Linux 5.19-c4 - A Bit Bigger, Also Fixes A Previously Reported Performance Regression
Linus Torvalds commented in the 5.19-rc4 announcement, "So we've had a couple of fairly small rc releases, and here we finally start to see an uptick in commits in rc4. Not what I really want to see in the middle of the release cycle, but not entirely surprising considering how quiet it's been so far. And while 5.19-rc4 is a bit larger than previous rc's, and is a bit larger than we usually see at this point, it's by no means anywhere near record size. So more of a "a bit bigger than usual" than a "Oh my God, this thing is huge"."
As for the changes this week, they are all over the board. One item worth mentioning is what I wrote about earlier in the week with the kernel's signature verification code was made FIPS compliant by adding some basic self tests for this checking code used by module signing, Kexec, and other functionality.
Another change to find in Linux 5.19-rc4 is a fix for a previously reported performance regression on Phoronix. Back in March during the Linux 5.19 merge window I pointed out a big NUMA regression appearing within Stress-NG during the Linux 5.18 feature changes. In that article I bisected out that problematic commit. With Linux 5.19-rc4 there is now a fix in place.
Due to an email snafu I didn't notice a proposed fix for testing until the end of May. I confirmed the fix then took some time for it to go through Andrew Morton's "mm" patch queue to make it into the mainline kernel. The fix is in Linux 5.19-rc4 and presumably will be back-ported to Linux 5.18 too. The patch is the "mm: lru_cache_disable: use synchronize_rcu_expedited" change.
The rest of the changes this week is the usual assortment of mid-cycle fixes with nothing else really prominent catching my attention.
The Linux 5.19 stable kernel should be out around the end of July. See the Linux 5.19 feature list to learn about all of the shiny new features coming for the v5.19 kernel.