Linux 5.9 Will Be Delayed One Week
The Linux 5.9 release cycle has not gone as smoothly as earlier release cycles. There has been some unfortunate virtual memory management issues and slab corruption issues during the 5.9 release cycle which are, hopefully, fixed in 5.9 rc7. Linus Torvalds wants to delay the final 5.9 release just to be sure those are gone. He neglected to mention that 5.9 rc7 introduced a regression that prevents certain Intel integrated graphics chips from being used as more than coasters.
The Linux kernels release-cycle is usually very quiet after the first release candidate is introduces. New features go in during a "merge window" which slams shut since the first release candidate is set in store. The rest of the release cycle tends to be boring and uninteresting with minor bug fixes and minor adjustments.
The Linux 5.9 release cycle has been different. Linux 5.9 rc4 introduced a very unfortunate regression affecting the entire virtual memory management sub-system. A fix was introduced when 5.9 rc6 was introduced. This fix appears to be alright. But the woes did not end there. 5.9 rc6 introduced a slab corruption issue which was fixed with a one-liner shortly after it was released. You'd think this was the end of the problems surrounding Linux 5.9 but that isn't so.
Linux 5.9-rc7 has a Intel graphics regression that prevents Xorg from starting on most, if not all, machines with Intel integrated graphics. Linux 5.9-rc6 does not have this problem, it's new. Linux 5.9-rc6 did of course have the problem where Linux randomly hangs if on a lot of Intel laptops unless the kernel parameter
ahci.mobile_lpm_policy=1 is present and it did of course also have an issue where it randomly hangs without
intel_idle.max_cstate=1, meaning you need both to have a stable system, but those are not new. Those have been a problem since Linux 5.0 was released ages ago. This new 5.9-rc7 problem is unrelated to the already-existing problems the Linux kernel has with Intel graphics chips.
Linux kernel architect Linus Torvalds had this to say about this weekends Linux Linux 5.9-rc7 release:
"So we finally have all the issues I know about sorted out - the fix for the VM issue I mentioned in the rc6 announcement is here, as is the fix for the slab corruption issue that was separately discussed, along with another silly page locking bug one-liner fix.
But while I do now know of any remaining gating issues any more, the fixes came in fairly late. So unless I feel insanely optimistic and/or a burning bush tells me that everything is bug-free, my plan right now is that I'll do another rc next Sunday rather than the final 5.9 release. And btw, please no more burning bushes. We're kind of sensitive about those on the West coast right now.
Anyway, while the MM side is what kept me on my toes last week, most of the changes here are actually drivers and networking. And networking drivers. With a small smattering of documentation and filesystem fixes and other noise thrown in.
Shortlog appended, but what I really hope you all will do is to give it a nice good testing. One extra week or rc kernels will help, but only if people actually try this out.
The one-week delay is not a huge issue. Those who want support for things like the Corsair Commander PRO RGB and upcoming AMD and Intel graphics chips will just have to wait a while. AMD will not release the new graphics cards Linux 5.9 adds support for until October 28th, so the Linux 5.9 will still be released because those launch even if it is released a week alter than expected. And all Intel integrated graphics is broken on Linux 5.9-rc7 so support for upcoming Intel chips is kind of a moot point.
It would, perhaps, be prudent to delay the Linux 5.9 release with more than one week. There are many people out there who do use machines with integrated Intel graphics. It would be a shame if a "stable" kernel release where to break graphical output on all of those.