how do you make software that calls freebsd features? like if i made an interface manager would it just exec 'ifconfig' binary or is there a more direct API to the OS?

in my program if i use the FS api and create a dir, i doubt it just runs "mkdir ..." under the hood so that's an example of a more direct API to the OS no?

system()

is there any way to know when the next quarterly package release will be?

polyex: what is the "FS api"

and it does run mkdir under the hood

if (mkdir(path, last ? omode : S_IRWXU | S_IRWXG | S_IRWXO) < 0) {

man 2 mkdir

Good morning! I filed bugs.freebsd.org/280028 a week or so back, and am wondering what the next step I could do to help fixing it would be.

As I understand it, it must be a userland issue with graphics drivers. But I'm unsure of how to debug further.

If I understand correctly freebsd-update is only able to install all patches between 14.1 and 14.1p1?

Thus my next step is to clone the freebsd git tree and start rebuilding on certain commits to attempt to find the regression point?

+1 if you know a point where it worked, and where it didn't, I'd run a `git bisect` to find out where the bad commit is

When attempting to follow docs.freebsd.org/en/books/handbook/…ng-edge/#updating-src-obtaining-src it seems mv fails, because /usr/src is a separate filesystem.

I'm not great with zfs, but the general steps would be unmount, create new dataset, mount new dataset?

Running `umount /usr/src`, `zfs create zroot/src-git` and `zfs set mountpoint=/usr/src zroot/src-git` appears to have worked.

Will my new dataset conflict in mountpoint with the old one, or are there other gothas one should think of before attempting to add those three commands to the handbook?

Heh, apparently two datasets on the same mountpoint are a thing reddit.com/r/zfs/comments/12qldul/t…ounted_on_same_mountpoint_home_what

I'm no zfs expert, but maybe `zfs clone` is the way here

Or wait, make a snapshot :-)

zfs snapshot zroot/usr/src@bak

Anyway I'm heading to sleep, good luck fixing the graphics issue!

indigo: thanks, and sweet dreams!

I went with `zfs set canmount=noauto zroot/usr/src`, which seemed most reasonable at first glance

Documentation issue for my above problem created on bugs.freebsd.org/280186

I'm trying to setup an sftp server, but neither internal-sftp nor sftp-server exist

which is what I'm assuming I should be putting in ForceCommand in sshd.conf

am I tripping?

fengshaun: you mean you don't have a /usr/libexec/sftp-server file?

oh wait, it does exist

sshd still rejects connections though

I have

Match User myuser

ForceCommand sftp-server

ChrootDirectory /some/dir

and logging in kicks me out with auth fail

if I enter the password wrong, I get 3 tries, but if I enter it right, get kicked right away

alright, the problem was bad permissions on chroot directory. should've just checked the logs *facepalm*

The website doesn't reflect that 13.2 is EOL

but we know

Hi. Anyone using Bastille for managing jails and uses it with ipv6? All the docs show examples of either ipv4 or ipv6, but if you have both I don't see a way to configure it.

My first thought would be to create it with ipv4, stop the jail, edit the jail.conf to add the relevant ipv6 lines as well, and then restart the jail

scoobybejesus: yes, did it this way. Bit silly though that the tool itself appears single stack

Perhaps. I imagine it might be a breaking change to the syntax/usage if one were required to disambiguate the optional interface from additional IPs, so that may be why they haven't added it yet

yeah, that's how I use BastilleBSD -just edit the jail.conf directly for stuff I need done.

I tried to use CBSD, but it's using a sqlite database to implement the jail.conf at start, and I can't sort out how they want the IPV6 addressing in there. CBSD looks like a better tool overall, and I'm sure the IPV6 is great, but no idea how to get that to work.

Is there way to do a "smart" file sort. . . say if the files are named with non-zero-padded numbers?

CrtxReavr: sort -h will sort 1 2 3 before 10 11 12

Well, not by itself. ..

Not by itself?

Basically I had foo_1, foo_2, ... , foo_10, foo_11_, foo_12

Yeah, that complicates things

What I finally ended up with:

ls -1 c7* | sort -t _ -n -k 2,2

From a cow-orker

With gnu sort(1)

Looks like it works with regular sort too

CrtxReavr, Are you aware of sort -V? Maybe that is the feature you need here.

for j in $(seq 1 20); do echo foo_$j; done | sort | sort -V

This version of gnu sort(1) supports no such switch.

Do you need to use gnu sort?

Oh. . . actually, it does, just in a peculiar order.

What made you think that it did not work with both FreeBSD sort and with GNU sort both?

And also, me checks notes, I might see that this is #freebsd not #gnu here so one would expect to use a #freebsd solution, no? :-) :-)

what is the status of the ZFS bug where trying to 'zfs send' an encrypted filesystem random causes it to be marked as corrupt? will we ever get a fix for this?

Hello. I'm trying to run Sway in a FreeBSD VM. Host it Linux, I'm using libvirt with QEMU. I'm getting the following error: "Found 0 GPUs, cannot create backend". 0x0.st/XBn8.txt I assume that I'm missing some graphics drivers package. How can I fix this?

has that zfs bug been reported by freebsd users?

scoobybejesus: yes, by me several months ago (on fs@) and my mason in this channel

s/and my/and by/

ah okay. ISTR an episode or two of 2.5 Admins where AllanJude mentioned that the bug reports were generally coming outside the FreeBSD community, and that there didn't seem to be reproducibility or ability to inspect the data, but hopefully it's being figured out now

i don't know what Allan said specifically but it seems like the issue is not being figured out since there's been no movement on the bug since January: openzfs/zfs #12014 - unless perhaps Klara are working on this internally...

are you corrupting data, lw?

jbo: you're talking to me again now?

wtf? did I accidentally ignore something?

well you kinda told me to fuck off on Signal so i assumed i'd offended you somehow

I did what?!

since when is "alright then, take care" after a conversation equivalent to "fuck off"? :D

well... in any case apologies if it came across as such - that was certainly non-intentional. my doors are open (figuratively speaking) - A closed door is a happy door.

this might be my crazy brain talking, i interpret "take care" to mean "i no longer wish to talk to you, but i an expressing this in a polite way"

but in any case i am not corrupting data, it is syncoid doing so, via the aforementioned zfs bug

status: One or more devices has experienced an error resulting in data

corruption. Applications may be affected.

but this is a lie, there is no data corruption on the device, it's in-memory data corruption and rebooting will fix it

aye. I use "take care" as a closing statement when a conversation between friends comes to an end. I might be misusing it tho - not a native english speaker.

(i just can't reboot right now because people are using the filesystem... which is what makes this bug so annoying)

lw, regarding that bug. does the data corruption error happen on the source host or the backup destination host? I truly hope that this bug does NOT corrupt the source dataset.

I might be in trouble otherwise.

it happens on the source and it does *not* actually the dataset

it just causes the host to *think* the dataset is corrupted (= it will return I/O errors) until you reboot it

s/actually the/actually corrupt the/

so I can have a ZFS native encrypted dataset, do a zfs send (for example via sanoid/syncoid) and then I can no longer use the dataset on the source host?!

also, it only happens if the dataset is using ZFS native encryption

(GELI is not affected)

jbo: correct, but it only happens randomly and not very often - this host runs syncoid 4 times per day and has not reported the problem for 2-3 months until it happened yesterday. but it's guaranteed it will happen at some point.

do you really have to reboot the host or can you just export/import the pool or whatever?

on that i'm not sure, but since the problem seems to be related to in-memory data corruption, i prefer to reboot

whats th ebest way to backup data...

or well, what would you guys recommend

polarian: zfs + syncoid (part of sanoid)

+1 on that

jbo: also, it mostly only affects snapshots in my experience, so the dataset itself still works fine, you just can't send any of its snapshots

but since that means you can't back up the dataset anymore ...

lw, still makes me unrestful

yeah, i really dislike that a) this bug exists, and b) no one seems to care to fix it

i still think zfs is better than any other filesystem... but...

aren't ZFS bugs usually pretty high-priority items?

you'd think so?

i really don't understand why the bug hasn't been updated in 6 months

if i were OpenZFS, this would be my #1 higest priority issue to fix

time to ask/bump?

incidentally the bug was first reported in 2021

does allan know?

idk, he did just email me about a job at klara so maybe this would be a good time to ask :-D

maybe i'll reply to my fs@ post to mention this

nice - congratz!

jbo: you like music right? youtube.com/watch?v=tRI7LLkq7E4 (not related to zfs)

I do in fact also like music that is not related to ZFS, yes.

lw, I'll tell allan that you were mean to me

lw: GPLv3 :/

polarian: so? are you going to modify it?

what makes syncoid good?

it does exactly what you want and it's set+forget

it just uses zfs features under the hood, no custom formats, files etc.

i'm sure there are other zfs-based snapshot/backup scripts if you don't like sanoid though

hmmm

sanoid is just what i use and find it works well

you can even write your own, it's not all that difficult

that script is why I don't get job offerings from allan

jbo: i interviewed for a job at Oxford University a few years ago and part of the application requirements was to provide a Perl script you'd written, which they decided to ask about during the interview. so most of it was like "yes, this is not the best way to do it but it works for what i was trying to do". i didn't get the job.

lw, perl == WOL

Wake on LAN?

write-only-language

heck I haven't written a perl script since 2014

lw: well you must be knowledgable if they even bothered to interview you :)

jbo: also i didn't get a job offer from Allan, he just followed up on the mail i sent them like 6 months ago asking if they had any jobs going

so we'll see what happens there...

why did oxford want you to know perl though

couldn't it have been shell?

or python?

polarian: this was a Unix admin job like 10-15 years ago, knowledge of Perl was basically expected back then

ah, that aligns with my "I haven't even written a perl script since 2014" quite well then

tbh i'd still be a little dubious of a Unix admin who didn't know perl, even though i don't use it myself nowadays

sandoid+syncoid is great for my use case (auto snapshots and auto sending of snapshots, pull style in my case, for backup purposes)

+ retention policies for auto deleting

lw: right... so I would have been in primary school then when you applied :)

maybe secondary...

scoobybejesus: couldn't you just cron it?

also according to the handbook, snapshots != backups

you should also do full backups

right, snapshots are not backups, syncoid turns snapshots into backups by sending them to a remote host

polarian, you make snapshots (which are not backups) and then copy them to another host (which is a backup)

yeah, sanoid and various syncoid jobs are scheduled in cron

jbo: so if you lost the zfs volume

say you reformatted it

can a snapshot be used to rebuild the entire pool?

polarian: ... then you would just create a new zpool and send the remote snapshot to the new pool

polarian, then you pull the data back from the backup

why not just use zfs send/recv then

whats the point of syncoid

syncoid *does* use zfs send/recv

syncoid does zfs send/recv

the point is it does the sending for you automatically

oh so its basically just a useful script then?

it just gives you a nice config file to set "hey I want one snapshot every 24 hours"

to make it more simple

it works out what snapshots don't exist on the remote host and sends them

I thought it was more fancy

why is it like... 2000 lines then

and does things like adds pv for a progress bar and adds mbuffer, and things like that

yes, it's nothing you can't script yourself if you wanted to (like i said earlier) it's just a pre-existing thing that does this

when it comes to backup solutions, you don't want anything fancy - trust me.

it seems to follow gnu coding conventions of trying to have as many lines as possible

jbo: +1 and also you want something which is well tested

sanoid/syncoid is widely used and well tested, so you can depend on it

or I can try to use my peanut brain and make my own shell script :P

I have always wanted to write my own backup script

seems like, yes, a large benefit is that it responds well to a simple config file

lw suggested sanoid to me a few months ago (half a year by now if not more). I was doing zfs send/recv myself before with custom shell scripts but that was just a PITA to maintain especially for production machines that have 100 jails. switched to sanoid and have been happy since.

scoobybejesus: ohhh so a lot of it is parsing of data then?

(I haven't read the code only looked at the line count and was shocked by it)

polarian: i think it's worth writing your own zfs backup script so you understand how all the pieces fit together, but tbh, i'd still prefer syncoid

I will stick it on my todo list then

you don't have to use it. use something else then.

thanks

I am still a little confused about the handbook

we're just answering the question of "what would your recommendation be". if you don't like the recommendation that is fine. we're not a rust community - you're welcome to use what you like for your job.

it says snapshots shouldn't be used as backups... yet send/recv'ing them is considered a backup?!!?!?

no you're misunderstanding it

so does it mean, keeping it in the same pool is not a backup?

polarian, snapshots live on the same disk/hardware as the data itself. so it's not a backup. but once you send|recv it to different disks/hardware, it can become a backup.

it means in the sense that RAID is not a backup, because you can accidentally delete files and RAID doesn't protect you from that

jbo: but I would assume you would copy it to a different pool anyways

snapshots are not a backup because you could accidentally delete the pool or your server could be hit by a meteor

thats common sense no?

a backup requires the data be stored on a remote, geographically distinct host

you can *use* snapshots to copy data to such a remote host, which is a backup, but a snapshot *by itself* is not a backup

yeah like to a thumb drive that lw carries around when going to the disco

lol

I can't find any UK-based block storage

its all German or Finish datacentres

Finnish*

"snapshots are not a backup" means a snapshot *by itself* is not a backup, it doesn't mean you can't *use snapshots* to create a backup solution

thanks for the clarification

polarian: my backups are on Jump, who have by far the cheapest block storage of any UK VPS provider

it is in UK datacentres correct?

lw, what's your account password? I'm asking for a friend.

yes

ok cool

I don't like the concept of overseas storage of personal data...

even if its encrypted

polarian: also it's run by cool people who know what they're doing, i highly recommend them

polarian: irc.jump.org.uk #jump if you have questions prior to ordering

lw: any other alternatives to jump you recommend? just looking at my options

rsync.net for the win. you can just zfs send|recv directly to them.

jbo: where is it hosted

but i have 1 VPS, 8GB RAM, 1.5TB block storage, and i pay £21/month

polarian, I know you said you're lazy but c'mon...

polarian: the only others i can think of are Mythic Beasts (good, but expensive) and Bitfolk (also good, but they are hosted at Jump anyway)

lw: I can't reach jump...

Mythic Beasts I have already used in the past... they are outstanding... they are the definition of you get what you pay for...

you can't "reach" them, what do you mean

jbo: non-UK storage no thanks...

lw: can't connect to their website

times out

that's weird, is your ipv6 broken?

IPv4

my laptop doesn't have IPv6 (too lazy :P)

my servers have IPv6 though

mtr and traceroute seems to suggest jump edgerouters are dropping all packets

oh yeah, their IPv4 site does seem to have some issues

oh...

wait nvm its making it through the edgerouters, its the internal network dropping it

polarian: i nudged them on irc

Thanks

it is cool they have an IRC channe;l

lw: by any chance you used AAISP before?

you seem into the tech savvy companies, so if you haven't tried AAISP I will be shocked

polarian: AAISP is our ISP and also my mobile phone provider (sip2sim)

yeah thought so :P

pricy, but well worth it

it's not cheap, i pay about £80/month for internet + sim service, but yes it's worth it

they provide my server network, and have been extremely useful

plus they entertain my idea of running a BSD router (currently OpenBSD)

i hope they'll add some better options for sip2sim data though, at the moment the only option is the £35 'unlimited' plan which is more than i want to pay

iirc they were phasing out the sims

their contract with O2 is coming to an end iirc

wait I am pretty sure I have seen you in #a&a how could I forget...

that's old news, they phased out the old service, they have a new service with ONSIM/Mobile-X

which is based on EE with O2 fallback

thats much better

EE network is fast... even if the coverage isn't great

it certainly is for me as i get full 4 bars with EE, O2 only had 1-2 bars at best

I looked into getting a sim from a&a but for the amount of data I would use it would have costed me £50/month

gotta have all the bars

which is eye watering considering spusu gives me more than enough data for £9/month

and uses EE's network

and supports esims

polarian: i believe the new "unlimited data" SIM is £35/month vs. £10/month for voice-only

thats not bad...

it's still a trial service though and IME that's really the case... sometimes calls randomly don't connect

aiui they're raising this with ONSIM and working on fixing it

(at least that's what RevK and AA-Steve said when i complained the other day)

onsim uses the business backhaul too

which is a lot less congested

sounds cool though

yeah i've been pretty happy with it so far, certainly the customer support is a lot better than O2

not sure if you followed my O2 saga in #a&a but i had no service at all for 3 months and O2 literally did not care

I like how they limit hotspots

but its so easy to spoof it

you remove the inital hop to the phone...

so the packet looks like it has come directly from the phone

I assume this would be illegal though

yeah, i'm 100% sure that's not AA's policy though, it's something ONSIM imposed on them

lw: my friend uses O2... we both are in London and he doesn't seem to have any issue

so maybe its a regional thing

people tell me three works really well and to use smarty, I did use smarty once... the three service was so horrific I had to cancel

I would average 100kbps

polarian: i'm sure it is, or perhaps it was a me-specific thing because i'm sure a nationwide 3-month outage of O2 would have been reported in the press :-)

it didn't matter where I was in North London

I was lucky to get 100kbps

but point is, they didn't care at all or do anything to fix it

even worse it was ~15% packet drop

I will never trust 3 again

tbh i might order the £35/month data sim, it is a lot of money but... feh

someones got money to burn

I dual sim

i really don't have much money (i earn £20k pa) i just care a lot about functional internet access

I have a lebara cheapo sim which I use for calling and as backup data... (5GB of vodafone data) and then I use a spusu esim as a data sim which uses the EE backhaul and is FAST

vodafone has the coverage almost everywhere but is pretty slow everywhere (~10mbps)

EE where it has coverage can easily get into 200-300mbps on LTE

I assume 5G would be faster too...

lw: how do you afford AAISP then?

polarian: £20k pa is £1,600/month, AA costs £80/month?

i pay far more in rent and groceries than i do to AA

like if i moved to TT maybe i'd pay £20/month for Internet but i'd only be saving £60/month which is the cost of two takeaways

so i just eat less pizza and get better Internet

so... i have a system, and the primary drive just went out on me. i didn't make any backups because it is a pretty basic setup. but i have a set of drives in a zfs pool. if i drop a new drive into the computer for the primary drive, enable zfs, then run `zpool import -a` will it reconnect my pool? or are there other things i'll have to do?

desnudopenguino: you probably need to forcibly import the pool as the host will have changed - run 'zpool import' for a list of importable pools, then 'zpool import <name>' then it'll probably tell you that you need to run 'zpool import -f <name>' which will actually do the import

s/host will/hostid will/

lw: okay, cool! that seems easy enough

the -f is to override the hostid mismatch, which is intended to prevent a pool being imported on two systems at the same time

that makes sense.

phew that's a load off of my mind. i wasn't sure if i'd have to jump through other hoops to get it working with a fresh install. i should probably make a backup of my OS disk somewhere available outside of the box so i can just dd it to a new disk. then i'll be prepared for this

you could always zfs send your OS disk to the redundant zpool

then you could restore the OS by just booting from the install media and sending it back to the new root disk

now that's an idea! do you mean set up a zpool for the OS disk, with redundancy (like a mirror?) then if one disk goes out, boot it and use the backup disk?

desnudopenguino: well, ideally your boot disk would be mirrored, but i was thinking in the case that it's not, when the disk fails, you just replace it, then zfs send the backed up datasets back to the new disk

oh! that's a neat idea as well.

lw: if you are knowledgable, why aren't you in more skilled profession?

I doubt computer scientists are paid £20k

the entry wage is £30-40k iirc

also thanks for the help

"more skilled profession"? what do you mean?

i work in IT, my day job involves writing F# code for data transformation in local government finance

seriously

they only pay you £20k?

for part time work, yes. there's not a huge amount of demand for this work

hmm

sounds sucky

which is why i applied to Klara...

if you got so much knowledge how come nobody takes you on? competitive market?

i have crazy brain and can't work for people?

work for yourself?

sure, then i'd make no money at all

(I know its not that easy)

sure... you need a business stategy of course

do you have any idea what it's like to have a personality disorder? i'm frankly amazed i'm not homeless and that's mostly because of luck

ohhh its a group of disorders

well I guess congrats on being self sufficient :)

my particular PD has a 10% incidence of suicide so honestly i think i'm doing pretty well just to be talking here

polarian: [#jump] <james> www.jump.net.uk ipv4 problem is fixed, was a faulty static route on the host machine

lw: thanks a ton

polarian: this problem aside, i still recommend Jump... this was only about the www machine specifically

oo jump is in telephone house in the docklands

lw: iSCSI looks appealing but £25/TB is quite a lot... I get it is because of all the redundancy they provide... a vps looks ideal, £10 for 1.5TB of RAID 1 HDD for backups

I can then expose it as block storage to be mounted on my home server... or I can copy zfs directly over to it...

polarian: ignore the iSCSI stuff, that's mostly aimed at people with colo servers

lw: they do supports freebsd right?

on request?

polarian: i don't think they updated the website yet, but yes, they do support freebsd now (after i nagged Andrew about it)

lw: legend

this looks like a good solution as a cheaper alternative to mythic beasts

thanks a ton

lw: you ever been to *bsdcon?

heh no

ah... I am going this year... was just curious

is there one in the UK?

looking at company house filings jump network ltd is pretty small...

i could probably expense that...

lw: no... the closest is EuroBSDCon which rotates around Europe... it will be in Dublin, Ireland this year, hence why I am going

it wont be held in UK for at least 20 years

polarian: yeah it's a tiny company, it basically started as a couple of people with a server rack who wanted to offer colo to people

which, to me, is a positive thing

james is online on #jump now btw if you wanted to ask anything

small companies <3

I am going to go watch house of dragons, thanks a ton for all the advice

Is there any way to issue ATA secure erase on FreeBSD, or is hdparam used like on Linux?

hdparam is within freebsd man pages... but i cant seem to find it in the port tree or anything either

any ideas?