it does but not to the degree that zfs does, as for ufs ... a nocache option not sure might be under flushing support

might merge 9pfs into my local branch and see if that works better

can anyone answer my question about that tuning guide, or help with network performance?

well this is annoying, my VM is broken and i can't fix it because instead of dropping to the load prompt, it crashes the loader (or uefi)? le-fay.org/tmp/30d/1Ve858.txt

I'd use a sparse zvol on the host and use that as the VM block storage device, then format that with UFS.

_xor: i stopped using zvols when klara posted rather convincing number suggesting that raw files are *much* faster

Oh yeah, I think I remember seeing that article. Forgot about it.

i'm not opposed to using UFS, but i want to run poudriere on this VM and i'm not sure how well it performs on UFS

since it can't do its normal fs cloning thing

I tried that a while back, but I wanted to be able to send/recv datasets from the poudriere system, so I went back to ZFS.

_xor: "bhyve-equivalent to the jailed zfsprop" -> reviews.freebsd.org/D41844 (you can use this as root filesystem!)

Title: ⚙ D41844 Add an implementation of the 9P filesystem

ok it's not quite the same as a proper zfs delegation but it's still pretty nice

I think that was also when I was building host-bootable images & jail images using poudriere.

_xor: how was performance with poudriere on ufs? was it not quite slow building small packages with all the overhead?

I don't remember now. I don't think I ran it long enough on UFS to really notice performance.

oh this is fun, make meta mode has decided to just rebuild all of clang for no reason

virtio_p9fs looks pretty neat

I'm a bit confused why sh history gets written when I run poweroff, but gets lost when I run reboot

tm512: what about shutdown -r now?

meena: trying that

shutdown -r retains history, I'm not sure what the difference is between that and the reboot command, and why the latter apparently isn't waiting for sh to terminate before powering down

tm512: which FreeBSD Version you on?

i thought we had fixed reboot to… behave

14-STABLE, the snapshot from jan 25

on another note, I got wifibox set up, seems a lot more stable than the native rtw88 port. seems like the default RAM allocation of 128MB is pretty conservative and that I could probably dial it down to 64MB?

can Linux even boot in 64MB these days?

(well, the best way to find it is to try it, i guess)

gentoo cann

with specific hardware and kernel tweaks

basically compiling the kernel specifically for what you have ...

free -m in the wifibox console says total is 110, 29 used, 19 buff/cache

then 61 free and 75 available

I'm just worried that wifibox's memory demands could fluctuate

hmm, so I notice that apparently firefox does not support process sandboxing on fbsd. how much of a security concern is that in reality? like the web is where I'm most likely to run into malicious code

alpine? anyone?

meena, hi there

how can I search in vi using regexes ?

I'm trying :set extended

then /20\d{2} to search for something like 2024

but it doesn't match

Good morning

depends where you are, but yes

I used to leverage CARP before with no issue, but today I faced with surprising stuff I don't seem to figure out on my own. Dunno if it's my rusty understanding of CARP leading to misconfiguration and misuse or something else I can't clock. Here's my ifconfig and netstat outputs from the two servers:

Title: dpaste/AeSp (Plain Text)

My two problems are:

1. pinging even the gateway has more than 50% packet loss when CARP is up

2. the backup node fails to have net access when master's up (I can handle it with a `nat on vtnet0 from !($vtnet_ip) to any -> (vtnet0:0)` pf rule, but the first issue's what I shall resolve first, I believe

Any thoughts?

plenty. but you dont want to hear them.

I'm all ears ;-)

oh, they're not related to carp or freebsd however..

Haha, no worries, we can chat till someone comes up with a thought on my issues as well. As far as chats ain't off topic and relating to #freebsd-social, anything's welcome and warmly accepted by folks here, I believe.

nah. nobody wants to hear my thoughts...not even me

but if you are using carp to share 2 hosts and getting 50% packet loss...is one of your hosts misconfigured?

Well, 'tis a thought per se, innit?

And I listened ;-p

I powered off one to see if it solves the case, but nah, as far as carp's enabled on the interface, the loss is there killing the packets.

i dont use carp...but if you are familiar with ethernet throughput issues i've been hoping to get some help here for a couple days. youre the first person who hasnt ignored me entirely

Nobody ignores you, peeps just respond to questions they've got an answer for, so... yep. I know how it feels, but nobody's being mean, trust me.

I don't know if I could have a solution either, but try me, I might.

ethernet throughput issues.

:-]

maxing out one cpu core i believe

What "issue" do you have ATM? Explain it a bit further, and we'll see.

i'm getting 500mb/s from freebsd, 1gbps from linux

iperf3 definitely maxes out the cpu, but an internet download shouldnt and is still slowed down by my opnsense firewall

it was faster before the latest release

there's a tutorial i found but its for freebsd10 and i'm on 13...was asking if it is still vallid...one sec

calomel.org/freebsd_network_tuning.html this one

Title: FreeBSD Network Performance Tuning @ Calomel.org

its rather long, so i havent 'just tried it'

noneoftheabove: it's probably better to ask on the mailing list because people miss irc chat. are you using pppoe by any chance?

Try it, FreeBSD cares much about compatibility, so things would either work for you brilliantly or you'd easily find out what's changed. But it might be the interface that your box ain't supporting well, the module/driver shortcomings, for example? But just like dstolfa, I recommend asking such topics on mailing lists.

dstolfa: yes..yes i am

noneoftheabove: it's because freebsd doesn't support more than 1 NIC driver queue with pppoe

i'm using the realtek driver

see redmine.pfsense.org/issues/4821

Title: Feature #4821: PPPoE WANs do not take full advantage of NIC driver queues for receiving traffic - pfSense - pfSense bugtracker

its even slower without

ah, realtek as well. this should only be a thing with intel NICs i think

i'm not sure, i'd probably start poking with dtrace honestly. alternatively someone might know on freebsd-net@

i think youre onto something though

that bug report is 8 years old

and still relevant sadly :(

:(

i dont k now dtrace and it looks complicate

...d

forums.freebsd.org/threads/pppoe-rx…affic-is-limited-to-one-queue.89843 might be relevant?

Title: PPPoE RX traffic is limited to one queue | The FreeBSD Forums

either way, i don't *think* we have pppoe support in the kernel and you gotta go through userspace or something like that. been a while since i had to look at this

the problem is assymetric too

oh wait

iperf is hitting the same limit, without the involvment of pppoe

it is maxing out one core...but the speeds are the same as the download test which doesnt

that post is highly relevant..but no solution

well, they say resolved in 13.2

FreeBSD 13.2-RELEASE-p9 stable/24.1-n254969-8659880248c SMP amd64

is mpd5 an alternative to ppp? will opnsense accept this change?

its in their packages. gonna try it

i asked in #opnsense....also ignored

oh carp (lol) it is already installed

noneoftheabove: i don't know much how opnsense handles things so i can't answer much there unfortunately

how can i tell which one is in use?

presumably ps will give it away

/usr/local/sbin/mpd5 yeah so thats a dead-end

noneoftheabove: at this point i'd just ask on freebsd-net@

there will probably be better answers there

dstolfa: is that a mailing list?

noneoftheabove: yes, lists.freebsd.org/archives/freebsd-net

Title: freebsd-net⊙Fo

is torvalds on it? i hear he can be hostile

kidding aside, i find them confusing and hard to follow

noneoftheabove: that's because they kind of are... at least for me lol

you need to configure your email client if you want to follow more than just 1

well, you don't *need* to, but navigating things is an absolute nightmare otherwise

i can't understand how to adapt my old bind9 config for freebsd

what did i specify here in named.conf(5)? zone "shapovalov.website" { type primary; file "/usr/local/namedb/primary/shapovalov.website"; }?

should it be file or directory?

didn't i have to write zone "ns1.shapovalov.website"?

V-T60: the argument to 'file' should be a file, and it would be strange, although not strictly invalid, to have a zone called 'ns1.<domain>'

options { directory "/var/cache/bind"; } should be changed with options { directory "/usr/local/etc/namedb/working"; }

what this directory is, though?

This sets the working directory of the server. Any non-absolute pathnames in the configuration file are taken as relative to this directory.

that's it, as i understand

Anybody else unable to control lcd brightness on an older Thinkpad since Freebsd 14? acpi_ibm is loaded and working (thinklight, fan status) but brightness isn't.

lw: you remember my tcp routing issue? i suppose it's worth pointing out that both the hosts are in bhyve on the same bridge. web traffic to a physical machine seems to work fine. not sure if that makes things any more clear

V-T60: did you get it?

lw, it's finally done.

What's the difference between socket() returning <0 where errno is EAFNOSUPPORT or EPROTONOSUPPORT

This is xrdp-sesman running inside a jail where it's trying open AF_INET6 but fails and will only retry AF_INET if socket's errno is EAFNOSUPPORT.

So ipv6 is supported in the jail but....ipv6 tcp ins't?

ok.. I had to give at least lo0 an ipv6 address.

Seems like it should be checking for both EAFNOSUPPORT and EPROTONOSUPPORT