Delemas: your PATH will dictate which one gets used when you invoke 'openssl', using the full path will give you a specific version

Delemas: the information from pkg-audit comes from vuxml, you can find a fully-rendered version at vuxml.freebsd.org/freebsd/index.html which will do a better job of describing what packages are vulnerable

Title: FreeBSD VuXML - entry date index

parv: i usually start with shallow clones, and then give days later i have the full checkout, and three forks…

interesting, so a 12.2 ko does seems to load on 12.3. I guess all major version's are binary compatable?

That would be minor version difference

Or, did you mean major minor versions? ;-J

yeah, 12.x for example

I do not count on the being compatible

KBI and ABI isn't meant to change with minor versions, but without details it's hard to say if that's the case here.

I probably should play it safe and downgrade back to 12.2

s/the/the minor versions/

at least until I can get a stable driver for 12.3

It's for this peice of hardware areca.com.tw/products/nonraid-1320.html

What's the driver called?

arcsas

Doesn't look like it's in base or ports?

it's not

I have to grab it from the manufacture, they provide binaries

Depending on how much the server is used, or if you have another machine and another HBA to test with, it'd be great if you could find out which commit broke the compatibility.

That can done by starting with a 12-STABLE snapshot and then using git-bisect.

Yeah, I have a few of those HBA laying around, they are my favorite, perform really well, etc.. I got to see why their latest versions (2.x series) don't work for me.

I mean.. they load, but the system just goes bonkers and just hangs up

It should be relatively simple to isolate the commit; using an older 12-STABLE snapshot where it works, then using git to grab the source (I'd recommend turning on meta-mode, as this speeds things up rather a lot on subsequent builds). Then you use git-bisect(1) and just build and install the kernel - if it loads once you've booted it, you can build a new version, and once you encounter a bad version you

simply boot the older kernel that the installkernel target saves.

Isolating the commit will allow us to try and determine what happened, and if there should be a fix for it.

Hmm, strange. Network interface appears to randomly be going down, "Jan 1 08:55:37 router0 dhclient[49327]: send_packet: Host is down"

`service netif restart em0` brings it back up fine. I wonder what's causing it to drop out.

Does the difference between "used" & "lused" (used > lused) -- termbin.com/yxll -- mean that compression should be disabled?

parv, Disabled? Because it isn't very much difference?

What does this say for you? zfs list -o compression,compressratio,used,usedds,usedsnap,refer,name

rwp, There are no snapshots; "used" & "refer" are the same number: termbin.com/yzxh

So it looks like you are getting 21% compression. That seems useful enough for me that I would keep it.

rwp, I would not have minded if there were small difference iff "lused" were larger than "used" instead of vice versa

Hmm... I admit I missed that ratio direction.

Space usage for source, termbin.com/u5zp2 # the usual case of compression

I don't normally look at lused. I have one case of it being that direction. lz4 2.19x 1003M 968M 1003M 0B 1003M zroot/usr/ports

With lused=968M used=1003M refer=1003M

Same here re checking "lused" value; got curious after a run of "git-repack; git-gc"

Ok, had set the compression to "lz4" for datatsets with used > lused.

zfs-find-useless-compression.sh : termbin.com/crx1

Hi all, happy newyear

Does anyone know why zfs only releases space of deleted files after a reboot?

like, I had 200GB free on a pool, I deleted 800GB of data but only got back like 20GB

I rebooted, bam, over 1TB of free space on the pool

this is pretty meh as I'm in the process of replacing a big part of my linux iso collection with updated versions, and I keep running out of space while a reboot gives me back TB's of space

no snapshots used as linux isos are easy to get

using a zpool of 2 raidz1 vdevs. each raidz1 vdev has 4 spinning rust HDDs

wow that sux

yeah. I know I should add more disks, but that's for another time

mvanbaak: which FreeBSD / ZFS version?

freebsd 13.1

13.1-RELEASE-p5 amd64

zfs-2.1.4-FreeBSD_g52bad4f23

zfs-kmod-2.1.4-FreeBSD_g52bad4f23

I confirmed there's 0 snapshots with `zfs list -r -t all storage/multimedia`

forums.freebsd.org/threads/reclaim-lost-zfs-space.39294 people having similar issues, but i don't see a clear solution. zfs scrub didn't help

Title: Reclaim lost ZFS space | The FreeBSD Forums

NAME AVAIL USED USEDSNAP USEDDS USEDREFRESERV USEDCHILD

storage/multimedia 550G 35.4T 0B 35.4T 0B 0B

this cant be true as my linux-iso automation just deleted roughly 800GB of data in 30 to 50GB files

this kinda thing keeps popping up in my search for an answer to mvanbaak's issue, askubuntu.com/questions/1402578/how…ter-deleting-syslog-file-with-18-gb and it's just deeply fascinating to me

Title: 20.04 - How to restore the disk space after deleting syslog file with 18 Gb? - Ask Ubuntu

whehe

I've been googling for days

but haven't found anything that helps

ok, it looks like the files are in use or something

I 'fixed' it without a reboot but with: `for a in bazarr bazarr4k pmm pms qbit radarr radarr4k sab; do iocage restart $a; done`

which is strange as none of these services had any mention of the old files anymore

mvanbaak: what does lsof give you? Well, uhm, you'd have to do that before deleting them, i guess

yeah, the delete is automated

I have linux iso X-version.1 and the tool gets X-version.2, when download is succesful X-version.1 is removed using 'unlink' and X-version.2 is moved into the directory X-version.1 was in

mvanbaak: BTW, iocage has fallen out of support, iocage/iocage #1292#issuecomment-1368531213

Title: Pull `freebsd-update` from active branch of freebsd/freebsd-src by aWZHY0yQH81uOYvH · Pull Request #1292 · iocage/iocage · GitHub

you could truncate before the unlink, given that something is holding onto the files. a server maybe?

ugh

how are you downloading them?

most are usenet, some are torrent

so it's either the sab or qbit jail that grabs the data

torrent would hold on to the files for seeding, unless you explicitly say not to

yeah, I know. they are deleted

(not sure how Usenet works any more…)

if a file in qbit download directory has no hardlinks to it, qbit removes them after 30 days

usenet is just grabbing the files and moving them into the right location

it's like fetching emails

hum, I suspect something in plex

lsof is going to be the tool of choice to track this down

the moment I restarted the pms jail, I gained 450GB of free space

ok, I'll look into that

thanks a lot meena

yeah, Filesystems queue up operations. If something is still holding on to a file it's not really deleted, until that process releases the handle

the file was not used for active playback of the content

so it makes sense that a reboot would fix it. and the fact that rebooting the jails fixes it, too, narrows it down to one of the jails

there were 0 active sessions in plex

github.com/freebsd/freebsd-quarterly/blob/main/2022q3/lsof.adoc i wonder if this work has been released yet

Title: freebsd-quarterly/lsof.adoc at main · freebsd/freebsd-quarterly · GitHub

ah, lsof is a port

yeah, could well be it's released then already

it segfaults

lol

cool cool cool

`lsof -v` works

`lsof /path/to/file` segfaults

anything useful in the core?

(time to rebuild with debug symbols!)

I'll need to rebuild it indeed

lsof-org/lsof #263 this one?

Title: lsof segfaults in FreeBSD jails · Issue #263 · lsof-org/lsof · GitHub

263 – There is absolutely no documentation for the dialog library bugs.freebsd.org/bugzilla/show_bug.cgi?id=263

yups

I have it on the host as well

lsof-org/lsof #264 wonder if this is skladu relevant to your use case

Title: [BUG] lsof +L1 produces no output on FreeBSD after file becomes unlinked. · Issue #264 · lsof-org/lsof · GitHub

i'd stick with the host, given that lsof, in theory, should know about jails and other containers

ok, I'll use fstat

which comes with base

I just realised, the person who's work I'm very excited about, github.com/dfr/ocijail has been emailing me about my work alpha.pkgbase.live

Title: GitHub - dfr/ocijail: An OCI runtime using FreeBSD jail

yeah, I guess I'll have to go hunt for an iocage replacement

mvanbaak: how many jails on how many hosts do you manage? and what iocage features do you use?

pot, bastille, cbsd, are all viable alternatives that are maintained.

freshports.org/sysutils/mkjail might be an option, if you have few jails, and really just need a tool to bootstrap and upgrade them

Title: FreshPorts -- sysutils/mkjail: Fat jail creation script

35 jails on one machine, bootstrap, upgrade, snapshot/rollback, snapshot+zfs send to backup server

I create 'thick' jails as iocage calls them, so I can easily send them to my backup server (and I also used it twice to upgrade my homelab server)

is a gaol where only /usr/local, /etc and /var are writable a thin gaol?

(that's a wrong question, ignore)

mvanbaak: I also prefer thick jails.

anybody with some bectl knowledge around at this time of the year?

tct: some. try me

meena, when I first started using BEs I used them incorrectly. Back then, what I did is creating a new BE, activating it and booting into it. essentially leaving the 'default' BE dangling. Now I want to "fix that" (eg. getting back to always staying on the default BE and instead just creating a new BE when needed and only booting into it when I have to)

for this, I want to delete the current 'default' BE and re-create it from the current state.

here's my bectl: pastebin.com/nVACKCYd

Title: jbo@knox:~ % bectl listBE Active Mountpoint Space - Pastebin.com

tct: I don't follow. why? why don't you delete the other things you don't need?

when I try to do: bectl destroy -o default I get this:

cannot destroy 'zroot/ROOT/default/git': dataset is busy

unknown error

meena, I'd like to "always" be on the BE named 'default'

bectl activate ...; bectl delete ...; bectl rename ...;

s/delete/destroy/

V_PauAmma_V, the problem is that bectl delete doesn't work. apparently there is a busy dataset but that shouldn't be the case as the 'default' BE is not the currently active one

bectl destroy* indeed

Even after a bectl activate?

V_PauAmma_V, yes. as you can see from my linked paste the BE named '20230102' is marked both 'N' and 'R'

why is it busy?

Mounted somewhere maybe?

not with `bectl mount`

Maybe with zfs mount? (Or do you need to reboot after the bectl activate? I'm not very familiar with BEs.)

already rebooted. the 'Active' column of 'bectl list' also indicates that it's the currently active BE

I'll check for other mounts

mount, zfs list, lsof, fstat, etc

on it

my jails are all mounting datasets from zroot/ROOT/default although the 'default' BE is not the currently active one

should I be worried? :s

or does bectl somehow alias whatever BE is currently active to zroot/ROOT/default?

tct: how did you create your jails?

meena, using sysutils/cbsd

:D

evaluated it like 5 or 10 years ago - haven't looked back

mainly using that because it also allows to manage VMs (bhyve, xen, qemu...) all in one tool

and it's just shell scripts, ofc.

Title: Installing CBSD - CBSD

tct: basically, you told cbsd to use your ROOT as a base for creating jails, and now that's a problem

meena, that is my current theory too - and I don't know yet how to test this theory and how to recover from it (if it is the case)

might have hosed this one

meena, however, I have four different hosts all running a similar setup (FreeBSD 13.1 with BEs and sysutils/cbsd)

the only difference on this host is that I am not on the 'default' BE

tct: what's the "root" of your cbsd installation on this host vs others? what method do you use (clones vs snapshots) on this host vs others?

maybe all you need to do is a targeted "promote"…

meena, cbsd working dir is set to "/usr/jails" that is the same on all other hosts too

meena, I have some jail snapshots which are created using `cbsd jsnapshot mode=create`

meena, but I can't delete those jail snapshots either :D

but you might be able to promote them, and then you'll be able to delete the old "default" BE

meena, promote in which context?

zfs dataset?

meena, I don't have any ZFS clones, only snapshots

clones are created from snapshot

how can I verify that my jails are indeed using datasets from the "incorrect BE" ?

terminology might fail me on this one.

still don't fully understand the situation

tct: the core issues is that something (probably one of your jails) is still holding onto zroot/ROOT/default/git. it might be mounted, or a snapshot relationship, or something is directly accessing it

tct: `zfs list -o mounted,jailed,canmount,name,mountpoint -r zroot/ROOT`

meena, indeed. an excerpt of `zfs list` is shown here: forums.freebsd.org/threads/unable-to-delete-be.87616/#post-593576

Title: Unable to delete BE | The FreeBSD Forums

tct: looking at what you posted in forum, here's what I would do

tct: how did you create your BEs?

meena, bectl create <name>

zfs unmount zroot/ROOT/default/{blog.,cppproperties,emby,git,irc,nextcloud}

that won't shell expand, but my meaning should be clear

then, work out where /usr/jails is mounted (I guess its a subdir of / i.e. zroot/ROOT/$ACTIVE_BE )

zfs create -o mountpoint=/usr/jails -o canmount=off zroot/usr/jails

zfs create -o mountpoint=/usr/jails-data -o canmount=off zroot/usr/jails-data

this makes 2 container datasets (non-mountable, just for keeping your layout simple)

then you can move these 3 datasets under that dir

zfs rename zroot/ROOT/default/blog zroot/usr/jails/jails-data/blog

etc

dch, is this with the goal of getting the jail zfs datasets "outside" of the BEs?

I personally would skip all the -data suffixes, life is *much* simpler if each dataset maps cleanly to each path

after this, you can run `zfs snapshot -r zroot/usr/jails/jails-data@`date -u +%Y%m%d-%H%M` just before doing upgrades

I might have some knowledge gap here: Wouldn't the idea be that creating a BE also "includes" my jails? And when I boot into a different BE the datasets are "re-mapped" and the jails will also be as of the state of the BE?

tct: you can do either

but as somebody who runs a lot of jails, I tend to separate out the base OS and jails completely

so I typically do upgrade/update FreeBSD in base OS (all of zroot/ROOT/default )

check all things work

and then upgrade/re-deploy all jails one by one

dch, thank you for outlining your idea above. Before attempting any of that I'd really want to understand this situation better. I have no idea how this situation came to be. My idea/intention is that creating a BE also includes jails. cbsd puts everything under /usr/jails which would be part of what the BEs cover, no?

tct: you may find freeside.skunkwerks.at/~dch/FreeBSD/diy-jails/tutorial.html useful, its the jails presentation from EuroBSDcon 2022 btw. AMA.

Title: Jails from Scratch 2022

the /usr/jails is just a mountpoint.

dch, thanks, I'll check that out!

you can provide this mountpoint by a dataset under zroot/ROOT/default/messy

which gets you in the situation you are in, upgrading base OS also requires fiddling with all the jails

I highly recommend having separate dataset(s) for each jail

and keeping them outside the base OS tree

dch, that is what cbsd is doing

both jails, and the data inside it

tct: not the way you have it set up

dch, zfs list shows one dataset per jail?

dir /usr/jails is clearly "inside" zroot/ROOT/default

right

or is (in cbsd) /usr/jails completely empty

nah, /usr/jails contains all the stuff

technical term

dch, so we agree that as per my current situation, jail datasets are "inside" of zroot/ROOT/default (which is what I want (or wanted, at least)). This means that creating a BE also snapshots/clones the jails. so how could I end up in a situation where cbsd/jails use the 'default' BE but the system is on the '20230102' BE?

you can't

then why is it this way? :D

as most of CBSD docs are in Russian, I'm not familiar with it

tct, I don't want to give bad CBSD advice as I really am not familiar with it

but what I see here is that you have an awkward mess

when you create a new BE, the subsidiary datasets aren't recursively copied, I guess

dch, that is indeed what I see as well. and I don't have that awkward mess on any of my other hosts. This is my playground host but still...

dch, so basically from here the idea would be to move the jail datasets outside of zroot/ROOT, then delete the `default` BE, re-create a new `default` BE and moving the jail datasets back?

tct I think you're best off finding a place where the CBSD people hang out and ask them for more advice

dch, I'll do that. Thank you for your help - it's much appreciated!

I can give zfs tips easily enough, but I can't know if any of this would break your setup

dch, I fully understand that :)

oh cbsd.io/cbsd/What-you-need-to-know-about is now all in English (how times change) but a quick skim I'm none the wiser

Title: What you need to know about CBSD - CBSD

bsdstore.ru/en/about.html#contactus is probably a good start

yeah, cbsd documentation has improved notably over the past few years.

Title: CBSD — FreeBSD Jail and Bhyve Management Tools

already on it: cbsd/cbsd #722

Title: Cannot delete jail snapshots · Issue #722 · cbsd/cbsd · GitHub

722 – Off-by-one error in wbkgd() in ncurses bugs.freebsd.org/bugzilla/show_bug.cgi?id=722

i, myself, somehow overwhelme by numerous jail managers

same here. after testing many I eventually settled with cbsd for the fast that it also does VM management and migration stuff. haven't had a single issue with it that wouldn't have been caused by myself so far (+8 years of using it)

ditto. thats why I stopped using them and just have a bare single script now

cbsd has lots of cool features

I think I should have done that too. at least that's what I usually do. as little clutter as possible.

dch: thx for the link

dch, is that tutorial you linked on an IPv6 only network by any chance?

I get a timeout on my end

no, but it is on my computer here

I'm afk for dinner, I'll fix it later. the source is git.sr.ht/~dch/diy-jails-tutorial

Title: ~dch/diy-jails-tutorial - sourcehut git

thanks & enjoy dinner!

is there an automated way to update leap-seconds.list?

Yes. /etc/periodic/daily/480.leapfile-ntpd, done daily, activated by "sysrc daily_ntpd_leapfile_enable=YES"

That is, if you mean /var/db/ntpd.leap-seconds.list.

tct, link works fine here from outside. I see that VimDiesel also fetched it ok, so its not just a Works On My Machine thing.

V_PauAmma_V: that's it!

hi, my computer keeps crashing, I can't figure it out. dmesg outputs drmn0: [drm] GPU HANG: ecode 6:1:85fffffc, in WebKitWebProcess [100957]

drmn0: [drm] WebKitWebProcess[100957] context reset due to GPU hang

anyone have any ideas?

i seeing similar for intel card on laptop. i reported similar problem. i have disabled intel card on laptop

how do you disable the card?

in BIOS

this laptop has intel an nvidia and it's possible to run either one ( or both )

oh i see. i think mine just has an intel

this annoyance not fixed in drm-kmod ( for me, at least )

i periodically check drm-kmod ( from git ) but problem persist

no idea what is the reason.

upstream ( freedesktop.org, intel ) has numerous bug reports with "GPU hang" pattern and it is all unclear what was the fixes. i suspect, that drm-kmod missing certain things in linuxKPI ( or just lagging behind the source code )

hi folks! Anybody knows what ACLs are for in ZFS? FreeBSD defaults to acltype=nfsv4 when creating ZFS datasets, while the default on Linux is acltype=off. freebsd.org/cgi/man.cgi?query=zfspr…+and+Ports&arch=default&format=html

Title: zfsprops

NFS ACLs cause issues with ansible, so I am considering disabling them. But I have a hard time understanding the impact of that.

michelem: Linux doesn't support NFSv4 ACLs on ZFS, whereas it's what Illumos and FreeBSD default to.

If ansible has problems with NFSv4 ACLs, that seems like a problem with ansible? At least in so far as they're made to be compatile with lots of things including Windows ACLs.

debdrup, did not know that - thanks

BarnabasDK: it also doesn't look like anyone involved in Linux has any interest in implementing them, despite NFSv4 being nominally supported.

no it seems if you want support for that then you need to run behind samba

nonsense - I am thinking of my smb drives

nfs - yes I understand

Is it important on a high level to tell apart acls on zfs as opposed to for example ext

provided the infrastructure supports it

debdrup: thanks. I've seen from the manual that Linux lacks support for them. I was just wondering why they are enabled by default in FreeBSD. UFS does not seem to have ACLs enabled by default.

michelem: sure UFS has ACL support.

Specifically, it's also NFSv4.

See getfacl(1) et al.

ok. I did not know it's enabled by default. Thanks.

kern.features.ufs_acl should be set to 1 too.

Hrm. If "sysctl net.wlan.devices" is empty, is that all she wrote for FreeBSD wifi on a laptop, aside from things like virtualization with some other OS providing wifi?

To get POSIX 1e ACLs, you'll neet to set the acl flag via mount -o I think, but it's been too long since I've used them.

mason: there may be drivers in ports

they are actually not set by default, indeed.

Huh. Wonder when that changed.

the problem is surely with ansible. It's just a corner case which they'll never fix. So I was wondering if the cause of the problem is worth it. Because UFS-based systems go without ACLs by default, I was wondering why it was chosen to have ACLs in ZFS instead.

I suspect, but can't be arsed to go look, that it might be because Sun had NFS ACLs.

yeah, I guess you're right

anyway, thanks much for the help debdrup

no worries

Nice, I have a freebsd instance in Oracle cloud :D

rogersm: same

dang. i'm so close to getting hdhomerun_record running on aarch64. with 14.0-current it'll run in linux compatibility mode, but for some reason it's detecting its ip as the broadcast ip instead of the interface address