?

<vkarlsen> I went all in and installed FreeBSD on the cat5 cables

this isn't possible

you say that, but

anyone with access to linkedin can mirror a copy of the article "freebsd pot vs docker: a technical comparison" pls?

polyex: englepine.com/freebsd-pot.pdf

idk the difference betweens pots and jails polyex

pots is something new I think

ty!

greets

o/

So, I recently replaced the graphics card on my laptop

and now I have issues waking up from suspend to ram

the old card used nvidia-driver-340

the new one is an nvidia quadro M2200

which uses nvidia-driver

the problem is: when I wake up from suspend to ram, the screen is black

it stops being black when I restart the X server blindly

what could the problem be?

Obvious less than good workaround: Exit X11 Windows before sleep :-|

( Personally all of my work of any significance happens in a X11 session, I would then just shutdown the damn thing in that case )

parv: doesn't change a thing

it's starting the X server that makes the graphics work again

weridly enough

FUZxxl, What was your context for "doesn't change a thing" as I am confused?

it doesn't change anything about the screen being blank (off) on wakeup

FUZxxl, Thanks

awesome-client

awesome.restart()

ah wrong window

Make sure ACPI_PM=on in x11/nvidia-driver.

Also nvidia-modeset.ko instead of nvidia.ko (I think).

_xor: it is on

I have nvidia-drm loaded, which pulls nvidia-modeset

without nvidia-drm, wakeup doesn't seem to work at all

nvidia kmods, for whatever reason, only properly work for me when loaded from rc.conf, not loader.conf. I'm loading nvidia-modset.ko and nvidia-drm.ko. The latter is required for wayland (x11-wm/hyprland).

oh interesting

let me try that

brb rebooting

hm...

that didn't really make a big difference

the symptom is a bit different now:

instead of staying off,

the screen just shows garbage on wakeup

^ this is what the screen looks like when I swich to a console

This is helpful: tecmint.com/view-colored-man-pages-in-linux

Title: How to View or Display Colored Man Pages in Linux

(Works in FreeBSD too.)

CrtxReavr: I just use textproc/bat + MANPAGER="/bin/sh -c 'col -bx | bat -l man -p'"

Though LESS_TERMCAP is nice to use without dependencies.

Does that colorize random switches burried in paragraphfs though?

Dunno, lets see...

Yes, appears so.

I use textproc/bat for PAGER and MANPAGER, as well as just a general pager in pipes as well.

databases/pspg is pretty awesome and integrates well with databases/py-pgcli, which is my primary means for CLI db work.

Thank you, CrtxReavr, for the tecmint URL

you can give readwrite diskspace to a jail either directly mounted or through a network fs protocol. is that right? tradeoffs?

Do not see the point of adding an indirection layer of NFS if jail would be on the same host

polyex: I use nullfs to mount host:/mnt/${NAME} into the jail (e.g. /var/jails/myjail1/usr/local/www). host:/etc/fstab contains NFSv4 entries to mount shares under /mnt/..., and they allow sub-tree mounts. That fstab configuration is shared across my jail hosts.

So essentially, the jails are insulated from having to deal with network mounts. Hosts have a consistent view of the network mounts and whatever a jail needs is mounted into it at launch time.

what's nullfs?

Basically a pass-through dummy file system type that lets you mount one directory to another (e.g. /mnt/data to /usr/local/data). Links don't work because jails are chroot'ed.

ok

if storage is on another server from the jail so they gotta mount through a network protocol, how much crappier would that be?

than having a local /dev/foo to a local dir

Don't know what you mean by "local /dev/foo to a local dir", but you can network mount directly within a given jail using mount.fstab in the jail config, but running it through nullfs from the host made it easier to manage and handle failure conditions.

ok

In my case, I did originally created NFSv4 mounts directly in the jail, but there were times when they weren't unmounted for some reason when the jail was shutdown, so stale mounts/locks/whatever would build up over time.

mount just means make a device available through a dir path?

ahh

ya that flakiness sucks

crap

don't really want actual storage going on on the jail host

Also, if there was a temporary network connection issue, then it was a toss-up which jails handled it fine and which didn't. Whereas with host-nullfs mounts, either all the jails stalled/restarted or none of them did. It was more predictable.

does it work better if the jail host has the network mounts then gives local mounts to jails? or that nonsense?

Well, to be fair, the flakiness could have been something I did, I don't want to give the impression that you would run into it too.

would prolly add a lot of i/o overhead being network based

But there wasn't going to be a case where the host could have network issues but the jail wouldn't, so if there was going to be a network failure, it would happen to all containers running on the given host.

latency or whatever

That's what I mean by using nullfs, having the host mount NFSv4 directories and then locally sharing those as necessary with jails via nullfs.

I didn't notice any real latency, but I didn't benchmark it either. I'd imagine it's minimal.

oh so still doing nfs, just moved it from jails to host

so you're using network storage for jails too, not storage on the jail host itself

In fact, now that I think about it, it could be more efficient because there's a set of NFSv4 clients on the host instead of on a per-jail basis (though I'm not sure if mount.fstab for jail.conf polls connections or whatever).

*single set

ya network storage for this kinda stuff is so good. hate state being scattered around servers

Yes, for persistent data I use network storage. Either NFSv4 or S3. For ephemeral data that doesn't need to be saved or be available across hosts, I either just let it live within the jail (i.e. temporary files) or create a zfs dataset and make it available to the jail via zfs:jailed=on (i.e. cache data that doesn't need to be available across

hosts, but would be nice to survive reboots).

so where does the nullfs route come in?

/mnt/services/myapp /var/jail/myapp0/mnt/data nullfs rw 0 0

Example entry in mount.fstab for the jail configuration file.

does that work with the zfs dataset somehow?

Yes.

Check out `man 8 zfs-jail`, it's really short.

ty

Basically you can set the jailed property on a ZFS dataset and then call `zfs jail myjail` to give control of it to the jail.

then use the nullfs line to make use of it in the jail

really cool

so a quota on that zfs dataset is what would let you put a disk quota on a jail ya?

You don't need to use nullfs if you're giving the jail a ZFS dataset using jailed=on and `zfs jail ...`, you use nullfs if you want to mount a directory from the host into the jail (regardless of the file system).

Yes, anything you can do with a ZFS dataset on the host, you can do on the jailed ZFS dataset in the jail (the jail basically "owns" that ZFS dataset when you do that).

ahh

so i need to start using zfs on freebsd servers with only 1 disk

get out of the mindset that it's only for multiple disk boxes

You don't have to, but you can.

There's also ZFS volumes + iSCSI, but that's whole different beast.

i'm gonna start using zfs for freebsd bhyve vms instead of ufs

just use it everywhere except usb flash drives

any downside?

just change the recordsize for the vm-image-zfs, use raw-images and use the nvme diskdriver if possible

Title: FreeBSD Bhyve Virtualization | 𝚟𝚎𝚛𝚖𝚊𝚍𝚎𝚗

I have 64k recorsize for the vm's -> compressratio 2.06x (with zstd)

how much overhead in that compression?

polyex, What do you mean, of CPU?

a little bit more the lz4 but i dont really see a big differece cpu-wise

if it need to much power set it back to lz4 or zstd Alan made a pretty good presentation about it

hello

can anyone explain me why rsync in script mess-up with the user credentials ?

I have been using zstd-{9,15} on most of the datasets; use {lz4,zstd-[2-4]} where "LUSED" amount remains lower than "USED"

Title: dpaste/aWkT (Bash)

the rsunc tries to make connection by using wrong credentials, not the one is running the command

nerozero, Use: "user@host":"/remote/dir" -- do not quote the colon for rsync

here is the rsync output opening connection using: ssh -l "'user" server.host.org rsync --server --sender

parv

hm ok

can docker images be run on freebsd?

Perhaps in some Linux VM? Certainly not directly on FreeBSD

parv doesn't seems to work...

the rsync argument parsing issue

nerozero, What were the exact rsync command & the issue?

see the script I posted earlier

the e-mail is the user

nerozero, Paste the output of "log.rsync" file (feel free to sanitize user & host): { set -x ; rsync <whatever> ; set +x ; } >| log.rsync 2>&1

lets make question simpler

Actually run with "-vv --itemize-changes"

if I execute command directly in the script: /usr/local/bin/rsync -avvvzn --chown=www:www "user⊙so":"/path/to/src" "/path/to/dst"

it runs

if this command is in the variable and I execute variable it does not

the echo of the variable: /usr/local/bin/rsync -avvvzn --chown=www:www "user⊙so":"/path/to/src" "/path/to/dst"

You are having quoting issues🤷🏽‍♂️

exactly

and I do not know where

<parv> nerozero, Paste the output of "log.rsync" file (feel free to sanitize user & host): { set -x ; rsync <whatever> ; set +x ; } >| log.rsync 2>&1

yah, found it,

extra quote .... \"

parv, thanks for quick response and sorry for stupid question....

nerozero, Instead of \" you can use single quotes inside the double quotes (the variable will expand). Install "hs-ShellCheck" to find issues with a shell script.

thanks a lot

"shellcheck" would be command available from "hs-ShellCheck" package

hi all

if i made a daemon that modified the system like maybe added users or something, it wouldn't make sense to run it in a jail right? because it actually needs to change the host

anyone knows what happend to my certbot? dumpinen.com/Xdk5ec4OqI3

installed with pkg

Is this fallout from the py-cryptography bump?

Which version of py39-cryptography do you have?

It was recently bumped from 3.4.8 to 41.0.3

py39-cryptography-3.4.8_1,1

ok

ill wait some more then

thanks

When did you last do pkg upgrade?

today

nothing new

im using 14-beta1 if that matters

polyex: yeah, we unfortunately have no concept of namespaces like Linux, so jailing System (host) daemons makes no sense, if all you wanna do is… contain them

using /${ABI}/latest"

sopparus: bsd.to/De62

Title: dpaste/De62 (Bash)

sopparus: It should be in the latest repo

hi! I know poudriere uses jails to perform builds. However, I need to use it in a server with some constraints and I need to containarize poudriere. Can I run poudriere inside a jail, I mean, create a jail and install Poudriere inside it, or is it better to use Bhyve?

vkarlsen: did pkg update -f and there it was

but got another error, dumpinen.com/IYM6BX5BnBt

Yeah, you'll need to upgrade py39-openssl too

py39-openssl-23.2.0,1

?

Yes

i have that one installed

Which version of py39-certbot do you have?

py39-certbot-2.6.0,1

Hm, that combo works for me

ok, odd

but you are on 13.2?

couse it works for me on my 13.2 machine

I am on 13.2

perhaps bugs.freebsd.org/bugzilla/show_bug.cgi?id=273813

Title: 273813 – ports-mgmt/fastest_pkg: RuntimeError: OpenSSL 3.0's legacy provider failed to load

Does the CRYPTOGRAPHY_OPENSSL_NO_LEGACY workaround work for you?

how do I set it?

no export.. :)

Depends on your shell

its tsch

env CRYPTOGRAPHY_OPENSSL_NO_LEGACY=1 certbot (if I remember correctly)

yes that did it, thanks!

:)

Hrm. So, barrierc(1) no longer works under FreeBSD, installing from latest ports (pkg) and connecting from the Debian workstation that's been the server for years.

I kind of feel like the answer is to ditch it and find some other, simpler equivalent. Maybe a ground-up minimal new effort. Maybe input-leap.

there is no reason it does not work

Well. It used to mostly work, but now it's stopped being a lottery, and it breaks every single run.

last I checked it was working (this was a couple of month ago, there was no strong changes on freebsd side

My data is more recent then.

it still use the synergy protocol

and I do run other software that supports the synergy protocol

barrier server on linux and wayland synergy client

It's conceivably an issue only on FreeBSD then. It's a failure condition that used to assert itself periodically, but now it's utterly consistent.

without data other than it does not work

it is hard to figure out the issue

Symptoms I've observed: modifiers are asserted and never released, or similar. Need to explore that. The most immediate issue is that ...

I'm reporting live, and gathering data.

any error message from barrier itself,

The most immediate issue that comes up is that when I type a character, it's immediately followed by a newline, and then more, infinitely, such that I can't get more than a character in before "return is hit". I see single characters that scroll off the top.

Then "meta" appears to lock, so my meta-alt, which brings up dmenu, never goes away, and I can't actually escape it. This time I've killed barrierc and dmenu via ssh and I'm exploring the current state.

run barrier from the terminal and what error does it spit

I only run it from a terminal. No error whatsoever.

X is hosed up fiercely, but I'll snag a photograph.

bapt: no error: 0x0.st/HO-R.jpg

barrierc -l meh.log

and check what is in the meh.loh

Will it show more verbosity than what's in the photograph?

you photo shows a home built version

not the one from pkg

bapt: You are incorrect there.

barrierc -d DEBUG

barrierc -d DEBUG -l meh.log

Will snag that. Rebooting so I can.

The script shown sets up an SSH tunnel, because I trust Synergy's homebrew crypto not nearly so far as I can throw it, and then launches the client, FWIW.

failed again, will share log

bapt: It doesn't seem to realize anything is wrong: termbin.com/o9w5

The server likewise shows nothing out of the ordinary.

Client running FreeBSD 13.2, I'd see this happen every third run or so, but if it didn't break early it wouldn't tend to break ever. Now, after installing 14BETA1 on the laptop, the issue happens every boot, making it unusable.

X on both sides here - no Wayland - FWIW.

Maybe I'll see if throwing in "xkbset nullify lock" in a loop helps, although the repeated injection of returns doesn't feel like a lock.

bapt: Any other data you can think of off the top of your head that might be useful?

This is the sort of thing that slaps some reality back into me when I try to switch to FreeBSD as a desktop.

I just installed i3 and barrierc on this laptop, and it just work fine

Works for me! Closed! Good enough.

just an information to tell you it can work, so their must be something wrong on your side

xdm, openbox here.

bapt: MUST BE

can you netcat the socket ?

What it is is that I installed FreeBSD on the laptop. Didn't present as an issue with Debian on it.

(disabling the crypto on the server side of course

)

don't draw conclusion too fast

until you know what is making it fail for you

I don't use the built-in crypto, I just run it over an ssh socket, which I've done since it was a new project and still called Synergy.

Right, this is why I'm bringing it up here, rather than giving up in disgust.

can you netcat the said socket?

what is the server saying in its logs ?

bapt: The server just notes connections, clipboard updates. Nothing at all indicating anything wrong.

Half a sec, rebooting again.

thnen sorry I am clueless to help you debug that

What would I get from the netcat? I haven't tried to get a pcap yet.

Anyway, Synergy has gotten flakier over the years. I'm not sure if input-leap is any vast improvement. I suspect a simple, back to basics solution would be best, but I haven't dug into writing one yet.

Which see, github.com/input-leap/input-leap, a fork of Barrier as Barrier was a fork of Synergy.

Interesting. So, running "xkbset nullify lock" in a loop once a second has worked around the modifier issue, and firing up a new urxvt and then closing it for some reason is letting me type normally again.

bapt: You have any interest in more data about this? A pcap or something? The issue is completely reproduceable. In fact, without the xkbset loop I can't escape it.

nope ;)

I was just trying to help

If not, I'm going to trash the install and get back to my actual paid work, from which this is a pretty unpleasant diversion.

kk

I do use waynergy personnaly ;)

because wayland

Yeah, I have no interestin in Wayland. I use remote X somewhat often, and a protocol that doesn't even try doesn't seem interesting.

s/tin/t/

I can ssh -Y just fine when I need to

But only for things not written in terms of Wayland, or am I missing something?

no

ssh -Y is work X11

Anyway, that's enough time sunk tilting at windmills. Back to work.

no wayland things will accept it

but keep in mind there are noone working anymore on xorg upstream (or close to noone)

Hi. On my new FreeBSD Desktop, I would like to move whatever will likely need most diskspace to my second harddrive. I used the ZFS guided partitioning in the installer. Now i tried making a new zfspool on my HDD, which works fine but i wonder which mountpoint would make sense to transfer over there and how i would best go about it. Is moving /usr/home a good idea?

./home

:P

Or should i not bother with moving any freebsd mountpoints? My SSD is 256GB, will FreeBSD run out of space there?

The HDD is 1TB

i don't know

better put it at /home

how much space freebsd itself will need will depend on what packages you install, etc.

so it can always expand

. /home + bla bla

. /home + bla bla

splitting /home away from the rest of the system is good

For the procedure, i boot into single user mode, copy the files to the new zfs partition and then edit /etc/fstab? Anything else?

for moving /home you shouldn't even need single-user

and you probably don't need any fstab changes if both are zfs

how can i delete a file named 'malformed PPS.csv' with quotes in it?

use more quotes :-)

So.. I am trying to move /usr/home following threads like forums.freebsd.org/threads/move-usr-fs-to-other-disk.3855 cyberciti.biz/faq/freebsd-move-usr-filesystem-to-new-disk where they basically say to use the dump command to do it. But i get 'dump: /usr/home: unknown file system' 'Tape is not a dump tape' :/

Title: move /usr fs to other disk | The FreeBSD Forums

e.g. "'malformed PPS.csv'"

KingShark: dump is for ufs, not zfs

RhodiumToad: it says the file or directory does not exist. i thought there were more characters needed

KingShark: if your /usr/home is already a separate dataset, then I'd think zfs send/zfs receive might be the easiest way to transfer it

CCFL_Man: perhaps there are non-printing characters? you might try finding a wildcard that matches it

i think there are nonprinting characters

try ls -lB

ahh

-rw-r--r-- 1 pi pi 2440489 Apr 29 2022 'malformed PPS.csv'

that doesn't show any nonprinting characters

does ls -lB "'malformed PPS.csv'" show it?

RhodiumToad, thanks. Says target must not be mounted and umount says device busy so i guess its single user mode after all :D

Or would taking a snap and sending that help? Never did any of that :)

RhodiumToad: no, it says it's not found again

it's a stubborn one

does ls -lB *alformed?PPS* match any other files besides that one?

just that one

i may be able to use that

then rm -i -- *alformed?PPS*

that did it! thank you! i'll keep that string in mind. i'm going to put it in my notes

I moved my /usr/home data successfully and set a new mountpoint pointing to it on the new zfs pool. How could i remove the pre-existing mountpoint, which is labeled as ' inherited from zroot/usr' ? I tried 'zfs set mountpoint=none zroot/usr/home' which gives 'pool or dataset is busy' and i can not umount it either (Device busy)

KingShark: you'll have to unmount /usr/home; where is the new mountpoint?

rtprio, thanks, umount -f /usr/home worked. The new mountpoint is my second zfspool named 'data'. I added it via 'zfs set mountpoint=/usr/home data' which apparently worked fine. 'zfs get mountpoint' shows it correctly. I removed the previous /usr/home mountpoint doing 'zfs set mountpoint=none zroot/usr/home'. Was that the good way to do it? zfs get mountpoint still lists it, is that a problem?

if you don't need zroot/usr/home you can also destroy it so it won't interfere

thanks that looks cleaner now

i will reboot and see if my system survived the operation :-))

rtprio, everything worked fine, thanks :)

that was a fast reboot

even took my time confirming in thunar and via zfs get mountpoint :P

KingShark: you have freebsd with a gui?

yeah its my desktop computer

oh nice

nothing odd about that :-)

i was on freebsd loooong ago on my back-then desktop but i forgot about everything :) i hopped back to linux after a while because freebsd did not have some packages that i wanted to use.. i hope it works out this time :-))

I used Windows on my desktop for a long time and FreeBSD on a laptop next to it. Then a couple of years ago I jumped ship and switched to FreeBSD as my desktop full time and re-purposed the laptop as a server (still running FreeBSD).

Definitely been a net positive. Gave up some hardware support and availability of some apps for better stability (though Windows got a lot better after Vista/7) and more consistent workflow between projects since I often deploy on FreeBSD.

Though Carmack had a good point a long time ago when he said that developing on a different platform than the runtime target helps to surface bugs.

He was asked why he developed DOOM on a NeXT cube.

RhodiumToad: Apropos to yesterday, if I did hypothetically switch to /usr/sbin/daemon to launch background services for my user session on login and assuming I can automatically invoke a command on logout, what would be the best way to kill those background services without having to know about each one explicitly?

RhodiumToad: I know I can iterate over a directory of daemon (supervisor) PIDs and SIGTERM them. I'm also wondering if there's a way to use process groups or something for it (don't know too much about them yet).

_xor: it helps in that case that Next was m68k

I want cool riscv64 and aarch64 Mini computers

_xor: I don't think daemon can handle arbitrary process groups, it'll put each process into its own session and process group

_xor: are they running as you?

meena, I'm not sure "mini computer" means what you think it means.

CrtxReavr: i mean an SoC, not one of these en.m.wikipedia.org/wiki/Minicomputer

Title: Minicomputer - Wikipedia

our standards for what mini means have greatly shrunk in the past forty years

meena, well. . . yes, but the mini vs. micro distiction was about single chip CPU vs. board full of chips CPU.

tell that to usb

parv: now?? I've always been this cool

USB plugs are like Mac SCSI cables in the 68k & PowerPC era.

(I have actually only used computers since after 2000.)

CrtxReavr: lol

meena, I know, I know😅. Sorry. I was waiting for the clarification of "mini computer.

I started working at an EMC facility in 2000. . . though it was formerly a DataGeneral facility.

There were still two old DG minicomputers in use.

Saw the decomissioned and rolled out one at a time.

CrtxReavr: i did just see on that wiki, that DataGeneral named a computer after my daughter

All the old time DG/UX coders walked 'em out like they were a relative.

i had a mac classic with a 20MB scsi hdd

still had to boot system 5 from a floppy

You know. . . the funny thing about those Macs in the 68k & PPC era. ..

hm, maybe it was system 6.

Everyone thought they were superior to PCs for graphics, I guess just because they had a GUI when most PC users were still ratting around in DOS. .

But they in no way did. . . the graphics hardware on Macs was garbage.

They had to reserve a parity bit for per-pixel color depth. . .

So if you enabled 16-bit color, you only got 32k colors. . . for for 24-bit color, you enable 24-bit color, you only got 8 million colors.

Plus, before OSX, their OS was garbage.

Cooperative multi-tasking?

Really?

*shrug* all i wanted to do was play stuntcopter and dungeon of doom, so that was cool. also eventually object logo and some smalltalk interpreter that i guess we had

I remember ratting around in DOS, but at least we had 4DOS

I did a lot of scripting with REXX on IBM PC-DOS back in the day. Still have fond memories of REXX.

wait, isn't REXX amiga?

the amiga dialect was calles arexx

oh gotcha

"I started working at an EMC facility in 2000. . . though it was formerly a DataGeneral facility."

CrtxReavr: Lol, I still remember years ago when you helped me out with that girl at EMC. Though we don't keep in touch as much as we used to, I'm still friends with her to this day.

Holy crap, that was in March of 2007. Almost 16 years ago.

REXX on PC-DOS? I thought that was a mainframe thing

", and has been "back-ported" to VM/370 and MVS." ← oh… that's where I used it on

well, was forced to try to use it at school.

meena: The daemon supervisor process or the main child process? Either way I think they'll run as me, so yeah pkill for processes owned by my user is an option. any processes by me.

pkill -u $USER as goodbye

RhodiumToad: Ah, and I take it process groups are static and can only be set at launch? (haven't read up on them yet)

_xor: i really want to have APIs for what you're talking about, but right now, we don't have them.

For what, querying processes or launching background processes?

_xor: also process groups can be easily escaped: github.com/freebsd/meetings/blob/ma…ision/README.md#tools-and-languages

Title: meetings/supervision/README.md at master · freebsd/meetings · GitHub

(by design)

oh

sysctl hw.snd.vpc_0db=1 this seems to not work very well in firefox

vol 100