What are the issues again with having an over-sized swap?

Anybody have suggestions for gigabit ethernet card with open source driver

yetoo: anything inteL?

The standard Intel GigE ethernet cards have good free software driver support.

I'm looking for 4 port 802.1Q. Are they all open source. I'm weary of botnet from intel, but I suppose there's no choice given I can't find anything but Intel for my criteria on bsd-hardware.info

Botnet from Intel? Reference?

rwp: backdoors in amt/ME cus gov. I'm wary of silicon backdoors or bad firmware in some soc on the card.

i know amt/me is motherboard, but i'm pointing out gov overreach

cooperation

_xor: only issue with over-sized swap I know of is that the kernel will whine about maxswzone, and may be unable to use all of the swap space.

I'm just going with 4G for the default size in this installer image.

what arch?

amd64

...though I do have a question about this really quick: cloudinit.readthedocs.io/en/latest/…tml#dmi-specific-kernel-commandline

Title: NoCloud - cloud-init 23.1.1 documentationMenuExpandLight modeDark modeAuto light/dark mode

"Cloud-init performs variable expansion of the seedfrom URL for any DMI kernel variables present in /sys/class/dmi/id (kenv on FreeBSD)."

How does that map in kenv? Is it literally just sys.class.dmi.id?

no idea

I looked at a list of vars on my system and saw that they're named differently.

Hmm yeah it's a bit ambigious.

Like, dmi.system-uuid listed in the table on that page maps to smbios.system.uuid on my system.

dmi.system-serial-number == smbios.planar.serial

Wait, maybe not. There's also smbio.system.serial but it's left at the original value.

smbios.system.serial="To be filled by O.E.M."

Yup, mine too. I have what looks to be a valid serial string for smbios.planar.serial though.

(that's what it says on mine :-)

smbios.planar.serial="To be filled by O.E.M."

hmm

What's the syntax at the boot loader to set kenv variables?

set blah=whatever

oh that's simple.

That's equivalent to adding my.variable.value=blah in /boot/loader.conf?

yes

Ooooh, I get it now.

Those variable names are correct, the docs are just written...less then greatly.

cloud-init knows that on FreeBSD the variable names are different. I grep'ed the source and found where it's being read.

Title: cloud-init/dmi.py at 6d09df5e4786a2a6c79d6098ab413c93b205221c · canonical/cloud-init · GitHub

RhodiumToad: Does your system have a valid UUID for smbios.system.uuid?

... for some value of "valid" - it has a uuid-like string, but the version and type fields are invalid

smbios.system.uuid="031b021c-040d-056c-4106-100700080009"

valid uuids have xxxxxxxx-xxxx-Vxxx-Rxxx-xxxxxxxxxxxx where V is 1,3,4,5 and R is 8,9,A,B

(and V=3 or V=5 are special purpose non-random non-unique UUIDs)

RhodiumToad: I was just checking my other systems and noticed the funky IDs too.

I found out the hard way a good while back about how close UUIDs can be generated.

I think remember seeing something in the changelog for 13.2 which addresses that.

"The uuidgen(1) utility has a new option -r to generate a random UUID, version 4. 8fd1953b7eb2"

which are time-based like v1, except the mac address is replaced each time with a different random multicast address

Good.

the kernel uses a random multicast address for uuids on boot, but it generates the random value only once

I had an issue with some template generator a while back where one of my templates was generating UUIDs in a loop...except I noticed all of the UUIDs were the exact same. After reading up a bit, turns out that there's some kind of caching behavior some UUID libs use. It wasn't just in this template generator either, I noticed it with uuidgen too at

some point.

It was really strange. Ended up doing some reading on various id schemas, including the various versions of UUID. I did kind of like ULIDs, but they're not widely supported.

caching uuids seems... weird

Yes, very.

Ah, this was the issue I was running into I'm pretty sure...

Title: Ansible random UUID generation - Stack Overflow

Oh, the caching was in the template rendering engine. Not the UUID generator.

Still, kind of dumb when there's a uuid() function. You would think that would be blacklisted from being cached.

Title: GitHub - ulid/spec: The canonical spec for ulid

"N+1 competing standards"

I have a system which is using zfs and I want to mount a disk to my home directory videos directory. I have this is the fstab. Everything works once the system is up and running but fails at boot time.

Anyone have any ideas

fails how, and what is the fstab entry?

the mount is /dev/ada1p1 /usr/home/bdavidson/videos ufs rw 0 0

once i log in if I do a mount -a everything works

fails as in I have to enter single user mode and comment that line out and then restart

its just a normal ssd drive with a freebsd-ufs partition

change rw to rw,late

see if that helps

the issue with that is then I have to do a mount -o later right

no

right

what does late add to the party

the "late" flag just means it's mounted by the "mountlate" service on startup rather than "mountcritlocal"

ahhh

cheers

will try that

thanks heaps

(it's the "noauto" option that stops it being mounted at all on boot)

cheers ill do some of my own googling for my own knowledge now

thanks RhodiumToad

yw

exit

Hello I am having an issue within kde(x11) and I know this issue can happen on linux as well, but when I scroll up on my trackpad a 'back' signal is also sent. meaning that when browsing the web, when I scroll up, the browser will go back a page. Anyone have experience with that?

I found a fix many years ago but can't seem to find it anymore

xinput is likely the command you want to fix the issue.

Run xinput to view you inputs then run xinput -test <id>

As you scroll your trackpad you'll see which input is being triggered.

Title: dpaste/a0f7 (Plain Text)

that's what I'm getting

not sure if a[0] and a[1] are different inputs

This is what I have for my mouse: xinput set-button-map <id> 1 2 3 4 5 0 0 0 0 0 0 0

My issue was when scrolling the back button would somehow get injected.

The above command disables the side buttons.

Similar to your issue

[6~

.

good moring everyone im running into an issues here im trying to change the font color of my freebsd box when it boots up instead of having the traditional white fg i want to have the color green fg on black bg ive added the following to the kernel confiuration file options SC_PIXEL_MODE , options VESA

ptions SC_NORM_ATTR=(FG_GREEN|BG_BLACK)

options SC_KERNEL_CONS_ATTR=(FG_WHITE|BG_RED)

and then did a make buildkernel KERNCONF=NEWKERNEL && make installkernel KERNCONF=NEWKERNEL but when i reboot its still the same traditional white color and im not sure what else to try

the SC_* options are for syscons, which is the old console driver; I don't think the vt driver respects them

for vt, there may be some usable options for loader.conf

man vt for info

Yeah that is what I want to change is the sysconsole

Color

Hello, I have a problem with a recently updated FreeBSD 13.2 system. Is this a correct place to ask?

yes

Good. The OS has become unbootable. During boot after the "Loading kernel..." and the "/boot/kernel/kernel" strings an error message is displayed which says "elf64_lookup_symbol: corrupt symbol table".

After that the system hangs

Any tips?

can you get to the loader's OK prompt?

I think so, give me a moment

In any case the error happens after the boot loader so I should be able to

try and boot from a kernel.old if there is one

(i.e. /boot/kernel.old/kernel)

no, same error. i'll try to boot from a freebsd iso

I've booted from a freebsd bootonly iso. ldd'ing /boot/kernel/kernel shows an error about ld-elf.so.1. ls'ing /libexec shows that the file is there but all commands on it, such as stat, file, etc. say that the file does not exist

moreover, I tried the same from systemrescuecd and it says that the file is a stale file handle

this seems to imply that the file is corrupted, doesn't it?

no

ldd won't work on a kernel, it operates with rtld direct-exec to do magic

ld-elf.so.1 must exist or you're running entirely out of /rescue

(ld-elf.so.1 == rtld)

so what is the problem in this case, the kernel or rtld?

Hey Dianora - LTNS.

:-)

kevans: which talk are you in?

loader themeing

you?

ah I am in porting FreeBSD to Firecracker

I did some porting work at BSDi

so it interests me

ah, nice! colin was nice enough to offer sending his slides last night since I was going to miss it

felt obligated to sit in on loader talk

it's being recorded

yeah, that may be a bit before it goes live, tho

I remember BSDOS crashing during boot due to a race condition in the serial driver...

it's.... annoying

I think I had to blink LEDs to find it

hah

yes the good old outb(ch, 0x3f8)

I'VE DONE THAT!

One does outb('A', 0x3f8); etc. for each location you get to

Dianora: thanks for the talk :3

debrup: was it ok?

I am very rusty at giving talks. Been a long time.

Dianora: aside from the mic being upside down to start with

sorry about that.

I was still warming up anyway.

I wouldn't expect anyone to be professional about giving talks, because I'm sure I'd do poorly.

FreeBSD doesn't run very well without a timer early on

16 chars is a hw buffer

weird

ah yes

sorry, how do I invoke single user mode from the OK prompt?

* nvm, I can use the menu

blu256, "boot -s" IIRC.

yes

or if you really like pain, you can do it in two steps: set boot_single="YES", then boot

the menu is nicer, but thank you

I solved the original problem by copying the kernel from the bootonly medium to the system

now I *only* have a corrupted ufs to deal with (sigh...)

thank you everyone

Dianora: You're porting to Firecracker?

Where should I look if I want to keep an eye on that?

What is firecracker in this context?

Title: Firecracker

Title: Announcing the FreeBSD/Firecracker platform

Someone isn't watching BSDcan, I see.

I wish I had time for such fun. .

Oooooh - FreeBSD in an AWS Lambda. . .

wonder how that integrates with the whole IAM auth maddness.

laaambdaaaaaaaaaa

complex

did they steal that from Black Mesa and Half Life by any chance

No, that would have been a lot more fun.

realistically the half-life lambda is from physics and the aws lambda is from lambda calculus in computer science

well, this is my guess, who knows how they decided :)

it's all Greek to me

My first exposure to lambda in computer science, is essentially a function without a name in Python.

And an AWS Lambda is essentially "serverless" function.

anyone using an IDS that can block an ip/subnet just for sending syn packets without follow-up ?

last1: check out syncookies(4)

See also: net.inet.tcp.syncache.rst_on_sock_fail, net.inet.tcp.syncache.rexmtlimit, and net.inet.tcp.minmss and some other OIDs in the net.inet MIB in sysctl(8)