no OS is good, but FreeBSD is among the lesat crap

The OS is the least important part of what you do daily on a machine. :)

is there a specific channel for pkgbase stuff?

I'm attempting to pkgbasify on 15 b5 on aarch and I'm getting segfaults when installing /some/ packages - gist.github.com/nreilly/a43b6b7eb94ece81800faf6b39e89eca

entrop: There's a mailing list: lists.freebsd.org/subscription/freebsd-pkgbase

i installed FreeBSD UNIX 14.3 on a 2020s ThinkPad and tried to boot it with Slackware 15's GRUB2 which worked on previous. This doesn't. Is that because of 'soft journalling' I need to opt-out manually in the installation (older GRUB2 from 2022)?

darwin: wiki.freebsd.org/MasonLoringBliss/BootingFreeBSDfromUEFIGRUB

Multibooting is easier with UEFI, so might as well embrace it.

i've rarely had problems multi-booting since 1997... not going to use UEFI until it's the only option

the problem appears to be some update to FFS/UFS2 GRUB2 didn't have three years ago

Hmm, I compiled pkg manially from ports, getting me to 2.4.2 (vs 2.3.1_1) and it seems to be working now

Reinhilde: you aren't wrong!

yeah

A question. If I use en EFI partitiom and boot with UEFO, do I also need a partition of type freebsd-boot? Or if the kernel and all necessary stufff inside the EFI image?

OK, I googled a bit and it seems that I don't need a freebsd-boot partition when using EFI.

You can try that in a VM to be sure, if your hypervisor can emulate (U)EFI.

But this guide, c0ffee.net/blog/freebsd-full-disk-encryption-uefi, seems to suggest having a separate freebsd-boot-partition containg kernel and kernel modules and copy /boot/boot1.efi to the efi partition, so I'm a little bit confused.

andreas303: freebsd-boot is an extra partition for when you're on GPT to accomodate the fact that GPT doesn't have reserved space after the MBR. It's not a terrible practise to include them AND an ESP so that you can boot either way.

andreas303: It's the FreeBSD equivalent of en.wikipedia.org/wiki/BIOS_boot_partition

andreas303: I talk about partitioning some here: wiki.freebsd.org/MasonLoringBliss/ZFSandGELIbyHAND

afk a bit, but I'll check back before bed in case there are questions

mason: Thx! I will check out the wiki.

good morning folks! does brave browser and widevine work together in freebsd or must I run chromium to get widevine to work?

nwe: works out of the box with just the www/linux-brave port

kenrap: today I have installed chrome with pkg install chromium and foreign-cdm

kenrap: will install it from linux-brave then :) thanks!

Sure thing. Amazon prime video works fine on it.

Just make sure it's enabled in the settings.

so this might be a little nutty, but theres something I want to solve. I have sshd set up to essentially only serve sftp, this is so I can offer basic file access to a couple of users. I have it set up to chroot the user into their own directory. However, I also want to give each user access to the same read-only filesystem.

would it be crazy to use nullfs mounts in each user directory?

crazier than that, would it be stupid to write a script that automatically mounts nullfs for each user directory on startup.

specialbomb: sounds like you'd want to install filesystems/sshfs and use that remotely mount each respective user directory

nullfs mounts sounds ideal for this

nullfs?

entrop: pkg 2.4.1 or newer should address your segfault.

man.freebsd.org/nullfs

use it like `mount_nullfs /usr/ports /some/where/else/ports` or put it in fstab so its always available

ugh, i'm running Home Assistant on FreeBSD and i was able to fix it only today. they started requiring PyNaCl 1.6.0 and it wasn't building on FreeBSD. today they released 1.6.1 and with MAKE=gmake i was able to build it...

i hate that project. they are worse than linux monoculture. they target a specific OS (their own OS) and they have versionitis too. every package is always bleeding edge, latest version only

hjf: hah, I'm literally doing this right now

I did it in a venv with pip and didn't run into this PyNaCl issue somehow

but I know the 'official' way is to just use their Docker container at this point, not the Python package

I seem to have things running okay in the venv, but it complains about a few things: Python 3.11 is deprecated, webrtc-noise-gain doesn't like FreeBSD, they're overly permissive with josepy versions and it has a breaking change in 2.0.0

specialbomb, why not just set those users' shell to scponly and add them to the same group for the shared filesystem?

People make shit too hard.

hodapp: no, the container isn't supported either AFAIK. ONLY the "HA OS" version on a raspberry pi for now.

you can check my blog, i've documented some stuff on how to make it work

hjf: the container isn't supported? wtf.

I'm gonna keep trying to use the Python package but we'll see how that goes

hodapp: blog.hjf.com.ar/en/2025/01/10/home-assistant-issues-on-freebsd

thanks

also this one for py 3.12: blog.hjf.com.ar/en/2024/10/19/fixin…g-home-assistant-2024-10-on-freebsd

also this one for today's fix: blog.hjf.com.ar/en/2025/11/11/fixin…ome-assistant-and-pynacl-on-freebsd

i can't believe how fast Wordpress runs now

I wonder why you ran into PyNaCl issues and I didn't

hodapp: are you in the latest version?

i'm running 2025.11.1 now

and you need py 3.12 for that one

oh, no, pip install homeassistant got me 2024.3.3 it turns out

yes, it's the newest you can use in py 3.11

that must be why

hodapp: create yourself a new venv with py312 and start using uv, follow my guide here blog.hjf.com.ar/en/2024/10/19/fixin…g-home-assistant-2024-10-on-freebsd

you can keep your old HA venv

uv was gonna be my next attempt. it's not a big real to recreate the venv though

real? deal. whatever.

yeah but you need to do things in a specific way for it to make HA happy

anyways after tuning mysql and installing memcached, my blog now renders in 80ms

i run woocommerce for my shop too. the admin page was doing like 250 queries for every page load.

after installing memcached, it's down to 40-60 queries

CrtxReavr: this is what im currently doing. the problem I want to solve really is to transparently give access to a filesystem for everyone, while simultaniously having separate chroots.

in the curren set up, lets say this share dir is called share, and each user A B and C has a directory of the same name. A, B, C and share all have to be in the same root dir if I dont use nullfs. I think this is a problem because now each user that logs in can see dir entries for other users, and that is not ideal. I can set up permissions obviously amd I have, but I wouldnt want an individual user aware of anyone else being on the system at all

if I nullfs mount media into A/media, B/media C/media, then each user gets their own chroot and has no idea anyone else is present on the system

hjf: okay that's a pro for you man

hello, I know there is an sql lite database por ports... I can't see any log file tough. I wonder if I can query that database to know when pkg was used to upgrade packages, the previous version and the version installed. Is that possible?

I think the whole point of sqlite is that it's a user application, not a client/server setup with a daemon writing logs.

If you want the daemon with logs, then you probably want MySQL or PostgreSQL.

Otherwise, then I guess you need to do the logging in your application.

I thought he was asking if FreeBSD uses an sqlite database for tracking package/port update dates/times that he can query.

'pkg info <package>' will show the install date of the current version at least.

Maybe.

With my logging setup, I can just search 'pkg-static' in /var/log/messages to see dates.

I just noticed that he ping-timed-out anyway

Capio|Yaazkal: Are you sure you checked /var/log/messages? Did you grep for 'pkg', for example? In my default syslogd setup, (un)install events are logged there.

I think 'pkg info <package>' will at least show you the install date of the current version.

psychonate: oh, didn't check there, was specting something like /var/log/pkg.log for whatever reason

If those events are getting to your syslog daemon, I imagine you could check your syslog daemon's config to write those events into a different log file if you want.

s/check/update/

psychonate: /var/log/messages has the date when the package version changed. It is what I'm looking for, thanks !

Is 15 release released once builds begin or after the announcement?

zfs(8): OpenZFS has been updated to zfs-2.2-release(2.2.7)(2ec8b6948070).

I thought it was updated to 2.4.x

what i do wrong? pkg install firstboot-freebsd-update, add firstboot_freebsd_update_enable="YES" to /etc/rc.conf, finish install and first boot... nothing happens

kerneldove_: firstboot means firstboot ever

ya that's what it was

there's a sentinel involved that, iirc, the rc framework wipes after it executes anything marked for firstboot

it was the first boot

(so if you `pkg install` this on the running system and reboot, it's by definition not the first boot)

it's not on a running system it's during the manual steps part of a new install

installer-installed things also don't go through a firstboot

generally VM or embedded images

huh

Is there an official freebsd.org guide on how to setup Wireguard (with examples) for FreeBSD 14.2? I understand that 14 has it in base, no ports needed. However I don't see anything in the handbook.

tuaris: it's all the same as other platforms

but perhaps wg-quick is the package that may help you

anyone else seeing crashes on freebsd 14.3p>3 ?

back when I was running it, no

kerneldove_ how?

im getting regular OS crashes on 14.3. fatal error mutex xxx own xxx is on list xxx xxx at line 151 in file freebsd-src/lib/libthr/thraed/thr_mutex.c errno 35

it's p5. haven't seen the problem on p3 or earlier 14.3 versions

trying to search bugzilla for any other reports but its' hard to use

what platform are you running on? im guessing its some kind of multithreading issue

ryzen 5700

ryzen 7 5700G specifically

and ya prolly multithreading related

thats definitely weird, Im gonna go on a limb and say it could be worth booting into a different OS and updating your microcode, but I dont know enough about how freebsd's kernel works :(

i wonder if it's related to lists.freebsd.org/pipermail/freebsd-current/2016-March/060264.html

that's old but similar part of the code

shrug

is this just freebsd without any extras installed?

or are you attempting to run specific software

freebsd + a rust server app i made

running on a bunch of boxes with basically identical versions and configs, no others having prob

exact same setup huh. weird

bios + microcode update would be a good start, should be able to use the cpu-microcode-amd to test new microcode without having to flash the CPU from another OS?

are they also running your app?

ya i use scripted bsdinstall so i can say it's identical

hm

ya

cpu-microcode-amd package*

will that tell me if an update is needed?

i would boot into a diff OS personally, if its gonna be unstable

I beleive it just loads the microcode into the CPU on each boot?

I dunno, I just wouldnt trust it :D

ok if it happens again i'll try updating the bios/bmc then try again

sucks because it was working reliably until i changed from p3 to p5

any way to see if there were changes to that file post p3?

freebsd-src/lib/libthr/thraed/thr_mutex.c

github.com/freebsd/freebsd-src/blob…aster/lib/libthr/thread/thr_mutex.c looks like 5 years ago so can't be that

hmm

difference between -pX and -pY should be very minimal, literally only he source files mentioned in the errata notice

maybe SO_REUSEPORT_LB breaks connect(2) for UDP sockets. [SA-25:09.netinet]? my server app does lots of udp

im very curious about this, I hope you sort it :(

ty me too hehe

cpu-microcode-* are necessary evil imo, not all hardware (motherboard) vendors provide BIOS updates for long so the microcode updates cannot be loaded from an updated BIOS on boot, the OS must take over that duty

is it recommended to run cpu-microcode-amd for any server using amd procs?

the cpu vendors recommend it

ok i'll set all of my boxes up to use it ty

hm just got the crash again!

this time it was rust tokio libp2p error "expected some established connection to peer before closing" think freebsd/freebsd-src 058bcb5 looks related?

i rebooted after setting up cpu-microcode-amd in /boot/loader.conf.d/. any way i can check and see if it was applied?

dmesg | grep micro

i put its config in /boot/loader.conf.d/cpu_microcode_amd.conf

nothing

service microcode_update start

pkg install cpu-microcode

first

i already ran pkg install cpu-microcode-amd, is that not enough?

i get this on my intel systems (with an up-to-date bios!)

dmesg | grep micro

CPU microcode: updated from 0xf0 to 0xfa

you either use /boot/loaders.conf* OR rc.conf, the first being preferred as its loaded before the kernel

ya that's what i did

the first method

i installed cpu-microcode-amd, then to /boot/loader.conf.d/cpu_microcode_amd.conf i added cpu_microcode_load="YES" \n cpu_microcode_name="/boot/firmware/amd-ucode.bin"

need i do more?

that looks correct to me

i wonder why not working then?

there is nothing you can do more