Hello; I'm trying to add open (public) Wi-Fi networks to /etc/wpa_supplicant.conf alongside my WPA-enabled networks. I've read that this should work as long as I make sure key_mgmt=NONE in that network= block. But I've tried on three different public networks and not been able to get an IP address from DHCP. My configuration: bsd.to/wGZK/raw

Title: wGZK

The "authmode WPA privacy MIXED" seems like it might be relevant? I should add that the WPA-enabled networks in my wpa_supplicant.conf work seamlessly.

i would set ifconfig_wlan0="dhcp", i don't think wpa belongs there unless you have a citation on that

rtprio, but then I'd need to edit rc.conf to add and remove it every time I switch between secure and insecure networks, right?

This is a laptop, so I'm trying to make switching between networks I've set up (at coffee shops and the like) fairly seamless.

s/set up/specified the configuration for/

fran: no

key_mgmt=NONE in the wpa_supplicant section

for that AP

I have that, yes.

it should work

With "WPA" removed from my ifconfig_wlan0 line in /etc/rc.conf, I can connect to that non-secure network. But now when I uncomment my WPA-enabled network in wpa_supplicant.conf, the system doesn't even try to connect to that SSID.

there's wpa_cli to interact with wpa once it's running; if you need to change a non-trivial times, i'd suggest figuring out some of its basic usage

you should be able to leave all network uncommented in your wpa_supplicant.conf

I was under the impression that "WPA" had to be in that line for wpa_supplicant and its automation to get involved at all. Without it, I think it's just defaulting to the last SSID I specified.

I had my secure network commented out while I was trying to get the nonsecure network working.

ther's no point in doing that

what's the open wifi section in wpa_supplicant.conf ?

I included it in the paste above.

It looks like I'm able to get the nonsecure network working with wpa_supplicant *if* I manually run `ifconfig wlan0 authmode open`.

So it appears that the problem is that wpa_supplicant isn't changing the authmode to open when it connects to a nonsecure network.

Title: Solved - wifi/wlan0 will not connect to open network | The FreeBSD Forums

huh

i swear i had this working before

It looks like wpa_supplicant *should* be setting the authmode: github.com/freebsd/freebsd-src/blob…/wpa/src/drivers/driver_bsd.c#L1180

Title: freebsd-src/driver_bsd.c at releng/13.1 · freebsd/freebsd-src · GitHub

I'm guessing maybe this is a bug?

maybe; see if it's in bugzilla

Nothing that I can see.

Is there a temporary holding place of links here? Mark M noted the issue for "FYI: upcoming 13.2-RELEASE vs. 8 GiByte RPi4B's based on U-Boot 2023.01 recently in use, given UEFI style booting", lists.freebsd.org/archives/freebsd-stable/2023-February/001129.html

Title: FYI: upcoming 13.2-RELEASE vs. 8 GiByte RPi4B's based on U-Boot 2023.01 recently in use, given UEFI style booting

... in case people would ask about it

Hey

I have a ROCKPi SE single board computer, trying to get ethernet to connect. I can see in ifconfig dwc0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500, but no IP address. Have ifconfig_dwc0="DHCP" in rc.conf. Any suggestions?

Have you set "defaultrouter" in "/etc/rc.conf" or added "default" route otherwise?

I haven't, and I can only see `defaultroute_delay="5"` in /etc/rc.conf. It's a somewhat prebuilt custom image for the platform from personalbsd.org, so not sure if it could be set elsewhere

Either you would need to add "default" route or have DHCP server send one to you

or have SLAAC do it for you for ipv6

if you do have dhcp

try 'dhclient dwc0' from console

or

ifconfig dwc0 inet LAN_IP/24 && route add default GATEWAY

to check its talking to the network properly

then add what you found in rc.conf as parv suggests

See the "dhclient.conf(5)" manual page

you mean if my router has DHCP? Would GATEWAY be something like 192.168.1.1?

Perhaps

Do you have any other computer connected via DHCP where you could check various configuration?

No DHCPOFFERS received from dhclient

Yes indeed

as a long shot try marking the interface up explicitly

ifconfig dwc0 up

also check if there are lights on the rj45 port

just to prove the cable its self is fine and in a working port on the switch

I tried a debian image before and it connected automatically. I didn't check if it was DHCP unfortunately, but I would assume

It's up

Check the values for "default" or "gateway" (along with DNS servers) on the other computer. What is OS on the other 'puter?

netstat -rn on most computers will show the rotuing table

including the default

ifconfig itself should tell you if the link is up

daemon, Hey that works on Windows 10!

yup :)

ipconfig /all

would as well though

simernes: what does ifconfig say about link speed?

I have one Void Linux and one Ubunutu machine. It looks like the default gateway is 192.168.1.1

there should be a line like "media: Ethernet autoselect (1000baseT <full-duplex>)"

so your lan is likely to be 192.168.1.0/24

daemon, I know about "ipconfig"; just never had tried "netstat" before

Remilia: `media: Ethernet autoselect (none)`

simernes: you do not have a link up

check cable

parv, wonder if -rn4 would work too

just to limit the returns

The cable is blinking and as I said it was working when I booted a more preconfigured debian image. Could it still be faulty?

simernes: your dmesg probably has no 'dwc0: link state changed to UP' messages either

it might also be a problem with the interface driver in the kernel, of course

you could try manually specifying link speed

you're right, there's only "lo0: link state changed to UP". But I ran `ifconfig dwc0 up`, I thought that would up it

see dwc(4) for possible 'media' options

daemon, "-4" did not work

simernes: ifconfig iface up means the interface goes up in the OS

media is *physical* link

when it says (none) it means no link was negotiated

netstat -rn shows "0.0.0.0/8 link#1 U dwc0"

try unplugging the Ethernet cable and plugging it back in

netstat will do nothing, you do not have any physical link

'media: Ethernet autoselect (none)' means 'cable is not connected or faulty' or 'the NIC driver does not work correctly'

well, I mean, it also can mean 'remote side has the port disabled or is just turned off', of course

Hunh. The online manual page accessor, man.freebsd.org/cgi/man.cgi?query=d….1-RELEASE&arch=default&format=html, does not know about "dwc" nor did for 12.0-release

Title: Apropos dwc

check `man 4 dwc` for possible options for media and try something like `ifconfig dwc0 media 100baseTX`

simernes: check `grep dwc0 /var/run/dmesg.boot` for any error messages

FWIW forums.freebsd.org/threads/rockpro64-network-adapter.83789

Title: Solved - RockPro64 network adapter... | The FreeBSD Forums

thanks parv, I tried the `ifconfig dw0 media 1000baseTX mediaopt full-duplex` before but also didn't work. Maybe I should boot into the debian image and see what's configured

I am inclined to say problem could be with the driver

If my router has gigabit I should set 1000baseTX, right?

"ifconfig -m" would also show the supported media options

Yes, re 1000baseTX if that would work; else may need to try 100baseTX or whatever else is supported

omg it's working

What did you do, simernes?

as per your suggestion `ifconfig -m` the top entry was `media autoselect instance 1`. The second being `media 1000baseT mediaopt full-duplex,master instance 1`

I tried with "autoselect instance 1" and it has selected "... 1000baseT full-duplex,master"

Sweet

Did you need to add "default" route or did that work out via DHCP?

Thanks so much. No I didn't add any default route

Ok

what does that do?

p4x639: query or set the media type

simernes, What does "route show default" list for "gateway"?

meena: media type?

p4x639: such as 10G full-duplex, 100G autoselect. just try the query on a bunch of interfaces to see

meena: arch user here :(

how is performance of archlinux in bhyve?

any loss?

also win11

parv: Gateway just shows the name of my 2.4GHz wifi network for some reason

simernes, Alright

Do not know why/how that would be the case, but ok

p4x639: ip should have an equivalent, or else, it's in /sys/class/net/<device>/some-file

Can I put this ifconfig in rc.conf or something somehow so it's enabled by default?

simernes: ifconfig_foo0= can take pretty much any option you can give to ifconfig

(in addition to DHCP, which is not an ifconfig option ;)

meena: ty

meena: and WPA, if we're going by things not strictly part of ifconfig

ah, i forgot about WPA

I need to revive my laptop

but, since yesterday, I'm thinking of putting haiku on the thinkpad, and see if I can get FreeBSD running on the system76 laptop

my main machie is an x230 with freebsd ;]

@Demosthenex how's that run, any drivers missing? Battery life?

Demosthenex how's that run, any drivers missing? Battery life?

nice size screen, 12.5" but I'm seeing it is old, pre-installed with Windows 7 😬️

is wifi modules white listed, can you upgrade this?

i slapped on a x220 keyboard ;]

runs fine

low temperature

haven't tested the batt

nice, what window manager/desktop environment?

stumpwm. no icons, no desktop.

just tiling

pmb: may I recommend wiki.freebsd.org/Laptops ?

Title: Laptops - FreeBSD Wiki

anyone knows which linux version is tracked in drm-kmod's master?

yuripv: I think it's "stable" whatever that is or means

yuripv: i think it is 5.15 now

with, maybe, some addition from 5.16 ( git log is unclear, only reduce with 5.16 )

i till getting artifacts, rectangular shapes, anf gpu hang, but less frequent and no kernel panics

i still much disappointed

hey all. possibly an obvious answer, but question regarding upgrading jails with freebsd-update. host is running

13.1-RELEASE-p6. installed a new jail with bsdinstall which installed 13.0-RELEASE for some reason. went to upgrade to 13.1 with freebsd-update but it says one of the installed components is kernel/generic which seems odd for a jail?

yuripv: ( aacd0e4bee Reduce diff with Linux 5.16 ) is what i meant, that's in master

angry_vincent: makes sense, thanks

Can anyone help me on this topic? Or is it out of scope?

Title: Is there any chance for me to read a DVD disc on this device? | The FreeBSD Forums

what happens to resource limits set by rctl on a process when process forks? (i have my guess, but I want to confirm it)

ngortheone: I didn't understand the question.

with rctl you can set limits on a process, using process ID (PID)

when process forks, the forked child has new PID

are resource limits inherited?

or is it a legit way of breaking free from limits?

I have rctl disabled

Clockwork: he's not talking to you

greetings

I have been asked by a user if there is a way on FreeBSD to manipulate the “embedded controller” of some laptops.

Title: Which library will let me write (AND/OR Operations) EC register on BSD (FreeBSD/ GhostBSD)? : BSD

Do we have anything like that?

rtprio: I understand

Clockwork: are there any other slices? /dev/cd0a etc?

rtprio: No. cd0 only

rtprio: > he's not talking to you

what makes you say this?

in that instance I was in fact clarifying the question for Clockwork

perhaps i didn't scroll back far enough; i apologize

no worries, I was thought maybe I said something incorrect (english is not my native)

Hello, I try to find a mean to roll back a newly built kernel after a "make installkernel" as thi new kernel panic on my setup. Is there a clean mean to do that ?

a way*

I mean

Midjak: after you do "installkernel" the old kernel is copied to kernel.old and is available for selection in the bootloader menu, when you see the bestie

read the menu options, if I remember correctly you need to press 2 to switch to old kernel

yes, I am hesitating about just rename old kernal and drop the new one

book kernel.old from loader menu docs.freebsd.org/en/books/handbook/boot

Title: Chapter 14. The FreeBSD Booting Process | FreeBSD Documentation Portal

I succed to boot on the old kernel

Midjak: mv /boot/kernel /boot/kernel.yikes; mv /boot/kernel.old /boot/kernel ;

I just want to roll back to the previous one now

it works fine, i've done it many times

just rename the old kernel back to :)

ok cool

thanks!

loader will boot once, but then you risk a 2nd bad kernel and no good kernel

if that makes sense

and you can always copy the /boot/kernel.old /boot/kernel.keeping (or something) so you don't overwrite it

otherwise what should I do to understand what is going on with the new one?

I've got kernel.13 kernel.13.1.orig kernel and kernel.old sitting here.

I am thinking about a bug report but I am not sure what I should give as information.

what panic do you see?

You'll need to rebuild the kernel with debugging symbols, then use devel/kgdb to examine the data.

chances are, a config option has changed and you missed something, I'd go through both configs line by line and compare them both with the defaults. I'd also boot with the new kernel with the verbose option to get a much better look at what/where it's failing

at boot up on the kernel.new I have built from the last checkout

A kernel without debugging symbols won't give kgdb information to be able to see what's wrong.

ok

thank you both I going to look that

docs.freebsd.org/en/articles/problem-reports this describes when and how to submit a good bug report

Title: Writing FreeBSD Problem Reports | FreeBSD Documentation Portal

what subsystem paniced, Midjak ?

My native language is also not English.

jimmiejaz, I am not on a specific configuration . Is it possible something is changed with GENERIC config which could be the cause of the panic ?

rtprio how can I check that rtprio ?

Midjak: the functions in the panic message

Yes where I can find the logs of the previous boot ? not the last boot because I have reboot on the old kernel but the previous one

Are the logs kept from one boot to the next ?

the panic is generally not saved, iirc

so I should reboot and look the message ok

thanks for the link, but it's been a while since I ran freebsd on a laptop, nowadays I go the lazy linux way... freebsd on servers ftw!

it could be as simple as "you included one kernel config item without the required other one"

the kernel panic after amdgpu kernel modsetting enabled

Fatal trap 12 page fault while in kernel mode

did you rebuild the drm-kmod after updating the kernel?

yuripv: ++

nope

please try that before submitting a bug report

I would to switch on the new kernel for that by the way. I have just follow the user guide

So when I rebuid a kernel I must rebuid drm-kmod, ok. What are the package I should rebuild ?

you don't need to be running the new kernel when rebuilding the drm modules, you just need to have /usr/src corresponding to the new kernel

Midjak: whatever port you installed that provides amdgpu

I dont want specially open bugs yuripv

I have just follow the user guide...

anyway thank you

disable the amdgpu module and see if the new kernel boots

then rebuild amdgpu and reboot again

you should be good to go

rtprio, these are drivers right ?

ah ok. they are drivers from linux.

yeah, the video driver

yuripv, I have rebuilt graphics/krm-mod and reboot same problem. Is graphics/krm-mod I should rebuild ?

graphics/drm-kmod is a meta-port, rebuilding it won't do you any good

you need to see which -kmod it depends on and rebuild that

man, i remember in the old days rebuilding a kernel for specific cpu optimization

and then came time to change the system, and you could no longer swap the disk becaus the new system was amd and the old one was intel or a different generation of processor

so you'd spend another two hours rebuilding the kernel

good times

How canI know what is my -kmod vishwin?

use pkg to get the info of drm-kmod

there is a flag to print dependencies

pkg info -a -o |grep kmod

ok I have rebuilt the package I didn't mange to build on the previous kernel (it was graphics/drm-510-kmod) and now I can boot on the new kernel.

Where drm-kmod is documented ? I have tried man drm-kmod without success

"pkg info drm-kmod"

man -k drm?

To search for installed packages, sans "grep": pkg info -x -o kmod

meena, Thanks. Just learned "kms", in "i915km" for example, stands for "kernel mode-setting"

s/i915km/&s/

parv: i relearn this every time. I cannot seem to remember

meena, Eh, there is no point in remembering everything that one comes across for all times

parv: especially stuff i have no interest in…

Indeed

I do not, even almost, care about the interna of graphics stacks

Thank you meena parv . I think I will have to build the drm-510-kmod. I am not sure to understand evenrything, but it's better now

but the things i know about virtualization, containers, bootstrapping OSes, and ELF could probably fill many, many books

(books you could use to stop doors from moving, or make a balance a shaky shelf, or do any number of other useful things you can do with useless books)

🤣

and it's still so broadly spread that it's still not useful enough. as you can see in #freebsd-dev with how much ngortheone and i are struggling lol

rtprio: `pkg -xo kmod` will do the same

i want to use lldb inside a jail; is it possible to set security.bsd.unprivileged_proc_debug=1 inside that jail specifically? i have Operation not permitted when i try

That'll do.

jmnbtsls1E: sysctl, as the name kind of implies, is unfortunately "system", and can't be broken down to jails or processes

debdrup: old habits die hard

there's some talk of this per jail here reviews.freebsd.org/D18319 , but the relevant jail attribute allow.unprivileged_proc_debug doesn't seem to be present on my system

Title: ⚙ D18319 Make security.bsd.unprivileged_proc_debug per-jail

ngortheone: I wonder if (fixing) this ⬆️ could / should be a goal

meena: well no, sysctls can be per-jail

for example net.inet.ip.forwarding, net.inet.ip.fw.one_pass

have not done much other than that in a jail

It's not a jail property, though - it's a sysctl(8) OID.

It even uses the SYSCTL_JAIL_PARAM() macro.

from reading that page so far, the jail property is allow.unprivileged_proc_debug , to allow the jail to change the sysctl

jmnbtsls1E: allow.unprivileged_proc_debug is a jail(8) param, and then inside the jail you should be able to set security.bsd.unprivileged_proc_debug=1

lol, took me 3 minutes longer than you

yeh, the problem is it's not a jail param though, i guess that code never made it , at least on 12.3

hmm, i thought i checked 13 but i have not checked it yet. maybe it's on 13

It might not be in 12.3, no.

It's in 13.0 according to cgit.freebsd.org/src/commit/?h=rele…54481a13a501bf9f315d160ff768174feff

Title: src - FreeBSD source tree

yeah, i didn't see an MFC

OK, it's in the most recent man page also, understood. thanks

You didn't check the manual pages for your system version?

i did, but did not check the ones for more recent version such as 13

can bhyve attache host filesystems to the guest?

crb: virtio-p9

meena, is that the only option?

nfs… cifs… sshfs…

Oh, yeah I suppose I could nfs mount

do you have any idea performance wise which option is best?

If you're doing performance-sensitive work, you probably shouldn't be using a hypervisor.

Ever since John-Mark G introduced me to sshfs (on Twitter or Mastodon), I prefer to use it over setting up a NFS mount.

parv: why?

crb, I can mount exactly the directory I need; also the "ssh" part of "sshfs"

parv: that's true I suppose but my network is fairly closed and it occurs to me that the two directories that I care about are already nfs exported across my network, that may make nfs the best option

parv: I'd argue that's one of those things that's more context-sentitive; for example, there's a good argument to be made that if both client and server are FreeBSD, you can do NFS over TLS - because that uses just one TCP port over which both the NFS traffic, authentication, locking, and everything else is sent. All of that, in turn, using AES-GCM, will be both encrypted and integrity checked (because

the cipher integrates both, and is optimized for a packet-sized chunks).

sshfs has a few issues that'll also affect performance; for one thing it's incurring an involuntary context switch every time you're doing an action because it's relying on fusefs, but in the event that you're on a network you trust (ie. one you setup and maintain), there not much gained from the actual security promises of using ssh.

Except it's not just one context switch, if memory serves - I think each action in fusefs is at least 3 because of the boundary crossings?

sshfs, like many other things, isn't a one-size-fits-all.

debdrup, Sure, re: context. Work mounts NFS without TLS & NFS v4 server on Rocky Linux; got no time to set up "idmap"

parv: so no NFSv4 ACLs then? :P

To be clear I do not care to set up "idmap" on my FreeBSD machine. That is set up on other CentOS & Rocky Linux machines

holy shit, installing/running a linux client under bhyve seems pretty easy and fucking cool!

ok so one more question, do bhyves have to be started by root or can an ordinary user start a guest OS?

The answer is that it depends.

There's a patch floating around somewhere on bugs.freebsd.org that lets you create unprivileged guests, but it means that it won't have PCI passthrough and there's a bunch of other things that won't work.

Title: FreeBSD Bugzilla Main Page

What you can instead do, though, is run bhyve inside a jail. That gets it a much stronger isolation from the host than simply running it unprivileged.

debdrup: thanks

I finished maintaining the laptop: wiki.freebsd.org/Laptops/HP_Compaq_nx6320

Title: Laptops/HP_Compaq_nx6320 - FreeBSD Wiki

There's also some issues with guest memory accounting, although the extend to which this is a problem outside of reclaiming the resources if the guest is shut down and/or using rctl(8) isn't clear to me.

But since it isn't properly freed, that could in theory lead to security issues.

Though whether this extends beyond an information leak isn't something I can speculate about after 2300 hours on a sunday ;)

Clockwork, Havd not seen "Compaq" for a long time until now

Isn't it just HP?

s/Hav/Ha/ # Ha!

I also wrote about /HP_Compaq_6715b

This name is written on the cover of the laptop.