Does etcupdate always try to build binaries when it goes to update its tree? Never seen this before: cc -O2 -pipe -fno-common -I/usr/src/crypto/heimdal/lib/roken -I. -DHAVE_CONFIG_H -I/usr/src/kerberos5/include -MD -MF.depend.simple_exec.o -MTsimple_exec.o -std=gnu99 -Wno-format-zero-length -Wsystem-headers -Wno-pointer-sign -Wdate-time -Wno-absolute-value -Wno-empty-body -Wno-string-plus-int

-Wno-unused-const-variable -Wno-error=unused-but-set-parameter -Wno-tautological-compare -Wno-unused-value -Wno-parentheses-equality -Wno-unused-function -Wno-enum-conversion -Wno-unused-local-typedef -Wno-address-of-packed-member -Wno-switch -Wno-switch-enum -Wno-knr-promoted-parameter -Wno-parentheses -Qunused-arguments

-I/var/db/etcupdate/etcupdate-JrGOU0J/usr/obj/usr/src/amd64.amd64/tmp/legacy/usr/include -c /usr/src/crypto/heimdal/lib/roken/simple_exec.c -o simple_exec.o

Oops. Among other things, like stuff in usr.sbin

amigan: how are you calling it?

i'm starting to think that zfs on an external usb drive may not have been the best choice

i must be going mad, does anyone else, fresh freebsd 14 install..using xfce4 NOT have the startxfce4 command?

fresh freebsd install does not have xfce4; which packages exactly did you install?

for Christmas I want FreeBSD 14 and NetBSD 10

zwr: FreeBSD 14 is available today.

(And NetBSD 10rc1 too.)

Ooh, sup removed from base. =removes hat in respect= netbsd.org/releases/formal-10/NetBSD-10.0.html

Title: Announcing NetBSD 10.0 RC1 (Nov 5, 2023)

pkg install x11-wm/xfce4

yuripv: correct, i "installed" it

i can get it to run with exec xfesktop but the configuration file for xrdp talks about a start xfce4 file.. which i do not see anywhere

never mind, looks like if you invoke installing xfce in multiple ways.. you get "different" results pkg install xfce this time

voy4g3r2, x11-wm/xfce4-session should provide it. See cgit.freebsd.org/ports/tree/x11-wm/xfce4-session/pkg-plist

Title: pkg-plist « xfce4-session « x11-wm - ports - FreeBSD ports tree

vbotka: thanks, xce install identified that package

anyone have any idea how to get the MM keys woring on an a1243?

are you using x or wayland?

or at the console?

voy4g3r2: package 'xfce4-session' has startxfce4

/usr/local/bin/startxfce4 I copied that to my .xinitrc and startx works to start XFCE4 here.

meena: just etcupdate, like I always do after installworld

well, I updated to 14.0 and now it can't find the ZFS root

weird, my zroot box was fine

tons of other little issues though

bugs.freebsd.org/bugzilla/show_bug.cgi?id=275336 for one, and lots of weird stuff in my kernel messages. And just now I ran ifconfig and I got "ifconfig: getnameinfo: Address family not recognized", and ran it again and it was fine

Title: 275336 – Processes exiting with SIGSEGV on shutdown on ZFS systems

"module_register_init: MOD_LOAD (sctp, 0xffffffff80a8c870, 0) error 17" on boot

something to do with the kernel. The old 13.2 kernel works. Btw I have to compile a custom kernel because my wi-fi card's driver is GPL. My kernel config is this: paste.debian.net/plain/1300309

did I miss a step for people with custom kernels? btw it's an encrypted install so it's ZFS on top of GELI

I see no reason it shouldn't work as it did before

But I have also never booted zfs on geli, not sure what is entailed in keying up the kernel so you can see the pool

I guess I stay on 13.2 for now, I only use that laptop for testing programs on FreeBSD

I broke apt in a linux jail :P

wait, it may be fixable lol

elsewise I want to see if I can install something like arch in a linux jail

amigan: try this man-dev.freebsd.org/etcupdate#B

Title: etcupdate(8) - FreeBSD Manual Pages (Dev)

otis: IIRC you know a bit about fibre...

dch: yes, a bit, yes :-) you mean storage or ethernet?

ethernet in this case

that's more of a simple case.

I planned a fiber upgrade at home here, and unexpectedly picked up a couple of 100G chelsio NICs at bargain basement prices

so now I am wondering whether that changes the fibre type I need

singlemode should do

OS2

nazxar vbotka

dch: what's the card type?

originally I was going for OM4 but it seems SMF is the way to go

Chelsio T62100-LP-CR store.chelsio.com/products/t62100-l…adapter-with-pcie-3-0-x16-interface

Title: T62100-LP-CR: 2-port Low Profile 40/50/100GbE Unified Wire Adapter wit – Chelsio Online Store

yes, i prefer SMF, too. when there's no other way, also MMF would do, ofc.

ok, qsfp28, thought so. you can buy qsfp28 with either mmf or smf, i'd suggest smf.

smf has longer reach, is not so prone to attenuation...

initially I am going to use 25Gb SFP28 because they are significantly cheaper than the 100Gb ones

fwiw fs.com/c/100g-qsfp28-1159

Title: 100GbE QSFP28 Transceiver Modules | Optical Transceivers - FS.com

the total length is ~ 40m

yeah my fs.com person already recommended some compatible parts, its very handy to check with them

nazdar otis

we even got some kind of a "partnership" for slovak republic. our sales person gives us slightly better prices.

hey vbotka! I've seen your really awesome ansible modules. nice to see you around here

yeah, vlado's ansible playbooks are neat, indeed.

some folks asked me if therw would be an interest in attending a FreeBSD Users Group meeting in eastern Slovakia during summer (they aimed to organize it in that case)

Thank you for comments. Nice to see you too!

I would come, if it lined up with my dates

(family holiday plans)

otis: I think my electricians could probably put in an actual patch cable directly, a 50m run would be ideal. there's a tube in the building already for this stuff.

dch: good for you then!

otis: thanks for the advice, I owe you a few beers!

few coffees :-P

otis: any other tips you can think of? I am planning 2 get a couple of these fs.com/de-en/products/177394.html?attribute=700&id=1786819

Title: 50m (164ft) LC UPC to LC UPC Duplex OS2 Single Mode PVC (OFNR) 2.0mm Tight-Buffered Fiber Optic Patch Cable - FS.com Europe

dch: that's good choice

care should be taken to not tear the connectors off while laying the cable

mason: what was sup used for in the past two decades?

either I am dreaming or p2 has fixed my vnlru woes

Remilia: i think you are not dreaming, looks related? bugs.freebsd.org/bugzilla/show_bug.cgi?id=275063

Title: 275063 – kernel using 100% CPU in arc_prune

yuripv: how do I get top to show that {threadname} thing?

mine just says kernel

top -H?

but otherwise percentages look similar, aside from vnlru taking 40-70% for me

top -SH maybe?

thank you!

never noticed that option before

ah ok yuri already responded.

-H also goes for ps

wow that is a lot of time in epair_task, I should probably see if netgraph is faster

Thanks otis

tsundoku: if you haven't solved it yet, i think it will be helpful if you can paste a full netstat -rnf inet output from the jail and host, ifconfig of the relevant interfaces, and tcpdump output from both sides of the epair, the bridge, and the vlan (and maybe the external interface), from the moment that you emit the ping

someone on LQ called KDE 'snappy' and I laughed pretty hard...

hello, any idea why colours disapeared in cmus when upgrading to 14 ? (tmux, rvxt-unicode, and TERM set to screen-256color)

mage: which command lost its colors?

ah ic cmus obviously

elliottp: yeah.

it's strange, neomutt works, cmus not

although they depend both on the same libncursesw.so.9

Address resolution failed for pkg.FreeBSD.org/FreeBSD:14:amd64/quarterly.

Consider changing PACKAGESITE.

why? WHY?

AumShivaya: ping pkg.freebsd.org

yeah, it doesn't

ping 1.1.1.1

root@ubuntu:/ # ping 1.1.1.1

ping: ssend socket: Operation not permitted

theres yr problem

yeah, perhaps p4.addr="192.168.1.150" is wrong?

does that have to point to my router address or is it just that?

I just copy and paste from the handbook that

if yr router caches dns then yes

well I set up my resolv.conf the same in side the ubuntu jail

but your problem here is "dns isnt working" and not "pkg is unreachable"

how do I change the settings once the jail has been created with jail -cm?

yes

i dunno about jails. but i think you have to set the jail up to allow networking

I am just learning, trying to follow the handbook

there might be a sysctl somewhere

sysctl -a | grep jail

what am I looking for?

Title: Chapter 17. Jails and Containers | FreeBSD Documentation Portal

security.jail.allow_raw_sockets i guess - but I don't know security implications of those

yeah saw that

Title: jails - Allowing raw sockets in jails | The FreeBSD Forums

nah I set raw sockets to 1

and chrooted back in

and still cannot ping 1.1.1.1 or pkg.freebsd.org

outside the jail, can oyu ping 1.1.1.1

sure can

im baffled then

aye

thanks for your help. Perhaps someone else may speak up soon

try subscribing to -questions and posting there? or the forums, or both

could do soonb

soon

brb

so the line ip4.addr=

I put in the address of my wifi

and it stopped all my networking I had to reboot

bear in mind more ppl will be on here later (seems most ppl are utc-5 to utc-8)

also try creating a freebsd jail 1st, and if that works then the issue would be specifically with the linux jail

which one? thin?

thick

one with all the bells and whistles. if it works you have a linux-in-jails problem. if not, a jails-general-problem

ok shall give that a go...bbl

Anyone else getting "No valid device tree blob found!" when booting -CURRENT on arm64?

how do you look at the not install packages

maybe ports.freebsd.org/cgi/ports.cgi

Title: FreeBSD Ports Search

i just use pkg dont use ports

freshports also lists the available packages

richard_: `pkg search` `pkg rquery`

e.g. `pkg search -Qfull <name>` would provide all needed details about *not* installed package <name>

richard_: pkgs are build from ports

hi all

does anyone use usb microphone?

i noticed volume issue. It is not loud enought

and as I understand it isn't microphone failt but a driver. Another guy says the same on his video.

wikan: i assume volume is at 100, when running 'mixer'?

wikan, sure it's not a hardware issue. . . ie. a volume rocker on the microphone?

meena: I bet sup has not been used in that stretch. Even I was using cvsup. But it still marks an era.

mason: I wonder how their transition to hg is coming along

I believe it's stalled, last I knew.

well it is not hardware issue

i am not only person with the same problem. Other microphone but usb type

volume is 100% but works like 35%

What's that CLI volume level tool?

Been a while.

Think it's part of base.

hmmm

mixertui

and it works up to "35%"

Oh, nope - just mixer(8)

What does it say your volume is at?

hi all. trimming zpool ssds is easy: zpool trim pool. How to trim a ufs ssd?

mixertui is a tui as the name says

wait, must check

mixertui is a port though.

Been a while, since I had a FreeBSD workstation, but I would toss a mixer command in my /etc/rc.local to set the levels where I wanted them on each boot.

mic/rec 96:96 95:95

f451: fsck_ffs has an -E option to trim unused storage. i've never tried it though

unixwitch: thanks

audacity prints my voice wave up to maybe 30%

hi, i fucked up and let freebsd-update completely fill my zfs disk. now i can't even remove files to free space?

if i am close

problem im having is a GIS returns old docs, and I know stuff has changed in the last couple of years

ascreen, zfs list -rt snapshot -o name,used,refer,clones

you may have a snapshot or two that will free up a bunch of space

i deleted a few snapshots, zpool list says i have 393M free

f451: trim(8)?

but the fs still say AVAIL 0B

What does df -i say?

zroot/ROOT/default 9612024 9612024 0 100% 568182 0 100% /

so inodes are full too?

but i think zfs doesnt preallocate inodes

Well.. .. probably first.

That I dont' know.

anyway, i can't even truncate some file i have laying around

zapata: checking ... didnt realise htat was there. old docs suggest its enabled at newfs time

also 500k files isnt that much

What does mount -p say?

zroot/ROOT/default / zfs rw,noatime,nfsv4acls 0 0

and some other stuff

Does ZFS not re-allocate inodes?

That'd be important to know for certain, 'cause 'df -i' says you're out of them.

well i guess it means it can't create new inodes

e.g. touch fails too

Well, every file, no matter how small, needs an inode.

So if you're out of them, the volume is "full."

yes, but a classical fs would have some inodes allocated and could create an empty file even if disk is full for blocks

but i think zfs doesnt do that

i just dont get why zpool list says a little free space yet it's not used

Can you delete a file and see if inode count decriments by one?

no i can't delete a file

else i could just clean up the old freebsd-update stuff :D

You in single user mode?

no, regular boot. but nothing else is using that machine

ok, i have an idea. perhaps deleting a file that is not in any snapshot helps

no :<

Maybe 'shutdown' to single user mode, make sure that / is still mounted rw and try 'fsck -y /'

f451: note that trim(8) will trim the entire device (i.e., erase all the data), at least by default, which probably isn't what you want

phew, deleting a newer snapshot helped

deleating a file that is part of a snapshot too won't help, as the snapshot prevents removal from disk, you should be able to delete anything that isn't part of a snapshot and some old snapshot could maybe prevent some big file from being removed from disk

yeah, the problem was that freebsd-update created a new snapshot :)

unixwitch: yeah. i think i can sort it by fsck_ffs -E -y then enabling tunefs -t then running the fsck_ffs -E again all while in single user mode

trim(8) seems to be a little different!

any dhcpcd users?

I'm struggling with the manpage to get it to do what I need.

CrtxReavr: what do you need?

CrtxReavr: also note that you can ask questions on the Github discussions, and it's usually very responsive

Well, I have an ISP facing interface that I need to do v4 DHCP on and v6 DHCP-PD.

Then I need to configure /64s from that PD on some internal-facing interfaces. . . which I don't necessarily need dhcpcd to do, but I do need the prefix.

*nod* *nod*

I think maybe you could pass that off to… something else.

ok, i have one more dumb question... the system says in zfs list: zroot/var 3.64M 7.73G 96K /var

however, /var isnt mounted?

ascreen: looks mounted

and thus the whole /var/db stuff ended up in zroot/ROOT/default

mount doesn't list it

Right. /var and /usr are usually not mounted. Run this: zfs list -o name,compression,exec,setuid,atime,canmount,mounted,mountpoint

The canmount will say off for some of those. That's by design and intentional. It supports Boot Environments.

so what is intended to be snapped by a boot environment?

But other datasets are mounted under those directories.

ah,

Yes. So that things that must keep moving forward keep moving forward.

And things that need to be able to snap back to a previous snapshot can be reverted if needed and everything works.

yeah /var is canmount no, mounted no

So /var/log for example needs to keep moving forward. So it is an explicit dataset mounted under /var

It's confusing at first glance until it is understood that /usr and /var set up an inheritance for other things to be created but those directories themselves are held by the zroot/ROOT

I saved this off a while back: forums.freebsd.org/threads/why-are-…var-not-in-a-boot-environment.59844

Title: ZFS - Why are /usr and /var not in a boot environment? | The FreeBSD Forums

And with that I must run afk but I saw this discussion in the channel and had to jump in! Sorry to barge in and then run off.

well thx :)

I have a 14.0p2-RELEASE installation on bare metal that can't boot until plugged into a monitor. xorg, sddm (did this give me plasma?), a couple xf86, and drm-kmod are installed, among others, if it matters. anyway, i tried in /boot/loader.conf setting console to EFI or even nullconsole. i searched on various dmesg items, and none seem related that I can tell. any ideas for making this boot even if not plugged into HDMI?

i wanted to learn to set up a desktop, and finally succeeded, but this will be a headless box

FYI for all the people who were puzzling over my bridge problem yesterday: the issue was that my rc.conf order was wrong.

I had cloned_interfaces=" " below ifconfig_bridge0=" " because I had some kind of problem earlier with it above, and I read something that led me to believe cloned_interfaces was not creating anything, just indicating to the OS which interfaces were virtual. that was not correct.

scoobybejesus: can you elaborate on "can't boot"? where does it stop booting? do you see anything relevant (like errors) if you connect a monitor after it fails to boot?

scoobybejesus: can it boot anything else without a monitor ?

I had a crappy motherboard that wouldn't post without video output

tsundoku: that's odd since order of directives in rc.conf shouldn't make a difference to anything... the whole is read in whole and doesn't affect the order services/devices are started

s/whole/whole file/

i wish i could be more specific. what happens is i am ssh'd in, so i do doas reboot, and then i can't ssh in again. so i plug in the monitor, and a few seconds later, things populate and it finishes booting

does it beep? :)

either i turned off beeping, or it never beeps. i don't recall. perhaps i could pop into the bios to enable it

scoobybejesus: if you disable sddm from starting, does it boot correctly then without video connected? just to isolate the problem

i will give it a try :)

tsundoku: rc.conf order should not matter, these are environment variables

(i don't know anything about sddm, but at least with xdm, the dm is responsible for starting X, so it's at least theoretically possible that if it can't start X for some reason, it might interrupt the boot process... this is just a wild guess though)

..

.

apparently the order does matter, at least for networking stuff

tsundoku: if you look at the rc.d scripts all they do is check variables

tsundoku: it really doesn't... it can't matter. the rc scripts in /etc/rc.d are run in a specific order based on their dependency requirements, and they always read the entire /etc/rc.conf... the problem is almost certainly something else

they source the conf file, then use values

well, putting cloned_interfaces above the other lines apparently fixed it, because it's working now

well unless you were setting and then using $variables in rc.conf in some reason, in which case order would matter, but...

one thing i noticed is that the way i have things now in loader.conf (console="nullconsole"), one section of booting populates a block in the center of the screen which ends in saying it's Loading /boot/loader.conf. Then FreeBSD logo. Then it uses the monitor normally, putting everything in the top left and prints down the lefthand side. This part starts with "Setting hostuuid." Continued..

When I reboot without the monitor attached, and it hangs, and i eventually plug it in, this is always the first line i see once the monitor comes up

tsundoku: chances are you first renamed the interface then used cloned with the new name

tsundoku: either you fixed the issue by accident, or you have some sort of timing-dependent problem that fixed itself by accident somehow

this might just be me but did 14.0 have more ENs than usual for a .0 release?

yes, disabling SDDM allowed for a normal, speedy reboot :)

<- still loves startx

scoobybejesus: i would be inclinced to just leave it disabled then, and use `service sddm onestart` if you really need it for some reason

I am impressed it goes in rcorder before sshd

although it is odd it breaks the boot process, even if X can't start it should just background and keep trying, not break the entire boot process

it would be nice to know more, but yes, this is definitely suitable for now. thank you unixwitch

my other suggestion would be to switch to xdm, which has worked fine for 30 years, idk why we need all these new fancy things (back in my day, etc. etc.)

unixwitch: So what would you do if trying to start it 5 times in a row failed? Try another 1000 times?

tsundoku: if you do something like: create_args_bridge0="name brj" cloned_interfaces="brj" it will not work

what I have now is like

it will create brj, but what kind of interface it will be is ???

cloned_interfaces="bridge0"

and then

armin: yes? i mean, i suppose it could also just log and error and quit, which would be fine too. but it shouldn't do whatever it's apparently actually doing.

unixwitch: There's something about trying the same thing over and over and expecting a different result...

ifconfig_ixl1="up" and then ifconfig_bridge0="addm ixl1 up"

that's it

tsundoku: and the order would not matter

it appears to matter

it cannot, sadly

hmokay hostit.tfnux.org/obj/wsz62a3y5exnozajqpo7m45qgu5j4jtf/paste.txt

armin: i really have no interest in how sddm handles X server startup failure, my point was that it should handle failure in some way that doesn't break everything else

I had a friend look at it who discovered the problem and what he said was that cloned_interfaces is what creates the bridges

tsundoku: rc scripts source defaults/rc.conf and rc.conf

as variables

no, /etc/rc.d/netif creates the bridge

cloned_interfaces is just configuration, by itself it doesn't do anything at all

the rc scripts do not read rc.conf line by line like some interpreted language

unixwitch: yeah, fair point, I've seen that on Linux, systemd trying to re-start X, X crashes, re-start X, X crashes, and so on. Try to switch to a TTY and log in while that happens. Fun.

the other thing I was told was that from FreeBSD 13, interfaces default to down on boot

are you sure it is from 13

and not like 2.2.1

*shrug*

though it has been a long while so I may be misremembering

tsundoku: `less /etc/rc.d/netif` and educate yourself

it is a shell script

yeah

well all I can tell you right now is that changing the order of things in rc.conf resolved the issue

Remilia: i'm pretty sure every Unix system since forever has defaulted interfaces to down... right? unless maybe there was some weird special case where cloned_interfaces would be brought up by default or something (but i honestly don't remember that ever being the case)

unixwitch: aye

tsundoku: switch their places and reboot and you will see that nothing changes

something something the interfaces were getting added to the bridge in a down state and I guess trying to bring them up after the fact does not work

tsundoku: it does work (i've recently tested that since i've been configuring some jails) but it is true that they will be done by default, which is fixed by ifconfig_xxx0="up"

you had the bridge renamed

s/be done/by down/

like, I remember very well that your bridge was called svcbridge

yes

so swapping the order was not the only change

well I'm still renaming the bridge

ifconfig_bridge0="addm ixl1 up" should not work if you renamed the bridge in rc.conf

but in troubleshooting we took out the rename for expediency and then put it back after it got sorted out

tsundoku: it's up to you, but i would suggest editing rc.conf to move cloned_interfaces after ifconfig_ (or whatever) and rebooting just to see what happens. i'm 100% sure it'll still work, if it works now.

so the rename doesn't appear to have affected anything one way or another

because by ifconfig time interfaces are already renamed

then why didn't it work before?

i have no idea, unless you have a backup of the non-functional rc.conf to compare, who can say

^

basically there is no magic in `. /etc/rc.conf`

well, maybe I'll lab it out sometime later if I get really curious

tsundoku: /etc/net.subr processes the cloned_interfaces in clone_up()

on this machine I kind of just want to move on

it literally goes `for ifn in ${cloned_interfaces}; do .....`

/etc/net.subr?

network

i recommend testing it just because this is fairly important to understand how /etc/rc works in freebsd. but like i said it's up to you so -shrug-

doesn't exist on my system

how can /etc/network.subr not exist

what did you do to it

oh, I misinterpreted what you were saying

if you look at that file you will see that it only accesses shell variables

and never tries to read rc.conf line by line

makes sense

kind of related, i've always avoided renaming interfaces because i've never been entirely sure what rc.conf variables are used prior to the rename and which are used after... i should probably go work that out at some point

yeah, I really wanted the rename because I was like... if this is a virtual object that I'm creating deliberately on my own, I want the name to be desctiptive and not just like bridge0

unixwitch: cloned first, then rename

I mean, if you are renaming via create_args, it would be after cloned

Remilia: so ifconfig_ is already after? (like if i want to rename then add to a bridge)

s/already/always

ifconfig_ lines operate on existing interfaces

unixwitch: look at netif, it goes # Create cloned interfaces \n clone_up $cmdifn

then # Rename interfaces

then # Configure

yeah, that makes sense. ok, now i just need to sit down and come up with an overengineered and needlessly complicated if naming scheme

you usually want somthing like

cloned_interfaces="vlan0"

ifconfig_vlan_name="myvlan"

ifconfig_myvlan="inet 100.64.0.1/16"

but the order of the lines doesnt matter

yep

what do people like for automatic zfs snapshot management? zfs-auto-snapshot, zfs-periodic, zfs-snapshot-mgmt?

tsundoku: the reason I was trying to explain this to you is to prevent more weird voodoo spreading

i used sanoid for snapshots. the config file is easy enough for me. put it in cron. done

scoobybejesus: thanks, never heard of that one

it comes bundled with syncoid which makes off-site replication easy as well

my kernel spends 3% of a core in epair_task hahahaha

mine is spending 700% cpu in "idle"! what a terrible waste of CPU

Remilia: fwiw i remember reading recently that epair sends all packets through a single taskq which seems not great for scalability

that's odd, why does my 2023Q4 poudriere have 209 ports queued? does an update to pkg rebuild everything?

unixwitch: no but if you updated a jail it kills everything

oh, that's probably it then since it just updated to 14.0-RELEASE-p2

maybe i should disable jail auto update on build, i doubt it makes any difference to the built packages in 99% of cases

I think the reason is this

nimaje yeah i see that