Linux kernel developers continue building out the support around the [1]Compute Express Link (CXL) specification for benefiting modern high performance servers. With the in-development [2]Linux 6.16 kernel there are more CXL features now in place.

One of the main areas of work for CXL in this next version of the Linux kernel is on enabling CXL Reliability, Availability and Serviceability (RAS) functionality. CXL is now tied into the Linux kernel's Error Detection and Correction (EDAC) subsystem code for leveraging that functionality for RAS. Patrol Scrub Control, Error Check Scrub, Perform Maintenance and Memory Sparing are the CXL RAS features to be enabled for Linux 6.16.

Patrol Scrub Control is part of the CXL 3.2 specification for proactively locating and correcting errors. This CXL feature allows setting the number of hours for which the patrol scrub must be carried out as well as limits for the scrubbing in trading off error rates and performance. The CXL Patrol Scrub support for Linux was worked on by Huawei.

Error Check Scrub "ECS" is another CXL feature and part of the JEDEC DDR5 SDRAM spec for memory devices to perform error checking and correction and count single-bit errors.

The Perform Maintenance command is used to request the CXL device carry out maintenance operation(s) on that device.

CXL Memory Sparing is for replacing a portion of memory within a portion of functional memory at the same DPA. This Memory Sparing works for devices supporting the CXL.mem protocol and allow cacheline/row/bank/rank sparing.

Plus there are documentation updates and various other improvements to the Compute Express Link code for Linux 6.16. All of the details on these changes can be found via [3]this Git merge .



[1] https://www.phoronix.com/search/Compute+Express+Link

[2] https://www.phoronix.com/search/Linux+6.16

[3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=29e9359005dd1ac5f9683608891718e6a32a20a3