is there some specific driver i need for the intel I211 ethernet controller

i tried installing freebsd earlier (on a dual boot) and freebsd could only find "igb0" and "wlan0"

both of which did not work

but i have freebsd in a vm right now and the network works fine on there (although i guess its because the host machine has a working network)

thegman: is the intel card not igb0 ?

i thought it might be but i couldnt get it to work

thegman, freebsd.org/releases/14.0R/hardware says Intel I211 is supported by "em" driver

Title: FreeBSD 14.0 Hardware Notes | The FreeBSD Project

huh, the man page says that I211 should be the `em` driver, so em0

i could never get em0 to show up

this realtek RTL8125 showed up as none0

is there anything in /var/log/messages about em0 not being able to attach?

i would have to reinstall freebsd

why?

it is cathartic

i lost all modivation and uninstalled it earlier

so il have to reinstall it later

ok then

daemon: indeed it is

same as any os when the crap builds up to much

now that i think of it i definitely shoulda saved some logs

daemon: thats why ive had to reinstall linux infinity times

use freebsd and zfs to run bhyve then use gentoo as the guest :D

days of fun

the main reason i wanna use freebsd at least on a dual boot is because of that fantastic zfs support

i would like to do that

well do it?

but theres a specific program that i cant get to work on freebsd

use freebsd as the hypervisor and linux as the guest

rdp or vnc into it

problem solved

im not sure if the program would work in a vm

id have to check

not a bad idea actually

it will

i sat down for about 4 hours straight trying to think of a workaround and that never came to mind

freebsd does way better as a server than as a desktop

you can even bhyve a freebsd guest to get that 'freebsd desktop' without the punishment :)

I will give you a tip though, using bhyve to run gentoo requires logging into the vm via virtual rs-232

which is fun

but it absolutely is possible

remember the virtio drivers

and the alamo

never forget that alamo

how well does bhyve gpu passthrough work with an amd cpu and amd gpu

using FreeBSD as my primary desktop for three years now and haven't looked back one.

once*

jbo, what DE are you using?

does bhyve have a UI and can you do stuff like add mouse or hard disks to it?

if so then that would be a huge reason to just use it to run whatever other os you need

does it have virtio support?

Title: bhyve(8)

Looks like it does

hmm, ok. i'm running freebsd in libvirt right now and i wasn't sure how to get virtio to work right.

I can't say im fully familiar with livirt

RoyalYork__, I am not using any DE at all. I use x11-wm/bspwm as a window manager and that's it. no DE.

lw, still up?

johnjaye, libvirt works

ah ok. that's reassuring. i wasn't able to find the right documentation though

the best i could get was a fedora manual about the virt-manager GUI

johnjaye, you can use deskutils/virt-manager

johnjaye, docs: libvirt.org/drvbhyve.html

Title: libvirt: Bhyve driver

what does the bhyve driver do exactly. i run virt-manager and it invokes bhyve for me?

libvirt is mostly an API, the libvirt bhyve driver is what interfaces with bhyve under the hood so any libvirt compatible tool works with bhyve (such as virt-manager)

i see

other than that there are tools like sysutils/cbsd, sysutils/vm-bhyve and others

personally, I'm not using libvirt because both privately and professional stuff is all bhyve only so there aren't any real benefit over interfacing bhyve directly

what do you mean professional stuff

Well, this is going to be fun: news.ycombinator.com/item?id=39438908

Title: WebKit switching to Skia for 2d graphics rendering | Hacker News

as in, people using freebsd use bhyve? or am i missing something

johnjaye, stuff I do at $dayjob

oh i see

i thought you meant, in general bhyve is commonly used over libvirt

I have a patch that builds skia on FreeBSD. Currently working on patching skia-rust so that I can use NeoVide on FreeBSD. If my patch helps fix probably soon-to-be-broken ports and/or introduce new ones, then happy to oblige.

libvirt is "just an API". it doesn't do any virtualization. it's just trying to define a common interface that then works with "all" hypervisors

and with the normal issues that come with solutions like that

right. if you use the tools in enterprise you have a good understanding.

I don't :)

whereas I feel like, i'm just reading internet descriptions and manuals. so it can be misleading sometimes.

johnjaye, what are you looking for in general? just something to have easy-to-handle VMs?

like i didn't even understand that cloud-init is something that runs inside a vm image for awhile

I assure you there is a myriad of things I don't even understand

basically just to use vms well so i could maybe run freebsd on the metal with linux vm or something

example of something i didn't understand: apparently I can ssh into my libvirt instance

johnjaye, have a look at sysutils/vm-bhyve. it works well.

Title: GitHub - churchers/vm-bhyve: Shell based, minimal dependency bhyve manager

i had 0 clue this was possible, I couldn't find it in any documentation, I just randomly tried it and it worked. that's my level of understanding of VMs.

ok

well the VM is an M - but with a V. so networking stuff still applies, including SSH :)

yes. linux puts all the complicated stuff like that under something called a 'bridge' called virbr0

same here

everybody is cooking with water

mostly i just would like to understand how the hardware emulation works and how to do basic things. ssh in. vnc in. setup storage and configure network.

well, it has an ip, it has an sshd, so why not :-)

i guess it's a matter of understand the networking in freebsd as well, then combining them

how the hardware emulation works? maybe just skip to basic things

oh no

oh yeah

in the porter's handbook example 1 where it says to create a diff for a new port... would that diff simply be all additions?

i think it's saying there's nothing then you submit a .diff file which creates all the files.

johnjaye: yep

i see

you making a port?

i am reading the manual on how a port would be made hypothetically.

it also said to check the list of open ports to use bugzilla i think. and i don't have an account yet

you shouldn't need one to search

oh i see

i wouldn't have guessed you'd search the bugzilla

unless it's like. this port needs a maintainer

you can search as a guest without an account

but you won't be able to participate in discussions, making new reports, or patch updates

what does 277086 mean? it says user Cy reported bug, then user Cy assigns the bug to himself, then says it will be assigned once "the problem" is figured out

yes most of these i see are about ports and packages

johnjaye: cy@ submitted a fix, then realised the fix didn't work, so he reassigned the ticket to himself while he works out a proper fix, then he'll assign it back to jgh@ (the maintainer) to commit the fix

jgh: depends, are you jgh⊙fo? :)

I doubt it :)

Hey guys, a question if you would want to share files in an local network, NFS or Samba using FreeBSD as a Server?

saohh: i would normally use NFS for Unix clients, SMB for Windows clients. you can run both on the same filesystems

perhaps its worth mentioning that network will be used by not just windows devices but, linux and BSD too

if you macOS clients, it's kind of a tossup, the Mac NFS and SMB clients are both a bit broken in my experience

oh thats an interesting aproach i had thought about that too

so i could share the same directory with NFS and Samba?

if you only want a single protocol, Windows does include an NFS client but it's a bit more effort to set up (and might not be some editions? not sure)

i had thought just go all out Unix (NFS) and dont use Samba at all

yes, you can

ik i have it installed on my windows machine xD

also if you have FreeBSD clients, you have to use NFS, since FreeBSD's SMB client is old and useless. i think there's a FUSE SMB client in ports, but NFS is going to work better

oh ok

i thought of the same as you mentioned, some things might be shared with visitors you know and tinkering with everyones windows machine isnt my thing (some might even not want it)

so having Samba additionally came to my mind as a solution for that situation

you could run NFS for your own systems and have a read-only public SMB share for guests

that was the idea yeah ;)

i tried to get the ZFS with NFS to work but my windows didnt recognize it

said the share list was empty not sure if i had the "exports" correct

not sure about zfs sharenfs, i just use /etc/exports

but i asume right the zfs dataset (filesystem not device) is a sharable (NFS) folder right?

e.g. i have the pool named data which exists in /data and a dataset called "almanach" so i could share \\ServerIP\data\almanach no?

yes, but if your shared dataset has child datasets, make sure to share those explicitly as well, and depending on the client you may need to mount them by hand (this should not be required with NFSv4 though)

i just made one dataset thats it

i read that about the subfolders you have to share them too, otherwise they wont show up

the information on the web is overwhelming do you know of any guide that explains NFS v4 with a windows machine nicely (in one place)?

i feel i get pieces everywhere but propably not all working correctly when put together

i don't know about Windows specifically, does the NFS client even support NFSv4?

ah yes it does learn.microsoft.com/en-us/windows-server/storage/nfs/nfs-overview

Title: Network File System overview | Microsoft Learn

saohh: for NFSv4, you want something like this in the server's /etc/exports: le-fay.org/tmp/30d/BICb0j.txt (if you're using zfs sharenfs, you still need the 'V4:' line), then you should just be able to mount server.name:/public on the client. (it's not /data/public because /data is the NFSv4 root already)

ah wait... that page says Windows only supports NFSv4 server, not client

which would mean i can only use NFS3 with windows hmm?

i mean what the heck windows can share with v4 but not access share with the same version oO propably to make people prefer smb?

should amdgpu load on a WITNESS+INVARIANTS kernel? for me it locks up the system as soon as it loads during boot... will do some more debugging later but i'm wondering if this is supposed to work

Man page links on man.freebsd.org are a bit broken at the moment. Links to pages that have '_' in name are incorrect as only the last part after the '_' is interpreted as a link. (example man.freebsd.org/cgi/man.cgi?query=netgraph&sektion=4#end)

Title: netgraph(4)

is there a way to load a specific kernel and its modules at the loader prompt? or does it that automatically when you load a new kernel?

Hi all, it seems I cannot login to my local file server at home anymore via root, never changed its password, weird. Rebooted into single-user mode and the shell prompts for "root password or ^D Multi-user mode"

I did not really, really set insecure tty at all.. Anyone would give a clue what's going on, and how to reset root password in that case? (ZFS)

tercaL: did you edit /etc/ttys to make the console insecure? that will cause single user mode to prompt for a password

ah

lw: Not really that I remember *feels too old

So "Enter root password, or ^D to go multi-user" is the indication that the console insecure is set?

tercaL: usually, yes. try setting "kern.init_path=/bin/sh" at the loader prompt before booting, although i'm not sure if this can be set from loader

lw: How? Tried to set it after choosing "Escape to loader prompt", it said unknown command.

Would a bootable FreeBSD 14 memstick help in that case?

tercaL: sorry, you need 'set': set kern.init_path=/bin/sh

yes, if you can boot from install media you can import the pool and check/edit /etc/ttys

(or mount the root filesystem if UFS)

lw: Thanks for that, it's ZFS and pool name is default, as zroot. It has a single SATA disk. So how could I proceed after booting within memstick?

tercaL: "zpool import zroot" should find the pool. you'll probably need to mount the correct root fs by hand, e.g. "mount -t zfs zroot/ROOT/default -rw /mnt", then check /mnt/etc/ttys

you may need to use import -f if zfs thinks the pool is already in use

sorry, correct mount command: mount -rw -t zfs zroot/ROOT/default /mnt

lw: I think I'm done with it, the "console" line was marked as "insecure" in ttys file. Changed it to secure and before I reboot, just to make things safe, tried umount /mnt and it said "Device is busy"

at this point, is it safe to reboot, or any suggestion you might have to unmount the ZFS system correctly before rebooting?

make sure you didn't cd into /mnt which will keep it busy, otherwise, it doesn't really matter as long as you do a clean reboot (not power cycle)

tercaL: actually for zfs, unmounting the filesystem doesn't do anything by itself anyway, you may want to export the pool though

not even call sync?

lw: booted into single user mode now, passwd root gives

passwd: pam_chauthok(): Error in service module

tercaL: check /etc/pam.d/*, you may have obsolete pam_opie listed, or some other bad module

lw: single-user mode is read-only, right? That might be the issue?

As the filesystem is read-only, I tried zfs mount -a said; "failed to lock" /etc/zfs/exports.lock: Read-only system

mount -rw -t zfs zroot/ROOT/default /

then zfs mount -a

lw: After both commands, it worked, thanks so much!

unable to start my freebsd installer from USB. for some reason it get stuck on the following row.. cpu0: <ACPI CPU> on acpi0

havent had any problems earlier. I managed to screw my zfspool over after doing an zpool upgrade and lost my boot info.

trying to save it, by starting up an freebsd installer - but now giving me this... I dont understand what has happened.

drobban: first, you have two problems: 1. zfs access; 2. live boot; I'd take the 2nd then move to the first; try downloading the last FreeBSD live CD and DVD - try them both; when you uncompress, make sure there is no error; also make sure there are no errors when you write the images on flash; then, see if you can boot - I hope you use the right arch for download, right?

drobban: after you boot the OS, try importing read only; if not, try reverting the Zpool using the previous transactions... post here how it goes

Hello there

Is there a way to get vim freebsd port with clipboard enabled ?

the stock port doesn't seems to have the option to enable it

clapont: yea.. no errors with the image..

and I would love to stay on ports rather then compile vim by myself

clapont: still problems... just to get something going I even tried to boot the pfSense installer.. the exact same problem. halts at cpu0

drobban, you may also have issues with EFI video driver

in this case the boot process may continue nut the image on the screen will freeze solid

drobban: to reduce the odds, try to use a core FreeBSD image, not other distributions based on FreeBSD

this is so weird... It has worked like a charm until now,,, until i fucked up my boot settings on the pool.

clapont: yea. using the the freebsd 14.0 amd64 image.

same problem. no difference.

Ha ... :) If you have beadm enable welcome to rootf mount issues from image :D

is there a way that i can tell the boot loader to use my disk to boot from instead

try to disable efi boot

if you dd image to a usb flash - it is there

drobban: your boot is on zfs... and you said that you tried to upgrade the zpool; the /boot was clearly affected so try one step at a time: 1. boot with live; 2. try to access/debug the Zpool

also disable a secure boot in the bios

nerozero: why would that magicly start causing problems from thin air?

nerozero: install the x11 flavour of vim

in case you cant boot it - secure boot may be the issue

drobban: you said you tried to uprade the Zpool; that has affected everything, the boot included

nimaje, Vim has flavors ???WOW

thanks !

zpool upgrade is what i did to be precise

clapont:

freshports.org/editors/vim indeed

Title: FreshPorts -- editors/vim: Improved version of the vi editor (console flavor)

* vim-x11: Console Vim only, with xclipboard support

nimaje, THANKS A LOT !

drobban: I did understand that; be patient and take steps one at a time if you wish to solve

not sure why that isn't the default, no idea why wouldn't want to have clipboard support

drobban, first you should check, is a boot partition / efi partition has support to a zfs version you running

clapont: well.. the image of the installer isnt the problem..

what can cause the installer to hang on cpu0: <acpi cpu> on acpi0

drobban: so you did boot? you are live? do you see the disk?

drobban, As I told you - efi bootloader issues with intel graphic cards

the modern onces

nerozero: using nvidia. Hasnt been a problem before

so try not to boot in efi mode

nerozero: I am in efi-mode

anyway try that, I had this issue quite a lot

once system completes boot - the image magically appears

but sometimes it requires additional drivers

with that frozen screen I where able to connect to the problematic pc by SSH

clapont: im able to see the disks with lsdev in the bootloader. im also able to see the pool.

with fully loaded system

in the bootmanager

or what to call it

nimaje, maybe for headless builds, with no x11 ?

clipboard do require x11 ... yes

drobban: I suggested to use a _live_ CD/DVD not an installer; if you are with install, try booting with ACPI disabled - although this problem should been present at the original install time

+1 clapont

drobban: you succeeded to import the zpool in read-only mode?

and you can try to import your zfs pool

nerozero: but that could be a nox flavour

exactly ...

nerozero: I'd prefer "clapont++" ;-)

:)

clapont++ :D

clapont += 1

pythonish way :D

which reminds me python's spacing problem when coding... I prefer C although python/sh have their advantages sometimes

yah Never understood why did they do indent sensitive language

clapont: ooh, was unaware of freebsd having usb-live images. where do you find them?

wait a minute

just do a dd or better ddrewcue to see the progress

yea..... well have that one...

ddrescue -F /path/to/image /dev/daX

dudes

Made a rookie mistake. I uninstalled unbound and deleted /var/unbound instead of /usr/local/etc/unbound - How bad did I miss up? :-O

I preffer bind and kea,

I cant recall if /var/unbound is important or not to the base system

delete package and reinstall it - it should recreate default directories

RoyalYork: it's for the local-unbound resolver. should be created automatically by the rc.d script if you use it, i think

and/or follow the log message, something like $ tail -f /var/log/messages

clapont: yea... im using the memstick image. even tried to use 13.3 instead of 14.

still the same problem.. I cant understand what has happened and why it halts on the line saying cpu0: .......

drobban, is it notebook or desktop ?

nerozero: .....

im using a ryzen9 7900X. 64gb ram. asrock b650.

I have recently joined and didn't follow the prev. conversation, sorry

the system has been rolling with 13.0 and upgraded to 14.3.

disable efi bot first from bios

I just fucked up the boot. just trying to recue it... i havent made any other changes to system. just a reboot. for the first time in 30days

hah ...

your ZFS version mey nnot been supported by current boot loader

so 1 you need fresh boot bsd

nerozero: .....

to address the freezing image issue - again try to disable efi boot, in this case at least you will see what is happening

hell not even sure im able to boot in legacy mode

My Intel I9 machine with Nvidia card never shows a boot screen at all, only bios and sh login ... ;

ooff.... looks like i missed up with removing unbound. i run it now and i get: could not write builtin anchor, to file /var/unbound/root.key: Permission denied

btw if you make bsd14 - this issue might be fixed

so try first

maybe change the user

i'll try that

to unbound:unbound

Ok, that did it. I chowned /var/unbound to unbound:unbound and it works

nerozero: uhh, the bootloader shouldnt break with an older zfs feature set

pheeww

for some reason the download speed for freebsd image is limited to 300k/s to me

miko, I had that issue 2 times

while upgrading in bsd 12

well... it didnt break for me.

that really doesn't sound right

out of curiosity, who is the user:group owner of your /var/unbound directory?

is it unbound:unbound or unbound:wheel

RoyalYork: le-fay.org/tmp/30d/LgIocj.txt

lw, merci!

find with xargs ... brutal :D

find -exec ls -ld {} \;

nerozero: xargs is quicker to type than the awkward -exec syntax

it shorter :D

find | xargs has more ambiguous syntax though, where as find -exec is pretty straight forward.

Plenty of rope, either way.

don't people have better things to do than criticise the specific shell command i used while trying to help someone?

you know, its like a vim

nix hass 10000000 way to o the same thing, and that it is why I love it

lw, welcome to unix.

CrtxReavr: nothing to do with unix, more like welcome to irc

lw yep, no ugly animated emojies ...

IRC <3

I dunno. . . I've spent a lot of time discussing unix with people both in persona and via plenty of other platforms.

good luck guys !

need some sleep

tried to remove all extra hardware.... still same problem - hangs on cpu0

okey. im able to boot clonezilla... any hints what to try to disable just to be able to start the bsd-installer.... Im phreaking clueless of why this problem has started to present itself.

tried to boot installer in verbose mode.. last line shown now is pcib0.

and some complaining about ppc0. cannot reserv and failed to probe

o_O

Oh, so you were booting clonezilla just as a test?

CrtxReavr: yea. just to establish baseline.

You can't boot with a FreeBSD 14.3 CD/usb ?

14.3?

12:15 < drobban> the system has been rolling with 13.0 and upgraded to 14.3.

I guess 14 isn't up to .3 yet.

14.3 gibt es nicht

although, it would be fun to edit the kernel version to show random versions like 16.2 to confuse people on irc

CrtxReavr: aaa... 14.0 and updated .

No wonder drobban can't boot - he's trying to boot an OS from the future.

:P

Try holding down shift-control-quantum while you boot.

o.0

funny guy.

Im about to snap on this freaking machine.

So what happens when you boot with a FreeBSD 14 image?

drobban: you said it booted before you did the zpool upgrade, right? if so, all you need to do is replace \EFI\BOOT\BOOTX64.EFI in the EFI partition. could you remove the disk and connect it to another system to do that?

it halts... last message printed is cpu0: <ACPI CPU> on acpi0

when running it with verbose.. it halts on another message

lw: that is actually something i could do.

it's just a FAT filesystem, so a Windows or Linux system would do fine

I have my ampere altra machine spinning that I could use.

i mean, it's odd that the 14.0 installer doesn't boot if you have 14.0 installed. but at least this would get it working again

Im close freaking throw this machine to hell and replace it with a arm64 system instead.

lw: yea exactly. that could be a solution.

Chance are, but the best set of tools to fix a no-boot FreeBSD issue will be. . . FreeBSD.

CrtxReavr: yea. that was my plan. just to spin up a bsd-image and start a shell.

but you are totaly right. I could actually just move the disks to my other machine and try to fix them on that one.

lw: can you go into more detail about that? why would you assign a bug fix to someone else?

wouldn't it just be something like, the maintainer assigns the bug to themselves if that's it?

johnjaye: jgh@ maintains the port. when cy@ opened the bug, Bugzilla automatically assigned it to jgh@ because, as the maintainer, he's the person who will need to deal with the bug. but cy@ realised the patch was broken and should not be committed, so the bug is now waiting for cy@ to submit a corrected patch. therefore, the bug is assigned to cy@, since cy@ is the person working on it right now.

it would not make sense to assign it to jgh@ as jgh cannot do anything to progress the bug, only cy can do that, by providing a working patch

that's strange because it sounds like jgh has no role or input on the process

is this how bugzilla works normally

which part seems strange?

well as an analogy say i run a lemonade stand and someone submits a bug my lemonade is too salty. then realizes their salt patch is wrong then reassigns the bug to themselves. then reassigns the bug back to me.

and the entire time i'm completely ignorant of what's happening. that's my mental model

johnjaye: ok, if we're doing analogies... let's say i think your lemonade recipe is bad. i submit a bug with a new recipe, and assign it to you, so you can start using the new recipe and fix your lemonade. but before you can do that, i realise my new recipe is wrong. i don't close the bug, because the root issue (bad lemonade) is still present, but assigning the bug to me indicates that i am currently working on it.

ok. so the "ownership" is more about the person who opened the bug then.

so it would be like having a bulletin board on the street about local businesses and you can take those notes and put them on the business door or back on the board or take them home.

bugs don't have "ownership", only assignees. the assignee being the person who is currently expected to look at the bug and do something with it

can anybody change the assignee of any bug? i'm confused

no, only committers (or other people with special bugzilla privileges) can reassign bugs

oh. is jgh not a committer?

jgh is a committer

ok. so they can both reassign the bug then. is that right?

yes, cy and jgh are both committers, so they could both reassign the bug if they wanted to

im so stupid..... now I have to pools with conflicting names

oh ok. that makes sense

i actually don't think that specific policy is very sensible, but it's how freebsd does things, so -shrug-

and the general expectation is the person currently with assignee status is the one with the obligation to update the status or reassign to someone else

yes, if a bug is assigned to you, it means you need to do something about it. either fix it, or close it (if it's not a bug anymore) or assign it to someone else if you no longer want to be responsible for it

i see.

in the latter case there are various meta-users like fs⊙fo and ports⊙fo for bugs that no specific person is working on. if you filed a bug against an unmaintained port, it would be assigned to ports@

what happens if someone maintains port X and everytime anybody posts a bug about it they resassign the bug to themselves and immediately close it and never give a reason?

i.e. they just want to sit on the port and never let it change

i don't really understand the question? if someone kept filing bugs and immediately closing them i imagine they would be banned from bugzilla for spam

oh, you mean if the maintainer did that

yes because you said people with commit access can reassign each other's bugs.

well, if a maintainer is closing valid bugs, you could mail ports@ and ask someone to remove them as maintainer

but that would be a very strange situation, people don't become committers (or maintainers) if they do things like that, usually

oh so ports@ is an actual account, i thought you meant it was like, a token for the bug system

ports@ is both a bugzilla account and a mailing list

if you subscribe to the mailing list, you would get notifications for bugs assigned to ports@

More like a list alias, but. . . sure.

i'm just trying to think of edge cases to understand the system a bit better.

as you say it is the way freebsd "does things"

like if the maintainer changed the software to put their name in a big flashing banner everytime you open the program that would obviously be rejected by someone further up the chain.

that would be the job of the committer

the maintainer can be different from the committer?

in this case the maintainer (jgh) is also a committer, but that isn't always the case. i maintain ports but i can't commit them, i have to open a bug and wait for a committer to pick it up and do that

oh ok.

although what i actually do most of the time is nag jbo to do it on irc

so what if jgh redefined port X to have his name in big flashing letters in the UI? Does that just get committed to the next STABLE release?

what's up?

ports aren't in releases, they're entirely separate from the base system. i suppose if he did that and someone noticed, he might lose his commit privileges

oh right. you have to manually git clone the ports tree and build it.

jbo: oh nothing, sorry for ping. just explaining to somehow how ports bugzilla works

what needs committing now? I am still digesting the +1k commits that just happened

lw, nah, all good mate :)

yeah i'm just asking questions about how ports maintenance works

i'm reading the porter's manual atm

johnjaye: becoming a committer is a very long-winded process that requires you firstly do a lot of work, then go through a mentorship period, then eventually you might get the privilege of committing by yourself. no one would go through all that just to make a joke commit that would be immediately reverted and then lose their commit privilege

and being a ports committer is the opposite of fun tbh

ah i see.

like, if i somehow become a committed and pushed a commit that deleted the entire src repository... it would be at most a minor inconvenience because someone would immediately undo it

Hm, guessing the Firefox thing has yet to be fixed...? "Crash Annotation GraphicsCriticalError: |[0][GFX1-]: vaapitest: ERROR (t=0.745326) [GFX1-]: vaapitest: ERROR"

right

it's not like random people off the internet get commit bits :-)

it's like that scene in the sopranos with the starbucks.

ah, I might have misunderstood. becoming an src committer is much easier than becoming a ports committer

jbo: well i used src as an example but this applies to both

i think the original question was about ports

lw, except that the consequence on mishaps in the ports tree are much more grave

jbo: only because ports process is broken :-)

do many people even use ports? most of what you'd want is in the base system anyway

(this is a joke don't @ me)

says the guy swimming in python crap

ah :D

by the way since we're talking about ports. how are the files in the Mk/ folder used? i don't see where they are included in the ports.

unless bsd.ports.mk in the makefle is somehow referring to the one in ../Mk ?

johnjaye, there's also #freebsd-ports if you have more specific questions

johnjaye: every port's Makefile will have .include <bsd.port.mk> at the end (or sometimes a more complicated things that includes several mk files)

why doesn't that mean the one in /usr/share/mk though

johnjaye, stuff in Mk/ is mostly re-usable infrastructure that the real port makefiles can use to make things easier to maintain

johnjaye: it does, but /usr/share/mk/bsd.port.mk is just a stub that includes the one from ports/Mk/

ahhhh you're right.

i see the trick now.

Becoming any kind of a committer is not exactly easy, because not only do you need to put in enough work to get someone to punish you with a commit bit because there's too much work for them to handle in committing it all, you also need to be part of the community to the point that the person looking to mentor you knows you; it's both a technical but also a social challenge.

indeed

also a medical challenge: each time I am about to push commits my heart rate & blood pressure go through the roof

Oh, that doesn't stop even once you become a committer.

ok. i just read lucas books and manuals so the social dimensions of the various terms is usually not apparent.

ports really need a CI run, something more useful than manual exp runs

debdrup, yep - I am 6 months in and it has indeed not changed :D

like, you submit a commit, it builds your ports and all dependent ports on all supported releases/T1 archs, then tells you it's ok to commit

from what I can tell, ports also has a lot more politics happening than src. Not sure how the situation is over at docs

jbo: ah you saw that thread too :-D

lw: it's a nice thought, but it's non-trivial if you're asking to build almost 40k ports for a change.

lw, that CI suggestion is difficult

debdrup: i very much doubt most commits touch 40k ports. but if it does, wouldn't you do an exp-run anyway?

which is basically the same thing

It's simple to describe, and anything but easy to implement.

+1

lw: I assume you realize you're not the first person to make the suggestion, including people on the CI team.

debdrup: does that mean no one else is allowed to suggest it?

i'm sure i'm not the first person who works on ports and heard of CI :-)

on top of the overall challenges, you'd also need an extremely large resource pool to get any benefit out of this. that is why exp-runs are done manually (or at least part of the reason)

having a CI tell you after 4 days that something didn't work and you got 400 commits in the meantime won't help much

lw, not saying this is what you suggest (because it's not) but in case you didn't know: portsfallout.com

Title: Dashboard - FreeBSD pkg-fallout

> You are receiving this mail as a port that you maintain is failing to build on the FreeBSD package build server.

yep, that e-mail always makes for a good day when you get up in the morning.

lw: not at all, just that a lot more thought has been put into it than "it'd be nice if we had this"

i don't get it. how is that different from CI and it telling you there was a failure?

portsfallout is just a way of sorting through the ports fallout that's sent to mailing lists, developers get the fallout from when ports start failing once the build happens. What lw is talking about is a pre-commit hook, I think?

Pre-commit hooked CIs have been considered, but for reasons which will be rather apparent, they aren't very popular.

lw, how is your networking "homework" going?

jbo: new switch is configured, should be able to install stuff tonight

assuming PPPoE over a VLAN works and doesn't like, flood the traffic to every port or something

neat!

it's so tiny le-fay.org/tmp/30d/aFiyNJ.jpeg

that plug was incorrectly assembled

</electronics_engineer>

correct

i have no idea where this cable came from, i wouldn't make one like that

Agreed. That RJ-45 was not assembled according to spec. </also_EE_here>

I assume the information is somewhere in the scrollback but so I don't see it immediately but what model of device is that and is it running a BSD? Looks interesting to me if it is.

rwp: Mikrotik CRS305, and no, it runs RouterOS which is based on Linux. wouldn't be very useful to run FreeBSD on it because the CPU is slow and there's no driver for the switch chip

You and I already discussed this but I am running two ZimaBoard small machines with FreeBSD and it's been a useful "pi-like" machine for me with dual NICs for routing. Now my house router for one and a testing system for another.

I am just about to purchase one of these protectli.com/product-comparison with CoreBoot installed and try it. Probably one of the multi-gig 2.5G NIC ones.

hmm, pricey

I love their page display there and think a lot of web sites could learn a thing or two. It's a just-the-facts list without a lot of fancy glitter on the web page.

They do come with a price tag. Which is why I haven't bought one yet. It would be 2.5x the price of the last similar machine I bought for the 2.5G NIC one. But the regular bottom end machine would only be 1.5x the last NUC-like system I bought so not quite so bad and a much better bit of kit hardware wise so probably worth another 50%.

if i was going to buy something like that i'd want SFP(+) cages... trying to move away from copper as we migrate to 10GE

ARM64 bhyve committed to main: cgit.freebsd.org/src/commit/?id=47e073941f4e

Title: src - FreeBSD source tree

I could do with some Arm sponsorship

this is to run only arm64 VMs on another arm64 system?

rtprio: yes

Cool to see ARM64 progressing! Speaking of hardware I have been wanting an ARM64 machine and I would really like a RISC-V machine.

Wanting machines with the ability to put in a workstation amount of memory. So... At least 16GB. But I am also a cheap craphound so that's an impossible combination right now. Being a cheap craphound scrounger is why I am sticking with other hardware. But I think ARM and RISC-V are the future.

rwp: i'd love to move to ARM64 but the mid-range hardware options just don't exist. you either get tiny embedded systems (like RPi) or huge servers like Ampere... nothing in between

like, i want something with a reasonable number of reasonably fast cores that i can put a SAS HBA in and run ZFS on for a NAS.

(i mean, i'd like to move to RISC-V too, but that's even further behind)

i may repurpose my mac mini into a server once freebsd runs on it usefully though

Yup! That's the state of things right now.

fricking rust _mktemp: mkstemp failed on /poudriere/data/logs/bulk/main-main/2024-02-21_17h39m26s/.tmp-.poudriere.snap_loadavg.YmI8eLYM: No space left on device

:>

love to build rust for three hours to be thwarted by running out of disk space

jup

I'm considering getting a hardware upgrade

but not sure whether I should upgrade my desktop or a new build server

upgrade how, like the storage?

nah, at this point it would be completely new machine

i should probably increase the storage on this VM from like, 128GB to 1TB or something, i guess 128GB was a bit optimistic. but still. fricking rust.

i stuffed a huge nvidia 1080 into my old PC and it's still usable.

still amazes me modern cards take 2 screws to hold down

my build servers are Xeon E5-2690v4 CPUs

and my current workstation is an i7-8086K

johnjaye: sometimes 3 screws :P

rtprio, How much space was not enough space yet?

(!)

I am considering to build a new workstation with w5-3435x

at that rate you will have no room for any peripheral cards, just the video card

Yeah, 3 to 3.5 slot cards are about the normal now, especially on the high end

johnjaye: On some setups that is already a thing

lw, How much space was not enough space yet? (Sorry for the misdirect to rtprio.)

how much do these things weigh. one wrong step and you could tear out the pci express slot

rwp: 128GB, but that's shared between poudriere and my src builds, package repository, etc

have you not seen the latest in pcie-slot technology? moulded-in-place-metal-shields

That's still a significant amount of space.

which I doubt help much when you bump it :D

johnjaye: I don't use those type of fat graphics cards, I'm just saying, they make them so thick because of needing a bigger cooler for an factory overclocked GPU

jbo: no my board is from the days of ddr3. are pcie express slots like steel reinforced fortresses or something?

johnjaye, these days yes

johnjaye, this is the board I'm considering to get: asus.com/ch-en/motherboards-compone…ds/workstation/pro-ws-w790e-sage-se

Title: Pro WS W790E-SAGE SE｜Motherboards｜ASUS Switzerland

my desktop's graphics card takes 3 slots and it's not even high end (RX 6800 XT). but fortunately the motherboard has an x4 slot at the bottom to put a NIC in

jbo: what does it mean by "safeslot"? those massive clips on the bacK?

johnjaye: generally no, a correctly built graphics card doesn't need reinforced PCIe slots, it just needs a correct mounting system. but you do some such slots, usually on gaming boards because i guess it looks cool or something

yeah, I doubt they help outside of marketing

> SafeSlot is a PCIe® slot reengineered by ASUS to provide optimized attachment and shearing resistance. Manufactured in a single step using a new insert-molding process, SafeSlot fortifies the slot with metal for enhanced strength and durability.

lw: exactly that

most likely way less than 0.5mm

tbh i'm more worried about mounting CPU coolers nowadays since those are generally not attached anywhere except the board

what does 14+1+1 power stages mean? the terminology is advancing faster than the tech

jbo: i like that all the slots on this board are mechnically x16. everyone should do that

lw: meaning it could tear the board or cpu somehow?

johnjaye: that refers to the configuration of the VRM that provides power to the CPU

but again... it's mostly marketing nonsense. the only way to find out if a board has good power delivery is to test it

lw, they are not just mechanically x16. they are also electrically x16

ah ok. i'm sure motherboards of the last decade (i.e. when i built my pc) had those. just didn't have a marketing term for them

jbo: ah interesting

maybe gpus should just come in a small box like the power supply. and you connect it to the motherboard with huge cables

jbo: odd choice though because what would you use that many x16 slots for? but i guess if you have spare lanes you may as well connect them somewhere

lw, the CPU that goes into that board as 128 lanes

112 lanes, sorry.

jbo: idgi, if it only has 112 lanes total, how did they make 7 x16 slots? that's all 112 lanes used before considering any other peripherals like ethernet or usb

i thought about building a new pc but i can play new games with just a monster gpu upgrade

unless it's like, you get 7 x16 slots but you can't use all of them at x16 at the same time, which would make sense because what would you even put in them

lw, the CPU as 112 PCI 5.0 lanes and a 16 GBps x8 link to the chipset which exposes a shit ton of 3.0 and 4.0 lanes (and USB)

lw, tpucdn.com/cpu-specs/images/connectivity/intel-ws-expert-spr112l.png

by the way if anybody here is an expert in overclocking or knows a good tutorial please let me know

lw, it also has octa-channel memory :D

ah the southbridge does usb and sata... that makes sense, i vaguely remember reading about that when the last generation of intel stuff came out

jup

just a sh*t ton of money :(

6k USD is not a drive-by buy for me

jbo: hard to believe you get all of that and still no RS-232 port on the backplate :-P

:D

(this has been annoying me this week because i need to run kdb on my desktop and no serial port... bleh)

lw, you can get very cheap expansion cards or USB based solutions that work well

lw, just stay away from the pesky CH11 chipsets

jbo: i'm going to try a USB adapter but it's not clear if those can be used as console for kdb

(prolific)

lw, help me out - kdb?

if not maybe i'll try a pcie card, but i'm not sure i have any spare pcie slots

jbo: the kernel debugger, that you break into on a panic

aye - I have no knowledge on that. only been doing userland stuff with serial

trying to debug why i keep getting random hangs for no apparently reason, i've tried enabling the kernel watchdog which is supposed to panic (and dump a core) on panic, we'll see how that goes

haven't done that in many years - but couldn't you "just" inspect the dump?

if you end up in a panic you can dump, right?

there isn't a dump, it just hangs. that's why i want a serial console, so i can forcibly break into kdb

lw, not sure what hardware you're rolling with but the z370 board I currently have has a com port

i know, some boards still do, my desktop doesn't

jup

lw: you... use rs-232 ports? how. i must know more

and that's probably still UART, not RS232

i have one but don't know how to use it

johnjaye: i plug a laptop into them and run kermit to talk to whatever is on the other end

lw, I guess if it's only a software thing you could build a vm loadable disk and nmdm it?

as in: not hardware related/triggered issue

jbo: and what, pass the graphics card through to the VM? ... i suppose if nothing else works i could try that

lw, not saying it's pretty, but...

i've just rebuilt current main so i'm hoping someone has committed a fix for whatever the problem is in the mean time :-P

the ppt is fairly non-intrusive (compared to everything else)

jbo: pro-ws-w790e-sage-se really good mobo it seems, I might get that even for server/NAS purposes which I need nowadays, thanks for that model.

jbo: you know, i'm now wondering if just running my desktop under vmm might not be a terrible idea. if i passthrough the storage, network and graphics, the performance hit probably isn't that big, right?

and it would make it easier to debug things like this

tercaL, jup, seems to be one of the best in that category at the moment

lw, that is what I am planning to do after the xeon upgrade. right now I do that for windows & linux desktops

lw, I am running stable/14 on metal and got one of those fancy PCIe USB cards that has one controller per port. this way I can have four VMs with USB pass thru

works lovely

lw, startech.com/en-us/cards-adapters/pexusb3s44v

Title: 4 Port PCIe USB 3.0 Card w/ 4 Channels - USB 3.0 Cards | Add-on Cards & Peripherals | StarTech.com

jbo: how do you tell the number of usb controllers on the card? is it from a visual image?

lw, compared to other cards here you get actually four USB controllers (one per port) so you can pass them individually to VMs without needing extra cards

(i have a pcie card with 4 usb ports...)

johnjaye, pciconf

johnjaye, but unless you purchased something intentionally, you almost always have just one per card

i see

this is a "specialty" item

but like. *can* you tell visually or is that more a software thing

jbo: ah good ol' startech

it's a hardware thing

lw, indeed

i think my pcie usb card is startech as well but it's not that one

lw, you now their slogan, right? "Hard-to-find made easy" :D

yeah :-) half the crap i buy is from them... including the usb serial adapter i just bought

lw, incidently, they have the same card with two ports per controller

is that part where it says "4 dedicated 5Gps channels" the tell for having 4 controllers?

lw, I also got some USB hubs from them. all good gear

the only bad product i got from them was an 8-port pcie sata controller that had two different controller ICs, one of which isn't supported by anything

should have just bought a SAS card from ebay tbh

yeah, I'll probably not bother with SATA again

made that mistake 5 years ago in some servers

not worth it

this was a few years ago and i think at the time i didn't know you can use sata disks with a sas hba... oh well, live and learn

that was exactly the same here :D

lw: i don't understand your sentence at all. but idk how to explain that i don't understand it

lw, I did the stupid thing and bought this: delock.de/produkt/89384/dokumente.html

Title: Delock Produkte 89384 Delock PCI Express x2 Karte > 10 x intern SATA 6 Gb/s - Low Profile Formfaktor

lw, instead just a decent SAS HBA

yeah, you can get 6/12Gbps LSI cards for peanuts on ebay now

i was looking at an Epyc board the other day and it had something called "MiniSAS", i couldn't work out what it actually was, seems like one port with 4-8 SAS ports in and you use an expander cable with it

except the port was different from the one you get on a SAS HBA

If I have a device receiving packets from off network through multiple NATting gateways, is there a way to have the device route replies through the particular NAT gateway that originated the connection, rather than a default route?

thanks for the help this evening. The system is alive'n'kickin again

Hi, I am trying to install Freebsd 14 in a Thinkpad x280, using ZFS and I need encryption with /home encrypted, another directory with a different encryption key as it is more sensitive and will be rarely used, plus another directory/partition without encryption as it will contain unimportant media files. I don´t know much about ZFS and this cannot be done easily from the installar, any help is

appreciated!

i would recommend reading the chapter in the handbook on zfs. and then buy the book about it

Also, my aim is to have native ZFS encryption but I would use Geli if that is easier

the book being "FreeBSD Mastery: ZFS (It Mastery)"

I would buy that book..

johnjaye: thanks, but if buying a book is needed for this, I would give it a miss, I think my requirements are not that weird

how do I change name on the labels for each disk in a zpool?

well you get what you pay for. the freebsd handbook is free and has a chapter about it

somehow one of my disks in the zpool changed name and I dont know why x)

For nut to be able to work with my ups I need to detach the driver from the kernel before starting the service. Where can I execute (usbconfig -d 0.3 detach_kernel_driver) without modifying the startup script? Should I create a new script and make nut's script depending on mine?

what are the common reasons for freebsd installer to be unable to boot?

I already have freebsd installed on my ryzen9 b650 machine.... but for some reason Im unable to get installer running and cant remember what I did last time to make it run.

drobban: did moving the disk to another system not work in the end?

lw: I managed to fix everything... now I try to find out why my first solution didnt work

it annoys me that I dont know how to start the freaking bsd installer

lw: thanks for your help btw.

I wonder what I need to disable during boot when I want to run the installer

so you have 14.0 installed and that works fine, and you're booting the 14.0 installer and it doesn't work? that is weird, it might be worth asking the mailing list about it

lw: yea,, its super weird. Perhaps I did something / disabled something in the bootmanager.. I cant remember. I know I installed 13.x version and then updated from there to 14.0.. But the funny thing is that it wont boot 13.3 installer neither.

drobban: check your /boot/loader.conf on the working system? if anything there looks relevant you can try setting it on the installer's loader before it boots

lw: yea, took a look. but didnt notice anything special.

lw: I mean, there is only stuff about zfs and some other nonsense that hardly can be related

I mostly run a generic kernel as well. the only thing that differs from generic is some modification a made to UART driver.

CrtxReavr: btw. what I was trying to say was 14.0-Release-p3....

Cool story.

tercaL: interesting, looks like expanders only though

also looks like it uses weird long-running commands like security erase

CrtxReavr: =) you seem pretty cool

oops, wrong channel

drobban_: obviously you would now claim you meant 14.0-RELEASE-p3, since you're trying to hide the fact that you came here from the future

;)

/bin/sh: arithmetic expression: expecting primary: "2**31+1"

Doh!

anyone able to spot what might cause problems with the installer from this probe? bsd-hardware.info/?probe=54d6c96d25

Title: HW probe of ASRock B650 PG Lightning Desktop Computer #54d6c96d25

when the system reboots because of a watchdog(4) timeout, is it supposed to write a crash dump? or do i need to enable a kernel option or sysctl?

lw: are crashdumps enabled, in general?

should be

# dumpon -l

ada0p1

meena: this has been plauging me all month, constant lockups or hangs and i can't get it to write a crash dump

that sounds extremely frustrating

maybe i should build with DEADLKRES

eh? That's enabled in GENERIC?

is it?

meena: ah but it's not enabled in MINIMAL which is what my kernel is based on

lw: we should change that.

it's even on in FIRECRACKER

maybe because it's not in GENERIC-NODEBUG... this is so messy, there should be MINIMAL and MINIMAL-NODEBUG which are identical to GENERIC(-NODEBUG) for things like this

GENERIC-NODEBUG should just be GENERIC with NODEBUG

oh, in the mean time my pkg builder vm made a crash dump, let's see what's going on there

Fatal trap 12: page fault while in kernel mode

you're having all the fun

i really am

interestingly that VM is running the same commit as my desktop

ed27ae8df4b10f67289a32458d89d143e758f6d4

that's like 2 weeks old, let's try updating before opening a bug

anyone recognise this stack trace? le-fay.org/tmp/30d/WdkR3T.txt

crash in tcp makes me *very* suspicious because i noticed my desktop crashed every time the internet connection went down when we were having issues with that the other day

looks like a few tcp commits went in around that time

meena: we should ask imp@ about this minimal thing because all changes are meant to go through him, i guess

meena: or i could just do this freebsd/freebsd-src #1124

Title: sys/amd64/conf: unify MINIMAL and GENERIC debug by llfw · Pull Request #1124 · freebsd/freebsd-src · GitHub

lw: we need github.com/freebsd/freebsd-src/blob/main/sys/conf/std.nodebug but for std.debug and just flip flop between the two

Title: freebsd-src/sys/conf/std.nodebug at main · freebsd/freebsd-src · GitHub

yeah i was just thinking about that

I wanna be able to build debug kernels on STABLE/RELEASE without pain

none of this is specific to amd64 so it shouldn't be in sys/amd64/conf

imp was saying he wants an "exclude" in addition to an include, so that would flip whatever is in the file listed

i think amd64 is just the prototyping ground for MINIMAL, because aarch64 is a bitch to boot

let's see if we can get this in first, then we can add std.debug, then we can apply it to arm64 and riscv

also just found an SFP in my skirt pocket, where did that come from?

meena: i think imp will commit this because he wants minimal to be "generic, but without all the builtin crap"

which this is in the spirit of

skirts are like tshirts for your legs, why wouldn't people wear one

meena: what do you mean by a separate commit, like one to add std.debug and another one to use it?

i'm from the balkans. It's too cold here for me

yes

oh now wtf is going on, my music player is broken

did someone break OSS

another review comment, sorry it was staggered in two reviews;

there's probably some shit that can be removed from MINIMAL, because it can be loaded as module, but I need a clear head to go through those, and right now I'm working on something else

hrm…

lw: n.b.: i386 has MINIMAL too.

meena: i pushed a new version, what do you think?

i will ignore i386 and arm64 for now, let's fix it for amd64 first

also i don't know what NO_UNIVERSE is