https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=283903
--- Comment #22 from Bjoern A. Zeeb <b...@freebsd.org> --- (In reply to Guillaume Outters from comment #21) The 16M ring + 1rsvd page would match as a start. Thanks a lot for the detailed answer. I just realized why we do see the 4K increases but I did not see any 4k allocations. contigmalloc in linux_skb due to the memory limit as-is in the tree. There are changes to come to change that for certain behavior. Can you also keep an eye on the mbuf-tag numbers (as that'll tells RX vs. TX apart). On main I use (to alos get the header; stable/14 might still be different): vmstat -m | egrep -e '(Use|lkpi|mbuf) -- You are receiving this mail because: You are on the CC list for the bug.
