i was in here crabbing about freebsd-update to 14.2-betas, RCs and RELEASE all left me with a system stuck at single user..

well pkgbase worked great

:)

fuck sake, 14.2-RELEASE has taken my laptop out :((

kinda strange watching pkg grabbing OS files.

mrelcee: I'm guessing that's the reason FBSD is moving to pkgBase.

The documentation will certainly need to be update for this, though. And fast.

*updated

polarian: do you have beadmn installed? revert to the snapshot

after reboot, during boot just after decrypting the root the screen resizes as I assume it hands over to the kernel as init is done, anyways this results in what looks like a blackscreen until I realise all the text is in about 3-4 pixels at the top of my screen

yeah docs were a bit leap of faith.

mrelcee: freebsd-update can fallback but I cant get into the system to do so

polarian: If you using freebsd-update, it should've automatically created a snapshot prior to the update. Use that to rollback.

ek: see above :)

lets see if I have a 14.1 media lying about

is your system ufs or zfs on root?

zrs

zfs*

should be a option in the boot menu to pick a boot target/snapshot and you can revert to the snapshot freebsd-update makes

ooo that might be useful

let me see

oh shit im an idiot

you can literally boot kernel.old

Yeah. Should be super easy to rollback.

thats why you havent read about Michigan Man Hulking out 4 times in one month when it took my server down. :)

well its almost as if the devs anticipate kernel updates going wrong

thanks freebsd bootloader <3

was annoying needing to drag a display over to it...

ugh now I need to go complain to the ML

which ML to scream about kernel problems?

I hate how freebsd has 101 different ML and I never know which is the right one

poilarian: this system has gone through 10.1 to present. most of the betas since 11.3, and this is the first time freebsd-update has ever crapped the bed on me

haha... so you have experience

I've...been around!\

my laptop was installed on 14.0, 14.0 --> 14.1 went smoothly, but 14.1 --> 14.2 has issues

anyways what ML to report issues?

or is this something which should be bugzilla'd

ah shit the revert didnt work

freebsd-update rollback didnt rollback the kernel

make sure you pick the right one. it can be tricky

polarian: You can report it on the ML or Bugzilla. Either is fine. You could use "freebsd-bugs@, freebsd-pkgbase@, freebsd-questions@, freebsd-reproducibility@, or freebsd-testing@ for the report. Any of them will likely point you in the right direction.

mrelcee: Same goes for you. I believe we spoke about this a day or two ago and you ran into similar issues?

ek: you see what I mean

why so many mailimg lists

OpenBSD has bugs@ and misc@, the rest are pretty dead

polarian: Well, they get pretty specific, but if you post to one and it isn't correct they'll tell you where to post it.

Generally, I would use bugs.freebsd.org since they can just move to where ever they need to.

ahhh shittttt

zroot cant be found after reverting

polarian: You should be able to tell it where to find the bootloader in zroot.

fuck it im just rolling back to the snapshot

I did like 10-20 mins of work after it which I will forgo

Also, are you using any of the i915 drivers, by chance?

errrr yes?

lol by chance has it been reported?

mrelcee: Were you not using the same? I saw something recently where people were saying that was a problem. You just had to rebuild it after the upgrade But, something unexpected, for sure.

polarian: Not sure about the reports. I think I saw it on forums.freebsd.org yesterday or the day before.

repost: i upgraded to 14.1 -> 14.2 on 2 different intelgpu machines. one worked fine and the other gets stuck when loading i915kms. by stuck i mean if i leave it enabled on boot, it gets stuck on the "Loading kernel modules:" phase. And if i kldload i915kms it attempts to load some _bin and then gets stuck

I'm not aware of any "official" reports.

HER: That sounds about right.

thank god, I typed some bs and my system booted with default kernel

lets hope the rollback didnt remove data

i had to rollback on the broken machine

sucks but it was nice the rollback worked

polarian: Which rollback do you mean? A beadm snapshot rollback or like a VM snapshot rollback?

i meant freebsd-update rollback

HER: Also, same question to you. Which rollback did you use?

HER: Gotcha'. Thanks!

and pkg upgrade -f

ek: freebsd-update rollback, twice the first time didnt work and the default kernel was still 14.2, this worked but then no zroot was being detected, so I booted kernel.old and that could see zroot so rebooted, booted default kernel with the autosnapshot from freebsd-update... that worked, so I zfs rollback to the autosnapshot and it works now

and I appear to have no data loss

Hrm. I'm wondering if this solution would help with the upgrade (not just a fresh install?): forums.freebsd.org/threads/boot-hangs-after-loading-i915kms.95904

But, it definitely seems like there's an issue. Hopefully, resolved soon.

polarian: A freebsd-update rollback would never lose any data (unless something failed, of course.) It will always be the snapshot taken just before the upgrade took place. It's the beauty of ZFS!

ek: well idk it didnt work so I manually restored the snapshot

but the snapshot restores the entire system no?

I didnt reboot immediatrly after install, I was working on some code

I see no noticable code loss

so I am not sure if my home dir is included in the snapshot or not

whats the scope of "default"?

polarian: It should be a recursive snapshot of zroot when using freebsd-update. So, yes, restores the entire "system" (nothing outside of /boot.)

ah yes I forgot zfs mounts /home/<user> separstely

I am not used to zfs on desktop :)

Yep. Any of the "system" changes would be reverted. Anything outside of that will still be included in the mounting and whatnot, but wouldn't be changed at all. Right where you left it.

tbh I barely used zfs... even after almost a year its still a headache

must take years for people to feel fluent and confident in xfs

zfs*

What!? ZFS is one of the easiest FS's to use ever! Granted, it can really get involved and things go really deep. But, basic maintenance and recovery is so quick and simple!

Look how easily it saved your booty and you didn't even know what you were doing. It's that good.

ek: well I can scrub, snapshot that stuff

but when things go wrong I am a little clueless :P

not to mention the shit ton of properties you can modify with set

polarian: Yep. Run a scrub whenever you want.

now to avoid me making myself look bad, during the boot process when the screen resizes

what is this... I am not a boot engineer :P

I assume this is the loading of i915

polarian: Don't even worry about the properties and modifications to datasets and such. Just use it as default.

and the handover from the init (as its just after decrypting the root partition) to the kernel

wait no the kernel is booted

ek: I know :P I am just saying zfs is much bigger

It's probably the kernel module just having fits (I don't know the reason.) You can choose to unload it, rebuild it, and reload it. See if that works.

Is it still doing the same thing after the rollback?

ek: my point is how to explain it

no

Okay. Good.

14.1 is fine... after I manually rolled back

freebsd-update rollback fucked it up somehow

or maybe I fucked it up

If you're talking about reporting the issue, just explain what you're seeing. You don't need to go into crazy detail about anything. What they need from you, if anything, they'll ask for.

anyways I am trying to figure out how to explain it, during the booting of freebsd just after disk init (if you are using geli, the disk being decrypted) the screen resizes and starts with running rc init

services, wifi etc

I'd definitely mention you're using the i915*, though, as it seems to be a culprit for many. Other than that, if you have logs or something, great! If not, it is what it is.

so I assume that switch is where the graphics driver is loaded...?

Yes. Likely.

imma sound like an idiot in this bug report :(

I know none of the correct terms

hey ho, lets be optimistic, at least it isn't the openbsd mailing list I have to report to :(

polarian: "If you disable that graphics driver and try the upgrade again, does it work? If not, try pulling *log* *log* *log* etc..." will likely be their question.

hmmm

it could also be libreboot which is the problem

There's no such thing as a bad bug report. Any feedback is good feedback. This isn't Linux where you'll get flamed for being "stupid" or something. If they can fix or make things easier, even if it's for beginners, they appreciate the info.

I have coursework to do I really dont have time to debug kernel issues

Understood. That's kind of the expectation of a -RELEASE. :)

dmesg.today and dmesg.yesterday only contain logs from 14.1 hmmm

does .today --> .yesterday every boot or something?

oo might be something in messages

ahhh....

I was right, drm is loaded (along with the driver) right after disk decryption

dmesg and messages should actually have info from current boots. They're likely parts of the /var ZFS dataset/vdev/whatever you have setup.

so I think its safe to say i915 might be the issue

polarian: That's my guess based on other reports. You could disable it, retry the upgrade, and then rebuild/re-enable it and see if it works.

... provided you have the time at some point.

welp I see all 14.1 boots, not a single 14.2, rollback rollbacked logs?

but logs have their own volume

ek: well I just have to remove it from rc.conf

well not even remove, just uncomment it

polarian: Right. Just disabling however you're including it in the boot procedure should be sufficient.

welp I got coursework, no time to piss about right now

14.2-R might come right up. However, re-enabling it in 14.2-R might still be a problem (based on what I'm seeing in reports.)

-R meaning?

Not sure if they're rebuilding from ports, using the binary packages, or just re-using the binary stuff they had from 14.1. Hard to say. It's still a really new issue.

polarian: -RELEASE? Is that not what you're upgrading to?

Sorry. Should have been more specific.

ohhh

-R means RELEASE

duh... I am stupid

-S = STABLE, -R = RELEASE, -C = CURRENT, etc...

yeah I realised now

lol

Nah. Not stupid. Confusion happens.

anyways I will look into it tomorrow, thanks for the help though :)

and for curing my stupdiity...

its 2:34am and I got to get some of this coursework done before I pass out :P

Haha. Any time. Hopefully, everything goes well. G'night!

gn

wait just as I said that

ek: lists.freebsd.org/archives/freebsd-…questions/2024-December/006045.html

haha... I am not the only one

Most certainly not! Hrm. I'll have to test this myself when I get a chance.

Obviously sc(4) shouldn't be used anymore in favor of vt(4)

polarian: Are you using any special settings for vt()?

... or, you know, be responsible and check it all out later. :)

ek: actually no I realised its not the same

theirs is failing in early boot

I dont think it ever enters the freebsd bootloader

ugh fuck I got excited there :(

polarian: Yeah. I would think a sc() / vt() issue would happen pretty early. I'm guessing you're just seeing the same i915 driver issues others are seeing.

yes probably

I saw i915 in the dmesg

and thought "score this might be the same thing"

buttt no :(

I think once they sort out their vt() issue, they're going to also see a i915 issue. Haha

yikes...

thats going to be a hell of a night for them

Well, with any luck, they'll both be sorted out soon and contributed.

s/contributed/committed/

ehhh for now 14.1 works :)

the joys of freebsd, you aren't forced to update quickly

14.1 still has what? 6 months left until its dropped?

I'm happy with it at the moment.

It has a while, yes.

You should probably get to that coursework while you still have the chance. :)

EOL on March 31, 2025

Yeppers!

15.0 in 1 yr \o/

I'm looking forward to the 15 release, for sure.

I still haven't done the october update for libreboot as well...

You might wanna get your updates.

I like to compile it from source... I am fussy like that

Better safe than sorry. I compile anything I *REALLY* need to function as expected.

and it means getting my spare laptop out, which I have bubble wrapped up from my trip to Germany... and then I have to get my flasher out take apart laptop and yeah

I *could* internal flash. but I see that as ruining the fun

But, I also mix ports and packages. So, there's that.

im insane so yeah :)

I have wanted to get into porting... and also building some ports

just... no time :(

I'm a port maintainer also. It does take a little time to learn, but it's pretty easy (mostly.) Once you understand it, it's a lot of fun! I would highly suggest it.

I used to package for Linux, not very well, but I get the jist of the process, its just learning the different tooling :)

people.freebsd.org/~rodrigo/fosdem25 <--- anyone going to fosdem bsd track next year?

devroom whatever you wanna call it

15.0 is when pkgbase lands?

lands may be a misnomer; it's been developed in-tree for years now, but hopefully we switch to it in 15.0, yeah

Yep. pkgBase has been available for a while now. I use it all the time.

But, it's a lot more steps than freebsd-update (and not officially documented (as far as I know.)) So, hopefully, that's rectified soon.

l00py I used pkgbase for the first time tonight installing 14.2

documentation is just... meh

enough, though

any known issues with FreeBSD VM networking under Fedora 41 ? 40 was ok, but now both my fbsd 13 and 14 are failing (as does a fresh 14.2 qcow2)

jgh: maybe you should run Fedora under FreeBSD then

it should work

Fedora/kfreebsd

Security question: So as you would do in a VM, you usually would create a non-root user and give it doas/sudo/whatever access and only privilege escalate when needed, is this recommended to do in jails as well or???

Everywhere, yes.

But the consequences of not doing it in jails should be lower than in a VM, so...

thought so

*common sense tbh*

Whats best for virtual interfaces, veth?

for jails that is

well jails have vnet jails, but I mean in general, what would be used if you needed a virtual interface which acts as a phy but sits on top of it

ah I think what I am trying to describe is epair...

A "standard" jail will share its IP stack with the host. A vnet jail will not. Choices :)

I have a few of both on the same host. Sketchy stuff gets its own vnet, trusted stuff does not

hmmm vnet by default wouldn't be a big deal... gives you more control with pf

It does indeed. Also more overhead to manage, of course

hmmm true...

(although jails shouldn't have too many changes to pf anyway, once it's running

I guess its the balance between security and ease of use

you dont want too much security that your device is a pain in the arse to use

I don't really notice it much, really; just the setup takes a bit more time

And if the host is internet facing, there's the IP allocation thing to consider

would epair be the standard interface you would use for virtual ethernet?

as described in the man pages, its an virtual ethernet-like virtual device which is virtually connected to the physical port

the analogy given ^

I'm not going to pretend I'm that much of a low-level BSDer. But the vnet jails I created (using Bastille) all use epairs it seems, yes

lol, good enough :P

thanks for the help

so... rebuilding drm-kmod fixed the issue for me

trying to install 14.2 on a box with 4 nics. put cable in nic "1". ifconfig says no carrier though. any ideas?

config is on bge0

0 vs 1?

nah on the ports it's labelled 1-4 but ifconfig shows them as bge0-3

dell poweredge r720 turdbox

never assume the os sees them the same.... still test them all

ya you're right my bad brb

Fortunately x86 isn't as picky but I've been bit by this a lot in the arm world with u-boot and device tree mismatches

nic 3 is bge0 "nice"

will that be stable across reboots atleast?

Heh that "is" a good q

:-D

I would "think" so but I haven't worked on enough multicard boxes in years

i'll go try 3 reboots

mine are all multi nic single card

(sbcs)

ya mine too. 4 nics on 1 card

1 reboot good, 2 more to go

2 down

now what i'm wondering is if other boxes with the same exact hardware specs will also recognize nic 3 as bge0. that'll be fun to test

Yeah I'm thinking it "should" hold as long as you don't move cards or add another nic...

yep

not sure where it will "align" heheh

we're gonna find out hehe

any opinion on if i should let the OS handle cpu ratcheting (speedstep) or leave it to the bios?

I had to let the bios (I have a laptop that has some oddness)

this is a server

oh.. I just had a thought on your nics... I can't remember "how" but I think there is a way to do this for the wired similar to wifi to bind a specific one...

wlans_rtwn0="wlan0" in rc.conf

that made sure rtwn0 always ended up as wlan0

but I'm not sure if there's the equiv for that wired driver

I have 2 wifi cards in this box

ok 3 reboots all good

excellent

i guess there's no way to bind all 4 nics to 1 alias and have it just use whatever has a cable plugged in?

I'll defer that to someone with a bit more setup experience here

l00py: you can use a tap device to bind them all together

it's a virtual device

man tap ;)

l00py> i guess there's no way to bind all 4 nics to 1 alias and have it just use whatever has a cable plugged in? - lagg(4) failover

lagg yes sorry

ty mzar

FreeBSD netwoking stacky is pretty decent

i'll read on lagg ty

any opinion on using OS to manage cpu ratcheing vs bios? (dell poweredge r720 fwiw)

ratcheing ?

you are likely asking about powerd(8)

but for r720 it has to be first allowed in BIOS

ya i changed bios to performance per watt (OS) mode

powerd can now load

but which is better? let bios manage or OS?

I am always using powerd, BIOS is probably better if you run OS like DOS

there is another power deaemon in ports, so people say it's better

s/so/some

i'll just use powerd tyvm

sysctl dev.cpu | grep freq will tell you more

to monitor CPU temp for intel CPUs load coretemp.ko with kldload(8) and to check issue command: sysctl dev.cpu | grep temp

there is another module for AMD processors

do i care to monitor those or just let powerd work with the profile i give it? adaptive high iirc

I provided it in case you want to check

don't bother with this if you have conditioning in server room

l00py: are you evaluating FreeBSD or switched to it for any purpose ?

using it for everything i can

i just have a cold house hehe

ty though i made notes on those

mzar got linux servers at work i'm converting over to freebsd

have all freebsd at home

i did a fresh 14.2 install, got network, but can't get dns to resolve. i used my home router ip as dns server. my laptop (linux atm) is set up to use home router ip as dns server and it can resolve fine. any tips to debug why it's not working?

check resolv.conf, check connectivity and finally you can use tcpdump(1)

resolve.conf has 127.0.0.1, i selected local_unbound in install

btw the ufs setup in bsdinstall still has mbr as the default instead of gpt even tho gpt is what's preferred so why's that?

thats odd... mine setup as gpt

and I installed just 2 days ago using FreeBSD-14.2-RELEASE-amd64-dvd1.iso

gpart show

=> 40 976773088 nda0 GPT (466G)

40 532480 1 efi (260M)

532520 1024 2 freebsd-boot (512K)

..

was it the default selection in the partitioning menu tho?

yes

I took all defaults

i'm using disc1 and mbr was default selection

for ufs

is your "machine" uefi capabb;e?

er capable

(or enabled in bios)

It could still be turned on legancy/mbr

er legacy

Some bios are that bad broken

i think uefi is an option but i'm in bios mode

how can i recompile kernel with realtek lan drivers my freebsd 14.2 doesn not recognize LAN it's a realtek 2.5GHz i want to recompile it with realtek lan support?

and then to restart the network service

l00py: unless I am wrong.. I think if your bios is set to mbr mode the install default to mbr... don't quote me though

brutex: what bus? (pci/usb?) and do you have the major:minor if usb?

Tenkawa: it's a integrated LAN into motherboard

Ok that would be a pci then... what type of motherboard?

PCI

no.. vendor & model

without that no way to tell what driver (realtek has numerous drivers and models of devices.... only certain ones work in freebsd... )

you can boot from GPT in BIOS mode almost all the OSes

Tenkawa:Realtek RTL8125 Gaming 2.5GbE Family Ethernet Controller PCI Connection Speed 1 Gbps MTU 1500 bytes this is - Intel Alder Point-S PCH - CNVi: WiFi Controller PCI wifi adapter with antena which also cannto be found

Gaming 2.5GbE Family Ethernet Controller - that's interesting

mzar: of course... but many bios's support hybrid booting... which could be why it defaulted his partition table to default to mbr instead of uefi like mine

or it could be turned completely off

I never trust the installer, just do it by hands

Nod.. I doublecheck the installer but its been good so far

brutex: man if_re

you need to enable that driver for the ethernet card

hmm.. I dont see 8125 in its manpage though

mzar

there is port net/realtek-re-kmod

oops

yeah I was about to ask you about net/realtek-re-kmod.

it will support Gaming Family

Ok good

brutex: welcome to FreeBSD family

mzar: specialty hardware in general.... I got an HP thats annoying like this for Linux too...

Tenkawa: i don't understand you how can i enable i-net

brutex: please issue command "pkg install net/realtek-re-kmod" and follow

Tenkawa: i need if_rl_load="YES in rc.conf?

nope

mzar: without net i cannot run pkg

then?

read the pkg message again

what is point of meaning

or read it here freshports.org/net/realtek-re-kmod

ok

21:43 < brutex> mzar: without net i cannot run pkg - that's chicken - egg problem, you have to bring this package somehow and install it

download it to a msdos formatted flash drive

you can then mount that on the machine and install it .. or is this your only os?

ok

thank you

if installer has a prob it should be fixed not just ignored tho no? i mean we want the noob onramp to be smooth otherwise linux will keep eating the cake and linux sucks

not saying it has a prob, just in principle

l00py: which problem?

just said there isn't 1 and that it's in principle

Ah ok

kldstat didn't show cryptodev, then i ran kldload cryptodev and it showed it. so does that mean it's supported or is it possible to load a kernel module that has no benefit?

i guess what i mean is how do i know if my hardware supports that kernel module?

without inet how can i use pkg

to install realtek-re-kmod

pkg install net/realtek-re-kmod

it's problem with DRM mod

ah, sorry

then how can i enable inet

to recompile kernel with rlt drivers?

make buildkernel installkernel KERNFAST=1 && reboot ?

is there even any point to loading the cryptodev kernel module anymore?

brutex: pkg add name.of.the.file

brutex left, but file for him is here pkg.freebsd.org/FreeBSD:14:amd64/la…t/All/realtek-re-kmod-1100.00_1.pkg

in a scripted bsdinstall with ufs, is SCRIPTEDPART="mfid0" not enough? bsdinstall prompted me if i wanted to use mfid0 or mfid1 but why prompt me when i already set the env var in the preamble?