Can sndio be used over oss ?

Not sure I understand the question.

is there something special required on FreeBSD to link in libstdc++ with clang++? I'm getting undefined references for basically everything c++-related...

so far I have tried to compile NymphCast on FreeBSD (14.1), current GhostBSD with both GCC(13) and default Clang, but I'm running into all kinds of undefined symbol errors...

if anyone feels like trying it, I got the current state up on GitHub, also following the patches for the v0.1 package that's up on FreeBSD Ports already: github.com/MayaPosch/NymphCast

requires the use of gmake, uses g++ by default, but NymphCast can be built with Clang using `gmake TOOLCHAIN=clang'

I'll be marking the FreeBSD built as broken for the upcoming NymphCast release for now, I think

hackkitten: freebsd uses libc++ by default, not libstdc++. are you compiling with g++ for a specific reason? it should still work as long as you link with g++ as well, not clang++

but then all your C++ deps have to be compiled with libstdc++ as well

I used g++ because that's what I use by default on Linux and MSYS2

but I support Clang in the project too

on freebsd, using clang/libc++ is much preferred if possible because it means user doesn't need to install gcc to build your port

that's something that I'm still working on with the build system, yes, as the Linux systems that I support have GCC installed by default

so it's not urgent there

well, if you need to build with g++ it should work, but you need to both compile and link with g++ and you can't have any build dependencies (e.g. other ports/packages) compiled with clang/libc++. can you show a few of the undefined symbols you ran into?

for g++ every single one in the libPoco shared libraries, for starters

with clang++ one in libPoco (SystemException), but 99.9% are basic_string and such related

i don't know what libPoco is, but did you compile it with g++/libstdc++ as well?

does libstdc++ need to be explicitly linked in?

this libPoco: ports.freebsd.org/cgi/ports.cgi?query=poco&stype=name&sektion=all

that depends... if you're using a pure C API, but your dependent library requires libstdc++, yes, you can link it in. but if you have a C++ API, that won't work as libstdc++ and libc++ are not API/ABI-compatible

yet g++ and clang++ should handle that part anyway

right?

e.g. you can't pass an std::string from a C++ application compiled with clang/libc++ to a library compiled with g++/libstdc++, it simply won't work

I thought FreeBSD used the same C++ lib for GCC and Clang

no, base system is clang/libc++, if you install gcc from ports (it's not part of the base system) that's libstdc++ and they are not compatible

I see

that does not explain why clang++ cannot find basic headers, unless it tries to somehow use g++ files then

"yet g++ and clang++ should handle that part anyway" - if you have a dependent library that uses libc++, and you're linking with g++/libstdc++, the linker *should* pick up the transienet dependency and link the resulting executable with both libstdc++ and libc++

i'm not sure how well that would actually work as i've never tried it, but, i would expect it to fail at runtime rather than link time

... this is certainly not something i would recommend doing though

well, in this case I'm doing all the linking with clang++

if you compile with g++ you should really link with g++ to make sure you get the right libstdc++ version

but it sounds like the right fix is to just support clang in your build system

Howdy Folks, I just got a SATA Hard DISK, which I am gonna mount in Dock Station I got to install FreeBSD, do I need to make a partition as in like MSDOS format or something, cos at the moment its not regonized by my machine

ivy: I am trying to do so, but clang++ cannot even find the headers for libPoco... thanks for the help, but I think I'll be giving up on supporting FreeBSD for now

Fresh installed on a VM. I don't have vim, want to install it. Bootstrap pkg install is telling me "libmd.so.6 not found, required by pkg". How to remediate?

quark: what freebsd version (uname -v)?

15-0 current.

15.0, rather.

i think this broke in 15.0 recently as libmd sovers was bumped and the pkg package hasn't been rebuilt yet

not sure if there's a solution currently, could you use a stable release (14.1) in the mean time?

Oh! Indeed I can. Thank you.

unfortunately using freebsd.org binary packages on 15.0 sometimes causes issues like this since it can take a week or two for the packages builders to do a new build

quark: as a workaround, you could also try using 'pkg-static' instead - however any other packages built against libmd also won't work

I will go for 14.1, is all good. Serves me well for always wanting the latest and greatest. LOL.

fwiw, i've been running 15.0 here for ages and it works great, but i build all packages locally as well to avoid issues like this

How will I go to do that?

the usual method would be via poudriere, which bulk-builds packages in jails

Ah!

hey dudes

quick inquiry

not a dude

do I really need X org and XFCE DE

oh sorry dudette

well you never "need" X.org

if you want a GUI you have two options, X.org or Wayland

excellent

well, I wanna utilize FreeBSD for few purposes, one is Apache2 Web Server (PHP and may be Java with Spring Boot but definitely JavaScript based web application with React.js in front and Node.js in Back end), but more so, I wanna utilize it as media server with Apache2, file server, print server and MySQL and PostgreSQL Database Server

well, none of those need either X.org or XFCE, so you should be fine

excellent dudette

Netflix is running in FreeBSD

I wanna develop similar clone

with JavaScript or Java

You definitely do not need to run X or Wayland on a headless Internet server. I never install X on servers.

how can i automatically renew my Kerberos ticket when it's about to expire?

ivy: pretty sure the only option is scripting with expect and some form of a cron job, which is certainly not recommended since you're putting your password in plaintext somewhere

kerberos tickets are meant to expire

dstolfa: why would i need expect? you don't need the password to renew a ticket

how do you auth to your domain?

i.e. when you do a kinit ivy@SOMETHING, what's the auth process?

password

and are you passing -r to it to get a renewable ticket?

if so, i'd assume a cron job would do it?

yes, i use kinit --renewable (i guess that's more or less the same) and i can renew the ticket using kinit -R

i suppose cron would work, but i thought there was some daemon that could do with for all users with active tickets... but i only heard rumours about it, i'm not sure what it's actually called

s/do with/do this/

i don't want to renew my ticket after i log out though, then it should expire. hmm, maybe i need some PAM stuff to destroy the ticket on logout

ticket cache is per user, so if you destroy your ticket, its gone for all sessions in that system.

yeah, i know. if i'm logged in twice i wouldn't want it to destroy my ticket because i logged out one session

ivy: if you figure out how to make kerberos renewal less crappy, i'd love to hear about it. the way i manage it now is a bunch of scripts that distribute my ticket through a number of machines and i have it last 12h (non-renewable). i do that every day i log in basically

maybe i should use forwardable tickets instead of kinit on the remote system

dstolfa: the reason i asked is i was having a discussion with Rick about Kerberized NFS and he mentioned such daemon existed

maybe i should ask him for more details :-)

ivy: feel free to forward them once you find out :P

it annoys me that Kerberos works better on Windows than macOS, tbh

MIT Kerberos for Windows automatically prompts me to renew my ticket before it expires, macOS doesn't seem to give a shit

i guess this is because no one except weird nerds uses Kerberos anymore unless it's with AD

bonus points if your login shell basically deadlocks if you have a kerberized nfs home directory. this is why i always use the default shell without any extensions

i ran into that and even posted to fs@ about it but didn't get a solution, i think automounted might just be broken there

s/automounted/automountd

i've had this happen on linux so i reckon it's not a freebsd specific issue

the specific problem i had was that ssh tries to read $HOME/.k5login but it can't read that until the home directory is mounted, so it just denies login

apparently this is intended behaviour because it's possible to *deny* logins in .k5login, but i've never, ever seen anyone do that in the wild

this is what Rick actually said: "4.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.1.0.1.d.5.b.a.a.0.b.8.0.1.0.0.2.ip6.arpa

uh wait that's not what he said

"You are probably aware that there are utilities out there

that run as daemons and renew TGTs if they are renewable."

so not very informative

thanks rwp

It's a security issue. Installing X pulls in a godzillian things that are not needed on a server. Yet all servers are an attack target. Best to limit the attack surface.

that's not a security issue

it's a security issue if you *run* those things but not just having them installed

(except a very small number of things that are setuid, but i'm not sure X stuff is setuid nowadays)

Those things pulled in by desktop environments are fluff and glitter and not really audited to the same level as say nginx+fcgiwrap and will certainly have more problems more likely.

not like the old CDE days what certain people could exploit dtappgather to get root access on the local uni's Solaris 2.5.1 system (no idea who this comment is talking about)

ivy, I disagree because these days successful attacks are almost always multiple levels deep. And they are making use of crazy things that most of us just don't think about normally.

jails.

rwp: those 'crazy things' are certainly interesting but very rarely, or never, have anything to do with a random binary lying around on the filesystem

Every year after Blackhat I look through the exploits and some of the cascading chains of exploits that are used are truly amazing.

everything goes into its own jail, with vnet and sysvipc isolation.

case in point: the latest bhyve exploits

they're more like some binary linked to libsomethingweird.so and libsomethingweird has a weird exploit in it

but the executable has to link to that library in the first place which means it automatically installed anyway

These days I keep hearing a new term. "Living off the land." That's where attackers make use of what's available. And so it's good to make as little available as possible.

there's nothing stopping you from putting bhyve in jail, and an exploit that lands you in userland as root (like this one) would be mitigated by landing in a jailed root

even better if that jail only contains bhyve, and no other userland

+1 debdrup!

i was actually *really* interested in the recent SAs about 1) escalation from bhyve guest to bhyve host and 2) exploitable bug in libnv which bhyve uses

this feels like something an attacked actually used to get root on some system

i try to minimize assumptions about hackers, and simply make things as hard as i can for any attacker

especially because both vulnerabilities were fixed/announced at the same time

ivy, I realize you disagree but I think that just speaks to your good character that you are not looking for ways to compromise systems. :-)

a state-sponsored threat will get in, regardless of what i do - but if i can make it harder for anyone else, that's the goal

rwp: i find your comment patronising, that you're saying i can't recognise security issues because i'm a good person

i was not always a good person, did plenty of bad things in the past, i just don't think having random (non-setuid) binaries on the filesystem is a common exploit vector

Well I screwed it up then. Because I was trying to be nice about it.

i mean sure if you don't need to install it, don't install it, no loss

ivy: the write-up at synacktiv.com/en/publications/escaping-from-bhyve is worth reading

but can anyone name an actual exploit (chain) that was only possible because of a random binary on filesystem?

Bruce Schneier relates a story where he describes getting an ant farm for a present. What he actually got was a postcard that he could send in to get an ant farm. On the card it had a form where you fill out the address and other things. Schneier then says he immediately thought, "I could send ants to anyone!"

almost all hypervisor escapes come in the form of attacks against the peripheral devices

debdrup: interesting but i don't think this is related to the exploit i mentioned as it's mitigated by capsicum

this has historically been true of hyper-v, esxi, xen, and every other hypervisor used

the recent pair of SAs allowed escalation from bhyve guess to full host root access

ivy: oh, you're right - the new ones haven't had their write-ups published yet

do you know if such write-up is incoming? like i said, would be interested to read about it

given that they did one for one they found previously, i would expect so?

was not clear if this is something a friendly person detected or the post-mortem from a malicious attack

if North Korea developed these exploits i doubt they will publish them :-)

they're paid to do this work by the freebsd foundation

or, i guess i don't actually know that, i just assume that to be the case

sorry i refuse to load x.com links, could you c+p the post content?

it's effectively just me backing up what i said, so it adds nothing

rather, it's "receipts", as they say

it's also covered on linkedin: linkedin.com/posts/synacktiv_alpha-…d-activity-7237354529585864705-YNfP

oh interesting, this does seem to be related

oh right, the actual funding is from alpha-omega?

also it's nice to know FreeBSD Inc. is paying people to look into this stuff, i didn't realise that was happening

it's not freebsd inc

feh it is though

the foundation is a 503c3 (not a c6 like the linux foundation)

i know what a 503(c)3 is, i used to work for one

in practice, if someone wanted to fork freebsd, they would have to make it a separate project because FreeBSD Inc. is responsible for so much technical and legal stuff

503c3 is specifically for non-profit within the sphere of public interest

(for anyone else)

ivy: no, someone has to fork the project because of the trademark, which obliges the freebsd foundation to sue if someone misuses it

otherwise, they risk losing the trademark

again yes i understand that, i don't know why you're having a go at me for saying 'freebsd inc.', i think it's pretty obvious what i mean

i wasn't having a go at you, but it's demonstrably not called freebsd inc any more, even if there's people from that time involved

the freebsd foundation is an incorporated organisation, a corporation, that owns the 'FreeBSD' trademark, so it's basically in practice 'FreeBSD Inc.'

i realise it's not actually legally called that

using that logic, you could call it BSDi - which i'm sure wouldn't confuse anyone

USL isn't around to sue over 1-800-ITS-UNIX though, so it might be fine? :D

but you could not call it BSDi since FreeBSD Inc. has nothing to do with BSDi, aside from both using the old CSRG 4.4BSD-Lite source code (and i'm not even sure if BSDi used that since they had a USL source license)

well, USL still sued the regency of of the university of california despite any legal agreements, and ended up settling over having taken code from BSD into Unix and stripping copyright

kirk tells the story better than i could, though

this is a bit pedantic since FreeBSD Inc. didn't even exist back then, afaik

pedantry, on the internet? it's more likely than you think

i'm curious though, are you annoyed because i added the 'Inc.' or because i didn't call it 'FreeBSD Foundation'?

i'm not annoyed either way

i'm slightly peeved that you're treating one of the only examples of in-the-public-interest-opensource-projects like it's a corporate entity, maybe? but that's your perogative

but it's all pretty whatever, really

... it literally is a corporate entity

Article 4. Purposes and Operation. The corporation is incorporated under the Colorado Revised Nonprofit Corporation Act exclusively for charitable, scientific, literary, and educational purposes within the meaning of Section 501(c)(3) of the Internal Revenue Code of 1986, or the corresponding provisions of any future United States Internal Revenue Law (the "Code").

from freebsdfoundation.org/wp-content/up…15/12/ArticlesofIncorporation-1.pdf

it's a 'corporation' which is 'incorporated' so i don't think calling it 'Inc.' is out of line

Greetings, I am having an issue with Samba and I believe system permissions. In my Samaba config I declare that I want Samaba to write with a 775, but the actual on the filesystem is a 764. I have checked the process and it is running as root, and I think that there maybe a umask thing going on that is adjusting the permissions. I am not sure what is going on and was wondering what some options I might be

able to take.

hello

so i have a system which is not accessible from the Internet, and indeed has no Internet access at all, but it logs this: le-fay.org/tmp/30d/icmp.txt - how would you go about diagnosing this?

what ports are listening?

rtprio: le-fay.org/tmp/30d/listen.txt

maybe log RST's in pf. that sounds like a lot of work

the system does not have pf.ko loaded... i suppose i could set up tcpdump as a daemon to log that

i was hoping there might be an easier way though

is the whole network airgapped? or is it software isolated from the internet?

duskmoss: it's not airgapped but the system has no route to the internet and internet hosts could not contact its assigned IP addresses (which are IPv6 ULAs)

[15!] hemlock ~% ifconfig|grep inet|egrep -v '127.0.0.1|::1|fe80::'

inet6 fd12:8247:3:1::5/128

[16!] hemlock ~%

Hi folks! I'm a longtime Linux user. I wanted to try out FreeBSD, dual-booting. I created a ~100GB slice using the "Manual" method, chose freebsd-zfs at /, EFI at /boot/efi or somesuch, and got presented with a message that it's not going to be bootable. What gives?

cizra: hummmm... I guess its because its a partial disk, which means it would have to make it vdev I guess. Im far from an zfs expert, but I assume that might be the problem.

Does ZFS want to own the whole disk? I might need to reshuffle my existing partitions, then, somehow...

not anymore

cizra: =) I would go all in. throw out that linux system of yours, buy two more drives and make it a zraid1... Then you got yourself a stable and fail safe system

it's hard to say without understanding exactly what was telling you it might not be bootable, or what happens when you launch our loader.efi

kevans: is it even possible to boot from a zfs vdev?

cizra: no, that's never been a requirement for ZFS

cizra: the *only* difference when giving a whole disk to ZFS is it will enable the disk write cache because it knows it can handle that, and other filesystems might not

which can make a significant different to performance, but you can also enable the write cache via camcontrol etc

drobban: you boot via loader.efi or the *zfs*boot for x86 stuff, which in turn understand a limited set of zfs bits

for example

for disk in da0 da1 da2 da3 da4 da5 da6 da7; do

echo -n "$disk "

echo 'WCE: 1' | /sbin/camcontrol modepage /dev/$disk -m 0x08 -e

done

enables the write cache on all of those disks

nice

Would it help if I tried to find some more detailed logs and post them here? or do partitioning manually (I'm comfy with manually partitioning my Linux (NixOS, Arch) installs, but haven't touched ZFS. I can read mans).

cizra: are you having a specific problem? i only saw your question about whole-disk zfs

oh, i see your other question

cizra: if you're installing freebsd on disk that already have another OS, that OS should have already created an EFI partition, so you don't want the freebsd installer to make another one

you need to copy the freebsd /boot/loader.efi to your EFI service partition, then somehow convince your UEFI firmware to boot it

iirc the installer handles pre-existing ESP better these days

maybe that hasn't made it into a release yet, though

this is honestly much easier if you can install on a separate disk

I do have an EFI partition, yes, which I intended to use for booting FreeBSD. I'm comfy with efibootmgr. I guess I'll go read the install guide on how to use shell to create partitions (it's probably "mount it to /mnt or /target" or such).

circuitbone, you may want to still use freebsd's bootloader because it implements geli and zfs

none of the other bootloaders including grub can do that

cizra: if you're comfortable with UEFI, install FreeBSD as normal and just don't let it create /boot/efi as a separate partition

grub's zfs implementation always lags behind on pool features

cizra: then copy freebsd's loader.efi to EFI partition however you look and boot that and it should just work

s/you look/you like/

Thanks for all the answers, I'll make another attempt.

I have another question related to this actually... if you go with mirrored zfs root in the installer, does it install bootcode to all drives by default?

yes, which you can verify by using "gpart show"

Hm, I'm baffled by something. I can have a mode 1777 directory on a Debian NFS server (v4) and it behaves as expected on a Debian client, with random users able to create and delete their own files. The same thing appears not to want to work on a FreeBSD 13.x NFS server. I'm sure I'm missing something obvious, but does this ring a bell for anyone?

As an added twist, the user gets "permission denied" writing to the file, but the file create got through, resulting in a zero-length file.

Anyway, I need to bail out for a bit, but if anyone has any ideas I'd love to hear them. I'll do some deeper debugging when I return later.

(I'll share what I find when I figure it out, unless someone beats me to it.)

mason, When it comes to this particular issue of NFS server permissions I have a similar problem in reverse. No sticky bit here though. On a Linux kernel (Trisquel) and NFSv4 it has rebuked every effort of mine to have file ACLs of either type function in just normal directories trying to give write access to members of a group.

On FreeBSD I believe the default will be NFSv3 unless you specifically configure things otherwise.

hi there!

hi there too

jaunty, are you sure? or maybe something changed since 12.0-RELEASE. but i tried mounting one of my mirror's EFI partitions, and i couldn't because it wasn't formatted

rwp: Hrm, interesting then. I briefly tried v3 but it didn't seem to impact anything. I'll poke at it more and share what I find.

anybody have information updates on this ? wiki.freebsd.org/WiFi/80211ac

also found out through testing freebsd doesn't support connecting to 802.11r networks

had to disable fast transition

Oct 7 15:34:19 lappy wpa_supplicant[63687]: FT: Invalid key management type (2)

also found out there's undocumented switches for the rtwn driver that enables ht40 channels, which by default is locked out

no idea why ht40 channels would be locked out by default and require a reboot

sfox: emperically i disagree, freebsd can connect to our wifi network which supports .11r

i did not try actually roaming, though

ivy, can you try it and see if you have similar error messages in /var/log/messages?

sfox, One problem with HT40 from memory is that it is legal only if you live on a desert island. If you live near other WiFi devices then HT40 is required to be avoided.

In theory if the receiver hears other packets on the associated channels then it is not supposed to transmit wide bandwidth HT40 packets.

So even if enabled if the driver is performing as per the specification then it won't use it.

nobody else is using 5ghz but me

and yes, i'm in the middle of nowhere

I wish that were the case for me. Near me is an apartment building. When I run a scan I probably have close to 50 SSIDs in the list. On both 5GHz and 2.4GHz bands!

I could totally use a 120Mhz wide channel if freebsd supported it

does freebsd support ht80?

I did some experiments some pre-pandemic years ago where I forced HT40 on 2.4GHz and experimented in my underground facility (basement) and I couldn't really get a significant performance boost out of it.

there is a ht40 2.5g network here but somebody else already claimed it

Therefore I was less excited about it. And have drifted away from trying for it now. But it looked like such a tease to me. The possible theoretical bandwidth is there but in practice in a city it is not possible to attain.

unfortunately, they didn't get the memo to only use 1,6, or 11 so I doubt their actually benefitting from it

however, it seems all of the 5g bands are completely quiet

You could set up an AP on that channel and it would cut the other AP's bandwidth by your presence there.

so i'm on 36/40

I already do. I'm on channel 1

their on channel 2/40

it's stupid. it's some century link box with default settings. the driver is supposed to disable ht40 when it detects other stations

but it doesn't

I mostly use 5GHz in the house because the signal strength of other APs are most attenuated. They are still there but at a lower level. It's crowded in town here and I am in a suburbia area not really dense at all. But everything uses WiFi now.

I recall being able to use 80Mhz wide channels with a atheros chipset on linux years back

using 802.11n

I wonder if that was a nonstandard atheros feature.

Of course the Atheros chipset is the best well known and most things work there. It's the reference standard for everyone. However I am using the Thinkpad Intel chips for the most part.

anyways, waiting on any word for 802.11ac support for freebsd. supposedly 80Mhz wide channels is standard now

I'm using the rtwn driver.

i just run an ethernet cord across the living room

It was the other wifi dongle I could find that claimed 'linux' support, had a built in preamp, and had 2 mimo streams

I also have this vague memory of reading that .11ac is somehow proprietary? I am not sure of the details and maybe you know? But that contributes to the problem with it.

11ac is not proprietary

hmm

maybe you would think that because there was a proprietary extra nonstandard mode in some high end broadcom chipsets before ac was invented which added QAM1024

but now 802.11ac makes QAM1024 standard

Wired is always best. Especially for TVs. It rather annoys me in my very crowded WiFi spectrum when I see that people always have their TV on WiFi. I mean you can wire that in easily and then it will be rock solid. But phones and tablets can't do that and it's just eating up radio bandwidth when people have TVs on WiFi.

I just don't remember the details of the .11AC issue. Which was why I asked.

rwp, when freebsd implements 802.11ac you should be able to use the 6Ghz band. There's way more bandwith in wifi6e set aside explcitly for people in crowded apartments.

The new grant in 6 GHz is 1200 MHz of spectrum, compared to 83.8 MHz in the 2.4 GHz band and 570 MHz in sections of 5 GHz (US figures).

rwp: i gave up on this argument and just allocated 80MHz 5GHz channel for every AP

go on, set up your TVs, have 500Mbps, bleh

That would be great! I'll look forward to it then. However, my devices that I currently own are 2.4 and 5GHz only. I would need to wait for newer 6GHz gear for me.

ivy how did you allocate 80Mhz? I'm looking in man ifconfig and can't find reference to anything higher than 40Mhz

sfox: freebsd is not involved in this, APs run Linux (MikroTik RouterOS)

sfox: i cannot get my freebsd laptop to use anything more than the 2.4GHz 40MHz channel which is much slower

i had a really long discussion here about this the other day but even with wifibox it just won't, perhaps a hardware limitation

rwp, alfa.com.tw/products/awus036axml?variant=39754360684616

unfortunately wifibox only works with pci cards right? usb wifi is not functional

isn't the usb controller technically a pci card ?

you mean if I open this thing up there's a pci card inside?

I haven't looked but I don't think so. alfa does make a mpci varient thoough

rtprio, oh you mean pass the usb controller. I tried that but that doesn't work either.

and there's no wifibox for the rtwn driver

sfox: does the controller show up, just the wifi not work, or does the controller not connect?

sfox, You know I do have a USB Alfa adaptor. I don't recall if it supports 6GHz or not. It might. That would be one.

rtprio, the device always say 'not ready' to byhyve

Mine does not: mediatek.com/products/broadband-wifi/mt7612u

rwp even my Chromecast is wired

+1 Wired FTW!

:) I specifically ordered the wired charger for this

I know this exists but it's rarely updated.

hoping someone knew where else to look

that page hasn't been edited since 2021

I also looked in /etc/regdomain.xml and found 6ghz entries

references to ac

sfox: fwiw, if you want decent wifi performance on freebsd, i think you need wifibox

how would I get wifibox with the rtwn driver?

sfox, "that page hasn't been edited since 2021" It's a wiki and you could change that. :-)

i installed it on my 10-year-old laptop and wifi went from 20Mbps to 100Mbps, which is still terrible, but like, it's 4x faster

Do I need to build a custom wifibox?

sfox: i don't know what rtwn is but you have wifibox-rt61, wifibox-rtw88 and wifibox-rtw89, do any of those sound likely?

oh wait this is your USB card, right

yes

i have this alfa.com.tw/products/awus036ach_1?variant=40319795789896

which is Realtek™ RTL8812AU

in that case i don't know, but i'd suggest asking the wifibox maintainer

you might get a more useful answer than here

according to the man page it isn't covered by the rtw88 driver, but the rtwn driver

(bhyve doesn't support USB passthru though, so if it's not a dedicated USB controller, the answer might be 'no, sorry')

or at least ask net@ which is where the wifi people hang out