I'm still debugging a 11x CPU slowdown in 16-current between 29 october and today.

One difference in dmesg is that "cpufreq0: <CPU frequency control> on cpu0" is printed for each CPU in the bad kernel. So I figure some changes in cpufreq went in.

Hmmmm. 3e6e4e4a0d42fa24f3b2a1c087e9ad25f9594081: hwpstate: add CPPC support for pstate driver on AMD

nimaje: ok that did the trick. thank you so much. i was about to give up. lol

Looks like I'm running at 399 MHz, which fits the benchmarks.

cracauer: was that it?

I am still building a kernel on 3e6e4e4a0d42fa24f3b2a1c087e9ad25f9594081 - 1. Takes a while now.

Yupp, that commit works fine.

f/6

derp. i was wondering if anyone has insights/mitigations for the atheros driver lock issues in hostap mode? wiki.freebsd.org/WiFi/LockIssues

awesome. i think i ironed out the major issues.

it really does hurt to not have mount.cifs though. maybe next weekend i'll investigate the possibility of going to nfs for everything

hodapp: nope, but I did change some flags since, so maybe it can work now?

lemme try real quick

I'm trying to do NixOS on 9p in Bhyve which is *probably* ill-advised and it's mostly working but running into cryptic things here and there

huh

Cryptic things like what?

some builds in nixos-install failing and trying to investigate why

Huh. hodapp: I dunno what I did wrong last time (maybe it is about flags in fstab), but postgres works 100% fine on p9fs.

(I *did* tell it to use 9p2000.L protocol, explicitly, so maybe that?)

(might've been using .u before or similar)

I have a /path/to/p9fs in Linux that's mapped from a FreeBSD fs, and it's fine. At least so far.

I wouldn't 100% trust it, mind, not when it's done like this.

DarkUranium, if you can mount something manually, there's an easy trick for getting it into fstab.

You mean mount -p?

Why yes!

The discussion was about a Linux guest (in bhyve), no mount -p there.

Yeah. . . . linux sucks that way.

But anyway, fstab wasn't the problem. The choice of flags for 9p was.

Acutaly. . .

/proc/mounts

Right, I forget about proc.

Kinda feels like a labyrinth, TBH.

It kinda is.

/proc/kcore is always such a huge file too.

ok. this is just odd.

Is anyone else seeing build errors with dns/unbound with python=on?