Before Linus Torvalds announced the release of the Linux 6.6 kernel, he said there’d also been a musical documentary made about it — which was actually Iron Maiden’s music video for ‘The Number Of The Beast’. While there are people who are both heavy metal and open-source fans, notably Jono Bacon, I didn’t know Torvalds was part of the crowd, too.
Be that as it may, Torvalds said: “I have absolutely no excuses to delay the v6.6 release anymore, so here it is.” And what we find in this release, besides the usual array of fixes and drivers, is a potpourri of new features.
Also: How to choose the right Linux desktop distribution
The one feature that caught my eye was the KSMBD In-Kernel SMB3 Server. KSMBD, which was designed by Samsung, is a new kernel module that implements the server side of the SMB3 file-sharing protocol into Linux’s heart. This module has been in the works for years — and it needed that time. The first versions of KSMBD were noteworthy for having more than their fair share of security problems. Thankfully, those problems have now been fixed.
The introduction of KSMBD does pose a question: “Why put an SMB server in the kernel when we’ve had Samba for decades?” There are two answers to this question. The first answer is KSMBD has a very small attack surface — and since file transfer protocols are often attacked, that’s important. The second answer is KSMBD should perform faster thanks to streamlined data transfer and file-sharing operations, and support for Remote Direct Memory Access (RDMA).
Many people are also excited by the arrival of the Earliest Eligible Virtual Deadline First (EEVDF) Scheduler. The EEVDF scheduler replaces the Completely Fair Scheduler (CFS), which has been around since 2007. These schedulers manage the allocation of CPU time between Linux processes, so no single process pigs out on processor time.
CFS has done a good enough job. But over the years, it’s become clear that, while some processes may not need a lot of CPU time, when they do need it, they need it right now. At the same time, other processes might need more CPU time, but can wait their turn. EEVDF addresses these latency issues more efficiently than CFS.
Under the hood, as its creator, Peter Zijlstra, wrote, EEVDF “completely reworks the base scheduler, placement, preemption, picking –…