is this set in the router advertisement? doesn't sound like it's going to work

that's what the VPS provider says to do, but it may be shorthand for something else

I can see a link-local IP if I ping link-local broadcast. I've never really seen it use *anything* beyond a link-local IP so far

::1 is the locahost address so you would send to lo

...exciting, why didn't I remember that

you could try to manual set the default gateway to the link-local address from which the router advertisements are send

I have once, seemingly to no avail...

I'll try it again I suppose, moment

don't forget to include the %interfacename for the route

This is what happens in tcpdump if I restart rtsold

ping6 on anything just gives "ping6: UDP connect: No route to host"

about your config, have you actually "accept rtadv" in rc.conf or "accept_rtadv" and it was a paste error?

(I did add the interface name)

accept_rtadv

just checked

I guess I can scrub out the identifying stuff and just paste my whole rc.conf, if you like

sorry I don't have time anymore, need to go to sleep

ok, that's fine

maybe it can be useful to someone else

linkedin.com/posts/freebsd-just-rem…e-of-share-7483715495494488064-2V3c fucking linux wankers

dont get me wrong I support its foss movement, but their linux way of looking at the world has just flung mud accidentally at FreeBSD

I am sure people here know that the issue with GPL is not because FreeBSD want to appease big corporations and feed the latest big corpa tech solution, its down to the fact GPL has "firendly fire" so many open source projects get tramped in the GPL movement, not just that, but the GPL is legally ambiguous and difficult to enforce (800 lines of FSF ramble)

doesn't seem like SLAAC is working at all

To be honest I dont blame its foss, or most the Linux users spreading this narrative. I blame the FSF. The FSF has campaigned for decades how the GPL protects consumers, and the only people who would hate the GPL are evil corporations or people who want to violate user freedom. The idea that GPL == freedom has been permanently engraved into many FOSS developers brains, and the idea of freedom

without GPL is just incomprehendable to many Linux users.

Sorry for the rant, it makes me so mad when people without knowing the history act like GPL eradication was a matter of appeasing companies.

mewt: show me your rc.conf?

has a couple things anonymized

but that's it

tcpdump output, also anonymized bpa.st/Z3OQ

that's from restarting netif service

I only have a link-local IP still

Also, no one *wants* to challenge/enforce the GPL because, whether it gets enforced or not will be an absolute *sea change* event

Kinda like how Google/Oracle API situation happened -- once courts ruled API implementation absolutely _does not count_ as infringement, bam, it's "settled law"

fucking c*nt

I swear if I don't close linkedin right now I will make myself unemployable for the rest of my life!

And we wonder why people dont adopt BSD when you have senior sysadmins spreading the narrative that FreeBSD is a "corporate slave OS"

that does seem extremely...sensationalized

it protects the time they invested in learning linuxisms

and also preempts having to learn anything new

mewt: whats with dhcpc6?

its one or the other...

one or the other, huh?

well technically not

sorry ignore that

im half asleep you can use dhcpv6 as well

also why the c?

is that a typo?

anyways for SLAAC you dont use DHCPv6 at all

mewt: can you verify that you have SLAAC support from the router

the router must broadcast router advertisements

The tcpdump log shows a router advertisement, I guess I need to check the content of that?

one sec

hmmm

the tcpdump seems fine

whats your ifconfig output?

[let me quickly anonymize and paste that...]

wavefunction: ugh copyright is a pain in the arse

is this helpful?

if that looks ok, maybe I'll poke the provider

mewt: wait is this a server?

yeah, it's a VPS

if so most hosting providers I know of use DHCPv6-PD

you sure they support SLAAC?

SLAAC is often used on client networks, not server networks (although I dont see any reason it cant be used for servers too)

let me double check

some others dont even use any sort of automatic IPv4/IPv6 config and instead use cloud-init to inject static allocations

but yeah I dont see anything wrong here, unless freebsd is bugged

rtsold will send solicitation, and afaik it also listens for adv

and the ifconfig for the interface looks fine

if you want to see my IPv6 setup (working) its as simple as this:

yeah, they don't have much documentation on this, honestly. Would be happy to see yours

(note lagg'ing wifi isn't supported, and causes issues, but thats how you do SLAAC)

this is client side not server

mk

mewt: contact them, I would assume its DHCPv6-PD or its static allocated

nope, another OS's guide for the VPS provider says SLAAC with no privacy addressing

err

VPS provider's guide for a nother OS

another*

mewt: not sure then, I just got back from a meetup so I am too tired to help out

hopefully someone whos got more energy can help out

I did find something interesting

when I manually dump the router advertisement, and then configure that address in place of accept_rtadv

somehow, that address gets assigned to lo0

at least according to netstat

linkedin has rate limited my posts, probably for the best, I kinda got a little heated and decided to reply to every post which discredited the achievement of FreeBSD becoming GPL-free

is it going to change peoples opinions? no. Is it going to make me seem like a arsehole? yes. Why did I do it? Idk... I love BSD and people lying about it angered me. Did I waste my time and labour? definitely!

first mistake was having a linkedin

bsdrobert: need work :(

money doesn't grow on trees

but making enemies with people is a bad way to find work too xDDDDDDD

lol

heh, I'll employ you polarian :P If I had any opened positions... and a business

wez: if only it was that easy xD

:)

no shortage of non-profit work to do ;)

but you want food on the table? good luck! :p

non-profit sure makes a lot of profit for the people up top

polarian: what do you consider your skill set?

kevans: breaking freebsd docs :p Jokes aside, im a rather useless junior in a market with very few junior positions. System administration, network, security and software in that order, the last I barely touched in the past 5 years. However it was a fun realisation when I realised the days of junior sysadmin positions are long over... did some contracting (mainly did networking though, got a

pfsense contract though which was a ton of fun!) and then back to being useless!

I should break a few more ports while I am at it too!

wez: hey still counts as experience! >:)

i was watching earth final conflict.. and after aliens come to earth.. humanity was doing video calls using.. mci

alright gn all!

dont let rs232 bite!

ok, I got it working

It needs static config with my eui64 as the IP. Somehow setting the default gateway in rc.conf didn't work, but seems to be working now without changing anything, so maybe I simply hallucinated one reboot/restart of netif

it first showed signs of life after setting the gateway directly with route. Still don't know why, but it's been good through multiple reboots now

┌────────────────────────────────────────┐

│ SYSTEM PROCESS CASUALTIES │

├────────────────────────────────────────┤

│ 💀 Killed (Zombies) : 0 │

│ 🩹 Injured (Suspended) : 0 │

│ 💔 Heart Broken (OOM/F) : 0 │

└────────────────────────────────────────┘

Press Ctrl+C to exit.

wtf, i've done better in wolfenstein

were you even trying

xway, 1. please never post that again, 2. where did you get that?

i'm impressed that something with emojis isn't f*cking up my terminal like it usually does

i like it

so, ports still has no updates. what kind of damage was done by that problematic commit? ( with big size blob )

angry_vincent: well rebuilding the entire port tree probably cost hefty amounts of cpu cycles :p

the port rebuild should be done now

angry_vincent: polarian: builds have been stopped to avoid putting out a pkg repo that will soon not be reproducible

kevans: what? elaborate? I don't understand

in progress builds would have started from commit hashes that are going away soon

polarian: you have to only rebuild the commits since the problematic commit, not all commits and it shouldn't be that many commits since then, also recreating commits isn't that expensive, my guess is that planning how to do everything else is the time consuming part, like stopping building the packages

(context sprinkle for others' reference: lists.freebsd.org/archives/freebsd-announce/2026-July/000294.html)

ohhhhhhhh

if you rebuild the hashes will change

I get it now!

I am curious about whats up with the fastly mirror though in #pkg

anyways thx kevans for the good explanation!

Reinhilde, it's all in the fonts you have supported in your terminal.

CrtxReavr, I never asked about that. I was asking where that user received that ASCII art drawing.

Neither those emojis or the box characters are part of the ASCII character set.

s/or/nor

"ASCII art" is commonly used to refer also to art that requires codepoints not part of ASCII. Perhaps "text art" might be better?

But please, spare us your pedantry. Thank you.

No. . . I think that would be called ANSI Art.

but it's not necessarily ANSI compliant either.

This page refers to ASCII as "Basic Latin.": symbl.cc/en/unicode-table

Not the fastest loading site, but it does have a lot to load.

for fuck sake

serial geli decryption still wont pop up with uefi boot either

i swear this has been the biggest waste of time in my life

I honestly have no fucking clue what else to try

fuck it im just going to bugzilla it

fucked off about this

polarian: submit after you cool off

polarian: did you try loader.env as I suggested?

kevans: heh, what was I meant to stick in there again :p

dnp1: dw im calm :p

polarian: boot_serial="YES"

maybe boot_multicons="YES" as well? I don't remember how the combinations work at the moment

isnt that just /etc/loader.conf?

sorry

s/loader/boot/

nooooo

s/etc/boot/

oh for fuck sake

holdddddd up

polarian: no, it's on the ESP.

the thing that you aren't trying to decrypt

I swear /boot is the ESP

s/swear/thought/

oh boy did my Arch Linux knowledge fuck me ;p

/boot/efi is where we mount the ESP these days

/boot is on your rootfs

yeah sorry I have been helping someone with arch linux too much recently

no, if you have recent fbsd, then your esp is likely mounted in /boot/esp

I just instictively thought /boot would be the ESP

kevans: thanks

hmmm

so in theory how do you do bios boot if /boot is encrypted?

then your stage2 will ask password

wait no because /boot contains the loader itself?

(gptzfsboot/zfsboot/... whatever else is there)

and efi is empty

tsoome: stage2 can't do serial without access to /boot.config on the encrypted partition, afaict

oh wait wrong terminal

kevans: so can you stick it onto an unencrypted partition then?

surely theres a way to do this for bios boot too

welp explains the frustration, I had a fatal misunderstanding about what was encrypted and what wasn't

see if /etc/fstab has entry for /boot/efi - if there is, mount it. if not, mount it manually by providing disk device too

kevans: wait how does bios boot work then

my laptop I am using right now for desktop uses bios booting with geli

kevans yes, thats too, but at least it can use local console.

there wont be anything in /boot/efi

so how does it boot the loader if /boot is encrypted?

polarian: bios doesn't have equivalent to the ESP

I know

thats my point

that's why we just read /boot.config off of rootfs

sure.

but if /boot is encrypted like you said it was

this is why i told you the setup just cannot work

how does geli decrypt?

bios boot can tell the partition is encrypted and know to ask for the passphrase, just like uefi

but that requires the first stage still?

the first stages aren't in the rootfs

oh...

first stage is in freebsd-boot ;)

cant you put a boot.config in there then? :p

oh wait there was a talk about this a while back I watched...

so /dev/ada0p1 is freebsd-boot, that contains the first stage correct?

yes.

right so in theory, couldn't you embed a config into that :p

what is your rootfs? ufs? zfs?

zfs

okay this makes sense now, I completely forgot about the freebsd-boot partition

and partitioning is GPT?

yes

so you use gptzfsboot for stage1

how do people learn all of this stuff, I always found low level a mystery :p

by using one skill humans have;)

learning yeah I know

reading;)

but I am convinced a lot of it comes from experience and tinkering

ok lets see

also is there docs on freebsd.env

man doesn't have anything and I checked the serial page

hmmm so if I want to geli decrypt using freebsd-boot, I assume its theoretically possible to hard code the configs in?

(an issue for another day because I have a bios server I want to repeat this setup with :p)

the problem with gptzfsboot is, the "console" is hardwired as common/cons.c:uint8_t ioctrl = IO_KEYBOARD; this can be overwritten, but to do that you either have to use config file - that wont work with encryption, or compile your own binary with default changed.

tsoome: for efi kevans said you can use loader.env but my issue is I can't find docs on this, and thats frustrating me :p

loader.env is usable in UEFI, it is stored on ESP next to your loader.efi file

it works because ESP is not encrypted.

pedantic, but specifically the /efi/freebsd one

how come theres no man page for loader.env?

feel free to add it:) someone has to write it.

I will be acquiring a mid-2010 intel imac mini with a nvidia dgpu soon. Im thinking of setting up a debian linux jail on freebsd with it to host jellyfin. would i have any issues with using the nvidia gpu for the encoding the media? It would need the r304 nvidia driver cos the dgpu is quite old. Im trying to guage how well this setup will work so i can make plans accordingly.

tsoome: wait you wrote a thesis on bootloaders, no wonder you are so knowledgable. I was trying to find if you are a freebsd dev or not and I instead found the paper ;p

ok, need to go afk for a bit. you can find reference about loader.env in stand/efi/loader/main.c

tsoome: thx

kevans: so you have been reading the srcs for this info then?

parts of it, but i also remember from when we added it

come to think about this, this would make a really cool talk!

everyone keeps asking "why dont you just use a kvm"

(friends that is)

and im like "but serial is cool :p"

im scrubbing through the src rn

wait isnt the loader meant to autodetect serial?

so VID_SER_BOTH will do both

autodetect - sure. but if your system does not provide serial console redirection, then we have no way of knowing if it is meant to be used for console.

read_loader_env("LoaderEnv", "/efi/freebsd/loader.env", false);

yes, thats the path from your esp

oh yes, sorry, I was wrong about the syntax

yes

thats what im figuring out rn

it should be the same as /boot.config

the sntax

and it was read_loader_env which I read actually worried about tsoome because I was finding where it was parsed to try to decypher the syntax

the option I want from the source is VID_SER_BOTH

case VID_SER_BOTH:

setenv("console", "efi,comconsole", 1);

^^^ thats why :)

I believe you want -hD

but wheres the actual syntax :)

reader_loader_env() passes it to parse_loader_efi_config()

wait no im wrong

efi_freebsd_getenv

but unfortunately my neovim config cant seem to find the definition of it...

kevans: where did you get that from? :p

./libefi/efienv.c:63:efi_freebsd_getenv(const char *v, void *data, size_t *len)

eventually the trail leads to sys/kern/subr_boot.c

ohhh thats why neovim cant find it, its because its launched within the loader dir

yes, we go everywhere :-)

xD

working configs also tend to help :p

gimme a sec while I figure out the src tree, not an osdev :p

nah my nvim dotfiles just dont work across files for some reason :p

hah

Now it makes more sense

help i'm being serialized

RB_MULTIPLE is -D, and its - because of if (*v == '-') and then RB_SERIAL is -h so thats how you got -hD :D

kevans: thanks for nudging me until I figured it out :p

appreciated!

~and preventing an existential crisis in the process~ ;p

i wish the answer wasn't "you have to use uefi for this setup," but alas- the ESP does simplify things

osdev really is a useful skill, which I should really learn :p

kevans: wellllll...

after this is done, im not giving up on bios :p

in theory I just need to hard code the stage-1 loader into multicons no?

yes

(whenever "in theory" is used, it always means "yes but not that simple")

question... why isn't boot_multicons default?

seems like a more sane default...

~especially when you cant configure it because of encryption~

iirc it's only semi-recent that earlier boot stages could do geli

ah... welp I guess the moral of the story is "Learn to hack src" :p

src hacking is good

yeah not when you're an idiot (like me) :p

I did want a src patch by EBC 2026

it was this years goal... but yeah... too unrealistic imo

ok, so it was yuripv who committed that github-copilot-cli

fuz offered to help but I just got caught up in other stuff

angry_vincent: and broke the port tree heh

I guess forgetting to check "git status" happens to the best of people then?

i do remember he is adding quite complex code into port's Makefiles

which looks like violation of port's phases

but it is just an observation, i do not blame

www/py-yt-dlp-ejs is an example

in do-fetch phase

ideally, do-fetch phase have to only "fetch"

and not doing any npm and extracting

there extract phase for extracting

watch this not work still :p

so, his commit bit must be suspended and he learned to not do complex code in port's Makefiles. gentoo, for example has this sorted

multicons is not default because of the zoo out there - one never knows which system will go bad on it.

lol I am reading the Github "managing large files" page

I can't stop laughing at how bad they are at this

what's the normal way to handle log data generated in a jail? like running nginx in a jail, do you put the nginx log in the jail or outside of it in the host?

this very much depends on the setup. at home I have a simple setup and log inside the jail. but as soon as the logs are a bit more important I ship them to a central logging server

good call tyvm

sure thing! good luck

kerneldove: I put the logs in the jail, then use syslog-ng on the host to send them to victoria-logs. Always, for me, logs generated in a jail are in the jail.

tsoome: do it for bios boot only then, most use EFI anyways :p

I can not claim what most use:)

well if its not configurable it ought to be default IMO :p

orrrrr

hmmm...

I see that you see;)

right now its hard coded

maybe a tool could be written to automate it

iirc settings can be overwriten by make

so a simple userspace tool could be written which takes in your disk eg: /dev/ada0 (or just ada0) and then builds the correct options you want for the freebsd-boot and then writes it to the partition for you?

and obviously when I suggest such a thing I suggest I should learn to be useful and do it myself xD

angry_vincent: yuripv did not, though

s/though/no/

I assume my idea was shit considering it was ignored :D

ok thx dvl i'm thinking the same

andddddddddddddddddddd

it still doesn't work

I added -hD to /boot/efi/efi/freebsd/loader.env

doesn't seem to have done anything

cat /boot/efi/efi/freebsd/loader.env

-hD

hmm

i bet we probe in `devinit()`, right before we parse args and read the environment

I dont know how to debug freebsd loader :/

im experimenting with packaging web services as jails. so let's say i need to get assets into a jail for the web service to serve out. but i don't wanna do 2 things: i don't wanna give every container its own public ip because that's a waste. and i don't want to use the jail host because i'm making jails as self-sufficient as possible

can i make it so you ssh/rsync into jail host's ip, but using a port that's mapped to the jail?

kerneldove: cant you just use vnet jails?

well the jails have vnet ya but so what's yor idea?

polarian

kerneldove: why not just sftp it in then?

where is this "asset" coming from?

sftp into them /how/? if they don't have their own ip

from my dev computer