is it possible to have the loader check for a keyfile, and only if it doesn't exist ask for a passphrase for a geli encrypted root? not seeing much in the docs on this.

I'm on 13.2-STABLE, and recently `uname -a` doesn't show the kernel build date anymore. Is there a way to get it without looking at filesystem timestamps?

fireglow: what does it show?

rtprio: right now it shows "FreeBSD localhost 13.2-STABLE FreeBSD 13.2-STABLE n255613-6322a6c9daaa TEST amd64"

iirc, that's a reproducible builds thing

WITHOUT_REPRODUCIBLE_BUILD=1 in src.conf might be what you want

RhodiumToad: indeed! Impressive! Thank you!

markmcb: where would the loader check for a keyfile?

i assume that'd have to be defined. ideally anywhere. a removable usb drive, a disk partition, etc.

for example, in linux with luks, i just say "look for UUID 1234". if it exists, try to load /path/key. if that fails, passphrase is the fallback.

judging by some posts i've read, i get the impression this isn't possible

/* XXX TODO: Support loading key files. */

I think that answers it

indeed :)

reproducible -- that's a new one to me

Hello, since I made a pkg update and now zfs list shows "zroot" using 41.2GB but (same as zpool list) but ncdu or df show only 28.4GB. I really cant find where are the 12.4GB. (freebsd 13.2-p1).

do you have any snapshots?

the pool is a single partition of 48G

I deleted the remaining sanpshots

zfs list -o space

I delete all snapshots except the most recent, which size 18G

I may have misread that this one is needed for boot

Can i safely delete it ?

are you using boot environments?

I dont know, i discovered boot env today ! I deleted the other snapshots with bectl

(you don't need any snapshots in order to boot, but it's common when using boot environments to keep the previous one around as a snapshot in order to revert to it in the event of problems)

I mean, if the default install use boot environment then I use it

And I'm a bit confused about it. bectl list show 2 environement, one called "default" which I presume is actually running and another with a timestamp in its name.

default was creation date is 2223-12-05 !

sorry for my bad english, I tried to sort this out all day

RhodiumToad: Thank you to point me out to boot environements. Reading the bectl man page wes enlightning. I deleted the remaining boot environements except the one noted "NR" and all is Ok now

except for the "default" creation date being in distant future :)

yes, this is very strange

what's a couple of hundred years between friends

possibly, when the install was done, the system clock was way wrong?

this a virtualbox VM, everything is possible

how do you merge conf files in /usr/local/etc/ if any?

merge them with what?

the general principle for conf files from packages is that things like startup/shutdown scripts are part of the package and those just get updated,

while config files have two versions: a .sample file which belongs to the package, and the actual conf file which belongs to the user, and is created by installing the package _only if it doesn't already exist_

afaik (unlike /etc) there's no general assumption that package config files can be updated in any sane [semi-]automated way

looks like there aren't many important conf files in /usr/local/etc

well that depends on what packages you have installed :-)

so, when a pkg introduce a new conf file, the .sample is updated, the user conf file is untouched, there won't be stuff like /usr/local/etc/NAME/NAME.conf.new like most Linux distro does

let's say, the package sudo has a conf file in /usr/local/etc/sudoers . when sudo got upgraded AND it introduces new syntax or add/deperecates some options.

how does freebsd handle this by default? or the admin needs to read the changelog and do manual changes?

the admin needs to deal with it

i got it, thanks RhodiumToad

hello

good morning or other timezone-appropriate greeting

good morning RhodiumToad

good <time of day>

Is starting a non-uefi VM (x86) still no thing in bhyve?

non-uefi in what sense?

booting using "bios" perhaps? it is possible, you don't need to use UEFI

althouhg, i'd recommend it.

I'd have a very old WinXP inst which I'd like to keep if possible.. don't need it often

have you tried booting it in bhyve using the CSM firmware?

RhodiumToad: loader="csm" ?

no

well as said, I don't know how to do it ;)

are you using bhyve raw or via some wrapper tool?

I use vm(8)

Churcher's bhyve tool

that looks a bit ... old?

even so, loader="uefi-csm" looks like the option you want

RhodiumToad: not sure if it's old, it just works ;)

well uefi-csm make's it start - but doesn't provide graphics_* ?

I knew there was a problem, it is just too long ago since I tried last time

graphics_vga="on" will be necessary, don't know if it's also sufficient

Juli 09 14:29:30: [bhyve options: -c 1 -m 1G -Hwl bootrom,/usr/local/share/uefi-firmware/BHYVE_UEFI_CSM.fd -H -U edcea0ad-1e52-11ee-a08b-ac1f6be8253a -u]

well graphics_vga="on" doesn't work either

but its booting, kind of

did it change the bhyve options at all?

no

do you also have graphics="yes" ?

no but thanks, that actually worked

graphical performance probably won't be great :-)

I dont need much, it's mostly for 1-2 IDEs

if I get RDP running thats great then :)

Hello

what's the definition of filesystem?

Anyone having issues getting texlive-texmf-20230313.pkg from FreeBSD repos?

Fetching texlive-texmf-20230313.pkg: 18% 216 MiB 14.3MB/s 01:31 ETA

Lets see if it can finish.

yes, Fetching texlive-texmf-20230313.pkg: 100% 1 GiB 13.7MB/s 01:31 :)

I had to disable fetch_timeout to get my pkgbase install to work

with pkg 1.20

apparently libcurl timeout is from request start and doesn't care if the transfer's still making progress, so anything longer than 30 seconds is borked here by default

Ok that must be my issue too.

Getting a size mismatch after 30 seconds

why libcurl though?

1.20 switched.

i see, still on 1.19.2 (and pkg install pkg says i have the latest)

a filesystem is a system of files, obviously

is there anything like pstree, but for jails? i.e., a hierarchy view of which jail initiated which processes?

ps -d -J <jid>?

yuripv, thanks! didn't know about the -J option, this is perfect

so now we have libcurl in base?

markmcb a lot of commands have the -j/-J flag, it's just perfect.

antranigv: it's a whole system

not a kernel + psmisc

uuidNuniq I know what FreeBSD is :) I was just wondering if we added libcurl into base or is pkg 1.20 using libcurl built into it

i want curl in the base

antranigv, ldd /usr/local/sbin/pkg doesn't show it using libcurl. Or libfetch.

V_PauAmma_V 1.20?

Ah, not yet. 1.19.2.

Should I be wary of that?

Of that version, I mean?

V_PauAmma_V I think we should do some good testing, since it's pretty massive change. I need to check if libfetch's env vars work properly as well. here's the file that does most of libcurl stuff in pkg github.com/freebsd/pkg/blob/master/libpkg/fetch_libcurl.c

Title: pkg/libpkg/fetch_libcurl.c at master · freebsd/pkg · GitHub

github.com/freebsd/pkg/blob/caaafa6…a2f607d/libpkg/fetch_libcurl.c#L152 Looks like a typoed environment var name.

Title: pkg/libpkg/fetch_libcurl.c at caaafa6d6130a39f27fb48070598a394ba2f607d · freebsd/pkg · GitHub

wc -w should give me a file count if a pipe an ls to it, right?

Nope. Filenames can contain spaces and other nonword characters. You may want wc -l instead. (Can you create filenames with \n in them?)

no, IIRC

I would say / and nul are the only two that you can't use

Or use one of the "display special characters as \... escapes" ls options.

so use ls -l then

is there a way to list network interfaces that have been consumed by a jail, without querying the jail itself? i.e., ifconfig will show them absent, but is there a way to show the absent items?

CCFL_Man: ls is not suitable for parsing

use find -print0

ah

ohh , but wc does not have --files0-from=- option

for example, ifconfig -l, outputs: ixl0 ixl1 igb0 igb1 lo0 iavf7 iavf8 iavf9. i would like to see iavf0 through iavf6, and ideally which jail has them.

CCFL_Man: oh you want to show the words in file? then, probably using this: ls -l | while read file; do wc -w "$file" ; done

sorry, not ls -l , but ls -1. but ls will list dirs as well.

so not a good idea, probably `find . -type f -maxdepth 1 -mindepth 1 | while read ... `

uuidNuniq: no, just an actual file count in a directory. there are no other directories inside, just files

CCFL_Man: you want count how many files in a directory?

ls -1A | wc -l

yes, just a count

perfect!!

thank you guys!

there's also a zfs thing that "size of directory" shows the number of files, so dirty hack: stat -f %z dir/

thanks!

(don't use it, saying just for fun :D)

does anyone have a link to a guide/tutorial for installing FreeBSD using zfs but not on the entire disk ?

Do you mean "while leaving unused space other than the ZFS partition and the swap partition?

I think I need a boot partition, zfs and swap yes

and leave the rest unused

If you don't get an answer meanwhile, ping me in 2-3 hours? I need to install ZFS VMs (and can therefore refresh my memory of the details) after a nap.

alright, thanks

kevans: How do you get that timeout value down to a poudriere pkg? FETCH_TIMEOUT.

kevans: poudriere jail ... start it jexec into add /usr/local/etc/pkg.conf?

skered: you don't need to start / jexec for that. just copy the file in the right location

I was thinking a start/jexec might be needed if it was bind mounting something that was already in the mount /usr/local/poudriere/jails/<jail>/usr/local

Because it's empty right now

I'm following a guide off the FreeBSD wiki and I've booted a FreeBSD installation USB

I've gotten the partition installed but when I'm creating the zroot it says that it can't mount it because my install media is read-only

I've tried remounting as rw with mount -o rw /

but it won't let me

last1: you want to mount onto /mnt probably

I can't do it there either

also note that in the installer media, everything's read-only by default

I've tried zpool create -o altroot=/mnt

cannot mount /mnt/zroot failed to create mountpoint, the fs is read-only

Maybe zpool create -R /mnt

Oh, that should be about the same. -R also turns off the cachefile, so maybe that's the issue.

-R root Equivalent to -o cachefile=none -o altroot=root

Try with -R or by adding the cachefile option, and maybe that'll work out better.

nope, same error

Strange.

yeah, following this guide here: wiki.freebsd.org/RootOnZFS/GPTZFSBoot/Mirror

Title: RootOnZFS/GPTZFSBoot/Mirror - FreeBSD Wiki

Hm, that doesn't show that option. Wonder how that's supposed to work.

last1: For comparison, I do this somewhat often: wiki.freebsd.org/MasonLoringBliss/ZFSandGELIbyHAND

Title: MasonLoringBliss/ZFSandGELIbyHAND - FreeBSD Wiki

they just show this: zpool create zroot mirror /dev/gpt/disk0 /dev/gpt/disk1

which tries to create zroot on /

Yeah.

Weird.

which is a nono

Something else I do if I want to ssh in to do the install: wiki.freebsd.org/MasonLoringBliss/InstallEnvironment

Title: MasonLoringBliss/InstallEnvironment - FreeBSD Wiki

The normal installer itself isn't bad, although it doesn't do a complete job if you're on a UEFI system. I'm maybe going to fix that if I get a thumbs-up from one of a collection of folks.

s/one of/any &/

normal installer is great, but I want to install only on part of the disk

I fixed it by doing mount -uw /

and then zfs mount zroot

not ideal but it worked

That doesn't seem quite right, that method.

You'll have a hard time installing if 1) your pool is not mounted and accessible, or 2) it is but it's masking your actually root

There could be something I'm entirely missing. If so, you can modify what I do in my install guide to accomodate a partial-disk install. I've done so to get FreeBSD to be one of like five OSes I installed on my (mirrored SSD) laptop once, to see if I could do it cleanly.

I have nothing else to try :|

Well.

let me try to complete the install this way

see if it works

Sure.

I'm doing this because my drives are 4Tb SSD and I only want to use 120Gb for the OS

and the rest for the zfs special device

Sure, that's fine.

seems like an odd layout, but I just showed up

why is it an odd layout ?

when you say "I want to use 120Gb for the OS," what portions of the OS file system are you intending to place on that partition? Just /, /etc, /tmp, /usr, etc? What are you intending to use the ZFS pool for?

I have 200Tb of attached HDDs that will be the actual storage

the SSDs are just for boot/metadata

I have 200Tb of attached HDDs that will be the actual storage

the SSDs are just for boot/metadata

So the 200Tb's are going to be zpool(s)?

yes

Did you want your home folders and /var/lib/ in zpools?

no, those go with the os on zroot

last1: if i install an hba and disk shelf, as long as the hba is in IT mode, the drives installed in the disk shelf should simply appear as devices?

I guess that is what makes it a bit of an odd layout for me.

i can mount them normally?

Are the zpools for resilience? RAID? And snapshots?

Most of my zfs setups minimize how much of the OS FS I place on a static drive. I mean, rigtnow im running root on zfs across two SSD's, but I have a few where root is next to the EFS(boot) vol, and the rest of the disk(s) are pooled to handle home, /var/lib/(docker, etc).

but that's on an external shelf where I keep app data

I don't want to put together os/user data with app data

Aha.

CGFL: correct

mason: the setup worked fine with mount -uw /

last1: I bet it was the cache file handling. I bet once you made root read-write it was able to write out the cachefile, or at least fail differently.

could be

one last question

is there a way to see a list of all pools, even those that aren't imported yet

zpool list only shows imported ones

no; `zpool import ` will show not imported ones

can someone explain why zroot/var has the property mounted=no ?

rtprio:

assuming you haven't done anything strange in the interim since the default install,

hang on, there are good explanation of this, and it doesn't make sense for me to just give a hackneyed explanation

here:

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

DrKK`: right, default installer

thanks, that's helpful. basically "for bectl" and you can mount them if you want but then you're way off the rails when bectl tries to make sense of it

rtprio: thank you , that worked