'pkgconf' damn that's an annoying name, in every other os i've used it's either pkg-config or pkgconfig

tobias: did somebody point you at /boot/loader.conf?

add 'boot_serial': '"YES"', 'comconsole_speed': '"115200"', and 'console': '"comconsole,

I do think its odd that its not a default. Especially given that most remote management gives you access to one.

but then you know if your here you're sysadmining like its 1999

:)

feurig: I'm pretty, and sure you can use packer / ansible / puppet / etc… to write such a loader.conf these days

I actually had to look up my bootstrap freebsd ansible script for that.

wait was that a damned ai?

That seems like the kind of junk gpt would say.

welcome to the ai spam appocolypse.

but meena: you are correct the appropriate syntax would be boot_serial=yes

feurig: i have it with ="YES"

i think loader is pretty lax about the actual spelling

How do I report an error in a specific MAN page and a separate error in the FreeBSD Handbook?

Title: FreeBSD Bugzilla Main Page

Thanks

You can also submit the fix there if you have one.

HardenedBSD's version of the ports tree cannot build librsvg2-rust for HardenedBSD's version of 13-S. Is the mainline ports tree and mainline FreeBSD 13-S currently suffering the same plight, or is its librsvg2-rust building fine?

("Idk mate, fire up mainline on your own PC and figure it out" is a valid answer, though one that will be met with grumbling.)

pkg-status.freebsd.org/builds?type=package - you can check there. There seems to be a few failures across the board.

how do you usually automatically add ssh keys to the agent? i am starting the agent in xinitrc. should i just do `ssh-add...` after that?

trev: In .ssh/config, where you map keys to hosts, you can add 'AddKeysToAgent yes' and when after first use they're remembered.

ox1eef_: nice, thank you

No worries.

There are also Control{Master,Path,Persist} options

under Host ?

Yes

hello folks, upgrade to 13.2-RELEASE done. Is it me or not? It seem faster.

trev: you can also put it before any host to make it available for all the hosts.

A small portion of ~/.ssh/config : termbin.com/1fqp

Sorry, wrong channe;

I'm trying to have a freebsd server of mine compile a new rpi64 image for my rpi4 which uses ZFS for the rootfs. I've found this article: wiki.freebsd.org/arm/Build_image_us…ing_release_building_infrastructure

Title: arm/Build_image_using_release_building_infrastructure - FreeBSD Wiki

now i'm looking at /usr/src/release/arm64/RPI.conf, but i have so far not found where the file system setup magic for the SD card happens

Hi

syncthing package on latest freebsd has known bug, which produces a lot of sync conflicts

i sent a mail to the maintainer but he seems not available

does anybody have the ability to upgrade the package ?

13.1 latest

the problematic syncthing version is 1.22.0

The way to sanity is to file a bug... freebsd.org/support/bugreports

Title: Bug Reports | The FreeBSD Project

ok, thanks

if you send an email to an unresponsive maintainer, you have sent an email to an unresponsive maintainer. if you create a bug report at least there's something you and other people can refer to

specifically, the maintiner timeout only starts from when the bug is filed, because the bug being filed is how the project keeps track of maintainer timeouts.

Hey all, running 12.4, uptodate, I'm seeing regular all-core cpu spikes on a process called zfskern. I don't use zfs; what can I do to make this go away?

zfskern is the light weight process/thread in the kernel, but it shouldn't be present unless you've loaded the zfs kernel module, so check if you've done that with `kldstat -vv`, unload with kldunload(8), and remove it from wherever it's being loaded.

I'm definitely not loading it explicitly (i.e., in /boot/loader.conf)

I can blacklist zfs.ko? but I wonder why I'd need to go to such measures

you don't have zfs_enable="YES" in your /etc/rc.conf for some reason?

No, I just discovered something mad.. MAD

I am using 'btop'

When I start btop it actually (indirectly? why? no zfs used! only ufs!) loads the zfs module

That sounds highly strange to me, why and how could btop decide to just load a kernel module?!

just use top in the freebsd base system

it's much better

I like btop. Should I file a bug for this?

not with freebsd.

obviously.

I'll have a look upstream

a utility like a top modifying kernel modules o.0 thats not great

it's actively hostile.

it either means it requires suid or to be run as root, both of which shouldn't be necessary

the only thing that should require root for something like top is renicing it to -20 in situations where a system is so bogged down, even rendering top is gonna be a huge task

(which is what top -q does)

yea something like that just willingly jamming kernel modules in could easily be used for malicious purposes

do not want

Title: [BUG] btop loading zfs kernel module on FreeBSD, even when no filesystems with ZFS exist · Issue #515 · aristocratos/btop · GitHub

515 – Info command has no tutorial bugs.freebsd.org/bugzilla/show_bug.cgi?id=515

I suspect he's doing some hackery to be able to show details about zfs filesystems. Agreed, should not just load kernel modules, that's madness

can you try with zpool list -H -o name github.com/aristocratos/btop/blob/m…n/src/freebsd/btop_collect.cpp#L539

Title: btop/btop_collect.cpp at main · aristocratos/btop · GitHub

rubin55: you mean the details included in freebsds top, without the need to load kernel modules if they're not present?

it also doesn't look like btop understands freebsds vm classes, so you're losing out on a bunch of fairly relevant information

What are "vm classes"?

Title: Memory - FreeBSD Wiki

maybe bugs.freebsd.org/bugzilla/show_bug.cgi?id=256147 but there it seems to be zpool itself using one cpu core fully

Title: 256147 – zpool list -P causes high cpu load while using ufs

but one would hope that zpool list exits when zfs.ko is not loaded and not load it itself

I don't have a system without ZFS to test on, but I can't think how zpool list can load the module if it's not run as a privileged user.

I'm trying to script unmounting all of the file systems I have mounted to /media/$LOGNAME. But somehow, umount(8) isn't able to work with that: paste.debian.net/plainh/18cbdef6

nimaje: no, it (zpool list) causes zfs.ko to be loaded, also on a system without zfs mountpoints and/or modules_blacklist=zfs in loader.conf

I noticed mainly because at load, zfskern seems to spike for about 10? seconds on all cores (i.e., 100% cpu usage)

ok, then that is a freebsd bug (I think only zpool import has a valid reason to load zfs.ko, but maybe more in any case not list and get), but how can it even do that? it needs a privileged user for that as debdrup already pointed out

hey, is there an alternative to synth or poudriere in order to compile multiple packags?

(good alternative)

nimaje: well, in my case, I'm running btop as root, which in turn seems to unconditionally execute a zpool list (for which I filed a bug upstream at the btop repository). In such a case, zpool list would run with enough privileges to do the kldload. I would just not expect it to do so I guess (I would expect it to error out because zfs interface is not available in the kernel)

I consider it a bug that it does and that it has enough privileges doesn't really matter

Okay, I've found the problem: `umount /dev/da0p1 /dev/da1s1 ` breaks because of a trailing space.

That's not supposed to happen, is it?

With only one argument supplied, umount does ignore trailing whitespace.

nimaje: shall I file a bug (bugs.freebsd.org)?

yes

Oh, wait, it just didn't work again, and no trailing space involved…

umount probably doesn't care, but the shell may be doing something wierd with it.

Okay, let me use /bin/sh.

'Course. . . I'm not sure why you're running umount against a device. . . normally you'd run it against a mount point.

I'm doing that because the names are predictable.

And your mount point isn't?

In the sense that they are not freely choosable, not user input that could contain spaces, newlines, etc.

Why not script something that grabs the mountpoint?

Because the moinpoint is not a constant.

Exactly my point.

My script works like this: qmnt <device> <label>. And things will be mounted to /media/$LOGNAME/<label>.

That label could be anything.

mount_point = $(mount -p | grep <device> | awk '{print $2}')

I don't want the script to break because of a weird directory name.

is there a way to search using pkg which package contains a file, not just the currently installed packages, the use case is that I don't know which package to install to fix a dependency

CrtxReavr: Well, yeah, the script does grab all the lines from `mount` that have `/media/$LOGNAME/` dir as their mountpoint prefix.

But it does not rely on those names for unmounting because I see no way to safeguard that against weird directory names.

neirac, yes. . . though I don't remember the syntax offhand.

neirac: pkg provides (you have to install pkg-provides first, no idea if it needs any more setup)

nimaje awesome thanks!

Okay, /bin/sh has the same problem.

msiism: and if your mountpoints (or even the label) contains a newline?

what shell did you try?

Yeah, the way I do it, this won't be a problem.

FreeBSD's /bin/sh and Bash.

newlines will likely break your grep and cut

Well, the mountpoint prefix that I grep for is constant.

And then I cut the first field from the line, which is always the device name followed by a space.

(mount doesn't seem to have libxo support :()

And there are no funny characters in device names.

neirac,

$ pkg which /usr/local/bin/screen

/usr/local/bin/screen was installed by package screen-4.9.0_6

CrtxReavr: not installed packages

pkg which `which binary` is always a fun command to run

nimaje: bugs.freebsd.org/bugzilla/show_bug.cgi?id=270789

Title: 270789 – zpool list loads zfs.ko, even on systems that don't use zfs

Wroks for non-executables too:

$ pkg which /usr/local/share/cows/jkh.cow

/usr/local/share/cows/jkh.cow was installed by package cowsay-3.04_2

Though, the actual question now is: Why is `umount` not able to handle more than one device name now and then?

msiism, then iterate over a list of devices. . . this isn't that hard.

rubin55: neither irc nor bugzilla are twitter

you mean because of the @?

CrtxReavr: You mean like `for t in $target; do…`?

yes, put it at the end if anywhere

lol sure

for t in thing ; do umount $t ; done

Right.

You just gotta figure out where you're getting "thing"

I already got that sorted.

Still, I wonder what causes this behavior in umount.

I'm pretty sure if you read the manpage, you'll see it was never designed to umount more than one thing at once.

Sometimes unix gives you lots of rope to hang yourself with.

Sometimes, it tires to keep a little sanity.

Well, the synopsis says otherwise. But reading the whole thing might offer a different picture.

Maybe look at the `umount -F` option.

Build yourself a /tmp/usr.fstab file, then run `umount -F /tmp/usr.fstab`

Okay, I will.

mount -p | grep whatevs > /tmp/usr.fstab

(mount -p produces syntax in /etc/fstab format.)

oh

Well, I see what you mean now.

I guess, I'll just go with the for loop, though.

CrtxReavr: for me pkg which says "… was not found in the database" for stuff that is not installed (did you maybe have cowsay-3.04_2 installed at some point in time?)

You might even get away with some fancy shit like: mount -p | grep whatevs | umount -F -

nimaje, yes. . . it's only going to show you info on what pkg installed. . . you can't expect it to tell you things about random shit you installed by hand.

And yes, I do have cowsay installed. . .

Don't make me start piping cowsay into the channel!

/ex figlet debdrup? | cowsay -n -f jkh

ah, you wrote non-executables there, the question was still about stuff that isn't installed

I put on my wizards hat and robe.

I really don't see why you had to hilight me there.

heh

debdrup, a classic!

(order's reversed, but whatver)

How is SysV banner different from BSD banner, though?

Nope, not clicking that link.

Just the text of that thing you quoted.

CrtxReavr: Canonical link, will put people off less: bash.org/?104383

Title: QDB: Quote #104383

mason, that's just a small portion of it though.

Oh? Interesting. It's all I've ever seen.

I used to hang in an EFnet channel with the bash.org guys. . . in its heyday, they auto-rejected about 80% of submissions at random, 'cause there was no way they could field them all.

Wow, there's a bunch more. Thanks for sharing it. Heh.

When that first came out, it was funny AF.

Gah, bits of it are horrid. I think I see why I've only ever seen the subset. :P

You take that bake!

back

bake

a mojibake?

Thunar cannor remember list/icon/detail layout display settings. Got message: Cannot spawn a message bus without a machine-id: Unable to load /var/local/lib/dbus/machine-id or /etc/machine-id: Failed to open file “/var/local/lib/dbus/machine-id”: No such file or directory

nimaje, I just grocked what you meant by not installed. ..

If you have a ports tree installed/downloaded, you an do like: for i in /usr/ports/*/*/pkg-plist ; do grep -H screen $i ; done

'Course, if you have a ports tree installed, you can also do: cd /usr/ports && make search key=foo

or make search name=bar

CrtxReavr: some stuff isn't listed in a pkg-plist file in the ports tree and sometimes that stuff gets even generated automatically (e.g. (most?) py- ports)

hi, i am having some issues with xorg

nimaje, like which, for instance?

it's refusing to start with nv/modesetting drivers (the other drivers work but they don't have the correct screen resolution and i don't have the ablity to change that)

rani, been a while since I ran a FreeBSD workstation/laptop.

I'd use the nvidia binary blob drivers, kldload those. . . then run something like: sudo xorg -configure > /tmp/xorg.conf

stuff listed in PLIST_FILES, e.g. math/glucose has PLIST_FILES= bin/glucose bin/glucose-syrup

Usually had to hand edit that file a little but, then move it into /etc/X11/xorg.conf

Then test with startx

Once satisfied, I'd finish setting up login manager, window manager/environment, etc.

the proprietary nvidia driver requires me to install an old version of xorg since my gPU is ancient

the main issue i'm having I think is that nv states that a kernel module is already claiming my GPU, although I don't see anything that would be doing this

pastebin your /var/run/dmesg.boot and the output of pciconf -lv

Oh - no gui.

alias tb='ncat termbin.com 9999'

Then: cat /var/run/dmesg.boot | tb

pciconf -lv | tb

Hi all

(Yah, you'll need to have ncat installed.

)

is it possible to just download/fetch a package without installing it ?

nimaje, I dable in python myself. . . and very much not a fan of how we handle py-* ports/packages. ..

We should better leverage pip

I have ideas on how to handle that. . . I need to bring them up up with some ports developers I know.

last1, sure.

this is dmesg: 0x0.st/HXnF.txt

this is pciconf: 0x0.st/HXnC.txt

I am trying to code a simple utility that will get the CWD of a process, (I dont want to use any high level lib like procstat or sysctl) so my plan is: get struct kinfo_proc *processes from kvm_getprocs

get struct kinfo_file by using kinfo_getfile(PID, kfiles) which returns all open file descriptors

check if kinfo_file.kd_fd == KF_FD_TYPE_CWD

if yes then read kinfo_file -> kf_un -> kf_file -> kf_file_fileid, to get INODE number

Convert INODE number into file (somehow)

. Is this the "optimal" way to do it?

rani, kldstat

ElectricJozin, you might want to ask that kinda question in #bsdcode on EFnet.

CrtxReavr, Ok thanks

Problem is I cant connec t to EFnet for some reason

ElectricJozin, actually - you could try #freebsd-dev on here too I suppose.

Seems more of the oldskool devs are on EFnet th ough.

EfNET has a invalid ssl certificate, thats why it wasnt working

last1, I think what you're looking for is: pkg fetch <package name>

Um. ..

irc.efnet.org:6667 shoudl work in clear text.

this xorg thing is very annoying

modesetting doesn't work either but that's because no drm support

i have no idea why nv is mad though

"mode setting" isnt' a driver.

What driver have you tried?

xcfb & vesa work, modesetting & nv do not

scfb*

Try one of these: nvidia-driver-304/ nvidia-driver-340/ nvidia-driver-390/ nvidia-driver-470/

the proprietary driver (304) requires an old version of xorg

i tried compiling from ports but it failed at the patching phase and i didn't really feel like manually patching xorg

Try a newer one then.

the newer ones aren't compatible with my GPU

Stickwith VESA then.

ugh vesa resolution caps out too low

xorg sucks

heheh

That's not xorg. . . that's VESA.

Try your scfb (whatever the fuck that is) then. .

then try xorg -configure

CrtxReavr, Do you know of any other populous BSD IRC's?

I think you've found 'em.

You're aware that lsof has been ported?

CrtxReavr, yes and I used their source code

But this is a first for me

I couldnt really find where some of the definitions were coming from and it all just got kind of confusing

and if you meant: Why dont I use lsof, is because where is the fun in that

There's also checkrestart which I think does a better job.

ElectricJozin, the upstream lsof code, or the post-patch port code?

Hmm - maybe we dont' patch it that much.

yep the upstream contains a lib->dialect->freebsd directory

where most of the important stuff happens once the function the function process_file_descriptors is called

bugs.freebsd.org/bugzilla/show_bug.cgi?id=270789 this is a cool bug

Title: 270789 – zpool list loads zfs.ko, even on systems that don't use zfs

Watch out for ZFS "block_cloning" option in -CURRENT: lists.freebsd.org/archives/freebsd-current/2023-April/003440.html

Title: Re: CURRENT: Panic VERIFY(!zil_replaying(zilog, tx)) failed (and crashing)

ah, yeah, for zpool create loading zfs.ko is fine too

Where should I push ports patches?

Or should I email the maintainer with the diff patch

nimaje: I see no issue github.com/openzfs/zfs/issues here yet

Title: Issues · openzfs/zfs · GitHub

ElectricJozin: open a bugzilla bug with origin/port:

ElectricJozin, bugs.freebsd.org

its not a bug its a simple change

Ok, then

You can still use bugzilla.

ElectricJozin, it's not only for bugs, but the name does suggest otherwise. It's for tracking changes and some of them are bugs :o)

if a maintainer goes missing, we won't know you've been mailing them diffs. but bugzilla will

docs.freebsd.org/en/articles/problem-reports has a good description of how things work when they work best.

Title: Writing FreeBSD Problem Reports | FreeBSD Documentation Portal

ok and for example if I were to patch a base system dependency like sysctl

Under the system there is no way to highlight the command at question

soo.... git?

bugs.freebsd.org/bugzilla/describec…omponents.cgi?product=Base%20System describes the components

Title: Components for Base System

patches to sysctl would go in the bin component

Right ok

Thanks

as a follow-up to the thing parv posted, openzfs/zfs #14739

Title: Fix data corruption when cloning embedded blocks. by pjd · Pull Request #14739 · openzfs/zfs · GitHub

exciting stuff here, bugs.freebsd.org/bugzilla/show_bug.cgi?id=270796

Title: 270796 – [exp-run] Enable -Warray-parameter for clang

debdrup: nice! i was wondering how long that's gonna take…

Brand new url parsing code now in FreshPorts. Please report any oddities. bsd.network/web/@dvl/110187747070886865

Title: Dan Langille: "New URL parsing code into FreshPorts soon. Major …" - BSD Network