I've been using OpenBSD for a while now and I am considering running FreeBSD on one of my machines but I have one thing I want to clarify first. I am aware the FreeBSD does include proprietary binary driver blobs but I have found that one can compile their kernel use WITHOUT_SOURCELESS in src.conf. Would this paramater remove ALL the proprietary blobs to the same extent as OpenBSD or are there other things I need to do aswell? Thanks

so much.

Bumblehorse: is it true openbsd automatically downloads non-free software if the hardware requires i?

the option seems fairly explicit; it excludes all drivers that contain either binary native blobs or binary microcode

johnjaye: yes

johnjaye, as far as I am aware, it does download non-free firmware, not device drivers though.

i'm not savvy enough to appreciate the distinction

WITHOUT_SOURCELESS_HOST is the option you want if you're ok with binary device firmware but not binary host code

I think the option only impacts modules, not stuff directly compiled into the kernel; you might also need to check that you don't statically configure any unwanted drivers

/

oops

seems there are not many binary host drivers; hpt*, ce, cp, are all I can find

hpt is a highpoint raid driver, ce and cp look like synchronous serial cards

disabling binary firmware takes out a bunch of wifi and network cards

there's also the GPU firmware modules, but those are not part of the base system

The machine is a thinkpad X200 that I put a certain Atheros card in already; I think it should be fine.

Is there anyway I can use these parameters during my initial installation?

The installation section of the handbook doesn't seem to say much about that.

not if you're installing from the ordinary distribution

which doesn't recompile anything

if you are absolutely determined, what you could do is compile everything yourself in some vm, and then install from that

So aslong as I reconfigre/install the custom kernel, delete the unwanted drivers (if any were present),

oops

... then I shouldn't have anything to worry about with kernel modules, right? Only 'statically' configured drivers as you say.

l/win bal

at least one of the hpt* drivers is in the generic kernel, for what little that means

if you're building your own kernel you can strip almost all the hardware drivers out, and load only the ones you need as modules

I ran zpool upgrade after an update and got the message, "Pool 'zroot' has the bootfs property set, you might need to update the boot code." What's the right way to update my boot code these days?

what partitioning scheme?

Title: # gpart show ada0 ada1=> 40 3907029088 ada0 GPT (1.8T) 4 - Pastebin.com

and from what you said before, this is legacy bios and not uefi?

yes, `sysctl machdep.bootmethod` says BIOS

for legacy bios, gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 2 ada0 (the 2 is the partition index of freebsd-boot)

and same for ada1

thanks again

ada1 is also failing smart tests so I wanna get this right, heh

for EFI, which you should probably do to keep it consistent, mount ada0p1 as an msdosfs partition and copy /boot/loader.efi to efi/freebsd/loader.efi and efi/boot/bootx64.efi

and same for ada1p1

your efi partition is much too small, did the installer do that?

probably, but I don't recall for sure

it's a few years old

Ok, I think I'm getting it now. So aslong as I unload, delete and prevent compilation with WITHOUT_SOURCELESS, I won't have to worry about non-free MODULES. Then, as long as I configure a custom kernel using the info in conf(5) and comment out the unwanted drivers, then I won't have to worry about drivers being compiled directly IN the kernel and these two combined should cover everything. Am I confused or is this correct?

And thanks a ton by the way!

think I will retire this machine soon and exit OVH for good.

as far as I know you're correct

Awesome, I hope you don't mind one more question. Earlier you were listing off example non-free drivers that fall under the categories of WITHOUT_SOURCELSS_HOST/UCODE. Where did you find this information? Was it on the web somewhere or did you look directly in your source tree?

I looked at sys/modules/Makefile

it doesn't seem to be a common question, I don't recall anyone asking it before and I've been hanging around here for several years

you mean mine?

yes

Well to be honest I came from a fully libre GNU/Linux distro on a librebooted thinkpad. I always had interest in BSD systems and then one day my distro removed package managers like pip from their repositories because one could use them to install non-free software. I found this to be overkill and switched to OpenBSD right away. Recently I've been wanting to try FreeBSD for some extra software availability though.

removing a package manager seems like a strange thing to do

very very strange. I understand they want to be ideologically consistent and make sure the user doesn't accidentally install non-free software, but anyone who installs and uses an OS like Parabola obviously doesn't need hand holding to the extent of removing a the main package manager of one of the most popular programming languages...

you might aswell disable networking because one can use it to install non-free software

I think that's about the point where your OS converts from a non-toy OS to a pretend playground

and mounting any devices that could have non-free software written to them

exactly

I was a little hesitant to switch but as soon as I understood that firmware doesn't run the in the kernel and exists in ROM already, I had not issues moving to OpenBSD. I don't understand why the distinction between the firmware and drivers isn't meaningfull to the FSF crowd.

I think I do understand it, though I don't make an issue of it myself

nVidia sucks with their firmware blobs.

It's sad to see because when I was more into lower-level gfx coding in the 90s & 2000s, nVidia drivers were a boon to the consumer market.

3dfx was really pushing Glide and hap-hazardly provided a crappy OpenGL "driver" (which wasn't even a full driver, just enough implemented to let specific games look good).

OpenGL used GetProcAddress to load the parts of the API that it could find and then fallback to the software implementation for what it couldn't load.

Then nVidia started shipping more closed firmware blobs. Really frustrating.

"Most notably, binary blobs are very uncommon for non-wireless network interface controllers, which can almost always be configured via standard utilities (like ifconfig) out of the box; Theo de Raadt of OpenBSD attributes this to the work done by a single FreeBSD developer."

What and to whom is he referring?

haha what

i might not take too much stock in that

Looking for a working command line example for ffmpeg invoking quicksync hwaccel. Supposedly i have all the drivers firing for it on my system. /dev/dri and /dev/drm are populated.. but im Having a hell of a time with figuring out the frmpeg side of it.

Has anyone successfully built security/openssh-portable with blacklistd enabled?

It fails for me without adding this patch, even after the recent update to 9.3p1 -- bugs.freebsd.org/bugzilla/show_bug.cgi?id=269623

Title: 269623 – security/openssh-portable blacklistd extra-patch broken with 9.2p1

Hello, Thinkpad p14s gen1/intel is not listed in the FreeBSD wiki laptop page and I wrote this howto about making FreeBSD run on the Thinkpad p14s Gen1/intel.

Title: mco-system/freebsd-lenovo-p14s: Tips, tricks and configuration for running a Lenovo P14s (Gen1 / Intel) with FreeBSD - freebsd-lenovo-p14s - Codeberg.org

So how this can be added to the wiki page ?

As I'm a new FreeBSD user, comments and error correction would also be appreciated :)

doug_713705, I think you can make an account on the wiki and propose the change by editing wiki page, but I haven't done anything with wiki in quite a while so I might be wrong

meka: The Wiki about page states that that account must be asked to the Wikiadmin on IRC

I stand corrected :o)

oh, asked on the #freebsd-wiki chan

Good luck!

You make it sounds like an impossible mission :D

Nah, just ... everything I did for the first time was a disaster, so I wish people more luck than I had :o)

Hi! I have a freebsd 13.1 server with igb0 and igb1 connected to my switch in a lagg. I have a bunch of non-vnet jails with their addresses on igb0, but now I'm experimenting with turning them into vnet jails

I'm using the standard jail/jail.conf stuff. My problem is that my vnet jail cannot access the network. It can ping itself, but not the defaultrouter etc

I suspect that I neet do bridge lagg0 and the vnet ethernet device? (I haven't)

clbin.com/ICd0Z here the host net config

jail net config: clbin.com/jD3p9

and the jail config: clbin.com/nEupv

did you run `jib addm www lagg0`? not sure if that's quite right, as I let bastille do the heavy lifting for my jails.

jschmidt: I am wondering if I perhaps should add jailether to lagg0?

jailether is a 3rd interface?

yes, it is a third physical interface (that is not connected to the switch with any cable)

BTW, `jib` is located here: /usr/share/examples/jails/jib

you're end goal is to get the converted VNET jail to connect through lagg0, though, correct?

yes

I moved jib to /usr/local/bin

or well, copied

but yes, the end goal is to have the nvet jail connect through lagg0

ok, sounds like you're on the right track. you'll need something like exec.prestart+="jib addm www lagg0"; to connect the epair and physical nic, and exec.poststop+="jib destroy www"; to tear it down when you stop the jail.

so, I should scrap using the third interface at all, and just jib addm to lagg0 instead?

(I got the idea about using a third interface from mwl's freebsd jail book :)

I'm looking at page 161 right now. :^)

jschmidt: heheh! See how he suggests using a seperate interface which he calls jailether?

yeah, that's an option if you want. I do something similar in the homelab w/ an LACP lagg and have all my jails use it.

if you want to manage the host out-of-band, I'd use the dedicated jail interface.

in the book, he doesn't do anything special to "connect" the jailether interface and the "regular" external interface

it all seems to be taken care of by using defaultrouter="..." on the jail host?

I'm on my kindle, so I'm not 100% sure the page numbers match what's in the dead tree book. see the loghost jail.conf on page ~160/161? that has the addition of the VNET specific additions you need.

Yep, I see it (I also have the kindle version). But I think I'm using that config more or less, using a seperate interface for the vnet jails

hrmmm

you'll need to plug that 3rd nic in, then. is it up? I see you set it to up in rc.conf, but that won't apply w/out a reboot. `ifconfig jailether`

clbin.com/hBjjU jailether

clbin.com/4dv7J jailetherbridge

can lagg0 contain interfaces that aren't plugged in?

yes

that looks right. if you plug jailether in to your network, www should be on the network.

oh, wait... if you're asking if jailether can be part of lagg0, then yes, BUT... www isn't going to magically jump onto lagg0 - it's still tied to just the one (disconnected) nic.

and, if you want to use DHCP for the jails, don't forget to add 'add path 'bpf*' unhide' to a devfs ruleset in /etc/devfs.rules and call that ruleset out in jail.conf

(I usually forget that part when setting up a new jail host)

So, I just read through freebsdfoundation.org/resource/usin…eebsd-as-a-virtual-host-quick-guide and I was wondering if the base vbox in Ports has the option to record VM audio and video, or if that's a feature of the extension pack.

Title: Using FreeBSD as a Virtual Host – Quick Guide | FreeBSD Foundation

Is there a simple way to sync users and groups between two jails? Or am I going to have to figure out NIS or LDAP or something?

do you mean that users newly-created in one jail need to show up in the other?

yes

I can't think of any simple way to handle that.

I was trying to puzzle out some shared null mount and symlink-ing the files in /etc to it, but... passwd is still gonna regenerate stuff on update in local /etc

haroldp: Maybe just script up a sync on creating the new jail. Or do you need users unique to new jails to propagate backwards?

nah, it will be one-way

If just coping existing accounts into new jails... Yeah, I'd just seed the new jail with data from the old. Anything more complex, a directory seems reasonable.

syncing on creating the jail would be no problem, but users get added and removed regularly\

may be nullfs mount of specific files from the host somewhere?

can only do directories, not individual files, unless something changed

checking on current, it can do both

and it's also in 13.2

huh, do you have a keyword i can use to google that?

Title: mount_nullfs

I'd bet that that interferes with renaming the files, which I'm pretty sure pw relies on

ah heh

"The mount_nullfs utility supports mounting both directories and single files."

hot damn

woah

Ok, who wrote this: it serves as a demonstration of layering by providing a layer which does nothing. (It actually does everything the loopback file system does, which is slightly more than noth- ing.)

at least it no longer says BEWARE OF DOG.

haroldp: "BSD 4.4 Lite sbin Sources" ;)

RhodiumToad: i think that was about unionfs?

pretty sure nullfs used to say it too.

that's like calling up air traffic control on the radio and getting a surfer voice. the random tone is disconcerting.

Title: Snippet | IRCCloud

I'm trying to get the basename of the directory that the Makefile is in, as a variable to use elsewhere

finding info on POSIX make seems to be sparse

uh, posix make has no $(shell) feature

and the WAT:= line evidently doesn't get executed

its a gnu-pollution then?

ok I can drop that and just use pwd -P

and it all works

ffs

use pwd for what? posix make lacks that too

make is posix, but im running this on bsd ofc + linux + mac

they all have `pwd -P`

Could you not switch to GNU Make?

parv: no

it's almost always a mistake to try and do anything nontrivial in posix make

well the whole makefile is ~ 20 lines

usual boring stuff, clean, dist

you know that even := is not in posix make?

sigh

no

only = is supported for setting variables

TIL

Use separate makefile🤷🏽‍♂️

Using GNU make might be the most reasonable thing. Remember that FreeBSD wouldn't exist without the GNU tools being there early on.

As a workaround, you can pass the current directory as a parameter to make. For example, you can run make CURRENT_DIR=$(pwd), and in your Makefile, you can refer to $(CURRENT_DIR). This way, you're using shell features outside the Makefile, and the Makefile itself remains POSIX compliant.

or you can re-invoke the make from within itself

*shrugs*

Cool idea, RhodiumToad

jfc I thought this would be a trivial fix

I'll just hardcode the name and if anybody renames the repo good for them to figure this out

hah

hi all

this is making me feel less salty about coding /bin/sh

;-)

there are worse things :-)

Just make sure to avoid bashisms! =snicker=

I mean, I have to code some "classic" ASP when I am done with this, so there are for sure worse things. :)

I have seen shell scripts with "#!/bin/sh" on CentOS|Rocky Linux that use(d) "[[ ]]" & other Bash specific features. Frankly Bash behaviour as (plainer) bourne shell is not strict enough

I try to teach people if you need bash do "#!/usr/bin/env bash"

man env

I have a friend who's working on a visual basic 16 bit app atm

the business plan is to migrate it to perl

a modern language for modern times

I wrote my first and last csh script for zsh, and still get yelled at for creating a csh script

RhodiumToad++

I love perl, and I know it much better than a lot of langs, but I think most people think of it as a dead boomer language ;-) Python, doood!

python is horrible

Agreed... Only because of the whitespace issue

How so, RhodiumToad?

I prefer C style syntax and whitespace agnosticism.

Perl, Python, Ruby are similar enough🤷🏽‍♂️ (not counting issues with respective package|module management). Now Tcl ...

tcl is nice but flawed

hahah tcl... May only experience with that is eggdrop related.

ok, so /etc/sysctl.conf, is it key="value" or key=value?

I don't quote my sysctls.

I only wuote if it has spaces, and not sure if that is required or not

quote even

"It's all strings"

Demosthenex: quotes are not required in general

key = "value" or key="value"

All these personal questions...

i'm looking at all the man pages, and there's no formal spec. it's very loose

whitespace around the = is disregarded

I enter mine all in unicode.

the only spec is probably src/sbin/sysctl/sysctl.c:parse()

loader.conf says key="VALUE" explicitly

rc.conf says "sh syntax"

loader.conf isn't necessarily the same

rc.conf is literally read in by doing ". rc.conf" in shell

And mine there are quoted. Dunno if required.

use the source

what can i run to confirm the rc.conf, sysctl.conf, loader.conf are valid and don't contains typos or invalid values?

using only sysrc to update rc.conf can help ensure it stays valid, though it won't detect typos

Reboot and watch the console ;-)

sysrc does validation for adding and removing things, but if you've hand-edited the file you're assumed to be smart enough to spot mistakes or figure out how to correct them I think (meaning there's no tool to do a check; though I think making one might be a good idea)

it's almost impossible to detect typos in names in those files, since in many cases the names are not from a fixed list

something small in c or lua that can be added to base seems like an idea, if someone wants to run with it

and sysrc doesn't work on sysctl or loader.conf

Would be a nice tool to have.... Rebooting and checking dmesg or watching it is the only sure fire way right now that I know of. If you don't do this then you sometimes pay for it later when a reboot occurs by choice or otherwise.

right, because one thing that sysrc checks is that the name is a valid shell variable name

Erhard: Yeah, it's just rather difficult to get right

I think I have only had a problem once or twice in 35 years or so, but it is a problem when it happens.

Typo something, or misunderstand something, etc.

ct(1) - configuration tester; test configuration files.

Free realestate!

Demosthenex: in shell variables you cannot use key = value

Would be a good tool that would likely be adopted scross OSes.

`service foo enable` which is a wrapper around sysrc can at least help with enable/disable

across even

beastie: yep, spaces break it

you must stick the equal sign to both the variable name and the value.

Oddly, I prefer to edit the file myself rather than use sysrc. But I guess that is because I am old.

It'd have to be able to handle UCL, shell variables, and probably at least a handful of other things..

Which basically means.. Good luck! :D

There may be a reason it doesn't exist ;-)

it's required to distinguish a command name and argument list from a list of environment variables specified before the command name.

Maybe the tool could instead fire up a vm for tetsing or something.

Though I am not sure that would help with loader.conf

e.g: CFLAGS="-g -O0" make a.out

for a lot of shell, yes.

I hate that too... because whitespace should not matter.

dch: I'm much happier coding perl than VB. Much happier, heh

would be nice if augeas could do ucl

haroldp: I think everybody would be

wow, perl.... ufff

$_ baby!

have you tried python?

meena0: in the interests of scope creep and yak shaving, I think you should add it

I'm definitely a functional programming zealot

but I quite like lua

dch: yeah, i should

anyone know offhand if there's a simple incantation to disable aesni in a kernel that has it compiled in?

if we're standardising on UCL, standard configuration management tools should be able to process it

Maybe sysctl hw.crypto.disable_aesni=1

?

UCL?

Erhard: unknown oid, at least in 13.x

ah

Demosthenex, Configuration in Lua

almost,

bah, took me 3 goes to get the device hint right in loader.conf

hint.aesni.0.disabled="1" works

Cool

RhodiumToad: yeah, bizarrely a boolean value needs to be a string.

That always gets me.

first time I typoed "aesni", second time I had "disable" instead of "disabled"

Wait, you're right, it doesn't. I've got hw.nvme.use_nvd=0 without any quotation marks.

When did that change?

RhodiumToad: you were correct that null_mount-ing the various passwd files wouldn't work. I presume due to due to copying, as you suggested

what packages do i need if i want to compile arm firmware with clang on my freebsd? does the preinstalled version already support that? if so, where do i find the header files defining uc pins and stuff?

phryk: in theory, base clang can build arm, or rather, all supported architectures (with possible exception for the ones that still need gcc)

you can also use qemu-user-static(-devel)

meena0: okay, where do i find the header files that define the cpus supported for cross-compilation?

been find | grep'ing through /usr/include but i'm not really finding anything that looks like the right thing.

right. hmmm

cross-building stuff isn't installed anywhere by default

the files used for cross-building kernels can be found in, e.g., src/sys/{arm,arm64}/include

I don't want to build freebsd. i want to build raw firmware for ucs.

well, that may require tools not in base

RhodiumToad: not seeing any llvm/clang arm packages tho. does that mean i have to use arm-gcc?

the llvm packages include all supported targets

but again, that won't necessarily include things like header files with details of specific arm hardware

some of that sort of thing is in the kernel header files for arm, which I pointed out before

I've tried switching from mysql57-server to mariadb1011-server. However, running service mysql-server start does nothing (as does running mysqld_safe &). I must be missing something.

did you enable the service in rc.conf?

do you have any logs?

Yes. mariadb uses the same items as mysql, i.e. mysql_enable="YES".

Nothing in the logs.

same here, not using mysql since years

in vt can I change the resolution itself, or just the font?

postgres has too much overhead for my uses.

try running with --syslog

mysqld_safe that is

it could be permission issue. do you use the same user on your /var/db or /var/run directories now as for the mysqld?

johnjaye: which backend is vt using?

Or check mysql/error.log wherever that is on your system

er... i'm not sure. i just thought vt was the driver itself

A permission issue is a good idea, but mariadb uses the same user/group as mysql

i'm trying various .fnt files and the ones in vidfont, but they're either too small or too big for my screen

can you truss the mysqld_safe process?

johnjaye: vt can be backended by e.g. vga (for legacy boot), efifb (for EFI boot), drm (when a GPU driver is loaded), etc.

johnjaye: are you booted with no GPU driver loaded?

i don't know if any of the 12 things kldstat listed is gpu but i'm not in x11 if that's what you mean

ok. grep '^VT' /var/run/dmesg.boot # what does this say?

it says VT(efifb): resolution 800x600

i guess changing resolution to smaller is what i want here. then the default font would be the right size

ok. so you are booting with EFI. you can either change the resolution in EFI, or you can load a KMS driver and have vt set the mode

kms driver would be what?

i know the radeon driver is what i need for this laptop to run X

if you load the radeon driver, then vt will set modes according to the setting of kern.vt.fb.default_mode or kern.vt.fb.modes.OUTPUTNAME

and yes, it's the same driver that you'd need for X

mysqld_safe --log-error+/tmp/mariadb.err doesn't save anything to the error file

ok. how do i tell efi then

that may be more platform independent

if you don't want to do that, then efi_max_resolution in loader.conf can configure a maximum requested resolution

or you may be able to set things like that in the bios config

ooh. didn't know about efi_max_resolution yet. any other fancy features i can use to make efi boot more beautiful? can i pick a custom background? :3

although I get " mysqld_safe Logging to '/tmp/mariadb.err'." and "mysqld_safe Starting mariadbd daemon with databases from /var/db/mysql" and nothing beyond that.

phryk: there are loads of things you can do in loader.conf

plus you can supply your own lua code to make it as fancy as you please

is that new? o_O

ah wait, that's the other loader.

the lua version of loader has been the default since 12.x I think?

yeah

well, that's new for me. i think the last new setup i did when 11.0 was fresh. :P

in 14.0 you'll be able to be even spiffier with a loader.conf.lua rather than having to hook into config loading in local.lua

bah

btw kevans, can anything be done about D24324 ?

I've had it in my own ports tree for three years, but it's been sitting in phab for two

let me poke some people

I don't like to have my ports tree diverge too much from upstream

RhodiumToad: is efi_max_resolution="640x480" the right syntax in loader.conf?

i tried a few different ways but it stubbornly stays on 800x600

yeah, understandable

I believe so. kevans would probably know better than I

oh wait

try "480p"

yeah, it accepts WxH or, IIRC, some common ones line "1080p"

ok. brb

the default should be 1x1

i.e. we won't touch the resolution at all, whatever the firmware left us at

usually the lowest res gop mode on the hardware I was working on at the time

it's worth noting that efi_max_resolution can only result in changes to modes the firmware makes available to us (as seen by executing `gop list` at the loader prompt)

no luck. i'll have to try the radeon driver. i don't recall how i determined which one I needed though

what are the way to free up some space

pkg reports at least a dozen different radeon drivers

i ran pkg clean

some ideas ?

i remove /usr/ports

removed

johnjaye: there are only two you need to worry about: radeonkms (old radeon cards) and amdgpu (newer cards)

johnjaye: the package will figure out what to load.

eoli3n: what's using the space, and do you need it?

my backups, and yes

johnjaye: use pciconf -lv to identify the GPU

It looks like a "Warning" was really an error "secure-file-priv" in my.cnf caused the problem. On to the next issue.

johnjaye: for the actual package, what you want to install is drm-kmod

ok. what do i do after that

did you identify your GPU?

yes

what is it?

mullins r4/r5 graphics

pci ids?

i also have drm-kmod installed

is that the vgapci@0:0:1:0 part?

no

there should be a vendor and a chip id

vendor x1002, device 9851, that?

{0x1002, 0x9851, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP},

pciconf -lv | nc termbin.com 9999

pciconf just has numbers

right, what I gave is the line from the driver that matches it

have you had X working on freebsd on this hardware before?

if you do not have the correct drivers installed does it not simply fall back to vesa

i think so yes

but after i reinstalled i lost everything

daemon: it won't let me ch ange the resolution from 800x600

so yes it fell back to vesa

so should i add kldlist="amdgpu" to /boot/loader.conf

Title: Graphics - FreeBSD Wiki

johnjaye: kld_list="..." and it's either amdgpu or radeonkms

but I would stick with RhodiumToad all my bsd boxes have been headless since it was xfree86

i don't need x. this is a different issue

I haven't used any recent amd/radeon gpus, I had an old radeonkms one but am now on nvidia

looks like amdgpu is the one for this.

is the hack disabling console output still needed for amdgpu on EFI?

ah, seems that's fixed now

hmm. i added it to loader.conf but it doesn't load automatically

then i hard freeze when i do kld_load amdgpu

drm drivers are blacklisted in loader.conf

oh ok

er kldload

hm. kldload should work.

should kld_list="amdgpu" go into my rc.conf instead?

hold off on that until you can make it work with kldload

i already did

it still goes to efi when i boot

oh

did i do it wrong

no

add the kld_list thing to rc.conf and see if that works

it should change resolutions during the boot process

then you can add kern.vt.fb.default_mode="..." to sysctl.conf or loader.conf

ah it hard froze the machine

i had to get on a different pc to type heh

well we gave it a good try. i gtg. thanks for helping me out

Back to mysql57. Mariadb doesn't like the json datatype in two tables for nextcloud. I'll have to do a bit more research before trying this migration again.

If only mysql80 wasn't so darned slow. I tried it, and ended up going back to mysql57 because of all the wasted time waiting for queries.

I am trying to build a Perl module from perl --MCPAN -e shell

and the untarring fails with this error: Cannot restore extended attributes: com.apple.metadata:_kMDItemUserTags com.apple.metadata:_kMDItemUserTags

I tried untarring the file manually as well and even added --no-mac-metadata but the error persists

This is on 13.2

that should have worked

were you doing it as root?

also try --no-xattrs

yes, as root

no-xattrs worked, weird

I am trying to virtualise linux on bhyve and can't get networking working in the vm installer. I am following the handbook. Stuck :/ thankyou

what have you done exactly?

Title: Chapter 23. Virtualization | FreeBSD Documentation Portal

no, what exactly did YOU do

i created the bridge0 and tap0 interfaces

loaded the iso into bhyve, everything working as expected

although I was confused whether it needs to be ifconfig bridg0 addm em0 or wlan0

to the best of my knowledge, you can't bridge over wireless

on the command $(ifconfig bridge0 addm 'igb0' addm tap0)

WELL, this is all so I can use the 802.11ac drivers on Linux

ah

so you want to passthrough the wifi adapter as a pci device?

eventually

ok. so for setting this up, you'll need a wire connection

very new to BSD, thought getting a linux vm running in bhyve would be a good first step

i can do that

once you have that working, you can start thinking about wireless

will need to circle back to freebsd/freebsd-src #745 once i know what i'm doing

Title: tap(4): allow full-duplex and non-zero speed by igalic · Pull Request #745 · freebsd/freebsd-src · GitHub

I messed with tinycore linux but couldnt get it to boot

so, the bridge0 interface needs to be configured with your wire nic first, then the tap device

using void now

how far do you get with the linux install?

i am up to 'Networking' in the installer

its in ncurses

ok, and?

sorry i am trying to set up ethernet >_<

on my host

ok i am on ethernet

RhodiumToad: how should my bridge/tap look before I start the vm again?

bridge0 should be up; it should have the physical NIC as the first member and tap0 as the second

the physical nic should be up

I've always had better luck with configuring the host's IP on the bridge not on the physical NIC, but people have told me that's not supposed to be necessary

this is only temporary

i will have to change the network setting to pipe connection out of the guest

good evening my fellow freebsd family hope all is well with everyone

RhodiumToad: tap0 is my first memeber, and em0 my second

hello bsdbandit :)

that might or might not work

can I just use DHCP in the guest?

yes

if you have a dhcp server

yes you can

and I am launching bhyve with the example provided in the handbook. Is there anything I should check there>

what exactly is not working?

RhodiumToad: in the void installer, there is a stage where you establish networking, It gives a vague error messagee

I am trying it again over ethernet bridging

"Network is inaccessible, please set it up properly"

ok. you checked that the bridge shows as "up" ?

bridge0: flags=8843<UP, <-- see the UP there

also tap0 should show as up: tap0: flags=8943<UP, <-- there

does tap0 show as "status: active"

yes

tap is up

bridge is not

ifconfig bridge0 up

its up

still not working in guest, should I reboot the vm?

shouldn't need to, but you might as well

wait

it works!

thankyou so much

yw

its "hanging on unhandled ps2 mouse command"

on the reboot

that's harmless

it's hanging somewhere _after_ that

i wish i had a void install for reference

it might just be the huge linux initram

loads every compatible driver at boot so sometimes very slow

RhodiumToad: I can currently see the output of Linux's runit bootloader

which has start a getty terminal somewhere?

is there a way I can access the other serial terminals

what bhyve command-line did you use exactly?

also, did you try just hitting return? the usual setup is to use bhyve's stdio as the console serial port

bhyve -A -H -P -s 0:0,hostbridge -s 1:0,lpc -s 2:0,virtio-net,tap0 -s 3:0,virtio-blk,./linux.img -l com1,stdio -c 1 -m 1024M linuxguest

ok, so bhyve's stdio is presented to the guest as com1

you haven't configured a framebuffer or any other serial ports so there's no other terminal

do I need to change anything in the guest?

I don't know, I don't use linux

could I open a serial port from another terminal?