Is doing a major upgrade from 13 to 14 safe using pkgbase?

_xor: Can you be more specific? I'm not sure I completely understand the question.

"pkgbase" is currently considered experimental

If I have a system that's currently running 13.2 and is configured to use a pkgbase repository, which is already built and available, that contains 14.x pkgbase packages, are there any known caveats for upgrading between those two major versions?

Yes, been using it for a bit now. Upgraded 13.x minor versions without too many issues. But haven't done a major upgrade yet.

As far as I can tell, I should be able to upgrade the packages normally, write the new bootloader + copy new efi image, reboot, and then find/clean any remaining *.pkgbase files.

_xor: If you have a way to snapshot what you currently have, give it a try.

_xor, There is comment 3 here: bugs.freebsd.org/bugzilla/show_bug.cgi?id=273601#c3

Title: 273601 – freebsd-update fills up /var/db/freebsd-update/files and never cleans them

The wiki page has not been updated since release of 14 so no news means no problems, right? :-) wiki.freebsd.org/PkgBase

Title: PkgBase - FreeBSD Wiki

I have two disks: one nvme one hdd. The NVME has my arch linux, the HDD has my recently installed FreeBSD. That means (at present) I have two uefi partitions, and to select between booting linux or FreeBSD I currently have to change the boot order in my BIOS

what can I do so that I have a menu, such as move FreeBSD's uefi data to the Linux disks uefi ?

I really don't know how alll this works tbh

but I'd like a kind of boot menu/loader

Also for bonus question - I would like to know: Why does enlightenment show a dark screen (eg. backlight too dim and it could possibly be frozen also) after the setup wizard

brb

hold that thought

back

hi

:)

I need help getting sound working: I have a Realtek ALC892 onboard sound, running S/PDIF (optical cable) to computer soundbar

I've read the handbook and tried everything to get sound working - mixer channels are at 100

I've set default snd unit to 5 - which corresponds with pcm5 (digital output)

AumShivaya paste some where /dev/sndstat

dpaste.org or somehwere elese

jb1277976, sure thing. Though I have tried all the outputs :/

!paste

oh wrong channel

jb1277976, dpaste.org/O47Am

Title: dpaste/O47Am (Python)

AumShivaya putting kids to sleep but install gtk-mixer open it okay something like YouTube. Switch the outputs in gtk-mixer and whatever one works you can mark it as default

Play*

No sound through any of them, sadly :(

back - computer cord fell out lol

whoa it is working? how?

good thing that cord fell out

must have been something a reboot fixed

Do I still set up virtual channels somewhere for an audio device to have in-kernel mixing?

rwp: Yup, but that references /var/db/freebsd-update, which isn't used with pkgbase (afaik; wouldn't make sense anyway).

I'm just going to try it. Worst case scenario is that I have to revert back to this boot environment.

AumShivaya: Check /var/log/messages to see if there was a pid holding your previous sound device open. That happens to me sometimes (e.g. headphones get unplugged) and I have to SIGHUP them to release the dsp device to let it re-initialize.

I have a devd script that does that for me, but it's not always successful.

ok

I'm playing with mixing now

I thought there was a way to just enable mixing in rc.conf or loader.conf before?

ok so mixing isn't working but I have sound ! :)

still dunno

if I set vchans for this card, it won't play at all :/

brb

back

yeah, I have this in my sysctl.conf

dev.hdac.1.polling=1 hw.snd.default_unit=5 dev.pcm.5.bitperfect=1 dev.pcm.5.play.vchans=0 dev.pcm.5.rec.vcahns=0

line breaks removed to not flood

if I change vchans to 4, sound stops completely, or if I remove that line

gonna try compiling the linux kernel here on my freebsd setup

does it work in bitperfect mode?

not all cards are working, hence the question

hm?

this is for AumShivaya

ah

got this issue as well bugs.freebsd.org/bugzilla/show_bug.cgi?id=275361

Title: 275361 – php81-pecl-APCu: Undefined symbol "php_pcre2_match_data_create_from_pattern"

Is it a rare bug when xorg mouse pops up on the vt?

Happens really rarely for me

i don't use the xf86-amdgpu driver so xorg should be rendered through drm and not vt framebuffer. Still, happens sometimes. When it does i just get my mouse from xorg on a vt/tty that i can move around but not do anything else with.

Happens mostly when i'm compiling something huge and my system randomly freezes

...okay i'm bad at explaining things

How can I make a drive avaialble as NFS in freebsd while freebsd is a guest OS inside virtmanager?

in virtmanager, how can I pass a disk from guest OS (accessable to guest only) to the host OS? In this case geli encrypted disk readable by freebsd (guest) and data access given to debian (host)

what do you mean 'pass a disk'

if a GELI-encrypted partition is attached and being transparently decrypted by the FreeBSD kernel, you should not access the underlying device in any other way

not sure how can I get host OS see the disk that guest has acces to

NFS, SMB, scp, rsync, etc.

Remilia: well the device will be access THROUGH freebsd

Remilia: ok.

SMB and NFS maybe but :

Remilia: can I use that remote samba/ NFS to just symlink an existing directory and pose it like the file was local? Can NFS do that?

do you know what NFS is for

(among other things)

network file system

NFS is how you boot diskless clients

well I am a noob at this

ok then Ineed SMB. not NFS?

eventually I want the drive data to be access by debian OS but debain cannot decrypt geli encrypted drives. Only freebsd can

basically NFS is how you make your client use remote directory structure as local

ok then that is what I need?

the whole point of NFS is mounting a remote filesystem over network locally

but setting it up is massive pain

I am not sure why you cannot just use rsync

ok. not in a position to get pain

Remilia: rsync will require another disk. I have only 1

do you seriously plan to run a guest all the time and access its filesystem over NFS just so you do not need to copy

If I have no other choice. sure

I could use advice on how to fool-proof my router startup against 'no link'. pastebin.com/raw/BX1iNtAr Power just came back up, and it seems the cable modem or the network is slow, dhclient gave up and took an extra 2 minutes to assign an ip, which made other things fail temporarily.

buy a USB external HDD, dump the encrypted partition content there, reformat the original disk

Beladona: how much content is there even?

(also realized I shouldn't reboot while waiting for an answer...)

Remilia: 3 TB

on HDD

in raid 1 zfs

a 4 TB USB drive costs $100

and will save you incredible amounts of I/O performance

might do via clound instead of $100

you really do NOT want to run NFS or SMB if you can

I see

cloud + rsync. Upload, download. whats say?

whats the most reliable could storage out ther for such operation?

Backblaze maybe?

i think rclone port is worth looking at for a cloud backup.

but it is not very cheap

ok

Title: Rclone Power Moves for Backblaze B2 Cloud Storage

hi! first time freebsd user. as far as i understand all the headers for the base system should be pre-installed. i'm looking for libzfs.sh, but that includes libshare.h which is missing on my fresh freebsd 13 install. do i need to install something else?

I mean, Backblaze is probably the cheapest backup option but fetching data from them costs money

ok thanks

aep: libshare.h is part of /usr/src which is not installed by default

going out

you need the full source tree for that one

/sys/contrib/openzfs/lib/libspl/include/libshare.h

not /sys

it is /usr/src/sys/.....

poudriere% locate libshare.h → /usr/src/cddl/compat/opensolaris/include/libshare.h \n /usr/src/sys/contrib/openzfs/lib/libspl/include/libshare.h

ultramage: docs.freebsd.org/images/books/handb…ll/bsdinstall-config-components.png the src distribution

aha thanks. going to figure out what that is

you can install it separately if you want to

cd /usr && git clone git.freebsd.org/src.git

Title: src - FreeBSD source tree

should do it

fetch -o /tmp ftp://ftp.freebsd.org/pub`uname -s`/releases/`uname -m`/`uname -r | cut -d'-' -f1,2`/src.txz

oh just like that? hm thanks

aep: do not clone main

was expecting its managed in some way so that it follows system updates

if you are on RELEASE you need to check out the correct branch

that makes sense

or rather tag

use the fetch command I posted above to get the distribution

it will put src.txz in /tmp

then you just tar -C / -xvf /tmp/src.txz

my example needs to be followed by 'git checkout releng/14.0'

Hm I'm looking at the dhclient manual, and it says that it will reuse its saved leases if it times out. Mine isn't doing that. I think the 'no carrier' is at fault, guses it won't try to work with an interface that's down. I either need to force it to do that, or, to wait longer for the link. Currently my timeout is set to '300' (5 minutes), which should be enough (or not?), unless there's a separate

timeout for link wait and dhcp reply wait.

ultramage: you seem to be using Realtek NICs, of the 'lowest end' variety

I urge you to reconsider

Remilia: when installing src from distribution set, will freebsd-update recognize it?

elirco: yes

it analyses your directory structure

good to know, ty

elirco: all the FreeBSD installer does with these is extract them, there is no database of installed distributions afair

makes sense

note that the distribution sets do not contain patchlevel updates, freebsd-update should update those for you

so you can only install src from distribution set, as long as you are on -p0?

basically if you are on -pX and extract src.txz, try running freebsd-update

no, you should be able to update afterwards

I cannot say for sure but could test later (right now I really should get back to work)

ah okay it can handle that and re-implements -p0 to -pX only for src?

cool

I cannot easily test since I am on 14-p0 and probably am to lazy to set up a vm...

I am in the middle of a workday and my job is translation, not IT, so setting up a VM and such would distract me for at least an hour

I am just not curious enough at the moment...

Remilia: it's an intel atom potato pc with a re onboard nic. the other one is re just because the previous one failed. and this issue is not the nic's fault, it's the cable modem taking a long time to bring the link up. I wonder if it waits for its uplink to fully calibrate first...

okay, the linux kernel compiled fine on freebsd

hm looks like i accidently installed freebsd 15, which is CURRENT. does that equal to main in git?

I could try bumping up the dhclient 'timeout' from 5min to 10min, but that feels kinda dumb. Doesn't matter atm since I'm remoting, but during an outage, it delays the startup of sshd and other LAN services.

yes, FreeBSD 15 is git.freebsd.org/src.git main branch

you'll have to constantly git clone it and recompile the kernel probably

yeah just learning things for now, its fine

looks like /usr/src is indeed just the source, not the result of make install, so the zfs headers aren't in the directory structure that it expects. am i supposed to call make install myself? wont that mess up things?

never tried installing freebsd on zfs before, i'm too deep into ufs

but what i do is just rm -rf /usr/src && mkdir -pv /usr/src

then git clone -b stable/14 --depth=1 git.freebsd.org/src.git /usr/src

Title: src - FreeBSD source tree

aep: zfs headers should be in /usr/src and they will not be installed elsewhere

aep: afaik, there installing of header/src files per package concept only exists in Linux. I FreeBSD src sits just there in /usr/src.

hm but openzfs itself expects make install to be called. the headers have hardcoded paths that dont match the source

which openzfs?

can you pass the path with the required header to cc via -I parameter?

then i just copy my kernel config into /usr/src/sys/amd64/conf/ and run the make command, updating freebsd from source takes only ten minutes because of how tiny the sources are and how fast the compiler is

FreeBSD comes with it in contrib

elirco: no because the path includes a name that is only created when calling make install

aep: what name?

libzfs/sys/zfs_context.h . there is no libzfs in /usr/src

I am not sure what you are trying to build

and which 'make install'?

> there is no libzfs in /usr/src <- yes there is

in /usr/src/sys/contrib/openzfs/lib/libzfs

ok but thats not where the header is i mean :)

once again: what are you trying to build?

its created on linux when you call make install

FreeBSD is not Linux.

uh just a thing using libzfs

FreeBSD has /usr/src/sys/contrib/openzfs/include/sys/zfs_context.h

so I do not understand your <aep> ok but thats not where the header is i mean :)

yes but the header is expected to be in a path named `libzfs/sys/zfs_context.h`. i dont understand (new to freebsd) how this is created

it is already in /usr/src/sys/contrib/openzfs/include/sys/zfs_context.h

thats a different path. wont compile

you are supposed to use that

its #include <libzfs/sys/zfs_context.h>

are you unable to use the correct path?

are you not aware of the compiler INCLUDES envvars and command line parameters?

oh you mean edit the zfs header that included it?

no I mean use the correct INCLUDES path

i wasnt aware gcc had a way to change <include> directives, no.

see the man page for cc

it does not have a way to change them

do you know the difference between absolute and relative paths?

#include <libzfs/sys/zfs_context.h> means include libzfs/sys/zfs_context.h found in include paths

it does not mean include /usr/include/sys/libzfs/sys/zfs_context.h

aep: [12:26:29] <elirco> can you pass the path with the required header to cc via -I parameter?

this was directed at you

clang: -I<directory> Add the specified directory to the search path for include files.

but considering the specific path in your case you may need to patch the .h file

also <aep> uh just a thing using libzfs <- what thing? are you sure it is not in ports

linuxulator here is amazing. even Microsoft Edge linux binary works on it

(i don't use edge though)

in part that is thanks to Linux keeping its ABI compatibility ha ha

yeah

i switched to fbsd like a month ago and now i can't go back, everything is so snappy

bluetooth works much better than on linux too

that's actually a surprise, bluetooth stack doesn't get much love

quick sanity check: if i have a fileserver (ZFS, serving NFS + SMB) with 16GB RAM, which stores a mix of small files and large files (mostly 4-8GB, some 20GB+) which are only accessed infrequently, sequentially, and almost never more than once every few weeks per file, then it would make sense to set primarycache=metadata on that filesystem... right? since the files will almost never be served from cache, and caching them will evict other, more use

if you hae a fileserver, what other data would be more useful on it?

the small files (which include databases, and easily fit inside available memory), filesystem metadata to improve performance of things like 'find', and the OS itself

is there some e-mail me daemon if say for instance a drive dies for fbsd?

Macer, not sure I understand what you're asking for, but could you setip something like nagios for that ? There's cacti and other similar tools.

Macer, look at monitoring systems like nagios or icinga or zabbix.

mns: Cacti and other rrd tools are great for capacity planning, but threshold and event notifications are secondary. I like nagios for is extendability and simplicity.

Hello, tried to update from 13.2 to 14, and was greeted with an error ( pasteboard.co/X3yUquChInqF.jpg ) - all is fine when I reboot on old kernel, but not sure how to go from here...

Title: Pasteboard - Uploaded Image

Say, I'm seeing a lot of "pid 27955 (getty), jid 0, uid 0, was killed: failed to reclaim memory" in my log (other processes tooo of course), and `top` tells me I have most of my RAM dedicated to "Laundry". Any idea what this is?

I prefer nagios as well, its also what a lot of companies use internally. I was just throwing cacti out there as something I've used before and an alternative to nagios. Yes it leans more toward trend monitoring, but you can still do the basic thing of triggering some action below a certain threshold. I've not used cacti in ages though so could be out of date.

'31

Me too. I remember being impressed with how well Cacti's device discovery worked, but much must have happened in 10 years. :)

I'm sure much has happened since then, Its been over 10 for me

Macer: Easy setup 1-system setup? Probably something like netdata.

Macer: Specific to drives failing and that's it? smartd, and there's also another project who's name I can't remember off the top of my head, but it's dedicated to this exact scenario.

Macer: Multi-machine setup that can monitor a lot more than just hdds, has a dashboard, etc? Prometheus (or VictoriaMetricsm) + alertmanager, and grafana if you want a dashboard.

Macer, The only default disk notification I know about is "sysrc -f /etc/periodic.conf daily_status_zfs_enable=YES" which enables a daily email which includes "zpool status" among other things.

I have not found anything similar to mdadm's notification email from Debian-derived systems. But one could be created.

The "logcheck" program will report any unfiltered /var/log/messages entries. But... It's a firehose of data and needs aggressive filtering of uninteresting log data. I use it. But I aggressively adjust filters all of the time.

+1 for smartd and periodic short self tests as well.

All right, pkgbase upgrade time from 13 -> 14.

Good luck _xor! I look forward to your report.

Already hit a snag, but probably my own fault. It stopped after "[369/407] Extracting FreeBSD-wpa-man-14.snap20231126144041: 100%", but no error message.

Running it again gave me: ld-elf.so.1: Shared object "libssl.so.111" not found, required by "pkg"

Not super surprising. Bootstrapping pkg beforehand may have helped, but not sure, yet.

oh wait, why is it looking for openssl111? I built everything with 3, including the version of pkg on 13.x.

_xor: good luck

I'll get back to this in a few hours, gotta go fix my toilet right now. (no, I didn't destroy it...stupid water supply line is leaking)

_xor, You might need to use pkg-static which is statically linked and created specifically for such issues.

yeah, i always do pkgbase upgrades with pkg-static

_xor: because when you run `pkg`, you don't immediately run the version you built against openssl3

you run the version built against base openssl, which then invokes the version you built from ports

s/version/bootstrap shim/

rwp yah I’m using periodic for snapshots

But I’ll look into it once I get time to do it. I’m pretty much done with all my little projects.

Ok, toilet is replaced. What a pain in the neck.

Back to pkgbase.

kevans, rwp: Yeah, forgot about pkg-static for a minute there.

i only remember pkg-static when pkg throws an error

ABI=FreeBSD:14:amd64 pkg -n -x 'FreeBSD-*'

Oops, wrong window.

Heh, `pkg install` might have helped there :P

meena: How does pkgbase handle merging config files (e.g. /etc/*)? I know it creates {NAME}.pkgsave, but is it currently that merging is handled manually by the admin or is there something I'm missing?

I don't mind merging myself, I have my etc files in a git repo, so it won't be too much trouble, but want to make sure.

it should handle it the same way as for ports

Dumb question, but is *.pkgsave the new one or the old one?

old one

That's what I thought, and just diffed to confirm. Wanted to make sure.

It doesn't create *.pkgnew files, right? (I have a few of those, but I want to say I created those a good while back, just don't remember 100%)

kevans: What's the main argument against having a single pkg executable that's fully statically linked and not requiring pkg-static? (aside from the whole dynamic libs for updates argument)

I'm actually asking btw, as I'm guessing I'm missing the reason as to why it's done that way.

_xor I am right there with you on that question. The usual reasons given for shared libraries are that they save disk space, they save memory pages, they avoid hidden security vulnerabilities, all by sharing components which are separate and shared.

What's the command to re-generate /boot/device.hints again?

Interesting, so the EFI loader (/boot/loader.efi) shrunk from 886kb in 13.x to 644kb in 14.x?