I can't seem to get firefox working with sndio... I set sndio in about:config as media.cubeb.backend=sndio and restarted firefox, and about:support says the backend is sndio

but yet there is no audio output when I try to play video

chromium which uses sndio by default, works just fine

any ideas?

hm, I use pulseaudio atm

I dont want to handle both sndio and pulseaudio

it seems most things use sndio (plus its more minimalist)

apulse perhaps...i use it on linux..

im still old alsa/asound..use apulse for pulseaudio stupid

course probably linux specific

PulseAudio emulation for ALSA. yeah...figures

maybe sndiod is needed for firefox to work...

hmm

what I am reading is that firefox is compiled with pulseaudio support and the only way to reliable change it is to recompile the port yourself using the sndio flag

Title: Sound - FreeBSD Wiki

oss in the end i think..

ah I will try oss

Open Sound System used long time ago on solaris also

the problem is I dont want multiple sound systems to worry about

if I set the volume of one

I dont want a different volume for another

plus if I mute one, I dont want another to still be unmuted

imagine being in a lecture using firefox and forgetting to mute pulseaudio because you normally use sndio for everything else...

oss is the "real" interface (/dev/dsp* devices), check /dev/sndstat text file for info on devices, but sndio/pulseaudio use OSS

set it and forget it! --Ron Popeil

polarian: is pulseaudio process running when you start to play audio on firefox?

saper: yes

but I changed the backend from rust-pulse to sndio

oh interesting

setting pulseaudio volume doesn't affect sndio

butttt... setting sndio volume using mixer(5) does affect pulseaudio

I didn't get sndio to run with chrome and firefox so pulseaudio it is now

chromium depends on sndio too

so I have to pull both audio frameworks in order to use firefox and chromium

why i do everything in vmware-vmx, and host os is just for running vm's

freebsd bhyve got nice benchmark results

you virtualise your desktop?

that must be painful

the latency :/

i game in win10 vm's

this is vm

those pics of amiga where vmware-vmx one of them anyway

how do you game in vms

the latency is so noticable

Title: Proxmox vs FreeBSD: Which Virtualization Host Performs Better? - IT Notes

no its not

adults actually play video games

thought that was just under 18s

polarian: I got chromium and firefox both to work with pulseaudio, without recompiling

imgur.com/L0NdwRO old gaming vm win10 consoles

Rue: those are some olddd games

and thats my old amd 4core/4thread laptop, this 8core/16thread smokes it

saper: what if I want to use sndio not pulseaudio :)

portal 1 and 2 ?

risk of rain1

Guns, Gore & Cannoli is not old

imgur.com/1psrhvL 67days 4vm's 2022

and once again old craptop 4core/4thread amd

new laptop 8core/16threads

polarian: feel free to troubleshoot yourself

freebsd bhyve probably could build nice setup..

saper: the whole reason I was asking about it is that when I set the backend to sndio firefox produces no sound

for pulseaudio, I could give you some working commands to check

it works fine under pulseaudio backend

host os and guest vm's..seperation of things...

jails light weight or go full blown heavy os vm's

baremetal os, juggle jails/vm's. compartmentalize.

compartmentalize - To separate into distinct parts, categories, or compartments.

i have decade old vm's ..my freebsd 5.4 vm is ancient..

ran on baremetal 333mhz/384MB celeron m @ 1 time

dump/restore into vmware..from baremetal to vmware..back in the day.

I got lost like...

ages ago lol

a future project for you

rennj: SmartOS is a bit like that, designed to run from a read-only USB stick, creating VMs and zones (jails) on ZFS

yes, joyent smartos they got bought

solaris foo

oxide computer

oxide.computer same people same foo

Title: Oxide Computer Company

en.wikipedia.org/wiki/SmartOS pretty clear

Title: SmartOS - Wikipedia

pkgsrc!

blog.brianewell.com/smartos-in-2021 pretty good read

Title: SmartOS in 2021

but they where bought i know that..wikipedia doesnt have info

On April 11, 2022, Joyent announced that MNX Solutions would be taking over the Triton DataCenter technology suite.[39]

Title: Joyent - Wikipedia

Title: A New Chapter Begins for Triton and SmartOS | Triton DataCenter

meh, freebsd has same tech

jails, bhyve, zfs

vs linux/kvm/qemu or proxmox distro

promox is just lxc and kvm/qemu...

any idea why pool/ROOT/default boot environment mounts itself over / although pool/ROOT/13.3 is selected (and mounted as / first)?

(13.3 is marked as R with bectl list)

I found smartos very interesting, it is solaris but quite unlike one. all readonly, very small root, not much space to install anything, just launch vms to do _anything_

(and vm/jail management tools are written in nodejs which is part of the base)

yeah i build usb keychain, load os into ram, remove keychain...i have just enough os to bring up gui/x11/vmware-vmx

current os is 597MB.iso..on keychain

rennj: exactly that's the way

my last os was 350MB.iso

yeah but i was doing it before them

heh

they got fancy foo...virtual switch, san foo

like vmware does

wasn't first vmware esx on a diskette :)

vswitch, vsan

no vmware was vmware on linux,windows,and freebsd

in fact vmware 1,2,3 in ports

and that was before intel/amd added ring -1 vm foo

vt-x and amd svm

let alone iommu/vt-d

like the cpu's didnt have the tech

vmware use to eat the whole cpu..back in the old days

I think even xen needs ept these days

Bo Burnham - Welcome To The Internet

Could I interest you in everything all of the time? A little bit of everything all of the time?

saper, en.wikipedia.org/wiki/Second_Level_Address_Translation

Title: Second Level Address Translation - Wikipedia

Extended Page Tables (EPT) is an Intel second-generation x86 virtualization technology for the memory management unit (MMU).

Unix, Linux, OS/2, Windows NT are considered modern OS's as they switch the CPU into protected mode at startup, never return to real mode and provide all of the benefits of protected mode all of the time. 64-bit operating systems use real mode only at startup stage, and the OS kernel will switch the CPU into long mode.

48-bit physical addresses and so can address up to 256 TB of RAM.

Title: Long mode - Wikipedia

i know intel did the 5 level tlb lookup

Title: Translation lookaside buffer - Wikipedia

Title: Intel 5-level paging - Wikipedia

from 48 bits to 57 bits by adding an additional level to x86-64's multilevel page tables, increasing the addressable virtual memory from 256 TB to 128 PB. The extension was first implemented in the Ice Lake processors

i thought the toor account was removed in 14.x? i see it in pw user show -a

14.1

polyex: no toor is still there in main

really sure i read it was going

Title: ⚙ D40565 Remove toor backup account

you can make your own reality by forking freebsd and removing toor

I think at this point if ed wants to rip it out he's going to need to get so@ to declare it a security vulnerability in itself

do not be bound by what you imagine to exist

which there is some precedent for

so /etc/jail.conf glob includes whatever's in /etc/jail.conf.d/*.conf, and those conf files can include big wads of common config between jails! even config using vars and stuff works omg!

before 14.x only /etc/jail.conf would include. the jail.conf.d/*.conf files wouldn't include

the depend directive works in jail configs now too wooooo

hello

anyone had any issues with unifi controller since last update? i cant connect to it and it doesnt log to /var/log/messages

running in a jail

it COULD be something with 14.1 which i just installed but I think its the port upgrade

sopparus, it's likely the upgrade to MongoDB60

ok, do you know where i can see any logs?

ive checked all of /var/log

279029 – net-mgmt/unifi8: switch MongoDB dependency from 5.0 to 6.0

yeah i noticed that too

whops

meant to paste this

Title: 279029 – net-mgmt/unifi8: switch MongoDB dependency from 5.0 to 6.0

I don't think it's running properly, because it can't access the db

I rolled back to the earlier version, changed the feature version compat of the db, per that bug report, and then upgraded again

that fixed it for me

ok, so i cant do it on 6.0 directly?

hm i dont have the "mongo" command

unfortunately, I don't know how, but I'm no MongoDB expert

is it part of mongodb-tools perhaps

no..

also, mongo does not log to /var/log, it logs to /var/db/mongodb/mongod.log

thanks

may I ask what client did you use?

sorry?

for mongodb

I rolled back to mongodb50

but how did you change the feature version compat then?

Okay, so I upgraded unifi, which bumped it from 5.0 to 6.0

realized that it couldn't read the db

removed 6.0 and reinstalled 5.0

The db ran fine under 5.0

so I could bump the feature compat of the db when running 5.0

after that, I could upgrade both unifi and mongodb

got it thanks, but to bump the feature compat one needs to connect to mongo db with a client right?

from that bugs page, mongo --port 27117

yeah, that command ran fine under mongodb50

ah, I see. thanks

👍

yeah I got the command mongo available now, but it looks angry dumpinen.com/fXigLWiWCCL :)

So I started up unifi under the old version

yep, ive done that and it works

but cant access mongodb with "mongo"

as root?

yep (in jail)

the client seems to segfault

it states there's a memory lock issue

did you look at mongodb.com/docs/manual/reference/ulimit/#unix-ulimit-settings

Title: UNIX ulimit Settings - MongoDB Manual v7.0

hm found somethng

yes now it works, allow.mlock; in jail config solved it

yeah all is good now, thanks alot rafe

happy to help

Hello guys, i need some help, is there anyone here who is really good with pf and wireguard?

think i maybe found a bug in 14.1 jails

i have 4 jails. 1 depends on 2. 2 depends on 3. and 4 depends on 3. they all start and run fine if i remove the depend directive, but having it in, jail 2 fails to start because jail 3 is already running

can't depend handle multiple dependents of the same jail?

only way i can let a user in a jail run tcpdump is in /etc/rc.local i chgrp /dev/bpf* and chmod g+r g+w it. problem is daemon running tcpdump is started before rc.local is run. how can i do it better?

didn't i give you the devfs thing for this or was that someone else

polyex: le-fay.org/tmp/30d/rEejVU.txt

lw ya i tried it but remember the /etc/devfs.conf part doesn't persist changes?

in a jail

only solution to even get it working at all was putting it in /etc/rc.local

idk jails are dark magic

sopparus: the unifi8 mongo upgrade was seamless for me, sorry

polyex: I guess you can try running tcpdump from the host, the interface should be visible there, no?

dch no i need the jail to be isolated from host and not blend the 2

polyex: aah its a vnet one?

ya

and /etc/devfs.conf should be persistent, I use it here all the time.

in the jail?

hmm thats out of my paygrade

so devfs is one of those synthetic filesystem things

i tried it and it didn't work

I would set the permissions on the host /etc/devfs.conf instead

and see what is in the jail on reboot

eg `own bgf* root:network` is a good example

but then it needs to know the user/group info of the jail to set it to

polyex: you can experiment with groups that are always in both, like root, wheel, operator, etc.

but it will probably work with numeric uid/gid anyway

i want more independence of jail from host

at some point a vm is more independent

ya this and 1 other limitation are all ive found in jails/vnet

mac_portacl support in vnet would be great, that's the other limitation

polyex: vnet support firewalling and you can limit user network access this way

i have pf running in the jail. that's why i want the jail user to be able to run tcpdump

oh you were talking mac_portacl

nah that sucks mac_portacl support in vnet would be better imo

hhe.. you have to prepare patch then, to make MAC framework VNET aware

there any way to make a daemon not start until after rc.local has been run?

polyex: sure, take a look at rcorder(8)

what in there talks about it? is there a way to make /etc/rc.local be a REQUIRE condition or smth?

oh it's "local"!

i have 4 jails. A depends on B. B depends on C. and D depends on C. they all start and run fine if i remove the depend directive, but having it in, jail B fails to start because "jail: C already exists". can't depend handle multiple dependants of the same jail?

what's the difference between "sysctl hw.machine_arch" and "sysctl hw.machine"

/usr/src/Makefile says # TARGET="machine" to crossbuild world for a different machine type # TARGET_ARCH= may be required when a TARGET supports multiple endians

KNOWN_ARCHES in /usr/src/Makefile.inc1 lists some values

hmm interesting

saper: although it doesn't fully explain machine=arm64 vs machine_arch=aarch64 i think

aaarch64 could have been named armv8, but AArch64 sees to be the name of the execution mode, so be it

in the Sun days, you'd have sun4c. sun4u, sun4v "machines" and then sparcv7 sparcv8 sparcv9 "cpu architectures"

saper: ah okay. i wonder if this might cause problems when armv9 is out

I wonder who NetBSD does it, I think they simplified to just one name per "port". but this means they have multple "*arm" ones in addition to aarch64

makr: it's out there already and it is AArch64, too :)

--s,who,how,

saper: ah okay :D

makr:if you look at the kernel sources, FreeBSD also introduced "x86" directory name for the code that is common between i386 and amd64. In addition to that, you'll find /sys/amd64/amd64 as well as /sys/amd64/ia32

saper: o_O well that's tricky

it all makes sense, but probably you don't need to worry about it

saper: for sure, interesting stuff though

thanks

I'm trying to understand whether I need the `-A` (generate ACPI tables inside the guest) option for my linux bhyve guest VMs. churchers-vm doesn't seem to use it, but it seems like a good idea?

oh, churchers/vm-bhyve DOES use it it seems!

to spell out the unifi issue, you need to check that the database compatibility has been set to 5.0 before upgrading to 8.2.93 from 8.1.113. with 8.1.113 running with mongodb50, you run `mongo --port 27117` to get the mongo shell. then `db.adminCommand( { getParameter: 1, featureCompatibilityVersion: 1 } )` tells you the current ver compat, then `db.adminCommand( { setFeatureCompatibilityVersion: "5.0" } )` updates you

then you can upgrade to 8.2.93

i had to change my repo from latest to quarterly to revert back to 8.1.113, then go and do the mongo compat work, and then went back to the latest repo and everything was then fine

hi folks

I'm just starting with FreeBSD and I'm happily supprised

the operational flow is like a breath of fresh air

are the real freebsd developers here?

do you think I could use a loaded unbalanced antenna with an extremely high Q in place of a cavity filter?

the cliff of the standing wave ratio should prevent other energies from flowing

sfox: this sounds spiritual

it's physics

close enough

Howdy, folks!

nicholaus04: sup

i have 4 jails. A depends on B. B depends on C. and D depends on C. they all start and run fine if i remove the depend directive, but having it in, jail B fails to start because "jail: C already exists". can't depend handle multiple dependants of the same jail?

kenrap: Nothing much.

nicholaus04: cool, enjoy your stay

part

anyone know?

sounds like it will handle only the first case. you could always make D depend on A

nah that's garbage i just figured it out. i only had to remove jail_list from my rc.conf and NOW IT WORKS LETS GOOOO

often the best way to get an answer on the internet is to say the wrong thing and let people correct you

man jail config has been so upgraded in 14.x tyvm freebsd team!

now we just need jails to support /etc/devfs.conf and mac_portacl in vnet and they're perfect

there any way to start a jail and make it ignore its depend directive? i wanna be able to have dep info for rc startup, but still be able to somehow start a jail independently, check stuff, then shut it back down

?

i just wanna start a jail and have its depend directive ignored

there a flag or smth for it?

onestart?

still enforces depend

1

?