on the flip side, there's a few man pages that get into some scenario-based examples, e.g. geli(8), but don't quite go far enough on the full scenario (which would probably be too much for a man page)

the thing with wikis is that they are everything for everyone, and so it ends up very cluttered and often outdated, because it might have been used only for prototyping

meena, agreed. i could imagine a curated, limited scope wiki. or maybe not even a wiki, but something that's a little more fluid and less formal.

markmcb: if something is worthy of a docs chapter, we try to put it there, but i agree that easily findable / searchable community docs that are less formal and more goal driven would be very beneficial

(and i also have no idea how to achieve that)

totally. sesutil is a great example. not worthy of a chapter, but it's a useful quick mention, and exposes a great simple tool that someone might find really handy. it's also a good way to point people in a direction and dispell bad info. i saw in a forum people talking about no equivalent to hdparm's secure erase in linux. i was really disappointed. then i found camcontrol secuity on my own by

chance while looking for something else.

maybe i'll hack on a prototype doc and get feedback. i took a bunch of notes when i was figuring things out.

markmcb: that's fantastic. i hope you corrected that forum post!

Title: confctl(1)

Ah, the in-page link to sysctl(8) does not work because the URL is ports-specific

what config files look like that?

% confctl -a /usr/local/etc/pkg/repos/FreeBSD.conf

FreeBSD:.url:="pkg+http://pkg0.fra.freebsd.org/${ABI}/latest",

FreeBSD:.priority:=4

%

confctl -a /boot/loader.conf

confctl -a /etc/sysctl.conf

… and so on.

trasz/confctl e84c851 i feel like this overcomplicated the problem

Title: Use asprintf(3) in a Linux-compatible way. · trasz/confctl@e84c851 · GitHub

pretty sure glibc's asprintf probably just doesn't modify `*newprefix`, so really probably just needed to initialize it to NULL in confctl.c

i'd be surprised if it trashes the pointer on failure

kevans: maybe it did back then?

Does the FreeBSD installer not have a timeout configured for whenever it tries to fetch files from one of the mirrors? I think the ftp://ftp2.br.FreeBSD.org/FreeBSD is down/not working, so I've been stuck on this screen for a while now: upload.lurkmore.com/z/OjjvNSYL.png

I cancelled the install and switched to the default/first mirror in the list and it's working fine now.

i noticed the default /etc/pkg/FreeBSD.conf shows pkg+http (vs. pkg+https). Any reason it doesn't default to a more secure transport? i didn't see any rationale in pkg.conf(5)

may have been related to the historical lack of certificate management in base

package signatures are checked regardless, I believe

yeah, the default repo's all signed

there's been some discussion about switching it to https

oh, he actually submitted a review for it, even: reviews.freebsd.org/D40473

Title: ⚙ D40473 pkg: use https by default

nice, thanks

Based on this -> gnu.org/software/grub/manual/legacy/FreeBSD.html Grub can launch freebsd kernel, but not advised.and since grub can decrypt luks1, with limited luks2 support, is there any restritions for using a luks encrypted boot partition on freebsd?

Title: FreeBSD - GNU GRUB Manual 0.97

sozuba if it is what you need and it does work, no problem. Otherwise, I'd suggest to use native tools - just because the native tools have better support.

sozuba: We have no LUKS in FreeBSD. maybe in ports, but not in Base

Title: FreshPorts -- devel/libluksde: Library and tools to access LUKS Disk Encryption encrypted volumes

tsoome: i would love to use native tool, but due to my previous PR (271938) of excessive heating, I am not able to use Freebsd freeley yet. So I have planned to dual boot, with the usual encrypted setup i have for linux and using the same for freebsd. This way i can continue my work and stuff,while also trying towork on the PR.

Title: Chapter 19. Storage | FreeBSD Documentation Portal

sozuba: grub uses chain boot loading, so this should just work

meena: thanks for the link. Best way is to find a middle ground, so i can use grub to boot the loader and not freebsd kernel.

meena: yup chainboot is what i was thinking

thank you very much

dual boot is perfectly good use case. and yes, in general, we do suggest chain loading (first boot loader starting other boot loader) and less direct loading.

problem with using grub to load and start fbsd kernel directly is about corner cases - if you need to load some specific module with kernel, some configuration details etc.

yup, understood. I mostlikely will use freebsd loader to boot freebsd kernel, whil using grub to chainload freebsd loader

thanks tsoome and meena for your imputs ad advise :)

btw, is this machine doing BIOS boot or UEFI boot?

i 100 GB enough for freebsd parition thatmight include src ports, etc for debugging and may be development?

i don;tplan to use freebsd for anything else until this is sorted

yes, 100G should be good.

tsoome: UEFI

aha, ok, I haven't tested chainloading loader.efi from grub, so it would be interesting to hear back how it is behaving;)

Will definitely post back results and experience once done

I am having a real bad time with installworld step (14-CURRENT) when trying to update from 1cebc9298cf to b98fbf3781 (or from 9fbeeb6e3831 tofacc213643f8 or dfa1982352ee on a different host). The error is of "missing cc" during "installworld. Some log from the first attempt: termbin.com/f2eh

/etc/{make,src}*conf: termbin.com/7j8kb

bugs.freebsd.org/bugzilla/show_bug.cgi?id=269366#c1 hints at change in source or build option but that did not happen

Title: 269366 – installworld calls cc breaking nanobsd build

... just noticed "ERROR-tried-to-rebuild-during-make-install" in there; searching on that

Log of "installworld" failure on another host from 9fbeeb6e3831 to dfa1982352ee : termbin.com/74je ; /etc/{src,make}*conf: termbin.com/c8ll

what about buildworld, did it finish successfully?

Yes, I did not see any error messages.

buildworld && kernel completed without issues

(also emailed -current@ list)

it says something about ccache, misconfiguration?

Hi, I created a Bhyve VM (Ubuntu guest) using vm bhyve. The machine has grub as default boot loader, but it doesn't boot after installing, to boot it I must type a series of commands (set prefix..., set root.., insmod ..., normal) to let if boot. Can anyone help me saving those commands to let grub boot automatically?

yuripv, Hmm; ok will try soon without ccache

yuripv, Did you imply the "sh: cc: not found" message?

Need to leave for now ...

hi all

yo

'lo party people

Booted up the memdisk, wanted to do a manual install, do "gdisk destroy ada0", i get device busy. Mount doesn't show ada0 being mounted.what ami missingor doing wrong?

gdisk? I see no such command

gpart*2~ sorry habbit

"gpart destroy ada0"

you're trying to wipe the partitioning on ada0?

yes

clean whipe the device

what does geom -t show

but for now just using gpart without dd

hm, that doesn't look like it should report as being busy

This is Freebsd 14-CURRENT

jfyi

oh I know, destroy only works without -f if all the partitions are deleted first

so gpart delete -i4 ada0 and so on down to -i1

ah okay. Got it, Thanks :)

we need an -rf here, for extra danger

just -f is dangerous enough :-)

sorry it's -F

-rf is recursive force? :D

(-f is already claimed for -f flags)

okay

(it's possible to have recursive partition tables but most combinations require a sysctl to enable)

worked,now i can peacefully test/playaround

okay

in particular you can't have a gpt table anywhere other than directly on a physical disk without turning on an option for it

(an option I added, because I wanted to do some VM stuff)

honestly,my recursive was a rhetorical question, cause meena said -rf

but now i am curious."Recursive partition table"? sounds cool

is it like haveing a partition table inside partition table, like having lvm on luks on gpt?

well in my case what I wanted was to use a raw gpt partition as a VM backing volume, and obviously installing the VM would put a partition table on that volume

and I wanted to get at the partitions from the host

hence the kern.geom.part.gpt.allow_nesting (off by default)

and likewise kern.geom.part.allow_nesting for other partition schemes

don't turn them on unless you have a similar use-case to mine :-)

got it.This is interesteing. But i amnot going to try it.I alreayd have my plate full, don;t want more compications :)

thanks RhodiumToad

following wiki.freebsd.org/RootOnZFS/GPTZFSBoot for manual partition. Is it madatory for freebsd-zfs root partition to have disk0 as the label? This is the first time i amusing zfs

Title: RootOnZFS/GPTZFSBoot - FreeBSD Wiki

no.

cool, thanks :)

there's some merit in not putting labels on zfs partitions

why so?

btw, i just found this -> forums.freebsd.org/threads/labeling…one-right-on-modern-computers.69250 and i amgoing to check it out

Title: Labeling partitions done right on modern computers. | The FreeBSD Forums

(I normally think labels are a good idea since they're a form pf metadata easily visible to non-freebsd tools, but for zfs the drawbacks may outweigh that)

'drawbacks', now my mind is focused onto your next words

so that article tries to draw a distinction between gpt labels and geom labels. this is incorrect

oh

the LABEL geom class is responsible for both.

okay

LABEL recognizes two types of labels

in liux i mostly dealt with uuid for fstab

explicit labels, created by glabel label, which live at the end of a provider (reducing its size) which show up in /dev/label

and implicit labels, which are obtained by examining the content of a provider. This govers GPT labels (found from the GPT partition table), filesystem labels (found by looking for "volume label" metadata in UFS, FAT etc. filesystems), and so on

(also disk serial number labels, GPT uuid labels, etc.)

so the problem with zfs is this:

zfs recognizes disks by content, not by name. it doesn't care what name you used when adding a disk to a vdev or whatever.

and that means it doesn't see any reason not to attach the disk using the label rather than the partition name; it'll do whichever one it tries first

RhodiumToad: you looking at bugs.freebsd.org/bugzilla/show_bug.cgi?id=271989 ?

Title: 271989 – zfs root mount error 6 after upgrade from 11.1-release to 13.2-release

no

though with that bug, note that zfs doesn't as far as I know limit itself to searching in freebsd-zfs partitions

it'll find zfs metadata _anywhere_

tocontinue what you were saying, if we have labels, it will try that first?

so in relation to what meena shared, having labels lying around (example the nug meena shared), can cause boot issues?

or be adverse,even if noissues

it usually won't try the label first

also, a follow up question, is it better to not have labels at all on any type of fs / partition? like on fat32 / efi?

but sometimes it might, which is confusing

understood

no, it's only really zfs that is a problem here

got it. I'll keep that in mind.

for other filesystems you can use a gpt label, a filesystem label or both

the other cases that don't like gpt labels are gmirror and gstripe

Honestly, interactig with you, i learn a lot. This is some good knowledge that comes mostly through experience. Thank you for Sharing, I really appriciate it. And thanks meena :)

i only ever used mirror and stripe once.and had no use case, but i might going forward

a few month down the line

but i also often share exciting commits!

so this looks okay then? gpart add -a 1m -s 100G -t freebsd-zfs ada0

meena: haha, good for you and for the community. Would be great if you could add somecommits for the PR i rasied :P :D

I don't have a commit bit, and i am generally not allowed near hardware, for danger of it exploding

na, hardware willbe fine. You seem nice :)

sozuba: that command looks reasonable to me

exciting commit freshbsd.org/freebsd/src/commit/bcf…47144f3499976a623f92712ecd0aca71912

Title: FreeBSD / src / bcf9147 / EC2: Default to "uefi-preferred" boot mode - FreshBSD

1m is perhaps more strict an alignment than is really necessary, but not a problem

thanks, i laready entered :-p and therewa no explosion, so good :)

fixes bugs.freebsd.org/bugzilla/show_bug.cgi?id=264697

Title: 264697 – FreeBSD 13 AWS AMI can not use free tier because t2 instance cannot use UEFI boot

So, soon ™ we can use freebsd for free on aws!

You can use it on Oracle Cloud free tier now.

x86 and Arch64

Aarch64 even

free??

ooohhb

Absolutelyf free

how does that work :P

I have three VPSes on Oracle free

nice, all you have to do is sign in blood here: _________

two are single core amd64 and the other is a 4 core Ampere with 24Gigs of ram

how is that free??

Been running for two years now.

sounds slightly dodgy

what's on it for them

It's part of their free tier.

Get people used to ORacle cloud.

So they are mor elikely to deploy there.

Same reason AWS does it.

I need to get to testing it

I use them for Wireguard servers for road warrior vpns.

I use one as an additional name server.

sounds like I need to get myself signed up

Been very reliable. YOu get emails if there are any issues, just like a paying customer.

they must be losing a lot of money

termbin.com/tqug <- ami doing something wrong, why is there a freespace between efi and swap?

I doubt it. That is so little resources for them

our blood now belongs to Larry Ellis

You are not allowed to use it for mining or anything.

But, like I said I use one every day as a VPN for my phone.

sozuba: what command did you use to create the swap partition?

RhodiumToad: gpart add -a 1m -s 15G -t freebsd-swap -l swap0 ada0

sozuba: so the gap is there because of the 1M alignment

ah okay, i was followign this wiki.freebsd.org/RootOnZFS/GPTZFSBoot

Title: RootOnZFS/GPTZFSBoot - FreeBSD Wiki

sozuba: remember the GPT table itself takes 20 kbytes

Erhard: i need to see how cloud-init does on Oracle cloud then. Today i saw my code run for the first time in vultr. that was exciting

so to not have that, i shouud remove that 1M alignment for swap?

I have not tried cloud-init

or make the efi size 20K less than a multiple of 1M

I installed manually.

fair point, will do. Thanks :)

I think I had to use the trial period credits to actually send the ISO over there.

Since I think ISO storage is not included for free, but you at least sued to get a month or $100 or something for free

s/sued/used/

Erhard: didn'tknow oracle had a freetier. Is there any other freetier stuffs? :D

AWS had a free year, not sure if they still do.

Erhard: pkg install -y net/cloud-init-devel # for my bleeding edge commits

azure still gives you a nice 200 bucks to spend

aaa the Reddit blackout makes me realise just how much info it has

So yeah you get $300 to use within 30 days as well.

Get all your stuff setup with isos within that period.

Rayyan: yup, evey questio i find an answer to on reddit makes me realize it

nice info, thanks

so people are butt hurt cause they can use 3rd party apps without the reddit stuffs on it

let it go with the dodo and twitter its ok

Just ask ChaGPT, since it scraped it all anyway ;-)

Erhard: hahahaha :)

I see let us have a word with this chatgpt

sozuba, The other day we were chatting about power management and parv posted these two links after you had already parted the channel. I don't know if you saw them but I wanted to point you to them as I found them interesting.

sozuba, parv posted these: neelc.org/posts/optimize-freebsd-for-intel-tigerlake neelc.org/posts/freebsd-speed-shift-laptop

Title: Optimizing FreeBSD Power Consumption on Modern Intel Laptops -

iirc sozuba's laptop is an older one

rwp: That'svery thoughtful of you. Thank you and to parv. I'll checkthem out now

Ah, like RhodiumToad said mine is a haswell laptop. But again, its good to read about it, :)

Does gpart use 1024 or 1000 units?

IIRC everything in computers are binary powers of two but I am sure the man page for gpart will clarify.

i looked ar the man page, couldn't find it

will check again

thanks

afaik gpart uses only binary units

(SI can go fsck themselves)

hahaha, same here :D

but a discussion on binarry over SI would be good one day, without flamewars.

I despise computer numbering based on how many figures are most common on a human.

s/figures/fingers/

Disk drive manufactures found powers of 10 to be a marketing breakthrough for them. Numbers appear larger that way.

and progressively larger as sizes increase, even

Right. It would have been much more awesome if we had 12 fingers because multiples there are much more interesting.

12 is indeed a superior highly composite number

I can't mentally understand naming the Babylonian base 60 numbers but having a large number of integer factors is a huge benefit.

I grew up learning binary, and feel awakaward using SI,simple as that. My natual intuition when it comesto bits is always binary

binary units make sense for storage, since it corresponds to the number of bits in the address

I think if the people pushing that computers should use powers of 10 were to do some actual engineering work developing one that they would change their minds pretty quickly.

not so much for communications channels, where you're measuring a frequency, typically

who is pushing for computers to use powers of 10?

Yeah, I think that only applies to people dealing with 8bit bytes, or at least power of 2.

the units purists push for binary units to have ridiculous names, not that they not be used

Pretty much every other engineering would likely benefit from powers of 10 only.

It is important for kilo and kibi to be distinct.

Otherwise how do you know which is which and when? By memorization and convention?

It adds complexity to not use kibi

RhodiumToad, There are endless requests for things like df which says "42G" to say "42GiB" now instead. For example. Adding two extra characters to the display. And breaking decades of scripts that are handling the traditional format.

More important for kibi and kibbeh to be distinct.

I have a terrible memory, memorizing things is hard

storage should be measured in binary units, period, and the fact that drive manufacturers inflate their sizes for marketing reasons should not be permitted to interfere with this

Agreed.

Agreed. Megabyte is always 1,048,576 bytes

"ebi" or "ibi" anything can kiss my bits.

Initially when this mebibyte came about, it not only sounded silly, but seemed ridiculous

No past tense to that

i was like, "what? Mebi?" it's childish

Erhard: fair point

and kibi is even worse

It's unambiguous.

RhodiumToad, When it comes to serial protocols that is somewhat different. Since it isn't a physical thing. It's a possibly infinite stream and so units are truly arbitrary there.

why force people to play guessing games?

Kibi sounds like a cute little creature that you can pet around

Kibbehbits... It's what's leftover at the Lebanese restaurant... not a measurement.

Maybe to someone from another part of the world 'kilo' sounds cute and childish.

Saying mebi and kibi sound childish is top "I'll design the logo" energy.

they still break scripts though

My scripts have scripted breaks.

hahaha

xtile: fair point

They usually look like: Segmentation Fault (core dumped_

but honestly that's how it felt to me

anyway, i gota go guys

Later

good chatting with you all. And thanks for the knowledge

have a great day evening night

later