Last Feb I had this idea to experiment with freebsd in a vm. April was my first "on metal" install, so it's been roughly a year for me. Thanks to everyone in IRC who's been so helpful. The community is definitely part of the appeal. :)

Carbon666, Usually all installers would have their own boot loader. Meaning that GRUB would need to "chainload" to the next boot loader. And the second boot loader would have the kernel parameters.

But there are multiple ways to do it. Six of one and a half dozen of the other.

markmcb, i'm curious, what do you use your FreeBSD machine for? I am always pestering people about it, so no big deal if you can't share

jauntyd, i've got a 2U rack server in my garage. it's my everything server. several bhyve instances and jails running a wide variety of things sitting on 124 TB raw storage. it just feels so cohesive to me.

markmcb, Only one? It must be so lonely. :-)

rwp: two actually, but the second is only for backup. nothing exciting

rwp, I will be quitting here to try out some menu entries and will let you know if something has worked for me. Have a nice day buddy

hi, hackers! Trying to install wayfire and I got this: eson.build:36:17: ERROR: Dependency "nlohmann_json" not found, tried pkgconfig and cmake

tykling so you have one entry in authorized_keys per server who can access the backup server am i right in this assumption

aldcor: pretty self explanatory error mesage. Try installing whatever nlohmann_json is.

tykling your backup push process: requires my source server to be able to ssh backupuser@backupserver, in the guide you test and zfs allow the backup pool to a user, so im not really sure why you need the authorized_keys beeing set, unless backup server is locked down only for usage of pubkey from what i read

aldcor: try installing devel/nlohmann-json

the handbook (/usr/share/doc/handbook) isn't shipped with the OS anymore, right?

lw, It's an install option with the bsdinstaller. So I think officially that would be optionally available.

rwp: is it still? i don't see the doc set at mirrorservice.org/sites/ftp.freebsd…FreeBSD/releases/amd64/14.0-RELEASE

Title: Index of /sites/ftp.freebsd.org/pub/FreeBSD/releases/amd64/14.0-RELEASE

or does it grab it from somewhere else

12.4 still has doc.txz

I just ran through a 14.0R install yesterday and I recall seeing it there as an installer option. Which I declined as I was installing for a test and throwing it away and didn't need to move more bits around than I needed to move around.

hmm curious

i wish it was easier to just throw an installer ISO at bhyve and say "boot this please", maybe i should check out markjdb's bricoler

i'm pretty good at vm-bhyve now but stock bhyve is really not super user friendly with all the 'stock' options you need to pass

qemu isn't great here either but it is a bit easier to just boot a new VM from a disk image / cdrom

isn't it weird that /etc/rc.d/swap runs before /etc/rc.d/savecore?

i wonder if this is what's causing my crashing system to not create dumps from savecore

hm, i guess this is documented in fstab(5) which tells you to use the 'late' option on swap devices

apparently the specific problem here is that as soon as you add a swap device, if it's on an SSD, swapon immediately sends a TRIM to erase the entire storage

so it erases the dump even if nothing has swapped out yet

Oh, you have the trimonce option?

ah, no, i don't. so maybe it's not that

if using a dedicated dump device doesn't fix it, i'll go run poudriere on the system console and see what the problem is (poudriere seems to reliably trigger the crash)

hi folks :) I'm having some trouble getting the sfp+ nics working, wasa hoping to get some help/pointers, although it is opnsense, I do not seem to be able to get the sfp+ to recognize the plugged in cables

collected some info here forum.opnsense.org/index.php?topic=40087.msg196570#msg196570

Title: Odido ISP Netherlands Connection Guide

am new new to *bsd world, familiar with linux though

funkymon1ey: did you remove the SFP+ from the ONT and put it in your freebsd machine? in most cases, this won't work, as PON needs encryption configured on the fibre

i'm not sure what you mean by 'modem' in that post as FTTP/PON shouldn't involved a modem, just the ONT and a router

lw no, atm the pon is not going into the freebsd machine, just trying to get sfp+ ports fowrking

directly pc to pc lan

so the ISP isn't involved at all, you just connected two computers together? via DAC?

so I got my internet speed upgraded from 1 to 8gb and eventually want to have th epon connected to the baremetal but I noticed i get no carrier on the sfp+ ports soo thought let's first try to get something simple working pc to pc lan

tbh it's hard to understand your actual configuration / problem from your forum post, could you describe exactly what you're connected/configured, and what doesn't work?

main issue is what ever i do, ifconfig shows "status: no carrier"

which i think it means it does not see the plugged in cables?

does it show no carrier on both sides of the link?

no, only on bsd

link comes up on Linux side?

on linux it shows link deteced

that's strange

yes

did you try both SFP cages on BSD side?

yes

same result

you might try 'ifconfig ax0 media 10Gbase-Twinax' although i will be surprised if that makes a difference

that returns: ifconfig: SIOCSIFMEDIA (media): Device not configured

funkymon1ey: what about sysctl -w dev.ax.0.link_workaround=1 ?

that returns: dev.ax.0.link_workaround: 1 -> 1

does the link come up? if not, try setting it to 0 then back to 1

does not come up, tried setting to 0 then back to 1, same result

ok, i'm out of ideas then, perhaps this is a driver bug... do you have anything else you can try in the same port, like a pair of fibre SFPs?

funkymon1ey: you may try booting FreeBSD 14.0 install media (you don't have to install it) and see if the link comes up there, just to rule out opnsense-specific issue. if it still doesn't work, you could ask rajesh1.kumar⊙ac who is listed as driver maintainer in axp(4)

lw: many thanks in any case for your efforts, atm no other cables here to try

thank you also for that :)

will try with live medium

it is strange that link comes up on the linux side, this is what makes me think it is a driver bug, for some reason freebsd doesn't notice the chip has link up state

funkymon1ey: just ooi did you try running tcpdump on ax0 to see if it's actually passing traffic even if ifconfig says no carrier?

(that shouldn't be possible, but...)

lw no i didn't try that yet, but just did, here the output pastebin.com/atpnfNU4

Title: tcpdump --interface ax0tcpdump: verbose output suppressed, use -v or -vv for f - Pastebin.com

there seems to be some stuff coming in?

funkymon1ey: yes - try running ping from the linux side and see if the incoming icmp shows up in tcpdump?

i tried that is in the output on pastebin

though linux shows 100% packet loss and destination host unreachable

lw

ok, so if 10.133.120.3 is the linux side, there is traffic arriving there from that IP address, but no icmp pings

lw, I'm dealing with a CMake project that is trying to do things with something called "python site package path". what is that and how do I configure/set that on FreeBSD?

that suggests issue might be on the linux side, maybe wrong routing table or firewall? does tcpdump on linux show the icmp packets exiting right interface?

jbo: uhh i'm not sure, i've never dealt with python in cmake. what exactly is it doing?

ah maybe i configured linux not correctly, will try tcpdump on linux

jbo: python site package path is usually /usr/local/lib/python3.XX/site-packages/

lw, gitlab.com/kicad/code/kicad/-/blob/…/CMakeLists.txt?ref_type=heads#L920

Title: CMakeLists.txt · master · KiCad / KiCad Source Code / kicad · GitLab

lw, gitlab.com/kicad/code/kicad/-/blob/…ndwxPython.cmake?ref_type=heads#L34

Title: cmake/FindwxPython.cmake · master · KiCad / KiCad Source Code / kicad · GitLab

jbo: so it looks like it's trying to autodetect this from python binary, does it get the wrong value?

lw, I'm not too sure. I'm hacking around things. Right now it's failing at FindwxPython.cmake line 66

lw thx there is def. some funky shizzle going on here on the linux side

Unknown wxPython/Phoenix version string:

jbo: i am not the person to ask about this, sorry :-(

will figure that one out and see if that resolves the issue, many thanks :)

lw, alright, thanks :)

jbo: you have wxpython installed right?

lw, jup. py311-wxPython42-4.2.0_1

funkymon1ey: you should at least file a bug at bugs.freebsd.org if you can reproduce this with upstream freebsd since if the link is working it should not say no carrier

funkymon1ey: correct output is something like media: Ethernet autoselect (10Gbase-Twinax <full-duplex,rxpause,txpause>)

(that's from ix, but ax should work the same way)

jbo: i suggest some printf debugging in the cmakefiles: cmake.org/cmake/help/latest/command/message.html -- see what the variables are getting set to

Title: message — CMake 3.29.2 Documentation

wow, des@ is pretty... let's say "short"

body shaming alert :o

lw, is there a difference in python stuff from FreeBSD ports to Linux distors? i.e. is it possible that it's looking for 'python' but my executable is called 'python3.11' instead?

jbo: you should have 'python', if not, install lang/python

uskerine, I am curious if your firefox simply consumed all available memory. It does that here sometimes.

jbo: also install lang/python3 to get 'python3'

lw, hmm yeah I only have /usr/local/bin/python3.11 nothing else

well, python3.11-config but that's it

different systems handle how to get python{,3} symlinks different ways but freebsd way is to install lang/python, lang/python3

my repo doesn't even have that

you must build it!

I have lang/python311

it won't take long, it's literally just a couple of symlinks and a dependency on lang/python311

lw, alright, thank you!

good info

how come every other time I have a shit day it's something python

because [censored] [censored] freebsd [censored] python [censored]

gesundheit

Creating repository in /tmp/packages: 5%

yay

Creating repository in /tmp/packages: 71%

q__q

i've always been dubious about that /tmp/packages thing, i wonder how securely it creates that directory

wondering if I can see which bluetooths are available and pair one (I am trying to connect my bluetooth headsets to my Jabra Link 380 -usb bluetooth-)

rwp> yes I am lately having issues with firefox being slow

I see too many processes around and a lot of memory taken

jbo: did that fix it or are you still creating repository

uskerine, If you are not using uBlock Origin then I would recommend installing it as that will reduce memory requirements and provide for better security too.

rwp> what is ublock origin

lw, do I want to install lang/python or just lang/pyton3 ?

Title: uBlock Origin – Get this Extension for 🦊 Firefox (en-US)

jbo: both

jbo: lang/python bestows /usr/local/bin/python, lang/python3 bestows /usr/local/bin/python3

uskerine ^^ what lw posted, +1

lw, now it works <3

thanks!

jbo: awesome, i'll send the invoice later

I see it

regarding the pairing USB dongle, I guess that is not easy, right?

[140Ramd64-main] Fetching texlive-texmf-20230313.pkg: 23% 281 MiB 9.1MB/s 01:54 ETA

the gift that keeps on giving

i wish -b latest worked with pkgbase

jbo: why do you only get 9.1MB/s, does your fancy new motherboard only have a 100Mbps port?

ask upstream dist infra

wait what

why do all my speedtests say 100mbps

lw, did you not provide a git-format patch intentionally?

jbo: where?

Title: 276996 – [NEW PORT] databases/postgres_exporter: PostgreSQL metric exporter for Prometheus

i probably forgot, you can just commit it as you, i don't mind

unless you want me to make a new one

nah, I can set the author field manually

keep in mind that this needs to go through review first - on that now

i will not keep that in mind

how care you tell me what i should keep in my mind

s/care/dare

lw*

thanks! can i have the motherboard too?

so needy

lw, should the rc script maybe match the package name w.r.t. prometheus?

jbo: i prefer it matches the binary name, i think this is more correct

lw, reasonable argument

for some reason sublime-music has stopped playing music...

ValueError: ('Invalid value for mpv parameter', -4, (<MpvHandle object at 0x3cac77a99a0>, <mpv.LP_MpvNode object at 0x3cac77abc80>, <mpv.LP_MpvNode object at 0x3cac77abbf0>))

why can't there be a music player that just works and isn't awful :-/

what's wrong with cmus?

flatdog: i don't remember but something about how it does ncurses stuff made it unusable for me

like it didn't highlight stuff properly or something

musikcube has a nicer interface than cmus but just randomly doesn't play some tracks and development seems dead

using the web interface of navidrome for now...

I see. Well, cmus works well for me, I must be one of the lucky ones :)

*no pun

although musikcube doesn't support multi-disc albums properly either which is annoying

jbo: how do i run poudriere with rctl to stop it munching all cpus?

lw, AFAIK a reasonable approach is to run poudriere itself in a jail, then racct on that jail

you'll need to allow the poudriere jail to spawn sub-jails and ZFS access tho

that sounds like a nightmare with getting the zfs into the jail

i never got that to work

I know one of my mentors does that - I haven't bothered to set that up myself. I just upgraded hw

yeah yeah yeah

duckduckgoing for "poudriere rctl" is not really returning anything useful

i wish we had something like Solaris projects, that would make this easier

maybe i can do something similar with login classes

(solaris had a really neat thing where you could do fair-share cpu timeslice allocation between projects... used to use that on a multi-user hosting system)

that sounds nice

it was... i'd really like to bring a few Solaris features over to FreeBSD (contracts!) but it's a lot of work

i think we could do contracts via 'null jails'

I'm looking forward to it :)

don't expect anything too soon, i'm fixing pkgbase first, that's a big job

if we had contracts though we could do something like systemd except not terrible (i.e., Solaris SMF)

lw: there are no null jails… i wish we could port contracts over

meena: i know, i just mean if we did have null jails that seems like a good way to do it

is solaris officially dead or... ?

in solaris 'zones' and 'contracts' were separate things but i don't think they need to be

jbo: no, it's still around

why talking about it in past tense then?

opensolaris is dead

lw: i think it's nice that they are orthogonal

jbo: well, solaris and opensolaris are pretty dead

meena: hmm

i think these days, it's primarily Illumos that lives on

well, solaris maybe isn't dead, just dead to me lol

jbo: it's in the past tense for me because new Oracle licensing means i have no interest in using it

... which is basically why i'm in #freebsd now

if oracle hadn't bought sun i'd probably be happy using opensolaris

hrm… i think i'd still be with FreeBSD, unless opensolaris was easier to install lol

meena: what did you think of my modset idea

i'm not conviced devmatch solves this, despite what imp says

not all modules are device drivers

lw, does std::filesystem::is_regular_file imply std::filesystem::exists ?

jbo: i have literally no idea

i don't really see how it couldn't though

can't be a file if it doesn't exist, right

yeah, that would be my line of thinking too, but I don't know all the possible edge cases

lw: i'm still mulling over it. I'm probably kinda stumped with Apache httpd's all/few/most/none/reallyall

meena: that doesn't really sound like the good solution

who would want 'few' modules instead of 'all'? :-)

devices should work with devmatch, and that should be improved; but I think modulesets should be grouped on purpose, and those might be overlapping

i'm thinking of stuff like if_wg can never be loaded with devmatch because it's not a device

i guess rc.d maybe autoloads it in that particular case

yupp

all if_ stuff is autoloaded, IIRC, tho sometimes you need to preload them, or you're ending up with wonky behaviour

wonder if efirt and aesni gets autoloaded

if_bridge needs to be preloaded, if you want the bridge to copy the device mac, which is influenced by a sysctl

we do need to do something for block drivers like ahci, and filesystems... filesystems should be fairly easy but i'm not sure how to handle storage drivers

i think a good start would be a standard modset including ahci, nvme, nda, etc

then we could switch to MINIMAL as the default kernel (if the filesystem thing is fixed)

could even get this in for 15.0 we if hurry

2 years?

do you think that's too long or not long enough?

i anticipate a lot of bikeshedding

yeah, right, so maybe that's hopefully long enough

we should probably rename it to MODULAR first though

maybe i'll ask arch@ about that, my last few posts went fairly well (got my mailwrapper PR committed)

yeah, that was somewhat surprising

i was not too surprised as my PR was clearly correct :-d but yes, we might have expected more people to object

i might ask imp before asking arch, i know he had some opinions on MINIMAL/MODULAR/whatever

i feel bad about asking him things since he commits all my PRs but this seems like an unrelated thing

I like MODULAR.

it's a much better than name than MINIMAL, which has very different implications

i agree, i remember there was some confusiong about your VIMAGE change because of that

confusion

but, yeah, i think we need an arm64 version first, before we can proceed making it the default

otoh: mv sys/amd64/conf/MINIMAL sys/amd64/conf/GENERIC; git commit -am "amd64: fix kernel configs"

meena: my aim is to be able to boot either amd64 or arm64 without needing anything in loader.conf

which is quite a bit of work, i think

i think i understand the reasoning, but can you still explain why?

i mean the loader.conf part

well imagine i'm a new user, i install freebsd on my desktop, it doesn't boot because nvme_load=YES or ahci_load=YES is missing from loader.conf

we can't make MINIMAL default without fixing this

right now, if i setup a machine with ZFS, my loader.conf looks like this: gist.github.com/igalic/c755cdb472f1858fccb3348713d6864f and i didn't do anything for that

Title: gist:c755cdb472f1858fccb3348713d6864f · GitHub

meena: how do you load the driver for your disks?

but: If MINIMAL, or, MODULAR being default, means that it is also the default kernel on an installer ISO, and that means those need to boot, too

right, so this has to be done in loader(8) somehow

not just in devmatch/devd

lw: this is a virtual machine, and the virtio stuff is all built in

ah yeah, but it shouldn't be. and it's not for physical hosts

eggsagtly

(i removed virtio* from my local kernel config)

so, we need to fix quite a bit

i think all the info we need is in loader.hints, right? we can probe pci bus and load all relevant modules

this is more complex though, i thought modsets might be a useful stopgap approach

that's what devmatch does

right, but devmatch doesn't run at boot loader time, it can't load boot device modules

yeah, i'm a bit slow here

i did 1.5h on my bike today, after getting the tyres fixed

*after fixing the tyres

it's a bit more complicated too for e.g. Xen which doesn't use PCI

what does it use?

'xenbus'

xn0: <Virtual Network Interface> at device/vif/0 on xenbusb_front0

xbd1: 1536000MB <Virtual Block Device> at device/vbd/51728 on xenbusb_front0

i have zero idea how this works, i just load the modules in loader.conf

or actually i think i compiled them into the kernel here, 'options XEN' includes all the drivers

(which is also wrong but ...)

lw: EHeM in #bsdmips knows Xen, so you can ask them

i'm ignoring xen for now, i want to make this work on at least amd64/arm64 physical hardware and kvm

unless anyone has a better idea i'm going to hack up a PoC for my loader(8) modsets proposal

that would at least let us make MINIMAL/MODULAR the default kernel today

if we find a way for loader to use PNP hints we can start moving things out of modsets

i need to let that modsets idea simmer for a bit more, or you write up an email to arch@ where you detail your current thoughts and goals, because, as-is, it's too vague for me to not just make up some nonsense

anyway, time to bring doggo out and head to bed

yeah, i think it's easier to demonstrate in code though

nn

O_o

why write words when you can write C

i'm hoping it's C and not Forth

forth is dead, long live Lua

but if Lua doesn't have the bindings yet, then C it is.

maybe i will mail arch@ first though, at least just to find out if anyone else is working on this

welcome front kananter

lw: see here, hackmd.io/@jhb/ByWrxQmr2

Title: FreeBSD 15.0 Planning - HackMD

meena: what am i looking for here?

lw: "Boot loader support for devmatch"

ah

(i think)

that's useful but i don't think it fixes this problem as not everything is in devmatch

no, but it's a start

yeah, but i'm worried we'll get tied for 10 years with "we can't switch to MINIMAL because not everything is in devmatch yet"

we've been there already for 10 years

also when i had this discussion with imp@ earlier he seemed to be under the impression that we don't need devmatch in loader since it's in userland...

meena: right, i can't fix the past but i'd like to fix the next 10 years :-)

design rolls

'netlink for ZFS' what

hmm, i wonder how much i trust zfs raidz device removal in -current

or did that not get committed in the end? can't find it in the manpage now

ah, it was zpool attach for raidz that got merged