On 07/03/25(Fri) 17:30, Julian Smith wrote: > On Sun, 2 Mar 2025 18:06:17 +0100 > Martin Pieuchot <m...@grenadille.net> wrote: > > > Hello Julian, > > > > Thanks for the report. > > > > On 01/03/25(Sat) 17:57, Julian Smith wrote: > > > >Synopsis: High CPU load for pagedaemon and mouse freezing > > > >Category: kernel > > > >Environment: > > > System : OpenBSD 7.6 > > > Details : OpenBSD 7.6-current (GENERIC.MP) #477: Thu Dec 12 > > > 21:42:00 MST 2024 > > > > > > dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP > > > > > > Architecture: OpenBSD.amd64 > > > Machine : amd64 > > > >Description: > > > Over the last year or two my OpenBSD system has often frozen > > > (including the mouse pointer) for varying amounts of time, > > > usually for a second or so when opening a new browser windows, > > > but sometimes for a few 10s of seconds or even minutes. > > > > > > When it freezes, top shows high CPU usage for pagedaemon, > > > sometimes 150%. And systat shows a number of `@` characters in > > > the Proc display (which i think means something is spinning?), > > > i've seen this extending halfway across the available bar space. > > > > > > I upgraded from 7.6 to current a few weeks ago; this does not > > > seem to have changed the behaviour. > > > > > > I have 16G of ram, and probably 50 Firefox and tor-browser tabs > > > open; there doesn't seem to be particularly high CPU load from > > > these, perhaps because i use uBlock origin and NoScript. `top` > > > is currently reporting this memory usage: > > > > > > Memory: Real: 8593M/15G act/tot Free: 769M Cache: 498M > > > Swap: 7889M/16G > > > I've tried disabling swap and setting hw.smt=1; neither of these > > > changed the behviour. > > > > Have you tried a -current snapshot? I made a lot of improvements to the > > pagedaemon post 7.6. > > Ah, interesting. > > I was already using a current snapshot (GENERIC.MP#477 amd64); i've now > upgraded with `sysupgrade -s` to `GENERIC.MP#577 amd64`. > > Unfortunately this doesn't seem to have changed the behaviour. > > For the first few hours `top` reported zero swap usage and there were > no hangs or freezes, then swap started to increase.
How do you generate this behavior? Could you tell me how to reproduce it? > Then it started freezing in the same way as i've seen before. Here is > the output of top and systat just after one of the early freezes: Can you enter ddb when the freeze happen? If you can, what is the output of "show uvmexp" "show bcstats" and "ps". Thanks, Martin
