On Tue, Dec 23, 2025 at 10:00:40AM -0800, Gleb Smirnoff wrote: T> On Mon, Dec 22, 2025 at 01:00:17AM -0800, Gleb Smirnoff wrote: T> T> This is an automated email to inform you that the December 2025 stabilization week T> T> started with FreeBSD/main at main-n282678-88b04633c29e, which was tagged as T> T> main-stabweek-2025-Dec. T> T> Due to lack of human and computational resources at Netflix during the T> holiday period, we are unable to make our regular A/B testing in a timely T> manner.
Late update on the December stabweek. The performance A/B testing at Netflix discovered two performance regressions both related to traffic distribution on a lagg(4). This was fixed by two commits: - ac1cd655f647 (applies to Intel NICs supported by ixl(4)) - 21865c970888 (applies to all) Also, a panicing regression in ipfw was reported. If you try to add a rule with 'log' keyword and without a rule number (autonumbered), the kernel would panic. There is a temporary patch attached to the email: https://lists.freebsd.org/archives/freebsd-current/2025-December/009754.txt There is no yet patch in main addressing this problem. Also, there is a report of a kernel panic in in6_selecthlim(). While the problem is known for a long time, but very likely with the December snapshot the race became easier to hit. Both panics are related to my changes, and I'm working on addressing them before the January stabweek. -- Gleb Smirnoff
