sometimes if i restart a jail with an iov network device, e.g., iavf0, the network device never becomes available and it's not possible for the jail to start. it used to be slow, e.g., might take 5-10 seconds for the device to become available. but now it's never. sound familiar? i'm not seeing any reported bugs so hoping there's something simple that might cause this.

and out of curiousity, is there a way to force the device to reappear in the host? like send some command to the pci device?

Ok... is there really a 5.38.0_1? This seems odd: pkg: perl5.38-5.38.0 conflicts with perl5-5.38.0_1

yes, it was bumped

but that also looks like inconsistent DEFAULT_VERSIONS

learned a new trick: devctl reset -d iavf0

that resolved my issue

Altq vs dummynet which provides a more efficient qos usage ? [ Can anyone share any document or information on this subject?]

any way to take a thumb drive freebsd installer is on and find out what version of fbsd it installs?

i can reboot into it and see i think but what about from another freebsd without rebooting and just mounting it

There is "/etc/os-release" which could be checked

... may need to be extracted from "base.*"

how can i check? i put thumb drive in, gpart show sees it, i try to mount it but mount says invalid fstype: invalid argument

Have you tried mounting with the file system type listed by "gpart"?

nah, don't know how. that in a man page?

Paste the "gpart list" output for the USB disk

Sorry, not "list" but "show": gpart show

mount has "-t" option to specify the file system type

parv termbin.com/0nxj

Manual pages: mount(8), gpart(8)

Try: mount -t ufs /dev/da0s2 /mount/point

same error

is it not allowed to mount a thumb drive with the fbsd installer already written to it?

What are the exact error message & the command used?

there was 1 more dev that worked. mount /dev/da0s2a ~/mount/point worked

da0s2a wasn't shown in gpart show. wtf?

why it hide that from me?

hello. anyone using the kitty terminal emulator by any chance?

Rue: yes, but not much on freebsd

starting to believe iTerm2 should be ported to all platforms

it's pretty great

closest ive come to that on windows is securecrt & putty but those in themself are just meh

im not one for pretty shit besides my wife but honestly functionality lacks hard for terms on other platforms that are not using iTerm2

kitty... just another rxvt

kitty supports images, for whatever that's worth

there were too many whistles and bells in iterm2. in a lot of them, really

yeah its basic at most

Just me over here on my 2nd day of going through port options screens :/

lol

CmdLnKid: tried new windows terminal yet? :D

haha powerhell ?

terminal, not shell, you can just `ssh` from it

not much for terminals on windows unless its specific to windows, otherwise install sshd on windows and use iterm from mac

;)

helps also having macOS installed on a VM just incase you are not on a mac

ill get there eventually

hackintoshing a bhyve vm sounds exceedingly difficult

yeah i don't think all the mechs are there for that

I can't imagine a good experience running a hack in bhyve

and now even more complicated as its no longer arch amd64

agreed

Even running a hack in proxmox is fairly miserable... I have one just for running some AppleScript services and I'm constantly thinking about how to retire it.

debian >>> vbox >>> macOS, but that in itself is still iffy now

Sorry... I don't have one.. but if I did, I'd be thinking about how to retire it :)

lol

I'd never virtualize MacOS on non-apple hardware

ffft

OS aside, my favorite thing about Mac is the keyboard. haven't taken the time to figure this out on freebsd yet, but the first thing i do on a linux install is map the keys with udev: markmcb.com/linux/mac-like-keyboard-in-linux

Title: Mac-like Keyboard in Linux - Mark McBride

markmcb, heck yeah!

dunno why but thats just so natural to me its uncanny

yeah, it just seems much more ergo by default.

i wasn't into mac before around 2015'ish. took me a week to get a grasp vs pc

maybe its how im wired but its just right

Other tip on Mac ergo: get Karabiner and map your capslock to a meta/macro key. then you can do things like capslock-h/j/k/l and have global vim-style arrows

if anyone knows how to do this sort of stuff in FreeBSD i'd be interested to know

most of what i find assumes you're in X

thats funny, spilled beer on this macbook pro 13'ish... have to use karibeener to do this capture.dropbox.com/1y27Z6AVWUyUPGEH

Title: Dropbox Capture

lol

it eventually dried out and need k to disable the left shift

cause it seems to activate randomly

passwords become a real bitch

unixman_home: i tried to tie it back to freebsd :)

unixman, has been on lsd

;)

i'm noticing a lot of extra plumbing required for jails to be useful, like creating epairs and so forth

no wonder a lot of folks use jail frameworks

It's pretty simple to have epairs set up automatically.

cloned_interfaces in rc.conf, right?

Title: MasonLoringBliss/JailsEpair - FreeBSD Wiki

Yeah, a bridge, and then something that spins them up per jail.

oh even better

i already have bridges since this host runs bhyve

does ${ep} in that example need to be incremended manually for each jail or will that autonumber?

oh nm, i see how it is now

derp

same way vnconfig used to hand assigning interfaces through a script ... VN="$([vn,md]config -c)", then then $VN to define the interface that is available for that ?

sorry i know ... disks vs network

same philosify

no, i didn't read the whole snippet before asking

fuck... and spelling

speaking of command line kid, i worked with a dude who typed the command line so fast it was quicker for him to ^A ^K and retype the whole thing rather than backspace a few times to correct his typo

why he didn't use ^r and fix his mistakes i have no idea

habbits my friend

not the same guy tho

my nick was created somewhere around 2014 after 29 years of linux command line usage after typing a 1 line perl live in front of a friend that was well involved with the command line and compared it to billy the kid

Title: ‎As Good As I Once Was by Toby Keith on Apple Music

mason: ok i like this so far, but packets are going out but not coming back in

firewall

no firewalls

the jail has the right arp table

nat ?

not even trying to escape the network, yet

k

i see the packets pass through epair to the bridge to the vlan to the lagg outbound

i wonder if i should set this up with a different interface

hmmm

i'm running freebsd 13.2. i have a network interface set up with 2 vlans. i'd like to use dhcp to obtain static ip addresses and have different hostnames for the ip on each vlan. it seems like it's sending the hostnames (?), but the macs are all the same. this is an example of my config reduced to 1 vlan: bsd.to/A77Z/raw

Title: A77Z

the effect is that i only see 1 instance of this pc in my router's client list.

and does the vlan create ok?

yes. i thought later that i should've included the ifconfig output.

do you control the dhcp server?

ifconfig: bsd.to/crNr/raw

Title: crNr

yes, it's a ubiquiti dream router at home.

how are the vlans configured on the router?

because to do this you need the dhcp server to distinguish between the same mac on different broadcast domains

(the mac addresses being the same is how vlans are supposed to work)

i do get different ips within the range of each vlan. that part seems to be working.

but for example, right now i can perform a lookup for foo-lan and get the ip of the physical adapter, but i can't lookup foo-office to get the vlan ip.

what kind of lookup?

host foo-office

is your dhcp server configured to do dynamic dns updates, or are you manually adding dns entries?

they're currently dynamic.

i was planning to make them static for the physical + 2 vlans after i get all 3 to show up.

how are the vlans configured on the router?

is that dhclient.conf correct for the vlan interface? should i use something else like interface "re0.20"?

chriswells0: i prefer that method over yours but they both should work

it has a web ui for creating networks. each vlan has a vlan id assigned along with a different /24 cidr and dhcp enabled using a range on that cidr.

i'm not sure (and may not know) the specific info you're asking about how it's configured since i created it through the web ui.

does the web ui have a way to show the current dhcp leases?

it shows me all the clients, but it's not showing this host 3 times. i believe that's because of the mac overlap.

mac overlap ?

I suspect that's not the issue

physical + 2 vlans with the same mac.

one could only expect a problem with a mac overlap

having the same mac address on all vlans is HOW VLANS WORK

it could just as easily be a bug in their ui for the same reason.

if you're using dhcp on multiple vlans it's up to the dhcp server to keep them separate

it doesn't mean it's "wrong."

aware of that. but... why would you do that.... its not scalable

what's not scalable?

convergence

consider converting from one vlan and converging into another to expand

do not ever duplicate anything

if you ever have that then go in a diff direction

there is no reason to have the same mac on a diff interface vlan or otherwise

maybe the problem is that each vlan is configured to have a dhcp server?

not likely

if it's running a different dhcp server instance for each vlan, it might now track them well across vlans.

it's not supposed to track them across vlans?

why would you

id redetermine why you are doing it the way you do

"(12:39:38) RhodiumToad: if you're using dhcp on multiple vlans it's up to the dhcp server to keep them separate" -> it has to know about them to keep them separate, right?

this coming from someone that has migrated clients from one vlan to another

try not to predict where you are coming from, but where you unpreddicatedly are going to

chriswells0: if there are separate dhcp servers on each vlan then they are automatically kept separate, no?

ok, i was only asking to be sure it shouldn't be managed at a single dhcp server.

if you go out and buy a linksys AP that offers up DHCP which one is prioritized over the other. this is a very admin overhead cost you do not want to afford

whichever one of the interfaces gets an ip last is the hostname that resolves.

wish that was true

i'm telling you what i'm seeing--not what you might expect.

ah

see thats what happens

let's assume that's a bug in the router. what would i need to change in order to send a different mac for the dhcp request?

the hardware config

is it dhcp-client-identifier in dhclient.conf?

worth a try

still only 1 shows up in the clients list, but all 3 hostnames resolve now. that's good enough.

don't settlle with good enough

i need to get some work done today. :)

makes sense

i'm off tomorrow. maybe i'll care more then.

thanks to both of you.

so a stupid question: I have a uefi linux vm running in bhyve using the vm package

I connected to the conole with "vm console myvm"

but I can't disconnect from it

the documented key squence does not work for me: ~+Ctrl+D

help

i recently got a microSD card that advertised as "U3" "Class 10" "V30" "SDXC". anyway, dmesg says "[1] mmc0: <MMC/SD bus> on sdhci_pci0" "[2] mmcsd0: 126GB <SDHC SDABC 1.0 SN AA000113 MFG 06/2023 by 111 0x0303> at mmc0 50.0MHz/4bit/65535-block"

my issue is that transfer speeds are like 10x slower in freebsd than in linux, causing it to overheat(?) and cause i/o errors after transferring like 1-3GB, whereas on linux it can transfer 50+gb

ngortheone: iirc it needs to follow an enter key

SymbioticFemale: is it the same filesystem?

no filesystem involved i'm talking about straight writes with 'dd'

writes to the block device

iirc linux has some assumed options with dd. are you sure it might not be ignoring io errors?

good question. i'll retry the operation with ddrescue and that might solve the non-speed related issue

Any read/write errors ought to show up in dmesg in Linux btw

Can I / how do I create loopback block device (from plain file)?

for the record, usage of ddrescue instead of dd not only resolved my issues with read errors but it also somehow resolved the speed issues

magic

indeed

I blame FreeBSD

For the life of me, I cannot build lang/rust in poudererie... 13.2-release

SpaceBass: ENORAM?

mason, what's ENORAM?

Error, insufficient RAM? It's big.

SpaceBass: Here, look at this too: freebsd/poudriere #888

Title: Rust + tmpfs - need a way to selectively not use tmpfs for some packages · Issue #888 · freebsd/poudriere · GitHub

888 – kernel doesn't compile with options DEVFS enabled bugs.freebsd.org/bugzilla/show_bug.cgi?id=888

Could be... I'll try again and watch

Check that out before trying again. Might be useful. TL;DR disable tmpfs maybe.

It's only building as a dependency for something else... can I just pkg install it inside the jail? I dont need to build it from source, per se

I think there's a way to do that but I don't know what it is.

There's #poudriere, perhaps both here and on efnet, that would likely know.

In fact, I should join those.

If you ask and someone answers, then I learn too. :)

SpaceBass, you can use poudriere-devel and "-b latest" where "latest" can also be "quarterly"

meka: Does that support binary build deps?

To pull a newer version of the port?

For example, this is the command I used today "poudriere testport -j 13amd64 -p local -b latest devel/py-factory-boy"

meka: It's likely an issue running out of resources more than a bug in the version of the port he's got.

the way it works is that if it can, it will use pkg to install binary, otherwise it will build it

I'll watch top and try again

aha

It's a vm, I can throw more ram at it if I need to

meka: This is new in poudriere-devel then?

mason, correct

Nice.

I'm using it for few months, not sure if it got to stable version

Ive got 500gb of disk space available, assuming the jail isn't constrained

So far, compile is using 5G out of 12G of ram

Is there a better way to monitor this than top?

SpaceBass, poudriere has web UI

Maybe it's sufficient for what you want to monitor?

I've got it

I wouldn't really count on it, but maybe ...

hrm, i can't get my jails to talk with epair

it seems the packets go out, but not back, it seems

anythings i can check?

rtprio: probably worth grabbing a pcap

Ive fallen so far down this rabbit hole and broken a server (thank goodness for a backup) ... I built freeradius3 with pouderie using heimdal, but whenever I try and install it, pkg insists on installing MIT kerberos which will really break things

SpaceBass: if you built the ports on the system, i suspect you won't have a problem

mason: i'll try that

rtprio, im trying to move away from ports and use pouderie

then maybe make config to force no mit kerberos

Did that too :/

And cleaned the configs for all the dependencies... I cannot find MIT anywhere... and anything with an option is set to heimdal

well something has to be doing it

mason: i can see the outgoing packets hit the wire, it appears but nothing comes back and i'm not sure where to locate that

How can I list valid root device specifications for loader/fstab?

ccx: what are you trying to do?

rtprio: set boot device in a way it doesn't get renumbered. Either by physical bus ID or some metadata. I know loader shown me a list on prompt, but I'm not sure how to get it from running system.

I've set label via tunefs -L but I'm not sure how to specify that as device name exactly

And mount/fstab/loader.conf manpages weren't of much help.

rtprio: I'd grab a pcap further down the line - see if they make it off the box.

brb

I just dont get it... if I install something from /usr/ports eg lang/perl5.38 and then try and use pkg add to pull a pouderie package (eg net/freeradius3) it tries to re-install perl5.38 and fails

They are the exact same minor version number

Its like pkg has no awareness of ports?

ports are just used to build packages, pkg only needs to know about packages

does it try to reinstall perl saying something about changed options?

It says it conflicts and then fails

ccx: i think you need glabel