Some theory: forum.netgate.com/topic/2322/calcul…for-ack-queues-for-asymetric-link/7

Title: Calculating the required bandwidth for ACK queues for asymetric link | Netgate Forum

mason: altq, yeah

mason: the firewall improves the more letters you remove from 'ipfw'

heh

I dunno. w doesn't look like much of one to me.

yeah, but f is off the chain

Anyone out here running unbound on opnsense ?

kevans: yes, f or p would be the next one

Hm, I'm reading about ack filtering, distinct from prioritizing them. Does altq do that too?

c.f. lwn.net/Articles/758353

Title: Let them run CAKE [LWN.net]

Looks like I can do ack filtering via Linux or ack prioritizing with altq. That's unfortunate, having to choose.

ugh

I was upgrading a box from 12.1 to 13.2 and ran the first freebsd-update install

due to inattention ran it a second time without rebooting

and all hell broke loose

I hard rebooted but now it says all kinds of libraries are broken

stuff like /usr/lib/lib/libmd.so.6 invalid format, etc

am I hosed hosed ?

last1: does 13.2 have /rescue/fetch?

if you can get to a rescue shell, you can always /rescue/fetch base.txz and /rescue/tar it

I booted to single user mode

tried running freebsd-update install here

it says: ld-elf.so.1 /lib/libmd.so.6 invalid file format

last1: Try what kevans is suggestion. It'll work.

suggesting*

is the single user mode the rescue shell ?

can I do it from there ?

last1: From anywhere you have a shell, /rescue has statically-linked binaries

you could do it from single user, but you do need to manually put your NIC up if you're going into single-user

yeah that's fine

ah no it's not

lol

?

ifconfig reports invalid elf file

/rescue/ifconfig

in general, /rescue should have everything to help get out of scenarios precisely like this

hmm, wpa_supplicant is notably missing from /rescue; that would be cool

darn, 13.2 has no /rescue/fetch

sh /rescue/fetch file not found

bah

i'd bet a -CURRENT /rescue/fetch wouldn't rely on any kernel behavior you don't have

but getting it on there is the tricky part, going to need some bootable media at this point

oh wow, pretty silly for freebsd-update install to allow itself to run twice and thrash the whole system

ok, so I got the bootable media

I boot with that, drop to shell, mount my / and drop base.txt in there ?

base.txz ?

yeah, pretty much; extracting base.txz over your / should generally be safe

wait

is this zfs or ufs?

ufs

ah, ok

that's unfortunate; with zfs this would've been just a "*sigh* /rescue/bectl activate blah; /rescue/reboot; profit"

Only wild-eyes fanatics use ZFS.

eyed - really kills the trolling when I typo

in the usb boot

do I choose shell or live cd

last1: live cd ought to be fairly full-featured

last1: you'll want to fetch download.freebsd.org/releases/amd64/13.2-RELEASE/base.txz

presuming amd64

yep

ijm fetching

I'm fetching, yay

so simply xz -d base.txz

then tar -xf base.tar ?

tar will decompress for you generally

k, fingers crossed

mount up root and any other filesystems that matter - probably just root, cd into it, untar

yeah yeah, I got it under /mnt

or untar with -C or similar

it says: can't unlink already existing object

Title: Imgur: The magic of the Internet

Well. tar has an --unlink option that unlinks before unpacking, which could be useful

I'm always hesitant to recommend anything that deletes stuff of course.

Title: Imgur: The magic of the Internet

now am I hosed ? :)

Hrm. Hopefully someone will chime in with better ideas, but I'd be tempted to unmount and give fsck a look at it.

"This post may contain erotic or adult imagery"

Also, note that you'll probably want to save some things in /etc, as I just remembered that unlike, say, NetBSD, FreeBSD doesn't ship etc.tgz separately.

kevans: Hot terminal action.

base.txz shouldn't wipe out anything pertinent in /etc

I'm thinking at this point I may just reinstall everything

kevans: It doesn't include an empty password, by way of example?

i have a quick question im running kali linux in freebsd bhyve is there a way to access the kali linux gui or is this not supported yet in bhyve ?

I don't believe so, passwd is considered a config file

last1: you're not completely hosed, just some weird stuff going on here

what does ls say about some of these? ld-elf*so and libc are probably noschg, but the error doesn't match what I would expect from noschg I don't think

those files are special

I tried removing them

didn't work

oh, right, flags

er, schg

hm, chflags doesn't have a recursive option

try `chflags noschg <path>`

find . -flags schg -print might be interesting

rather, find . -flags schg -print

I did chflags noschg on thsoe files

removed them

ran tar again

argh, sorry, meant +schg - tired

Ill reboot see if the old system fancies a run now

yep, came up lol

but users are gone

let me see what else is gone

last1: That'd be the passwd file I was mentioning. =cough=

and vi depends on libncursesw.so.8 which I no longer have lol

I just have libncursesw.so.9 now

last1: That's odd. The tarball ought to have been internally consistent.

is there another edit that lets me modify the /etc/group file ?

there should be a copy in /var/backups for some of these

kevans: You should push for a separate etc.tgz like what NetBSD does. Useful for these situations.

There's a lot of dancing around upgrades that can be avoided.

is there a way to redo the upgrade process in a more controlled environment ?

cuz now I feel some files haven't been updated and whatnot

last1: ZFS is the most common answer because you can snapshot. Unsure what the best way is with UFS.

so there's nothing like freebsd-update -r 13.2-RELEASE reinstall ?

Not as such, no. But the tarballs really ought to be internally consistent. That's worth exploring.

tar exits after a few errors

so it doesn't complet extracting everything

last1: More of the same sort?

yeah

can't unlink, etc, etc

schg errors

last1: As noted, find . -flags +schg -print

so now /bin/vi is the old version

while /lib has the new libs

hence that error

then if you're very confident with the list, you can swap -delete for the -print

but note that it's quite dangerous to do this

last1: And remember the statically-linked bits in /rescue, which should include vi

but you want to get that tarball unpacked fully before trying to use the system

mason: yeah, I tend to agree now. up until, oh, 5 minutes ago, I would've sworn up and down that base.txz didn't contain anything pertinent in /etc :-)

I guess etc.txz would contain both etc/ and the etcupdate metadata that goes with it

It's been a while so I don't remember precisely what they included.

I really care just about the /var partition on this machien

can I reinstall and somehow keep /var in there ?

or rather, can I fool the system that it's 12.1 again and run freebsd-update on it ?

last1: You can rename it - careful with targeting the find -delete if you end up doing that so that you don't reap through it, in case there's anything schg in there

Don't think that'll work, unfortunately.

you already updated base to 13.2, is kernel 13.2? if so, you don't really have much else to do besides maybe some config files that need merged and pkgs

oh, this was 12.x to 13.x

yeah, nvm, I got it somehow

I re-extracted base.txz in multi-user

it worked here lol

last1: It's usually not as painful as what you just went through, FWIW.

for sure, but it seems silly to go through this after running freebsd-install twice you know

I just did arrow-up enter out of habit, and voila

basepkgs will arrive at some point

rebooting now for good measure

yep, came back up. huzzah

\o/

thank you for your help fellas

much appreciated

happy to help

If I am running 13.2-release is there a way to just update if I download 14.0-current image?

prob shouldn't be using current if you have to ask that

Appreciate the input

llua: and yet, i have made it very easy for people to do: alpha.pkgbase.live

Title: Unofficial FreeBSD pkgbase repository

FIRECRACKER fails to link

meena: it should be just 'base hash'

(the help widget needs to be updated, yes)

yuripv: i tried several things, but the preview doesn't do shit

try base hash, preview will show it

oh, it probably does not work when *creating* PR, i remember some reports about it

works when updating PR though (the preview)

btw, the issue you reported is discussed here: lists.freebsd.org/archives/dev-commits-src-main/2023-July/016995.html

Title: git: 60381fd1ee86 - main - memdesc: Retire MEMDESC_CCB.

I'm surprised FIRECRACKER doesn't include CAM

grumble. why is the dev-commits-src-main list sending me malformed MIME messages

CAM has a lot of delay stuff at startup.

that's a good reason to not include it in FIRECRACKER. or speed it up

I think whatever's generating dev-commits mail has a bug; it seems to sometimes generate multipart/* messages with quoted-printable transfer encoding

you mean the replies to commit messages?

no

looking at the one I bounced, it's an actual commit

anyone by any chance have a copy of message 17037 on dev-commits-src-main including full headers?

lists.freebsd.org/archives/dev-commits-src-main/2023-July/017037.txt :)

the copy in the archives doesn't have the offending headers

ah sorry

i.e. whatever mangled it happened after the copy was archived

RhodiumToad: i reckon that our archiver sanitizes headers

I think it's the opposite, that the mailing list manager is sometimes adding headers

but I can't tell until I see the headers :-)

hey, ive been trying to set up a windows wirtual machine through bhyve, but it never seems to actually start even though i have the iso for windows and bhyve-firmware

what options are you using?

and are you using a vnc client for display? does anything show up?

options, i think i just copied an uefi config from a website

as dor vnc, im using tigervnc. doas vm list says the vm is stopped

Hi Is it possible to mount a subpart of my fs into a jail ? I have confusing info on the net. (As a possible security issue with jail-mount)

use nullfs

the security impact will be up to you to deduce in how you are using it. some automated jail systems already do this using the mounts from /

from outside the jail, you can nullmount basically anything into it as you please

(I have a vm running poudriere at the moment, and it has like 40 nullmounts just because of that)

personally i like to have a jail that mimics the rootfs / exactly and then remount with nullfs from there

if you have a good handle on zfs ... then this becomes very easy

thank you. Documentation about that is a little bit old. thanks again

np brother BSD'r

hi, any way to run old 32 bit stuff in a 64 bits FreeBSD modern installation?

applixware # ./install

ld-elf32.so.1: Shared object "libX11.so.6" not found, required by "install"

Title: Applixware Office for FreeBSD (Walnut Creek) (November 1999) : Free Download, Borrow, and Streaming : Internet Archive

uskerine: you'll need 32-bit versions of any libraries that the old software needs

Is it doable or is it going to be a hell on earth?

probably easier in a jail

are jails like containers for FreeBSD?

but I don't know where you'd find the necessary old packages

yes

ports have archives, but I feel like x11 hasn't changed much in the past 24 years

uskerine: yes, more closely related to Linux's OpenVZ than containers, but the same similar idea

interesting, libX11 is indeed still version .6

so maybe just installing 32-bit X11 packages in the jail would do

Well I think I will save the task for one of those days where you feel like learning something. Does not sounds like something to do in half an hour. I just found that CD and I was curious to see applixware in action (in all its Motif/X11 shine)

I didn't see anything inside jail. I am using mount parameter bpa.st/2XTQ. What I am doing wrong ? /usr/home/mik is empty

it seems mount option mount nullfs inside the host. Is nto what I want . I want to share existing directory on my host with the jail. Is it possible ?

ah ok got it

the mount line should be /path/to/jail/some/dir /home/whatever nullfs ...

er sorry

/home/whatever /path/to/jail/some/dir nullfs ...

and why is there no manpage for ucl

lol

dch: hey! do you have knot-resolver setup guide anywhere?

Yes RhodiumToad Did that indeed

I did*

angry_vincent, just made me lookup knot ... thanks man

Midjak: you can also do /home/whatever $path/home/whatever nullfs ...

I would like to use this jail for building stuff from amd64 to armv7. As I understand I juste need to buildwolrd into this arch. Right ?

do you want to build ports, or software of your own?

Well I gave the path from the jail which is an actual path in my system

a go program

I need sdl2

there's more to it than just having a jail

the best way to cross-build packages is to use poudriere, which knows how to mix native and qemu-static

well with go as soon we don't need an external lib no

I don't want package it's my own program

I have no idea if it's even possible to set up a go cross-compiler5

Title: GitHub - veandco/go-sdl2: SDL2 binding for Go

cross-building freebsd itself is amazingly simple, but that won't help you much

Title: GitHub - veandco/go-sdl2: SDL2 binding for Go

too bad...

I thought with an armv7 world it would be ok

there is something called binmisc

maybe it could help?

if go will let you cross-compile, do you even need an armv7 world?

ultimate goal is ?

binmisc combined with qemu-user-static is how poudriere can do cross-builds for ports

this is what I get into the amd64 env bpa.st/V55A

Title: View paste V55A

ah good. i did not know poudriere does it but maybe was on right track about it

/usr/include/sys

(best way to set up binmisc is to install qemu-user-static and do sysrc qemu_user_static_enable="YES", then service qemu_user_static start)

I found nothing about binmisc

Midjak: ah. ok. in that case, do make TARGET_ARCH=armv7 buildworld and then make DESTDIR=/your/jail TARGET_ARCH=armv7 installworld distribution

(in /usr/src)

I am using this wiki.freebsd.org/arm/crossbuild

Title: arm/crossbuild - FreeBSD Wiki

ah cool

what id distribution meaning ?

but the resulting jail won't run unless you install and enable qemu-user-static

I am doing buildworld right now

make DESTDIR=... distribution is responsible for installing most of the config files and some other directory setup

Thanks a lot.

do NOT run it without DESTDIR=, it can wipe out your whole /etc if misused

Where is these is documented ? I have a real hardtime to find information abotu that?

see the start of /usr/src/Makefile

ok that's was the reason I will do that inside a jail

though mention of the distribution target seems to have vanished

But I will get the semae result indeed

So I need to do that touside the jail right ?

do what?

the buildworld

no ?

your source tree is outside the jail?

because I cannot run the jail

no

where is it then?

but If I install world jail will continue working ?

in /usrc/src inside the jail

you use DESTDIR= to tell it where to install to. the place where you install the armv7 binaries will NOT work as a jail, unless you use qemu-user-static

well I don't need jail

(because you can't execute armv7 code on an amd64)

I have to find a way to give this DISTDIR to go build

DESTDIR

> because you can't execute armv7 code on an amd64

yes

you'll have to tell it to get all include files from there rather than from /usr/include

I don't know what else it might need

I think I need to look that for bo build I think

go is your problem, not mine

fair enough. But I am not sure it will be easier with another language :-)

if I needed to do it with something written in C or C++, I'd make a port for it, test it on amd64, then use poudriere to cross-build it

to be fair, the arm people are working on arm32 support for aarch64

yes, but that only helps if you have an aarch64 :-)

cool in order to have only one image running on various arm arch right ?

emulate it

ok

why emulate an aarch64 rather than an armv7?

I have read that corssbuilding freebsd for ARM v7 needs qemu. (this is what i try to avoid in fact)

no. crossbuilding freebsd _itself_ requires nothing.

crossbuilding _ports_ for armv7 on amd64 requires qemu.

hi

unless you're crossbuilding FreeBSD, in which case it does require qemu-user-static, but I'm not sure it actually uses it :D

I use this code in my installation script -> chroot $BSD_ROOTFS /bin/sh <<EOT

and I wonder why it doesn't work while installing freebsd

$BSD_ROOTFS is correct for sure

humm ok

Iget this error with buildworld disctriution bpa.st/5MBQ

Title: View paste 5MBQ

may i share my script with you?

Maybe something to allow into the jail

meena: why would crossbuilding freebsd require qemu?

meena: (I do this a _lot_)

wikan: "does not work" -- in what way?

RhodiumToad: qemu-user-static, not qemu. and it probably needs it because ports requires it, and whoever built the (Pkg)Base build feature didn't look any further.

Midjak: you'd need allow.chflags, or tell installworld not to do chflags

meena: yes, you need qemu-user-static to build ports, and NO, you do not need it to build freebsd itself

RhodiumToad: i know that. poudriere doesn't.

meena: if pkgbase is requiring it, I assume it's to run the target build of pkg

Title: PkgBase: why does poudriere require qemu to cross build FreeBSD? · Issue #1048 · freebsd/poudriere · GitHub

1048 – ep driver fails to detect card when told specific values bugs.freebsd.org/bugzilla/show_bug.cgi?id=1048

thank you I was looking this

Title: Mozilla Community Pastebin/TBhyajHz (Bash)

my installation script breaks around 217 line

RhodiumToad: ah, i didn't consider pkg… but, doesn't base come with pkg and pkg-static?

wikan: breaks with what error?

meena: nope

right, it doesn't. it comes with pkg-bootstrap…

ha, i am really sorry but I can't tell you what error because I don't remember now.

meena: there's a 'pkg' command in base, but it does nothing except try to fetch and install the real pkg

but it was shell error

wikan: then go away and reproduce it

and that might be fraught if we don't build ports for that arch

maybe I would try to install freebsd on virtual machine

you can't expect people to help you if you can't tell them the error you got

wikan: and… if you install in a virtual machine and it works, then maybe the error was specific to the hardware…

will try

wikan: btw, you don't need chroot to do that

pw -R "$BSD_ROOTFS" ...

right. i wanted to fix pw to work without root…

i should put that on a real todo list

hmmm RhodiumToad

echo "$BSD_CFG_ROOT_PASS" | pw -R "$BSD_ROOTFS" usermod -n root -h 0

is it correct?

should work, though I use -h 3 3<&0

(actually i use -H, but that's similar)

secure_data | awk -F':' -v user="root" '$1==user { print $2 }' | pwcmd usermod root -s /bin/sh -H 3 3<&0

so, what is wrong with chroot?

where pwcmd is a function that does pw -R "$rootdir" "$@"

wikan: how on earth can we know that unless you tell us the error?

wait, will install soon

downloaded

will back, must restart my X session

i don't have privileges to run VM :)

i haven't checked in on raidz expansion. how has that come along? is it in openzfs yet?

hm. I may have found another autofs bug

RhodiumToad: cool!

with a -media map I keep getting a subdir showing up named "-fstype=ufs,nosuid"

which, um, shouldn't happen

o_O

nosuid is an fstype?

nope

ufs is, though

That's kind of my point though. . . seems like there's more going on than misplaced syntax.

special_media outputs lines of the form "-fstype=ufs,nosuid :/dev/whatever" when run with a specific device param

automountd is supposed to parse that

Looks like auto_master(5) syntax.

yuripv: not sure I understand your latest commits, zpool iostat -w (first run, specifically) displays trims done since the system was booted, and all the pools I have on SSDs report millions of trim commands executed at around 8-16ms latency.

debdrup: it's only about the default value of autotrim

not related to trim itself

yuripv: but what would responsible for TRIM stuff, other than autotrim?

what exactly are you asking?

my change is not related to how/when/why trim runs, it's related only to default value of autotrim used when creating a pool (if not specified explicitly); it was supposed to be 'on' (same as before switching to openzfs), but due to mess with defines it became 'off'

yuripv: BIO_DELETE doesn't automatically send ATA_TRIM or SCSI/SAS UNMAP without autotrim, does it?

that's completely unrelated?

we are talking about default property value

I guess I don't understand what that means.

the one used when creating pool

you can set autotrim to on explicitly, and autotrim will work

Oh, I _did_ enable explicitly.

right, so it does something for yo

you :)

my change was removing the FreeBSD specific defines that were supposed to have autotrim default to 'on', but didn't work

Right, and it was probably appearing as off before then, due to IFDEFs - but then wouldn't it be better to ensure it gets enabled, to restore the functionality that we had before OpenZFS?

well, may be, but it's been quite some time, *and* the man page says default is 'off'

is anyone around ?

Don't see anyone.

is this the official irc for freebsd ?

Yes.

One of them anyway.

how can I get an official affiliation with FreeBSD for my software projects ?

Title: IRC - FreeBSD Wiki

randomphpdev: Maybe look at creating ports for your projects. That seems most straightforward.

randomphpdev: docs.freebsd.org/en/books/porters-handbook/new-port

Title: Chapter 2. Making a New Port | FreeBSD Documentation Portal

how can I get people interested in developing and writing programming code for my software projects ?

randomphpdev: Putting them up on something like Github is often a good idea for that.

randomphpdev: also advertising that you're looking to hire (and pay) people gets people very interested ;)

no I'm not willing to pay

It's all free and open source

100% free

Then give it away and enjoy when people use it. :)

That said, this channel is for FreeBSD support. You might prefer #freebsd-social.

randomphpdev: you don't really get your software affiliated with the project; even if the project uses a particular piece of software (and it uses quite a few, because it's wholly self-hosted for all things official), that in and of itself isn't really an endorsement.

Also, you probably shouldn't seek affiliation with any given opensource project, and instead aim to support as broad a set of opensource projects as your project allows.

^ good thought there, yeah

I'd like to get an official affiliated with FreeBSD and Arch-Linux

randomphpdev: If you pay them a lot, the FreeBSD Foundation will list you as a corporate sponsor.

Title: Donors – FreeBSD Foundation

o/ I have some extra machines lying around and I'm interested in PXE booting them likely with just FreeBSD or a Linux on them depending, are there any good web UI-based tools you all might recommend which could run out of a jail to make it easy to boot machines with a particular image?

No thanks. I want an official affiliation.

What I'm doing with my software projects here could change the world.

but it won't be for the world it will be for our own gains

that sounds like someone working on Cryptocurrency / "A" "I"…

for sure.