i did a pretty basic install of a freebsd vm in a freebsd guest. when i start it up, it says route: message indicates error: file exists, add host 127.0.0.1: gateway lo0 fib 0: route already in table, route: message indicates error: file exists, add host ::1: gateway lo0 fib 0: route alraedy in table. why am i getting that? i didn't even configure

any network settings

14.4 fwiw

just did a stock freebsd install, same errors on startup. wtf why does a clean 14.4 install error on startup

Looks like it's trying to add a route that already exists.

mason, Did you make progress understanding the missing sshd-auth new file? I am hitting it now with one bare metal system. Starting sshd. /usr/libexec/sshd-auth does not exist or is not executable /etc/rc.d/sshd: WARNING: failed to start sshd

ya but why does a stock 14.4 install ship with errors output right to the startup msg?

It looks more like warnings than an error. Just wondering if you configure a network interface and hostname, does the symptom subside?

no still happens. and it says error

kerneldove, Can you share the error you are seeing? And say the full 14.4 version?

when i start it up, it says route: message indicates error: file exists, add host 127.0.0.1: gateway lo0 fib 0: route already in table, route: message indicates error: file exists, add host ::1: gateway lo0 fib 0: route alraedy in table

I was rather hoping for a pastebin with the full page of errors... :-(

For example here is one from one of my systems that I am upgrading: bpa.st/IUKQ

man that's weird

i'll get a repro together

For some reason upgrades to 14.4-RELEASE-p6 are sometimes not installing /usr/libexec/sshd-auth and then sshd can't start.

I just reverted to a previous Boot Environment. Everything worked there. I upgraded. Error reproduced.

However other systems upgraded okay and did include sshd-auth. Weird.

kerneldove: Yeah, understood that it said error. Just the sitaution itself appeard more like it was meant to be a warning.

what pastie site should i use?

paste.debian.net/hidden/290215d3 that's a super minimal scripted bsdinstall for 14.4 just trying to get zfs in place and default everything else, but when running it toward the end it gives the error debug window but doesn't say what the error is?

does it put it in a file?

that you can then grab?

what path would it be at?

i dunno actually!

can you try it yourself and see locally what i'm describing?

i made it as simple as possible, it should work

oh i dont have any 14.4 images on hand

what's supposed to happen is at end of successful scripted install, it automatically reboots. but this error window shows when either there's an error, or when your installerconfig runs a command that generates output. but don't know what would be doing that ins uch a simple installerconfig

rwp: No, I haven't had a chance to look. I've only seen it the once so far. I've never looked at freebsd-update innards, but somehow the detail that a new file is required for sshd is missed in some cases.

nayone had any issues upgrading mysql80 to mysql84? just a drop-in replacement?

f451: Always take full back-ups before you try. That said, I'm more of a PostgreSQL person. But I still take full back-ups just in case. The time I don't take them will be the time I needed them.

oh yeah, i got htose ;)

good good

also tarsnap

f451: FWIW, they have what might be a useful page here: dev.mysql.com/doc/refman/8.4/en/upgrading-from-previous-series.html

hteres other stuff too like this server is on 13.5 so like *everything* has to be backed up

mason: thanks

It's worth looking at the equivalent docs for 8.1, 8,2, and 8.3 as well. I'm betting it's pain-free, but can't hurt to look.

a little light reading ;)

also have drupal to think about and it's on 10.1.x

f451: I hope it all goes smoothly for you.

and it's 0415 here lol im going to bed

o/

ty :D

\o

f451, I would consider switching from mysql to mariadb. It's the better development track. I would do an sql backup dump. Remove mysql. Upgrade the system. Install mariadb. Load the database. That has worked for me.

what is .SUNCTF_ctf section?

.SUNW_ctf

or rather why I started to get this during boot.

No idea. I am having other problems with the most recent 14.4-RELEASE-p6. sshd-auth missing. Me feels that this release may have been rushed.

I have this feeling with the entire 14.4 upgrade

I have a system that has upgraded but with a missing required sshd-auth twice now. Seems reproducible. Each time I fell back to a previous Boot Environment.

I am trying an experiment now where I emptied out /var/db/freebsd-update leaving it an empty directory forcing a complete new download of the update files. Testing to see if that fixes the upgrade problem or not.

I've been running jails for many many years now, and upgrades never were any problem, and now I have to switch off hardware checksums to reach them... makes no sense

Because many other systems upgraded without this problem. So maybe it was a failed corrupted download? Don't know. Trying this experiment.

Oh I saw that chksum bug go by in the bug tracker, someone else referenced it.

during the years I made some modifications on for instance pam, but it never gave me any problems, till now, where my cron wouldn't run anymore. no mention of cron nor pam in the release notes

This one I think? bugs.freebsd.org/bugzilla/show_bug.cgi?id=295057

Dan Langille, not the least, also committed one: 295986

In my case there's no NAT involved.

I've read somewhere else that ipfilter might be involved in this issue

ipfw, I mean

kinda crazy that there are these probs with 14.4. you'd think being so many patches into so many minor versions it'd be the most stable release ever

joemie, I think that chksum issue was tied to ipfw. So pf users are not seeing the problem. I think. I use pf and haven't seen that problem anyway.

kerneldove40, Does seem like something was rushed this release.

My experiment still failed. "Starting sshd. /usr/libexec/sshd-auth does not exist or is not executable /etc/rc.d/sshd: WARNING: failed to start sshd"

Strange that this upgraded just fine on a dozen other systems. Mason also has seen this problem. I have one system with the problem. What makes it different? No idea.

Too late for me to debug this further tonight. I'm rolling it back and going to sleep on it. Good night all!

sleep well

what would you guys say is the one filesystem that would work the best between freebsd and netbsd

read/writes OK without it being "risky"

im getting the impression its ext2fs even though both have UFS

from asking netbsd, it seems like its ext2fs

"BUMSRAKETE is a YUUUGE kernel bug in FreeBSD. Probably the biggest. Tremendous, really." bumsrake.de

Hi, why there is quarterly and release_x packages in pkg.freebsd.org/FreeBSD:13:amd64 ? When would pkg install use release_x and when it will use quarterly?

release is allways the newest versions, quarterly only gets updates when it's imported or after 3 month

it's a bit like linux roling release vs stable like distros

pkg install poudriere

satanist: when it is imported?

I did not understand that

sorry important

how do you select one or the other, and if you are using RELEASE you say that the packages are constantly updated? (until EOL I guess?)

you can select the repo in /etc/pkg/FreeBSD.conf

url: "pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",

so here I am using quarterly, if I would want to use release I would put there release_5 ?

no just release

but when I observe, for instance the dates for packagesite.pkg in FreeBSD 14, inside quarterly the date is 2026-Jun-09, but inside release_4 is 2026-Feb-15. So it seems like it is actually quarterly the one being updated

url: "pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",

you can see the dates for packagesite.pkg in both folders

ah sorry, I mixed up latest with release, I have no clue what the release_* repos are

I am trying to find in the documentation what is packagesite.pkg in release_x but I can not find anything

the latest vs quarterly seems clear

sort of new to bsd admin, still confused about ports and packages. i thought pkg(8) was for binary packages and ports are a different thing, yet the pkg repository is FreeBSD-ports? is there something i'm missing?

i've also found the handbook has a command that will wipe your existing /usr/local/etc/pkg/repos/FreeBSD.conf when it probably shouldn't and i'm not sure how the handbook is edited

it is a bit confusing IMHO, I have been using FreeBSD for like 7 years or so and now I am trying to finally understand in full the mechanism and different branches/repos

to my knowledge, pkg's are the binary distribution of the ports, and ports are just the ports to FreeBSD of third party software

okay that makes a bit more sense if it is the case

the base system is downloaded elsewhere

you can also compile from the ports (I have never done that, I always use pkg)

i know that pkg now does the base system in 15.0 which is nice coming from a linux user pov

I have no system in 15 yet

I am trying to freeze the last released version of 13 to use in internal jails, I am downloading all packages and setting up my local repo, so you will be able to use it both today and in 10 years (this is only good for internal systems)

yea compiling from ports has also been a bit confusing (haven't done it yet, just the concept). from what i've read online i should use ports XOR pkgs and not mix them?

I have never dealt with compiling ports, but it makes sense that you can not use pkg for that particular port at least if using the sources?

not a question, just an educated guess

luckily my use case is just a couple of servers which are running linux guests via bhyve. i am also a fan of running the latest and greatest for security patching so it was a shock to know quarterly is the default on FreeBSD (unless security patches are backported?)

yeah i presume you can't install the same package via pkg and ports. maybe i was getting confused when reading because i thought people were saying to not mix use of pkg and ports at all...

I think quarterly DOES include important security updates

thats what i assumed

I have many systems running in jails and all are internal, security is no concern, so to me it is better for those system to have a static snapshot that I can use today and in 10 years if needed

a live exposed system is a different matter

fair enough, these machines are supposed to be internal but since they're within a network i don't control i'm pretty much forced to assume they aren't

the only thing meant to be exposed are the guests. even then, with recent linux lpe and container escape stuff i don't trust the separation between the guest and the host.

if by exposed you mean it can be freely accessed from the internet I would ensure the host system is updated no matter what, in my case everything is internal and I trust the environment

yeah the guests are, hosts aren't freely accessible. they're all in their own vlan but i don't control the network policies and getting them changed is a bit of a slow process

well if the guest are exposed to the internet the host is indirectly is, I guess bhyve isolates it well but who knows, I would not risk it

to my knowledge the case there is to have it in quarterly and update it regularly, and you will have to keep upgrading releases and majors

that is why I am aiming to freeze a version for internal stuff, you do not have to update anything, but for that you need an internal pkg repository

yeah that's my mindset with it all anyways, assume everything is compromised and harden things to limit damage accordingly

i'm yet to see how pkg upgrading the base works. i guess i'll see when 15.1 is ready

I only use packages, except if I want something special: for instance I use postfix, but the package doesn't contain the cyrus-sasl part (at least, last time I checked), and I compile it from ports. To prevent pkg to meddle with it, I lock it after compiling.

joemie: okay that sounds about right then. i think i just got confused when searching online

FreeBSD actually advises against mixing packages and ports, but sometimes there's no other way (unless you use 100% ports)

yea

petercs: poudriere-devel mixes packages when it pre-fetches binaries from pkg.f.o

it compares shlibs and things like that though

easy to make a mess if it's manually done and not in cleanroom like poudriere

rwp maybe the devs resent maintaining 14 for so long and they're focused on 15 and kinda being sloppy?

hey i just noticed pkg.conf format is different now, when did it change?

kerneldove: ? pkg.conf has been UCL for a long time

i've been using my own custom file, guess i never caught the upgrade!

awesome that ucl is getting used more and more

kevans do you also like that ucl is getting used more and more in base?

yes

ucl is good

nice

anyone use the helix text editor? all of a sudden it now panics anytime i try to open a file. fresh install

panic? not messing around is it

no, piece of junk imagine panicking on opening a file

seems to be written in rust, can you run it with RUST_BACKTRACE=full in the environment and share the backtrace?

ya sec

sorry it's taking forever i was in the middle of compiling latest hx code

termbin.com doesn'twork anymore?

kerneldove: yes I use it daily on fbsd and linux

amlor cat file | nc termbin.com 9999?

cat what file?

any text file

that's how i've used it for years but it's not returning a URL anymore

what does it have to do with helix?

ah I got it now

lol ya

you were talking about two things and I didn't specify which one

I meant helix, which works on updated freebsd

version 25.07.1, installed from pkg.freebsd.org

ok figured it out...

can't run hx from vm console myvm connection

guess it hates vt100?

lemme check

which shell are you using btw?

I ran it in sh from vm console and it is working. lags a bit when starting and has no colors, but doesn't crash

FYI: termbin.com has been stuck the past few days. Hopefully it will get fixed. But the paste process has not been working.

although $TERM variable has xterm in it for some reason

added export TERM=vt100 to .shrc and it still doesn't crash

amlor echo $SHELL => /usr/local/bin/bash

and you're using the default vm console vt100?

`TERM=vt100 hx` is sufficient

t test that, amlor

I did that too

bash doesn't crash it for me either

wtf

latest or quarterly pkg repo

not sure if it matters but what's your hx --version?

latest

version 25.07.1, as I mentioned above

ya dunno

i even updated 14.4 to latest patch and same crash

ah well whatever i'll just ssh in after using console to install

ty for helping me find a workaround guys

Question about "dump" command. Before few days, I've been asking somebody here why does he use the dump over let say rsync. I was curious whether there is some speciality about using "old" dump command. And I've been told here, that basically there is nothing special. Now I have discovered, that first, it is about 2x faster than rsync (on linux),

and more importantly: there is the "-L" flag, which is ment to be used on Live system. With -L, it will first create the UFS file system snapshot, and then it will backup this snapshot. So, guys... This isn't a small thing, isn't is? This could be an old and overlooked life system backup tool. Even tools like restic can't do snapshots.

kerneldove: ahh I missed that you have 14. I'm on 15 latest

rsync is slop now, so that may be one reason to use dump instead

OpenBSD has switched to OpenRsync.

JurassCZ: people use dump with restore, and rsync is not dump replacement

what happened to rsync?

Guys, I'm now confused, it looks like dump is only for UFS, and according to manual, with the -L flag it can do the UFS snapshots!!! i've never knew this is possible, I thought that snapshots can be done only via ZFS or LVM. Heck, hjow is it possible that such old command and old filesystem can do snapshots? I'm just surprised.

but zfs-send(8) can be considered dump(8) replacement

yes, dump is for UFS and zfs-send is for ZFS - the same purpose

Ok, and nobody is surprised by this? I've spent several mandays to investigate some general purpose backup method on Linux, and basically all the tools the Linux has like restic, rsync, dd, etc., nothing can do snapshots of the life system! And heck, the FreeBSD has such old dump command, which is able to do UFS snapshots. It's just ridiculous.

amlor what makes rsync slop now?

kerneldove: forums.freebsd.org/threads/rip-rsync-ai-slop.102835

kerneldove: they started using claude in their commits. i /think/ the this is actually unrelated but happened at a similar time they started using claude, a load of regressions then ensued.

kerneldove: the main dev decided to let LLMs do anything with the code base, like rewrite the test suite into bad python neuromatch.social/@hailey⊙ho/116657392148745644

ok ya gotta switch to openrsync

i trust openbsd instincts

i think someone (potentially the author of rsync) wrote an article "proving" claude hasn't caused any more bugs via statistical tests, but i think that ignores the root of peoples' upsettance (if that is a word)

ya well he also runs a business (tarsnap) based on it

so he's biased

the noun form of upset is also upset

oh I didn't realize tarsnap was his

glad I didn't start using it

yep

also please don't post foreign post links private to your modondodon instance

use the menu and select copy the link to the post

ty for heads up on rsync becoming slop

modondodon lol

kerneldove: yeah that's exactly what i thought. apparently he had his wife (clearly no conflict of interest there!), who has some degree in stats from some university, review it.

ffs

Also statistics talk about numbers, not about the severity of bugs

"my mom sais that I did nothing wrong" basically

MelanieUrsidino: very true, brain is not braining today

lts: but what about bugs per 10 lines? isn't that a good measure?

weird move to use his wife as a shield for an obviously bad decision

(yes that's what they used in the post if i remember correctly)

mildly misogynistic tbh.

petercs: I prefer wtf/min

yep

that's what i was thinking melanie

unless she's financially independent of him, there's a conflict of interest there

and even then, ties of love even not strengthened by money can cause a COI

alexispurslane.github.io/rsync-analysis okay so i must make a correction, it does not look like tridge authored this post

ya for sure, he literally penetrates her

ostensibly

allegedly

i'm not sure who alexis purslane is in regards to rsync

Anecdata but Debian mirror maintainer complained about mirror sync problems this week and suspected it was related to rsync being updated to post-LLM version

s/Debian/a Debian/

kerneldove, that was not appropriate nor relevant, thank you

+1

oh sorry i was just thinking if someone was entering me there'd be some intrinsic bias

np

wtf doesn't clicking on a post in a mastodon thread go to the original server for that post?

it depends on the frontend

and your settings in it

if you're logged int

regardless of authorship, it's kind of a weird way to argue about the introduction of LLM use to a big, widely-used project. also still weird to use your wife's degree as a way of saying "i'm right :)"

the web one, is there any native frontend to read random mastodon threads?

had no idea about OpenRsync though so i guess i'll swap to using that :)

good thing rsync isn't included in base

nimaje: i suspect there are many implementations. there are at least multiple android apps.

nimaje, no, not if you're on a mastodon account of your own

it opens via your server usually

there are linux clients too. tuba or something. don't remember the others

kerneldove, again, that's not it's not appropriate to talk about contributors to a FOSS project having or not having sex here

I prefer phanpy.social in a browser tab on desktop

s/that's not it's/it's/

where's the bsdinstall source system hardening panel code? i wanna see what those hardening options correlate to

petercs: and any that on freebsd that is just for reading? Most I found require an account and/or are just the web implementation repacked with electron

you will have to somehow redirect all the possible domains there

I don't think it's possible

nimaje: not that i know of, sorry :/ i thought web instances would still show the post without login

kerneldove: i'm not sure but you can see some in your rc and sysctl configs on a machine configured with them iirc? maybe also /boot/loader.conf? if that helps

i found it

where was it? i tried to look for the bsdinstall source before but didn't find it with some quick searches

petercs: they show the posts, but they require js and are shitty and as I just learned they don't go to the original server when I select a post for viewing

nimaje: hm, you might have luck with some CLI/TUI clients? they might require signing in or an access token or something though

nimaje: there's also this no-JS web client but i've not used it and it looks unmaintained forge.spoons.technology/theothertom/planiverse

hm, the demo instance is currently down, maybe I try it later on my own system

do the majority of people use `pkg` or the ports collection or a combination of both? I'm thinking about using the `pkg` collection since it's precompiled and I usually stick with defaults in software config options.

you can do both

i used to do 100% ports but now mostly packages unless i have some specific config option (which isn't very frequent anymore)

rtprio: ok, thanks I'll use pkg then

also, depending on your system and patience, i know i don't have the time to wait around 22 or so hours for rust to build

rtprio: I was also thinking about using pkg since I'm usually fine with most distro defaults and I'm guessing it will be the same w/ freebsd

td123, Start with pkgs at least. It's the easiest and will do everything you need. However two caveats. Graphics modules from pkgs are sometimes out of sync with the latest kernel. That's been improved recently but hit me again recently.

Some complicated pkgs fail to build initially and the maintainer needs another build run to get through it. Build daemons are always busy.

This doesn't affect people running the system over time but upon initial install firefox may still be waiting in the build queue sometimes. This is only a new install issue.

rwp: ok, thanks for the heads up about pkg

Is there a way to auto clean or auto drop snapshots? I want 5, when 6th is created drop the oldest?

yes there is, I use a script on one of my servers that keeps the 4 last snapshots

...are you going to share? :-)

Well I know it could be scripted but... nothing builtin

or is it a user added attribute? Like a bolt on action?

Turns out when you let freebsd-update create snapshots over the span of 4 years things build up.

I use sanoid for periodically dropping snapshots, according to the ruleset applied to each dataset

for the freebsd-update snapshots, i delete those manually a couple/few times per year

It's worth using different patterns for snapshot names for backups, updates, etc.

Then you can mechanize deleting certain classes of snapshot.

at distinct schedules

s/at/on/

JurassCZ: FreeBSD is a designed OS

mzar: Huh?

JurassCZ: - answering why dump(8) is able to dump live filesystem

THB I used dump with tapes, worked like a charm

I had to recover from them a couple of times

later switched to dump over ssh

Do you know since when the Dump and UFS supports the filesystem snapshots? Is this a new added feature, or it could be like this in 1995?

Damn that zfs rollback works like a sharm, even on live system. This is something a linux user will not get so easilly. I suppose in case of lost laptop, I could just install new FreeBSD with ZFS, trasnfer the backuped snapshot, and just do rollback to it. And it will give me the exact same "/" and my system will works. Never tried it yet.

But I suspect it will work.

:-O

JurassCZ: I have :D works great, just needs a boot partition/bootloader too

I will write article about it to root.cz soon, should I tell everybody how good FreeBSD is, or rather stay silent so nobody will know it so easily? :D

aha, dobrý večer :D

yes you should tell people how awesome it is, we're not greedy :D

interesting

I mean the enthusiasm

JurassCZ: if you want to search for the time of implelentation of -L option in dump(8) you should check the commit history

but IIRC I was using it since the beginning of using it, so probably in early 1990s

hm, how does ufs' snapshots work? file based COW?

as the tool came with at&t version 4, I would guess the first version came after the big rewrite?

I can tell you that there's no "-L" in FreeBSD 2.2.8 (I happen to be in that in a VM at the moment)

It is in 8.4, so somewhere between there. (boots up 4.11...)

not in 4.11 either

spork_css : 4.11 was so close version 5.0 man.freebsd.org/cgi/man.cgi?query=d…ath=FreeBSD+5.0-RELEASE&format=html

nicotine (an interpretation of soulseek) does not run on xlibre ? (freebsd 15.0-RELEASE-p8) I don't know how exactly to understand the output pastebin.com/ydLuwJWc

solstis: i dont think that is the full log

wait

it just cuts off

well, it talks about a debugger :O

maybe just report the error as a bug, but with who :|

thx, will try that later

JurassCZ: when i found out you can just do stuff like `zfs send zroot/vms/example | ssh other-host zfs receive zroot/vms/example` with ZFS it kinda blew my mind. makes it so easy to just move a VM image or back it up to another ZFS host. i believe there's also a way to incrementally send snapshots too

petercs: I've been using zfs in production since 8.mumble when you had to do a lot of work to get it kind of stable. It was rough, but saved a ton of money on some standby db servers where we were able to get a combo hdd/ssd volume performing nearly as well wi/Postgres as on our ssd-only dbs.

And I have an old junk PC in my garage that has no UPS, runs a freebsd desktop, is on a power strip that tripped quite often and the way that thing always just came back after incredibly rude sudden shutdowns has me in awe.

petercs: sure, you can send incremental snapshots too

and these snapshots don't have to really refer to previous ones, but can refer to bookmarks (traces of nonexisten snapshots)