hi, im trying to resize a zfs partition and gpart warns: gpart table "nda0" is corrupt

i dd'ed a smaller disk into a bigger one. and now im trying to resize the partition

im doing: gpart resize -i 4 nda0

whats the correct way to fix and resize ?

What was the dd command you entered?

(The answer to your question is *probably* to use "gpart recover", but it being a whole data partition, I'd rather check there aren't other problems lying in wait to mangle your data.)

V_PauAmma_V ya that worked ty

jbo: pong

V_PauAmma_V: simple dd if=/dev/nda0 of=/dev/nda1 status=progress bs=1000M

it boots but displays wrong disk size/free disk space

jbo: pong

there a convention for what to call a cli that connects to a program using the program's .sock file?

just programcli or?

V_PauAmma_V: gpart shows the bigger (new) disk as corrupt, but when i run recover, it doesnt boot

throws some error

HER, you said the gpart error was on nda0 (the smaller, source disk). Was that a typo in reproducing the message?

V_PauAmma_V any upside to having rc daemons log to their own file or is it always better for stuff to log directly to syslog?

I have no basis for an informed opinion on that.

jbo: EXICTED PONG!

why does he keep pinging me then fucking off

V_PauAmma_V: the error is in the bigger one, after dd copy. it should show 1tb free space instaed of 500gb (smaller disk)

the copied disk boots, but shows wrong disk free space... and when i check with gpart, gpart displays the disk as corrupt

and its a geli disk

OK. Did you run "gpart recover"?

zfs/geli

V_PauAmma_V: yes, and it "works" but when i reboot, it shows some errors...

doesnt boot

OK. Try a "gpart resize" (with appropriate command arguments) now.

V_PauAmma_V: i will have to dd again

because resize fixes the corrupt issue, but then disk wont boot

throws some error

but thans, V_PauAmma_V , dont spend more time on this... i will try some standard way =)

instead of dd.. i will install a new system and copy over

bsdnow.tv/558 *listening*

Title: BSD Now 558: Worlds of telnet

Hello folks! I switched to Dragonfly Mail Agent. I noticed that emails from cron bounce because cron sends them to the username (e.g. "john"), without adding any domain. I suppose I could manually list aliases to workaround that problem. How are you handling this?

Good morning

morning

morning

has anyone experienced nfsv4 mounts (zfs based) not being seen on linux systems? and receiving a mount.nfs4: Protocol not supported message, this is occuring on debian bookworm

voy4g3r2: Firewall?'

I know I had something mounting zfs nfs at somepoint.

is it possible to mount a geli/zfs partition into a /some/dir ? I have 2 freebsd installs, 1 in each disk. I want to boot into live cd and mount each disk on different directories (/mnt/disk1 /mnt/disk2) and then copy data from one to other

Hey. How can I trace syscalls on FreeBSD?

truss(1) or ktrace(1).

Ok, thanks!

HER, "zfs mount -o mountpoint=/mnt/disk1 /dev/whatever"

I think the problem will be that both installs will have used "zroot" for the array name and therefore at least one of them must be renamed. I think.

If they have different names then one can be mounted at an altroot using -R. Probably needing -f too. Example: zpool import -R /mnt -f zdata

V_PauAmma_V: thanks!

I think the pool UUID will need to be used instead of the name since the pools will both be named "zroot".

rwp: right

i've set failover between ethernet card and wifi as described in handbook and also with some minor changes as can be found by cy@ answer on forums.freebsd.org/threads/help-eth…ifi-failover-link-aggregation.92949

Title: Help! Ethernet/WIFI Failover Link Aggregation | The FreeBSD Forums

it works

however my dmesg polluted with endlessly with: wlan0: link state changed to DOWN

wlan0: link state changed to UP

is this expected or something not ok with config?

Depending on which WLAN interface you have, you may need to set the ethernet's MAC address (to the same as the wlan's), not the reverse. That's the handbook current recommendation and what I did when I had an Ethernet at home. ISTR the symptoms for me of doing the opposite were failure to attach to the AP, which is similar to what you describe and may be caused by the same. If that doesn't stop the

flood of wlan0 bounces, the next question is: do you have the same problem with separate (unaggregated) wlan and ethernet? If you do, it's likely you need to move your AP or add another.

i have to try

and yes i have ethernet mac set, not wifi's card

*nod*

Just a clarifying question. The wlan0 logging of state isn't incorrect but just the logging of it up and down is annoying? And it's all working but you would like to suppress the logging of wlan state? Is that the same message that would be received normally using wifi too? (I don't know as I am not using wifi. That's why I am asking.)

i will try using only wifi and see.

Sadly I am not using FreeBSD on a laptop at the moment. I need to set one up and get experience with the native WiFi on 14. On the linux side doing link failover I have only ever been successful setting up OpenVPN which runs over whichever link is active. That technique works well there. But it's not turnkey and one must set it up.

from what i can tell it is still use 11g mode. but it is enough for me.

Hey folks. Does anyone here have experience with "NIST SP 800-171 Rev. 2"?

any upside to having rc daemons log to their own file or is it always better for stuff to log directly to syslog?

Have you posted this question to the freebsd-questions mailing list? I am behind reading it. But it's probably time to get a new audience and that list will have a broader base of readers and more experience than here.

reliablesignals, Have you completed your Cyber Awareness training? :-) public.cyber.mil/training/cyber-awareness-challenge

can syslogd be configured to send logs with tcp?

doesn't look like it

polyex, not to my knowledge. As far as I can determine FreeBSD's syslogd can only use UDP. You will probably need an alternate, such as syslog-ng, to use TCP.

if i'm logging to localhost udp is plenty reliable?

Unless you have a horrid network, UDP should be fine for log packets.

but if it's localhost:514 why would the network even matter?

Oh, I missed the localhost question while I was talking about UDP. :D Yeah, should not matter at all.

When the network is under stress (think switches and routers overloaded with max traffic) then routers will prioritize TCP connections over UDP connections (generally) and that means some UDP connections get discarded when the capacity overflows. Due to queuing quirks UDP packets might be forwarded out of order. And due to retries UDP packets might arrive out of order.

But on localhost on the loopback device? If the internal loopback network is in such distress that it drops a UDP packet then you will know the machine is in distress by how slow it is running.

i wonder if that happens on localhost too. like if the local box is saturated with localhost tcp traffic, does localhost udp traffic start getting dropped?

nice

It would. But I really can't imagine a case where UDP on the loopback device is so saturated.

ty

ya

Once in the work lab where we had an Extreme Networks Black Diamond switch/router (a Barney box, it's big and purple like Barney) which was big heavy iron at the time. I was working on a distributed networking using UDP in a program to collect data and control ops across our compute farm of 3k+ machines.

I found that if I cranked it up I could completely overwhelm the box, as one might expect with a compute farm, and then stop everything and over the next 2 minutes (magic number) UDP packets would drain out of the router and other places.

Two minutes is the max lifetime of an Internet packet so any packet older than 2 minutes is discarded by design to prevent complete breakdown of the network in the case of overwhelming and endless data.

max lifetime is any protocol like udp, tcp ...

?

Max lifetime of an Internet Protocol (IP) packet. IP is the base under which UDP and TCP are the two most well known types. But there are other types too.

3rd most popular after udp and tcp?

Hmm... Don't know. IPSEC protocol 50?

Oh, ICMP most certainly!

nice ty

in ancient theory a router is supposed to decrement the TTL once per hop *and* per second

if every implemented the second half the max. range for an IP packet would be limited by the max. 8 bit ttl counter times the speed of light

*everyone

I might be remembering the 2 minutes wrong but I remember that value strongly. But looking at RFC 791 it says 255 seconds (4.25 minutes). So... Maybe I am misremembering that somewhat.

Regardless routers and switches can't hold onto a packet forever and must discard it at some point.

then it becomes another interesting layer when 802.11 comes into the next layers up

crest: hap ax3 recommendation was awesome by the way.. picked up some cap ax access points.. now waiting for a switch to slowly replace my non-managed switches in network

what's hap ax3?

it is a mikrotik router

ah. with terrible config UI or?

polyex: a mikrotik wifi "accesspoint" with quad-core 64bit arm cpu and a gig of ram that doesn't break the bank

polyex: yes winbox is terrible, but the cli is quite decent and these days it's winbox that's lacking behind the cli

Oh! Searching I see references listing different systems using different TTL and Hop Limits and people were using it to identify the remote system OS. So likely at the time I was doing the experiments 2 minutes was implemented by the Barney box. Likely that or something similar.

crest: i even got CAPsMAn to work.. under wifiwave2, not perfect setup but central management of access points is nice :)

yes it is and it's a requirement for 802.11r

(to allow the APs to help clients roam faster)

yeah, the whole "seeing" access points thing, i am trying to figure out how to manage

make 2.4ghz not as "loud" and jumping to 5ghz but not staturating the airwaves 20/40/80 setups

Every time I see microtik my brain pronounces it "my kra tik" rather than micro-tick as I am sure they intended.

i have not touched network stuff like this in decades.. so in typical fashion.. just diving into the deep end and see what happens

iirc 802.11r provides a clients essentially the bssids and channels of accesspoints worth trying instead of having to search for them

i want to just have 1 SSID and still have not gotten it to work.. right now i have a <base> and <base>-5G

rwp: same

bssid != ssid

think of the bssid as the mac address of each access point's radio

i am trying to see if i can link "configurations" to radios on each device

each access point announcing a ssid does so under its bssid for the radio it's announcing it under

so if someone would try to add an AP to the network.. they will just be out in lala land

so i heard of bssid, enought to know i am clueless on how to actually getting it to work

you can do that e.g. to have one AP on the lower half and one AP in the upper half of the 5GHz band

Ah! MSL Maximum Segment Lifetime in RFC 793 says "MSL Maximum Segment Lifetime, the time a TCP segment can exist in the internetwork system. Arbitrarily defined to be 2 minutes." It's specifically a TCP parameter.

It's required to ensure that TCP sequence numbers are not duplicated within that time interval. As that would cause problems. Packets older than that are discarded ensuring that sequence numbers from older than that can be reused.

crest: that is the plan: 1drv.ms/i/s!Ag86nuiRCza3jbQmQowpZbJW5dyoqA?e=UFwCes

just breaking up the frequency ranges.. set it up for 2.4ghz the 5ghz i am still figuring out how

i have three ssids per accesspoint: just the name, with a -2.4GHz suffix and a -5GHz suffix

for each radio (2.4GHz and 5GHz) of each AP the frequency specific configuration is the master configuration

and the other ssid(s) are defined by slave configurations (their terminology in the config language)

so anyone can connect to any and it does its things

if the client is too dumb i to reliably pick the right frequency band the extra ssid's help

yeah, that seems to a common theme in these talks i am listening to

assume everyone/everything is dumb and remove as many variables as you can

but if you want your laptop to fall back gracefully to 2.4GHz when you're sitting in your backyard you have to allow it to pick the band

unless you covered everything with outdoor APs :-P

so i got opentelemetry collector configured to receive syslogs on localhost:514 udp. so i add a line to /etc/syslog.conf, *.* @localhost to make it send messages to the collector and service syslogd restart, but then syslogd logs: sendto: No such file or directory. what i'm doing wrong pls?

not there

polyex: are you trying to bind two services (syslogd and the collector) to port 514/udp on the loopback?

crest well i want the collector to start its syslog listener then syslog to send messages to it. and ya all on localhost

crest: not there yet.. i did find that the old wifi stuf just flooded the whole property.. but i do like the fallback things.. connect to the <base> and it _should_ switch

polyex: what's the output from sysrc syslogd_flags?

did you give it -ss to put it into secure more mode (aka not bind to udp)

crest -s

i even tried changing the ports to 5140 to get above 1024 but still not working

logger "test" shows up in /var/log/messages, but syslogd doesn't send it to the collector running on localhost

no wait

i had -ss somehow

ty!

only works when i move collector to listen for syslog on 5140, not 5`4

514 makes service collector start fail saying port is already bound to

polyex: does the collector start as root and drops privs or is it like so much of that stuff written in go and lacks a portable api to drop privs because you can't implement one on linux (because of linux braindamage aka EWONTFIX)

dunno prolly the later tbh

know how to check?

yes

how?

even if it can't drop privs you can use FreeBSD's TrustedBSD MAC framework to allow specific groups or users to bind low ports (<1024)

see mac_portacl

ya, done that. i just didn't think that was the prob because i don't get any error logs that it tried to bind to port but failed

if syslogd is stopped, looks like collector can start and bind to 514

so i guess by making syslogd flags -s instead of -ss, it binds to 514 even if it's not configured to listen or anything?

ya, default /etc/syslog.conf, -s flag, start syslogd, netstat -na | grep 514 shows it wtf!?

polyex: try sockstat -l46 -p 514

polyex, I think that's because it doesn't know, at the time it can bind to UDP port 514, whether it will have to *send* to syslogd on another host. Hence, it always opens it, unless asked not to (insecureish mode) by -N.

polyex: the double -s flag doesn't come from /etc/syslog.conf. the syslogd rc.d passes it to the syslogd process as argument when configured to do so in rc.conf

ya i know

V_PauAmma_V where is that -N documented?

which syslog collector are you using?

syslogd in base, and opentelemetry collector

how did you install it?

pkg install something?

nah had to figure it out manually

not officially supported

does the collector stay running or does it die quickly?

does it accept messages sent via nc -u 127.0.0.1 514?

dies right away. in the collector log it says can't bind to port

as which user are you attempting to start the collector?

otelcol

but we already solved it, its' because syslogd auto binds to 514

because on any *nix by default only root can bind to ports below 1024

so i need to fix that

fixing the syslogd is as easy as sysrc syslogd_flags="-ss" && service syslogd restart

-Ns did it

where is -N documented tho?

man syslogd

look for -N and -s

oh right -ss does even more

it configured syslogd not use network sockets at all and prevents it from forwarding messages over network as well

-Ns should prevent it from binding the listening socket but allow it to still forward messages using an unbound socket (that gets implicitly bound to a some high port)

unless you syslog collector requires something archaic like putting faith in the source port and expecting senders to send from port 514 it should work

you could try this: sysctl net.inet.ip.portrange.reservedhigh=513

iirc this is the sysctl to reduce the highest reserved port number (the default value should be 1023)

ok just tested. -N makes it so syslogd won't send logs to the collector on port 5410

by reducing it to one below the udp port of syslog (514/udp) any process is allowed bind to port 514-1023 (first come, first serve)

i guess -N disables all udp binding, not just the default

in that case try -ss

i't been a while since i had to tinker with this

or you could just bind them to different addresses

seems weird to have syslogd need to bind to 514 for no reason, in order to bind and send out on some other port

e.g. syslogd listens on 127.1:514 and [::1]:514 and the collector on 127.2 and [::2]:514

"nice"

i don't need syslogd listening at all, only sending

polyex: what makes you assume syslog binds port 514?

unless you see it bound by syslogd in sockstat -l46p514 it's not binding that port

sockstat -l46 -p 514 shows it

your service user is instead not permitted to bind port 514 by default because it's a so called low port in the range 0 to 1023 (inclusive)

nah i have that disabled

this is not that

if you still suspect syslogd kill it and test if that allows your collector to start

(just remember to start it again)

did

did that fix it?

hi!, i have a fresh 14.0 install. installed drm-kmod however when i try to kldload i915kms it hangs. what could be wrong ? 14.0 with i915kms has been used in the machine before, so i know it works

HER: you knows it used to work. you could be doing everything wrong if it's a regression

do you still have the config files that used to work?

can you still log in via ssh?

can you be more precise than "it hangs"? what hangs? are their any "signs of life" from the hanging system?

does the monitor stay lit? can you ping it?

crest: i have the config files, i tried to re-use them, but then it hangs during boot on "Loading kernel modules:" IIRC

i have phisical access to the machine

do you have access to a second maching on the same network?

crest: if i kldload manually it displays:

iic0: <I2C generic I/O> on iicbus0

often times drm/xorg fucks up the video is unusable, but the system is otherwise running

iic1: <I2C generic I/O> on iicbus1

and then hangs

so you can debug it via ssh (or serial)

i have access yes

what should i check ? dmesg ?

dmesg or /var/log/messages

it's also a good idea to reboot and try kldload in tmux/screen over ssh

ok

maybe it prints something use that gets lost on the failing video console

how long ago did you last had success using i915drm on the same hardware?

did you check freshports.org for changes to the relevant port?

did you already try both drm-510 and drm-515?

HER, Is this i915kms in base? Or in ports? If it is in base then everything should be consistent of course. If it is in ports then there are cases when ports is out of sync with base.

which freebsd version are you running? 13.x? 14.x? something different? stable? current? still on 12.x?

Though that seems like it should not happen with in the .0 release cycle. It's a problem I have now in a 13.3 point in time.

HER said "a fresh 14.0 install".

sorry i missed that

did it work on 14.0 before? 13.3? 13.2? 12.4? even older?

it worked on 14.0

i think i had 515

crest so basically, the only option is to use syslogd_flags -s, and port other than 514. the reason is -ss and -N make syslogd not bind to any ports to send, but taking them away makes syslogd autobind to 514 and fight with collector for that port. so only option is port other than 514

crest: it was working 2 days ago :) im upgrading the hdd, so reinstalled everything. i am pretty sure i used drm-515. and always install from 'pkg'.

really don't know why it's not possible to disable the syslogd autobind to 514

crest: kldload remote from screen hangs and disconnects =p

that makes it a lot harder/annoying to debug

rwp: the i915kms is from 'pkg', same as older install

HER: did you run 14.0 before as well?

yep

Then I suspect that the port is somehow out of sync with base.

ahh

maybe i used 'latest' in pkg settings

i will try that

instead of quarterly

rwp: normally i would suspect that as well, but there is only the 14.0 release in the 14.x major release branch

so the kernel modules *should* be perfectly compatible

crest, Right. I don't see how a .0 could get out of sync. But yet it seems like it must have.

I mean, right now there is only .0 for both base and ports and that's all there is.

can you run pkg query '%q %n-%v' | grep -i kmod

crest so basically, the only option is to use syslogd_flags -s, and port other than 514. the reason is -ss and -N make syslogd not bind to any ports to send, but taking them away makes syslogd autobind to 514 and fight with collector for that port. so only option is port other than 514

crest: paste.scsys.co.uk/27162

Title: magnet_web paste from Someone at 217.168.150.38...

changed /etc/pkg/FreeBSD.conf from quarterly to latest

HER: thx. looks correct. there no old freebsd packages hidden among the installed packages

and running pkg upgrade -f

right

you could try to install the older drm version

510 ?

yes

ok

or try to load the firmware kernel module for your gpu first?

what you mean ? its intel processor only

bah, same thing with 'latest' in pkg

i think you meant the intel module ?

removed 515, installing 510

kldload worked with 150

510

but x doesnt start =p

crest: how can i load the firmware kernel for the gpu first ?

hmm

check the firmware, install gpu-firmware-X then load it

i will try taht

can freebsd shrink a filesystem?

doing a ports install is a kitchen sink wow

Soni, Normally no. But even on Linux systems it is best not to attempt shrinking. I once had a machine spin 12 days 24x7 shrinking a file system. Now I just create a new one the right size and copy the data over. It's consistently able to conclude that way.

rwp: +1.

rwp: we recently shrunk a filesystem from 2TB down to 150GB, on linux

can freebsd do something similar?

filesystem stuff on Linux works well, especially when using LVM from the start

Last i checked its not possible to shrink on FreeBSD

I would have said so too. But I did a search and turned up this interesting recipe: askubuntu.com/questions/1231355/how…shrink-a-zfs-volume-on-ubuntu-18-04

Title: How can I shrink a zfs volume on ubuntu 18.04? - Ask Ubuntu

I assume this is ZFS we are talking about? Not UFS?

okay let us boot up the VM 1 sec

I still think the most reliable path is to create a new file system of the desired smaller size and then copy the data onto it. Do a backup test by doing a full backup and then restore on to the new smaller size.

rwp: we have nothing that can read UFS

I wonder if that askubuntu.com/questions/1231355 is using a new feature because I don't recall that one could remove a stripe in previous days. But if you can then yes that is one way to do it. I would want to test it out thoroughly first before increasing the size and making the problem worse.

we mean sure we can come up with space for where to shove the smaller filesystem

Soni, A FreeBSD system? So, is it a UFS file system?

but we need to figure out how to do this with qemu properly

Am I to assume this system is using UFS? And that it can't be booted up onto the network for the data to be copied off? It's an offline system?

that Ubuntu ZFS guide is not "shrinking" but rather backup -> destroy pool -> re-create pool -> restore

that can be done on any filesystem/os

presumably we can throw -device virtio-blk,drive=ext -blockdev driver=raw,node-name=ext,file=/dev/sdc at the VM?

rwp: crest: just for info, so i reinstalled 14.0 and now kldload i915kms works. So i have no idea why it didnt in the other install

mjp, I don't think so. It started with a non-striped array. It added striped disks of the new smaller size. Then removed the original larger striped disks leaving things back to a non-striped array of smaller size.

HER, Glad to hear things worked out for you! Regardless of whatever glitch occurred. That was an odd error.

Soni, Here is a reference I have saved in case I need it: gist.github.com/ctsrc/9a72bc9a0229496aab5e4d3745af0bb9

Title: Install FreeBSD 14.0 on Hetzner · GitHub

dont think what exactly?

I don't think it is a backup restore procedure.

In the LVM world it would be like adding a PV to a VG, then doing a pvmove to pour the data from the old PVs to the new PVs, then removing the original PVs. As the person said, all live while the system is running. (Which I still think is pretty cool.)

Soni, If it were me I would back up the data onto another machine using rsync and then resize and then restore with rsync. The system temporarily holding the data does not need to be a FreeBSD system. Which I say since it seems having another one is a problem.

And if I were doing that I would take the opportunity to create the new file system as a ZFS file system, because, ZFS FTW! :-)

the hardest part of this whole issue is figuring out how to use qemu

I am still not understanding why it would be needed?

For the Hetzner recipe pasted it's because there is no bootable FreeBSD available so one must make due booting a Debian rescue system and then going from there. But your FreeBSD VM boots already so none of that would be needed as far as I can see.

from experience and heartache, shrinking no bueno.. default to rwp actions