polarian: it is possible to enter GELI keys through the serial console, we use it on all our servers like this, nowadays mainly with EFI

you need to redirect video output to the console in your BIOS/EFI firmware and have the loader enabled to work with the serial console

If your using classic firmware, use comconsole_speed, boot_multicons, console and boot_serial

If your are using EFI, use boot_multicons and comconsole_speed

The only other thing that comes to mind is that sometimes in some firmware implementations, you might see some (awkward) automatism in modern EFI firmwares:

they do redirect serial console from COM1 to IPMI SOL

(and vice versa)

then you are sitting, waiting for output on cu and should instead just have fired up ipmitool with sol activate

;)

i'm no longer the only crazy one :) bugs.freebsd.org/bugzilla/show_bug.cgi?id=295057

iRobbery: You're an insane person! :)

ek: i dont deny that, but at least i'm not alone anymore B)

2 weeks busy with that, thinking i was messing something up

though in 15.0-RELEASE the result is even weirder

I have Intel X553 SFP+, but can't get anything out of it. is the driver supposed to say something when an SFP module is inserted?

all I see is the driver loading and then nothing, "no carrier" even when module and fibre cable are connected

hey, at least you see your NICs attached to a driver ;)

I was on 13.5 with an aarch64 system, and after upgrading to 15.0 lately, the PCIe NICs on one part of the bus do not seem to get discovered anymore :/

Now I need to somehow get back to 13, but the boxes are remote, I do have IPMI and serial access, but no networking off the sytem

No idea if 14 would have helped here

no ZFS snapshot to recover?

or "boot environment"

Anyone running aarch64 and ThunderX?

those boxes are similar to the ones aarch64 packages are built on, at least this is something I heard from around the NYC guys

I trust you've filed a bug already?

"Cavium's ThunderX platform is the primary reference target for the FreeBSD/arm64 port" <- says the wiki. you'd expect someone to attend to it pretty quick?

Koston: just did freebsd-upgrade -r -- tbh, don't know why I didn't care about a snapshot this time

freebsd-upgrade creates a boot environment automatically when doing major version upgrades, unless you specifically tell it not to

I did not file a bug yet, I just tried to get in contact with somebody, in here or over in #freebsd-embedded

file a bug, good sir. most productive people tend not to waste time in instant messaging

@"instant messaging": when I started with FreeBSD in around 1998, it was common to find somebody on IRC in the appropriate channels ;)

and even back then, mailing lists were much more useful

especially for the project

I didn't want to blow it up that much and looked for the error on my side, probably with a hint from someone, something that I have overlooked. But I guess I will go for filing a bug -- it is not that I do not have an account there

spekul: I cant do that

its a consumer motherboard

not server motherboard

by "classic" you mean bios booting?

if so maybe I should disable EFI and use pcbios

classic = BIOS

yes

honestly, they all do a BIOS post first, and instead of just jumping onto the first sector of the first disks, you nowaday hook up another micro environment (UEFI) with a bit more capacity (flash, adressable space) and jump into the next OS from ther

I didn't even realise arm does uefi now. jesus wept :/

well, theser are not just credit card sized machines, but instead full-sized servers

I think the guys over at netflix used them to spill out their HTTP?

had my first real run-in with uefi a while back. reading the official specs was very much a "if I pour bleach in my eyes and also drink some I might be able to forget ever seeing this"

I first tried to circumvent all this UEFI stuff, quite successfully, tbh

but then more and more .efi binaries came around, for e.g. firmware updates

so i made an oopsy, i thought subscription names should be unique on the subscriber, and i added a subscriber with the same subscription name. How do i stop that? On the publisher i see 'ERROR: replication slot "sub_pdd" is active for PID 91171'

where you did not have to have those boot sticks lurking around with oldest MSDOS on them

arg wrong

this is where my mind about UEFI changed a bit, because it makes remote maintenance of machines way easier

ridiculously over-engineered behemoth designed by genuine Windows People congratulating themselves for the innovation of a shell, shoved down everyone's throat with no small effort and at no point anyone bothers mentioning what's the benefit

all the other, windowsish features (secure bootloader) are non-sense.. when you loose physical control of your machine, you have lost. The only thing that might help are encrypted disks and a proper detection when to shut down

then again, UEFI brings something into the x86 world that all the others before already had.. e.g. firmware on sparc64, sgi, or apple kit

yep the uefi shell specification defines what for all intents and purposes is openboot from 90s

except it fails to recognise such a thing ever existed before, so it's rather amusing. I think the dildo who wrote it genuinely thinks he invented something.

full ack @dildo

(made my day)

btw. if you want to read something much more idiocracy heavy, check out ietf ipv8 draft

have you been going through IPv5 with it's "gateway systems" already?

Koston: I'm pretty sure there are lots of ipv8 drafts hosted(!) by ietf, which one do you mean? the one by James Thain?

nimaje: yep. didn't realise there were multiple

spekul: that looks hilarious, I take it's an april fools?

Koston: well, anyone can submit an Internet-Draft, that doesn't meant the IETF has throught about some draft one bit

fyi malwarebytes.com/blog/news/2026/05/…omes-silent-4gb-ai-download-problem

A reminder that Windows apparently also killed S3 in UEFI, which is why modern Windows laptops drain a shit-ton of battery in sleep. Learned this recently.

TL;DR CPU is running in modern "sleep". All so that Microsoft can turn the PC on by themselves in order to do Windows Update.

And the other major stupidity (out of many, I guess), is keeping time in local time instead of UTC. Though at least there's a DST flag nowadays, so yay or something?

DarkUranium: windows completely and utterly molested its power management w/ UEFI yes

that was a long time ago, too

just upgraded to FreeBSD 15, my intel I9/32 core now runs much faster with bhyve, system stall for some unknown reason just gone .... thanks !

does sound work hdmi? im trying to make a 'smart tv' and wondering whether it would be better wayland/xorg. ive had better experiences with multi-monitor with wayland but its been a while since it

it has worked for mme in the past, I have no idea how to make it work now

i wanted to just ssh into it and choose media then it pop up telly

hello, I need a hint, how can I pass-thru usb device ( scanner ) into jail

Does anyone know of a unified solution that can handle FreeBSD jails *and* Docker or Podman in Linux?

(I want to host most things in FreeBSD, but a handful of things only support Linux ...)

DarkUranium, this is my first time using jails

as I know docker using a virtualbox virtualisation ( quite possible that i'm wrong )

look into bhyve virtualisation

you can literally run any linux/bsd or whatever as VMs ...

Docker definitely doesn't use VirtualBox, they're not even in the same class of virtualizations.

Docker and Jails are both container virtualization (shared kernel, etc). The closest thing Linux has to bhyve would probably be qemu.

wiki.freebsd.org/Docker << take a look

"Currently, Docker needs VirtualBox installed and configured. "

That's ... not at all what I'm asking about? For starters, the whole point of what I'm saying about is running Docker *on Linux*.

ah sorry ... misunderstood

never used docker so, sorry I cant help you in this question

I heard some people had good luck with Hashicorp Nomad, but it kinda feels like overkill.

DarkUranium: look into podman and ocijail - may be possible to run some linux images directly via linuxulator as well

kenichi: Yeah, I've been using Podman to attempt exactly that, but had a bad experience with it, TBH. Turns out it *REALLY* wants systemd to run properly.

Otherwise, things can get stuck in a forever-wait :|

it's not the cleanest, but with some hacking, i've made a few things work

Same. But that doesn't mean I'm happy with it.

yeah, the systemd direction is unfortunate

i've been running a few linux images as illumos zones for about a year now, very manual/hack tho

Yeah, I've looked at SmartOS. It seemed nice, but the Linux stuff is mostly a compat layer. Lots of things need the real deal / kernel, sadly :(

true, i think i've been somewhat lucky. LX zones are absolutely a compat layer :D

Yeah, same for running Linux in FreeBSD jails.

What I have is an impromptu Podman Machine, basically a big Alpine VM in bhyve, and *that* runs all the Linux stuff.

nice. i think this ^^ is the oxide/helios approach as well: bhyved linux for containers, zones for infra

bhyve is great

hrmm any idea why hdmi over audio isnt working? im using mpv from a tty, which does seem to work to my suprise

im not sure, does normal audio work?

yes

without mpv

oh

one sc

sec

and ive tried changing hw.snd.default_unit/auto to my hdmi interface but it still plays over laptop speakers

if there's a commit to stable/15, how long before it shows up in releng/15? would it be a -p## release or not until something like 15.1?

maybe press 0 to turn up volume

oh nvm thats a hw thing

im not sure

kenichi: I'm not familiar with Oxide or Helios.

DarkUranium: it's bcantrill's (dtrace, etc) latest company

dont think you can do audio over hdmi on this card

:(

wcarson: depends, note that it is releng/15.0 and at some point there will be releng/15.1 too (if there isn't already), new releng/ branches get branched off of stable/ branches and usually no commits flow from stable/ to releng/ after that, but sometimes changes get classified as errata or security fix and land in releng/ at some point, new features definitly won't make it into an existing

releng/ branch

Oxide is a fun company. All Rust All the Time

As long as they follow the official Rust specification

Not sure I follow what you mean

There isn't one

nimaje: ah ok. both the changes i'm looking at are bug fixes, i had hoped for an ERRATA but i don't think it's going to happen (reviews.freebsd.org/D56764, reviews.freebsd.org/D54815)

so i will patiently wait for releng/15.1 :)

is it possible to search an unconfigured repo? like say pkg search -r pkg.FreeBSD.org/FreeBSD:14:amd64/latest mysql80-server

^^ returns no output (for any port)

so audio over hdmi used to work

but was removed because it was proprietory or something?

morpho: are you sure you don't need to just turn the volume up?

pcm1: <Intel Kaby Lake (HDMI/DP 8ch)> (play)

rtprio: pretty sure

so it's not in `/cat /dev/sndstat` ?

it is

this makes me wish i tried harder with my freebsd migration heh

maybe i'll try again soon

try obsd

Right, I remember why I'm not using 15 anywhere. The installer doesn't give me the option of setting up the network, and hence can't get distribution sets.

Then on retry it sees no disks.

Backing out if I geli detach, the disks become available again.

Yeah, so, I go right from partitioning to "mirror selection" but there's no network setup.

Hm, yeah, dunno. Even setting up the network by hand in advance, with a valid address and working resolv.conf, yields i.imgur.com/qnTLHIi.png - I wonder if I need to race it and get resolv.conf inside the filesystem being set up maybe. Seems like the answer should be "no".

Cute. It's overwriting my /etc/resolv.conf with a symlink to a non-existent /tmp/bsdinstall_etc/resolv.conf

I will open a bug I guess.

Hm, seems like it's a problem when running bsdinstall from the shell, as there's some effort to set up the network when going straight into the installer when booting install media.