recently I did some Reiser4 benchmarks on the Linux 3.5 kernel compared to the alternatives.

- While there's been patches on the kernel mailing list for months, Loongson 3 CPU support wasn't merged this cycle.

- There ended up being no major advancements to Nouveau, the reverse-engineered open-source NVIDIA graphics driver, this cycle. The main feature many Linux users have been after is proper re-clocking support so the GPU core, shader clock, and memory clocks can actually be dynamically adjusted to run at their factory-specified speeds. By default the Nouveau driver is running at the frequencies the video BIOS programmed them to at boot time rather than their optimal frequencies, which is leading to performance shortcomings of this open-source DRM driver. There is experimental support for Nouveau re-clocking on some generations of NVIDIA GeForce hardware, but there's still a ways to go to improve this support.

- VFS hot-data tracking isn't yet ready for merging. Hot-data tracking is initially implemented for the Btrfs file-system and will allow for potential performance optimizations. "The feature comes down to tracking commonly used data on the disk through the VFS layer and storing file statistics to create "temperatures" on data to find the most commonly used data (the "hottest" part of the disk). The warmest data can then be automatically migrated to a faster disk (e.g. a solid-state drive) or just dealt with in a more optimized manner for its common usage."

- While the Linux 3.8 kernel has the initial Tegra DRM graphics driver for supporting the Tegra 2 and Tegra 3 SoCs, the open-source NVIDIA 2D acceleration support for the GeForce graphics on their ARM SoCs wasn't in a state for merging. We'll likely see the 2D experimental work merged for Linux 3.9. There's also the user-space X.Org Tegra driver that needs to be written too for implementing the 2D support. There isn't yet any public / open-source 3D acceleration for Tegra outside of NVIDIA's binary blob.

- On a similar Tegra note, NVIDIA Tegra 4 "Wayne" support made it too late for Linux 3.8 but should be found in Linux 3.9. This is the next-generation NVIDIA ARM platform built around the ARM Cortex-A15 CPU.

- Zswap wasn't merged for 3.8 as a means of compressed swap caching.

- DRM Render Nodes is still being worked on for providing better open-source Linux graphics security through the entire stack, but isn't yet in a state for merging to master. This also requires changes to the X.Org Server, Mesa, and other driver components.

Is there anything else missing from this next Linux kernel release you were hoping for? Share your Linux feature desires within the Phoronix Forums.