For many years, Linux system administrators have gone out of their way to avoid swapping. The advent of nonvolatile memory is changing the equation, though, and swapping is starting to look interesting again — if it can perform well enough. That is not the case in current kernels, but a longstanding project to allow the swapping of transparent huge pages promises to improve that situation considerably. That work is reaching its final stage and might just enter the mainline soon.

The use of huge pages can improve the performance of the system significantly, so the kernel works hard to make them available. The transparent huge pages mechanism collects application data into huge pages behind the scenes, and the memory-management subsystem as a whole works hard to ensure that appropriately sized pages are available. When it comes time to swap out a process's pages, though, all of that work is discarded, and a huge page is split back into hundreds of normal pages to be written out. When swapping was slow and generally avoided, that didn't matter much, but it is a bigger problem if one wants to swap to a fast device and maintain performance.

One of the the most commonly repeated mistakes in system-call design is a failure to check for unknown flags wherever flags are accepted. If there is ever a point where callers can get away with setting unknown flags, then adding new flags becomes a hazardous act. In the case of mmap(), though, developers found a clever way around this problem. A recent discussion has briefly called that approach into question, though, and raised the issue of what constitutes a kernel regression. No changes are forthcoming as a result, but the discussion does provide an opportunity to look at both the specific hack and how the kernel community decides whether a change is a regression or not.

Back in 2017, several developers were trying to figure out a way to safely allow direct user-space access to files stored on nonvolatile memory devices. The hardware allows this memory to be addressed directly by the processor, but any changes could go astray if the filesystem were to move blocks around at the same time. The solution that arose was a new mmap() flag called MAP_SYNC. When a file is mapped with this flag set (and the file is stored on a nonvolatile memory device), the kernel will take extra care to ensure that access to the mapping and filesystem-level changes will not conflict with each other. As far as applications are concerned, using this flag solves the problem.

Please take eight minutes to complete this survey. The results will be shared publicly on The New Stack, and The Linux Foundation’s GitHub page.

Mesa 18.1.4 is planned for release this Friday, July 13th, at or around 10 AM PDT.

Another routine Mesa 18.1. point release is being prepared while waiting for the August debut of the Mesa 18.2 feature update.

Dylan Baker, the Mesa 18.1 release manager and his first stab at the task, has announced the Mesa 18.1.4 release candidate today. In its current form, Mesa 18.1.4 is comprised of just over two dozen patches.

It's rare in recent years to have anything to report on xf86-video-ati, the X.Org driver for the display/2D experience for pre-GCN Radeon graphics cards. But this week has been a large batch of fixes and improvements for those using this DDX driver with pre-HD7000 series hardware.

Longtime Radeon Linux driver developer Michel Dänzer has landed a number of commits already this week of various fixes/cleanups, some of which were inspired by the xf86-video-amdgpu DDX driver that is used for current-generation hardware with the AMDGPU kernel driver (unless using xf86-video-modesetting...).