Linux 6.1 will tell you when your CPU is crocked

Linux customers are set to get a brand new solution to determine defective CPUs, with Linux kernel model 6.1.

The brand new function will enable customers to print the socket and core that are probably accountable when a segmentation fault happens, which ought to enable customers to identify if a specific CPU/core is routinely inflicting issues.

This won’t imply a lot to the armchair Linux lovers, however sysadmins operating a fleet of Linux servers stand to learn.

How does the function work?

You possibly can allow the brand new function by including “print()” to “show_signal_msg()”, which is able to print the CPU, core, and socket on the time of the segfault.

This specific printk could be managed by going by way of /proc/sys/debug/exception-trace, and you may see the official documentation (opens in new tab) for extra data.

The authors of the Kernal documentation, Rik van Riel and Borislav Petkov, acknowledge that the function “just isn’t excellent” as “the duty would possibly get rescheduled on one other CPU between when the fault hit, and when the message is printed”.

Nonetheless, they did say that “in apply, this has been ok to assist individuals determine a number of dangerous CPU cores”.

For those who’re a developer and at the moment utilizing your Linux distro for demanding high-performance use circumstances you are unlikely to be contemplating utilizing this function in isolation. You are most likely going to need to additionally make use of widespread CPU stress testers corresponding to Prime95 or Aida64.

The replace will land throughout October as a part of the Linux 6.1 merge window. The earlier Linux replace, model 6.0, has solely not too long ago arrived for testing.

Although Linus Torvalds claimed that “there’s nothing essentially completely different about” model 6.0, the replace provided quite a few modifications, 60% of which have been driver updates, but additionally lined GPU, networking and sound.