michaeldexter, one thing to note, when i set -s 30,xhci,tablet, i get these errors in console when linux kernel in the guest is booting: pci_xhci: portregs_write to unpowered port 1

this gets logged 20+ times just after selecting kernel in the guest's grub

Are you willing to file a PR at bugs.freebsd.org, first searching for the issue?

sure

when i search for that error, only result in google brings me here: illumos.org/issues/14623

Title: Bug #14623: bhyve/xhci: Connecting device on port 1 failed - illumos gate - illumos

14623 – Some Netgear FS509s have problems bugs.freebsd.org/bugzilla/show_bug.cgi?id=14623

hmmm, to apply this patch, do i need to build whole kernel? illumos/illumos-gate 104fd29

Title: 14623 bhyve/xhci: Connecting device on port 1 failed · illumos/illumos-gate@104fd29 · GitHub

CorvetteZR1: That looks like it would be in the bhyve(8) utility.

right. so can i patch/compile only bhyve to test this?

when i try to compile bhyve, it fails: dpaste.com/89GRPXR8E

Title: dpaste: bhyve make fail

any idea what i might be missing or doing wrong?

CorvetteZR1: That with the patch? Maybe ask on freebsd-jails on libera. There are a few nvlist users there.

the 'patch' i'm trying is just modifying one line, not sure if it's the fix

CorvetteZR1: I have documented the known aspects of your issue on the bhyve weekly call doc so they don’t get lost.

cool, thank you!

CorvetteZR1: How hands-on do you want to get with this?

i'm not a developer...but i'll help where i can

You have a good… example bug. :)

if i can figure out how to compile bhyve, that'll be a start :)

That is a great start!

CorvetteZR1: Willing to take this to #bhyve?

sure

Probably best not to spam this channel.

CorvetteZR1: if you're building random things in the tree, you need to buildworld first unless you want to play 52 card pickup with the dependencies

CorvetteZR1: what you're seeing is becaue you need to build libnv and libvmmapi, but they themselves might have dependencies that need to be built, and so on and so forth

kevans, gotcha, thanks!

CorvetteZR1: there is also buildenv target that will get you set up with needed prebuilt stuff and put you in a shell with proper setup for changing/building arbitrary bits in the tree

(i just hate how zsh's CPUTYPE needs to be unset every time though)

interesting...never heard of buildenv before. will look into it

Guys, just thank you for colored diff :)

made my first comment on a bug :-)

mns, Thank you for being involved and contributing! :-)

mns++

i got a question on sudo rule. polyex ALL=(ALL) ... applies to me on any host but what's the 2nd all for?

doc doesn't make sense to me

The second all is the target user. It allows all users to be a target.

If the rule allows root then it might as well allow ALL since root can always become other users anyway.

well if 2nd all is target user why does rule start with my username?

The first field is the from user and the later field is the to user.

It's reasonable to allow a specific user to become another specific user, such as a role account, but not to become any of all possible users.

You probably don't need to restrict it for whatever you are doing but as a general purpose tool the capability is there. And I use it in my environment.

what's your rule of thumb on it?

My rule of thumb? Do the Right Thing. :-)

As I said if you want to allow yourself to become root then say ALL there. Because as root you could become other users anyway.

It's an optional field anyway. You don't need to include it at all. It will default to ALL if you don't include it.

I see I do mostly include it in my rules just as part of the template of use. But I can also see that I don't always include it. I should make that consistent in my files. Editing...

But for example if you were to say "polyex ALL=(root) ALL" then you could only switch user to root and would not be able to switch user to rwp for example. Which if you can switch to root it is silly to prevent you from switching to rwp.

i want to run maybe 10 OS in Bhyve and be able to open their programs in X

for something that requires root like 'service' we need ALL so you can become root right?

when i changed it to me it stops working

Yes.

or just root i guess

Here is a small snippet: bsd.to/Z2Tw/raw

Title: Z2Tw

darwin, What is "10 OS"? Mac OS 10? MS Windows 10? Debian 10 Buster?

polyex, Here is a snippet from a system using a shared role account for mailing list management "LISTHELPERS havoc = (mlmgr) NOPASSWD: /bin/bash" which is an example using something other than (ALL) in the target user. "sudo -i -u mlmgr" works and is expected but does not allow switching to any other user.

polyex, Another hint: I don't modify /usr/local/etc/sudoers instead I add /usr/local/etc/sudoers.d/sudoers with my own file as an override without ever touching the original file. That way I never need to merge files on an upgrade.

The packaged default sudoers file has "@includedir /usr/local/etc/sudoers.d" to include that .d directory of config files.

why do you run from havoc, and as mlmgr?

btw i could get away with putting (root) didn't need (ALL) so any reason to use ALL instead?

"havoc" is the name of the server where the mailing list is hosted. So the host name field says havoc to limit that rule to only work on havoc. That file is shared across multiple systems. Other systems sharing that file won't match on the host name.

ya sorry i meant what's LISTHELPERS? the from user right?

It's useful in my situation to have a shared role account. And since most of the people don't have a password, they only use ssh keys to log into the system, they can't use a password with sudo. They don't have a password. But the shared account is exactly as secure as their originating accounts, which is pretty secure since they can only use ssh keys to log in.

rwp, '10 operating systems (OS)'

So "mlmgr" is the mailing list manager role account. Note that I am not proposing that you need this or should do this. It's just an example.

ya i know just curious

darwin, Cool! I have never heard of this system!

so any reason to use ALL over root if you only need root?

what system?

polyex, Yes, LISTHELPERS is a list of user account names that can do this.

darwin, "10 operating systems (OS)" that you mentioned. You just said it!

no I didn't. I said systems

i didn't mention a specific one yet

You said "<darwin> I want to run maybe 10 OS in Bhyve and be able to open their programs in X" and I don't know what "10 OS" is.

In any case I am not myself very good with bhyve but maybe other people can help there.

yuripv: buildenv still needs the stuff to be bootstrapped

it only does the flag / env setup bits

if you got a bin that builds up a cache file as it runs where do you put it?

in man hier i saw there's /var/cache, but that docs implies it's temporary. but this cache file is persistent

normaly cache is stuff that can be deleted for space reasons, but only the user should delete stuff from /var/cache (or the program that has originally written it, when it notices outdated stuff)

ya it could be deleted, and it would be automatically recreated next time program starts, but it takes time and it's inconvenient. so i preserve the cache file across system installs

still the best place is /var/cache?

i was thinking maybe /usr/local/share too

not sure which 1 is better

/var/cache sounds better, as it directly tells the user that it can be excluded from backups

ok

if i make an rc script to run my bin when the server starts up i use daemon that way it keeps it running. but in my long command line, how can i pass some args to MY bin instead of it being more daemon args?

all arguments after command-to-be-run-under-daemon are passed to command

(see man 8 daemon)

also people.freebsd.org/~blackend/doc/en…/articles/rc-scripting/article.html

Title: Practical rc.d scripting in BSD

though I forgot if it actually mentions daemon

oh right

it does not

nimaje: well.. I can load i915kms if I move away the firmware as mentioned in issue 252 in freebsd/drm-kmod.. so maybe the firmware needs updating? Or something...

Remilia: I think docs.freebsd.org/en/articles/rc-scripting is a more up-to-date version of that article

Title: Practical rc.d scripting in BSD | FreeBSD Documentation Portal

does seem like it

It's also the right file to modify, in case anyone wants to contribute a section on daemon(8). ; )

I checked it as well and it does not mention daemon either

I would if I had decent enough knowledge :(

Decent enough knowledge of what?

of rc.scripting and daemon, in practice

that full stop is a mistype

Ah, okay - I was thinking if it was knowledge of writing docuemntation for FreeBSD, there's documentation for that too :P

haha I got like 6 years experience as a technical writer but I never write about things I do not understand well enough

now is the best time to begin!

well sometimes I do but only in my personal weblog or short length like fedi

what's the preferred way to configure a touchpad in FreeBSD these days? I see old wiki pages about synaptics and sysctls.. is that all superceded by libinput now?

you could port someting you need that has a service and write about it while you learn from man pages and other rc scripts and at the end we should probably have a useful port and a new section in the documentation

Hi all, how much will FreeBSD complain if I move all my disks (incl. system, full zfs arrays, etc.) to another board? (currently have a raidz1 array with 4 disks, and that's it)

x0.at/Bsay.txt <- /usr/local/sbin/zfsnap destroy -r zroot # didn't remove old/stale snapshots such as zroot/ROOT/default⊙b2_16.34.30--34d . Why?

moviuro: I am unsure if the developers of zfsnap frequent this channel

have you tried posting an issue in the github repository?

oh it seems dead

the repo looks pretty much dead (last open issue 2020)

last closed is Oct 27

moviuro: zfsnap/zfsnap #110

Title: zfSnap -d command is not working · Issue #110 · zfsnap/zfsnap · GitHub

110 – initial install leaves /tmp and /usr/tmp mode 755. bugs.freebsd.org/bugzilla/show_bug.cgi?id=110

interesting

zfsnap destroy -r -v -p before-13.2- zroot # seems to do the trick

are you using zfsnap or zfsnap2

zfsnap -V says 2.0.0.beta3

oh so you installed zfsnap2

Any zfs gurus around? Got two kinda old disks (running for 8.5 years) in the server. 120GB ad0 (GPT, freebsd-boot) and 1TB ad2 (GPT, freebsd-boot). Thinking about putting a 12tb disk in. Not running in mirror or raid at the moment. Should I do zpool replace zroot adaop3 NEWDISK ; gpart backup ada0 | gpart restore -F NEWDISK ; gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 2 NEWDISK ;

wait for resilver until done then ; zpool remove ada0p3 ? Can I do this to the second (1TB disk) too? Or do you guys have a better solution? Creating a new pool on the new disk and move the old pool to the new? Thanks in advance.

gendish: make sure that 12 TB drive is not SMR

gendish: by the way does your system only support legacy boot? if you can use UEFI boot it is a lot more convenient

Remilia: It's a CMR-disk according to the spec sheet. Think it's a non-UEFI board. Asus P6T SE

yeah :\

i'm not a zfs guru and i've never done zpool replace where the pool has only one disk, but i would just create a new zpool on the new drive and then zfs send -R, then destroy the original, unless you cannot allow downtime

i'd wonder whether an unsuccesful zpool replace would leave your pool in an undesirable state

(you will need recv -u when you receive it so that it doesn't automatically mount over your existing mountpoints)

jmnbtslsQE: ok thanks!

Remilia: So I must keep my 120GB disk to be able to boot if I understand it correctly right?

gendish: no, if you use gptzfsboot etc.

the bootcode part

refer to stuff like 14.0 release notes there should be instructions for upgrading MBR bootcode and they apply

or just read the gpart-bootcode manual page

Remilia: Cheers!

morning peeps

MORNING

HI!

good afternoon

gendish: afternoon :)

muuuurnin

hi hernan

mornin'

what are you all up to with FreeBSD?

jauntyd: morning

sup jauntyd

not much. browsing the forums

well, jauntyd I for one am waiting to pull the trigger on a NAS and can't freaking wait to install fbsd on it and play with zfs and possibly caching if I can afford it

Game development

whoa Zyxer, that's awesome.

I have always wanted to test run steam one of these days

sounds fun deacon426

Uh... I am using godot

Guess we will release for steam as well, no idea how to make it run on FreeBSD. But I heard it is possible.

Zyxer: and relatively easy :)

OpenBSD can download steam games from cli without the steam interface

cool

to the NAS!

ZOOOOOOMIES

I like the idea of having a local steam server just having it call out for updates on its own

Oh, that is what you use the NAS for

Zyxer: one of many utilities I'm sure

everyone talks about FreeNAS or TrueNAS or Synology or this and that, I always wonder way they just don't go the way of the sword so to speak (native fbsd)

I'm likely to go with a new i3 if it will be to remain striclty as backend storage, but if a hypervisor will be utilized on the same box then I'll been a beefier chip. The new i series have native ECC and aren't very expensive. Mainboards on the otherhand, :(

*need

plus decent hotswappable chasis, jeepers :( expensive

Mainboards are expensive due to PCI-E 4, which is overkill unless you go for the 8K gaming

yeah I think intel pushed for PCI-E 5, probably AMD as well which is ridiculous

Bigger profit

letters on a box

Profit margin is usually x%, but if the price is big then the x% ends up being a bigger sum

GET ME THOSE LETTER ON THAT BOX -- some CEO

Yea

*letters

Zyxer, download steam games without steam...explain

PCI-E 4 is already overkill. I know of one single use case for it that is aplicable for consumers

yea, i used godot on freebsd, works great. i didn't release anything, but what i had was playable on freebsd...that was with godot 3.5 though

I think AMD's SAM uses it, though I very well may be ignorant

but i still have godot 4 installed and the editor launches...but the game i was working on is broken now. because of godot, not because of freebsd

CorvetteZR1: It logins to steam and all using API, so it is a script that runs in terminal. Then use API to fetch a game in your list. I think it runs the steam API terminal program when you launch steam games to still work despite DRM, since the DRM is usually ping server to confirm you owe me

I bet the steam client is electron or something,

With extra privileges

Or well, I guess all electron apps get bonus privileges on system than if run in browser

what's it's called? steamcli or something? i guess it's not just an openbsd thing, you can run this utility on other oses too?

CorvetteZR1: steamctl

cool! will check it out

damn near everything is on git these days no doubt

'on git'? you mean in a git repo? why would that matter?

he want's to check it out, does it not make sense to visit the development of said program?

and yes nimaje, in a git repo. Thank you for the correction, sincerely. :)

but to see it you need to know where it is hosted, not what vcs is used, it doesn't benefit you to know that there are some git repos on my laptop, as you can't access them

hm, there seems to be some steamctl on github, not sure if you meant that, but it seems to do what you spoke off

Oh, the actual usecase for PCI-E gen 4/5, you can use fewer datalanes for more/same speed

Meaning my 4 NVME to PCI-e adapter could be shrunken down to only need 4 or 8 lane PCI-e slot

Now that I got an mATX I cannot fit it in my motherboard, since 1 16 lane is for GPU, and there is no other 16 lane that the PCI-e nvme adapter needs. Luckily the only reason I bought it was to be able to add another NVMe, but my new MoBo has 2 NVMe slots.

Sadly I bet at least one does PCI-e 4 without sharing lanes, meaning that unless I got some passive cooling using heat sinks with plates or whatever it is called in english, the NVMe bottlenecks due to thermal limits.

The early NVMe that supported PCI-e gen 4 performed worse than gen 3, because it performed better in a burst until it got hot, then it had to bottleneck so severely that the PCI-e gen 3 NVMe outperformed it

Zyxer: that is my setup, though my main rig is mITX

all pci-e 4 lanes. one gpu and two nvme

the main does have a heat sync and fan

(nvme)

I thought my PCI-E card was a 4 laner... Otherwise I would not have gotten an mATX (or at least one with two 16/2x8 lanes)

deacon426: Lucky you. My NVMe has nothing of the sort

So I connected it to the secondary M.2 slot

had to tweak the curve though because of coil whine

Hoping that it shares lanes with other stuff thus "bottlenecking" the NVMe so it won't overheat

I am scared the heatsink by itself will be to fat and collide with GPU

it is beefy

mATX you should be fine.

I fell into the SFF rabbit hole, and was A-OK

what is SFF?

also it is this motherboard I got

"small form factor"

Title: ASUS PRIME B550M-K ARGB - Moderkort | Webhallen

Oh, looking at it I realised the prioritized M.2 slot is above the GPU and should fit a heatsink on top of

I've built in these two cases thus far: dan-cases.com/dana4.php ssupd.co/products/meshlicious

Title: DAN Cases

ssupd being just, awesome.

fits a 280mm rad and a full sized 2.5 slot gpu

Zyxer: yeah, fabricate one, that would be cool.

I found a heatsink

A returned one tho', they claim it has no signs of usage except maybe instructions not following

be quiet! MC1 Pro

Oh you really went down the SFF rabit hole

Damn, the ssupd had a nice case

I almost went down the SFF rabbit hole. I found it interesting enough to stick my head down the hole, then saw the prices and noped out of there

Today the only mATX I bought were because they were cheaper for me, or I had found some chassi on black friday, and when upgrading MoBo only mATX would fit

but the SFF chassis are very sexy

I have a full tower but some of the SFF chassis can fit bigger GPUs than my full tower

A friend bought one of those, when it was a semi new concept afaik

An SFF chassi

He did not reciev one within a reasonable time, they costed around 400-500 bucks back then, so he was getting worried and emailed them and asked about it. They excused their mistake, they had forgotten to ship it.

I think they got stressed and almost panicked because when package arrived he got 2 chassis instead of one. Or that was their way of excusing themselves for the mistake

(2 different packages, they arrived one day after the other)

my reentry into the world of FreeBSD is coupled with a desire to go for a smaller, quieter, server, so I'm working on an mATX board in a smallish cube case. I'm pretty impressed so far

I think I'm developing an abiding love for the zany ASRock server boards that cram as many features as possible into a small space

Minimal OS, minimal (physically) hardware

tsundoku: I'm not sure but I think the 790-itx I was looking at was ASRock. ECC and all :)

yeah Zyxer: by the time the dan case arrived I think I paid just over $300 which is waaay too much. Darn niche products

same goes for mechanical keyboards.. :(

Split keyboards are the way to go

you get what you pay for I guess

I make my own keyboards... it's not that cheap but it gets me what I want

Nah, you don't always get what you pay for, but you never get more than what you paid for

My friend made his own keyboard, I think in total tops 70bucks

I'd have to add it up but it probably ended up costing a few hundred dollars at the end of the day

Bought his own microcontroller, programmed it, the PCB board, the switches and all. Even has 2 smoll screens. split keyboard

Wat

the keycaps are the most expensive part, especially if you're doing something less common. I had to mix and match from a few sets to accomplish what I was trying to do

Damn, yea ok they can go up· I only heard some keramik keycaps. They sounded so good it was ASMR

I have a beautiful gradient seafoam green set that didn't fit, darn standards

...just sitting in a bit now, taking up space

*bin

I gotta ask, why do many tech communities use libera?

I have... a lot of sets

Personally I really dislike having to sign up and give email for the luls

whenever I see something that has Mac-style modifiers, I jump on it unless I don't like the way it looks

> I gotta ask, why do many tech communities use libera?

as opposed to... ?

'having to sign up' is a barrier for spammers

it was freenode, but freenode fell victim to some drama, libera got stood up as a replacement, and it was widely accepted

I have an iMac G3, with original keyboard, running OpenBSD (FreeBSD was much more hassle to install on that iMac G3).

oftc / hackint and probably others

tsundoku: As opposed to unofficial servers I am in. No signup or anything required

define "unofficial"

Also how does the email prevent spam?

Zyxer: have you ever been to DALnet in late 1990s/early 2000s?

the thing I like most about registration with services is that I can make sure nobody steals my nick.

Unofficial as in not referred to in any official page. Freebsd.org or whatever the main site is for example

DALnet did not have services and you did not need to register

in return you had to run your own bots and hope they do not disconnect

I run an IRC network myself and even though it's tiny, dealing with spammers takes a certain amount of work

or your channel would be taken over in an instant

services packages go a long way towards making it less of a pain

Bro, I was still in diapers in the 90's

bro yourself

etc.

nothing wrong with being younger, but historical context is important

things usually are the way they are for a reason

yeah

If ppl self hosted it wouldn't be as susceptible to spam and such, I think

Yea ok. True, historical context matters

and by requiring a valid email for registration + a registered nickname/account means spammers have much harder time getting in

what do you mean?

'self-hosted' IRC is insanity

if it's on the internet, it's susceptible to spam. it doesn't matter if it's big or small, self-hosted, cloud-based, whatever

I guess I am insane then

I and a friend.

I've self-hosted irc before. No white coats showed up.

obscurity won't get you far in an era when spammers are constantly crawling the Internet looking for identifiable services to pester

But sure allocating nick names I get it

some recent spam wave seemed to just scan the ipv4 adress space for irc servers and spam them

first of all, 99% of hosting/dedicated/etc. providers will kick you out for hosting ircd

????????????

unless you disclose that to them and often pay extra

because IRC servers are risk factors

My friend has a VPS that doesn't care?

I self-host all sorts of stuff, and you still have to have reasonable mitigations in place to deal with spammers and casual malicious access attempts. they will find you.

And I only allow i2p tunnels

so I don't understand your logic of self-hosting making this less of a problem

if you have an IRC server that does not limit access to it, it will be used as a command & control centre for a botnet

Ok, and why can't the botnet owner setup their own IRC server for that?

why would they?

this conversation is getting very confusing.

I mean

other than that

They can, but they don't have to if they can use yours :D

Because many VPSes allow for anonymous pay and all that

Get their friends on and plausible deniability

I think you misunderstand the purpose and appeal of a botnet

I guess it is really great to have to connect to 15-20 self-hosted IRC servers if you participate in as many open source projects?

well yeah that's the other thing

that could be traced back to them, so using an open ircd is a better idea for them and it doesn't cost them

Yes. I rather have to do leg day than risk one actor fucking up or crashing or whatever and every community suddenly gone

it sounds like you're saying "why does this public IRC network for widely used software projects not work like the private one I hang out on with my friend and limit access to?" which doesn't make sense

what does 'one actor' mean

nobody wants to have to like, contact the host of every chat channel and set up a tunnel to access their special invite-only resources

Libera is an IRC network

nimaje: You pay with crypto for a VPS, then make a public IRC, and no one can prove the IRC owner was behind the botnet

Maybe I am misunderstanding how libera works or IRC, aren't all channels on libera hosted on libera server?

it's nice to have small friends-only spaces online, but the purpose of Libera and channels like these is fundamentally different from that. it's just like in "real life." different spaces have varying degrees of public/private nature, and are set up accordingly

[22:46:22] * There are 70 users and 33496 invisible on 28 servers

Or is libera only for verification of email?

LiberaChat has 28 IRC servers

all over the world

yes?

so I do not understand [15:39:18] <Zyxer> Maybe I am misunderstanding how libera works or IRC, aren't all channels on libera hosted on libera server?

And everyone being on facebook is not a problem? Because facebook has servers all over the world?

what is 'libera server'?

service liberad restart

oops

haha

I think you misunderstand what an IRC network is and how it functions

LiberaChat is not 'a server'

Someone is running libera. I mean, literally the issue with freenode could happen to libera, no?

Yea ok, liberachat is not one server.

Facebook is not one server either

the issue with freenode was a hostile takeover of an NPO I though.

Zyxer: Libera is a *network*

But it is one actor controlling all libera servers

no

Hmm. Ok

How not? So libera cannot be bought byu a mallicious actor?

No one can sell libera?

Title: About Libera Chat | Libera Chat

Zyxer: this is why Libera was created in the first place

please educate yourself

deacon426: I see, I thought it was similar to the freenode network

it is freenode 2.0 sort of.

Zyxer: you're in the right place

Remilia: I am trying. Remember I did not grow up with IRC or any of that. It is still new to me

My frame of reference is very different from IRC "veterans".

I just don't like centralisation. Sometimes it is needed yes.

But hosting an email server on FreeBSD or whatever, is also susceptible.

Doesn't mean I am comfortable having around 100% of all email servers being owned/run by 2 actors

I understand, in the dilemma of 'convenience vs centralisation' you choose the radical former

latter*

errrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr

From my perspective everything that is centralised in any way is that just to collect data or control your speech.

You meant decentralisation, lmao

yes

you hate being able to join multiple projects' channels on a single non-profit IRC network

most people prefer that convenience

that is the difference

basically, your approach is kind of against the principle of IRC itself

an IRC server lets you create as many channels as you want to, for any kind of topic of interest to you or others

by design it means there is value in people converging

Maybe. I am trying to learn. Remember that what I am experienced with bans ppl completely for claiming the government doesn't have your best interest in mind. And those that do not ban all collect data. That is the environment I grew up in

Remilio: <by design it means there is value in people converging> ok that was a great point actually

Thanks for encouraging me to read on liberas site and linking it. Most things I read about are due to controversies, so my look on things is probably overly pessimistic.

Ok something actually related. Can I read LUKS on FreeBSD?

I get ambiguous answers when googling

there is freshports.org/devel/libluksde no idea if it works I haven't tried it

Title: FreshPorts -- devel/libluksde: Library and tools to access LUKS Disk Encryption encrypted volumes

Thank you so much!

That's just a header file... ? Aw well, a pointer

the pkg-plist says it should contain bin/luksdemount (and other stuff)

Oh, thanks. Guess I was thinking in terms of linuxism

I tried to install luksdemount but to no avail

lmao

the pkgname is libluksde and you can give pkg the portname devel/libluksde too, you can't install an individual file from a package

Hi, again, how can I tail console? sudo tail -f /dev/console fails me. I get no output

I am trying to run xconsole in a terminal, and not in a separate window. The man page for xconsole mentioned -daemon, but it just detaches from the terminal so I can use it for other things

Zyxer: i usually enable console.log in /etc/sysctl.conf, then you can just tail /var/log/console.log (and it also saves the log, which is useful if you need to check earlier events)

So console_log_enable=YES?

no, edit /etc/sysctl.conf, uncomment the line starting "console.info ...", then touch /var/log/console.log; chmod 0660 /var/log/console.log; service syslogd reload (you have to pre-create the file because syslogd won't log to files that don't exist)

er, /etc/syslog.conf, not sysctl.conf

Oh, syslog not sysctl, yea I got very confused. Googled and they were all like "Just uncomment obvious line lmao"

yes sorry, i just made the same typo 3 times checking those instructions... apparently i'm asleep today

It is a day of rest afterall

Is a reboot needed or can I reload syslog.conf?

no reboot, just reload syslogd

It doesn't work. Perhaps I am missunderstanding something.

I want all things that output in xconsole to be read in terminal, with like tail -f or something

I reloaded it, and stil lempty file after unplugging and replugging USB

So I restart it, now it isn't empty it just says syslogd exiting on signal 15

But unplugging and replugging usb doesn't give any output there

Good news

tail -f /var/log/messages

Zyxer: ah, it only shows console output from userland programs, not the kernel (because that goes to /var/log/messages). if you want only kernel output, you could add a new entry to syslog.conf with "kern.info ..." and put it in /var/log/kern.log or something

/var/log/messages is good enough for me. Is that bad practice?

oh, no. i thought you wanted console output specifically, /var/log/messages includes a lot of other stuff as well

Oh, I see. So uncommenting console.info line and also make the kern.info output to same file will get me only console and all console output to that file?

Yes, that worked. Thanks so much!

Well, luksdemount is failing for me. Invalid master key size value out of bounds. and unable to read from volume handle, unable to open volume. Bah, guess I'll plugg in my loonix to transfer files.

for recent kernel messages you can also just dmesg

Yea but I wanted to... Hmm... I guess dmesg|tail -f might accomplish what I want?

Zyxer: yeah tailing dmesg is SUPER useful

I can tail it with -f flag?

try it, find out

Yea true lmao

No it didn't work

are you trying to pipe tail into dmesg?

or other way around rather

dmesg|tail -f

try $ tail -f dmesg

tail: dmesg: No such file or directory

tail takes input from stdout or a file afaik

will watch accomplish the same thing?

or maybe just $ tail dmesg

No, tail -f is superior because the watch workaround, the loop that accomplishes linux watch, spams every 2 seconds and then you wanna scroll up to check something then buffer is full of the watch workaround

I would read the man page for tail

dmesg isn't a file, can tail command really take a command as argument?

I think the issue is dmesg command ends

you can tail dmesg, I do it all the time

yes, tail it

tail -f doesn't work with dmesg because dmesg just prints the output once and exits

tail -f does something else

Yea, my suspicion and reason I didn't try to tail -f the dmesg, otherwise for one time thing latest dmesg lines I tail it without -f

Zyxer: it's $ tail -f | dmesg

"tail -f | dmesg" doesn't make any sense since dmesg doesn't read any input

ah

read the file then, not the binary?

tail -f /var/log/...

I never said you should use tail with dmesg, just that you can check the message buffer using it

better -F instead of -f as -f sticks to the file (inode) and -F sticks to the file name i.e. when /var/log/bla is renamed and zipped to /var/log/bla.0.gz and a new /var/log/bla created, -f would stop, but -F would continue to tail what is written to the new file (with different inode) named /var/log/bla  (see “man 1 tail”)

that's awesome thanks Kalten

:-)

deacon426 you might want to make use of /etc/syslog.conf (or even /etc/syslog.d/) e.g. a line “*.*  /var/log/all.log” and than “service syslogd restart”

I was literally just in /etc/syslog.d (for other reasons) but still oddly weird you just mentioned that. Will do, thanks for the clue

I'm already there might as well :)

boy, the bandwidth I'm getting from the official download site is all over the map.

12 MB/s, 2MB/s

BIND is supposed to send notifies to secondaries for changes from dynamic zone updates (nsupdate), right? i'm sure this used to be working but somehow it's stopped... i wonder if it's because i enabled inline dnssec signing

ah okay, you aren't supposed to use inline-signing on dynamic zones, it works as expected without it

does anyone have a recommendation for hardware for a home/office router with a CPU fast enough to handle 1Gbps of PPPoE-encapsulated traffic (with a few ipfw rules)?

unixwitch: something APU?

unixwitch, Uhm... There are many different machines out there but I happen to be using the small ram zimaboard.com here. 2-3 watts. It's an inside-out system with storage on the outside.

Title: ZimaBoard - World's First Hackable Single Board Server

But it has 2x GigE NICs making it good for my use as a router/firewall.

meena: from what i saw from other people's testing, the APUs' CPUs are too slow to do 1Gbps PPPoE

(PPPoE is awkward because it requires everything to be serialised to a single CPU core, so it can't benefit from SMP)

I am not using PPPoE so I can't say from experience how that would work.

I have an unused Acer Chromebook c83-431 and was looking to get FreeBSD on it. I think it is an Intel Atom CPU. Which kernel would that be ?

rwp: that looks interesting, thanks. do you happen to have any real-world routing throughput numbers at all, even if not PPPoE? (no worries if not, i might email them and ask...)

unixwitch, I just ran www.speedtest.net here and got 930Mps down, 878Mbps up, ping 4ms through the router to a neighboring town that is also on the municipal fiber.

$120 for the basic one... perhaps i'll just buy one and test it, i'm sure i could find another use for it if it's too slow. interesting that the lowest spec actually has higher max frequency...

It's typical for a dual core to be more speed daemon than a quad core. Double the bus load then must reduce the speed.

Same thing with max ram in a lot of systems. More ram means more capacitance loading which means more wait states for a ram cycle.

I'm looking at man.freebsd.org/cgi/man.cgi?fdisk and right here it says This command is obsolete. Users are advised to use gpart(8) instead. why is that ?

Title: fdisk

jb1277976, Do you see any mention of GPT anywhere in the fdisk manual?

let me look i just landed on that snippet.

nope

jb1277976: because gpart (part of GEOM) is how partitions are normally managed on FreeBSD, so that's what you should use. fdisk is a relic from the old MBR / BSD disklabel days.

ok

(you can still create MBR partition tables with gpart if you really want to, so there's no reason to use fdisk at all... i'm mildly surprised it's still shipped)

And if fdisk only handles MBR and does not handle GPT then it can't handle the new preferred GPT partitioning.

Thanks. looking at the examples it makes sense

It does seem like a distractor to have fdisk still in base. People coming from linux would get sucked into it. But that's why they have a prominent warning in the man page and a pointer to gpart.

Love the bugs section quote that it does not handle partition types introduced since MS-DOS 6. Okay then! It's definitely a museum piece command.

rwp: did you notice the HISTORY section which mentions it was ported to 386BSD? :-)

quite possibly it hasn't even been updated since then, too

unixwitch, I did see that! I was surprised it didn't have any subsequent history of ports. But probably because it hasn't been!

it might be the only thing left in FreeBSD which actually came from 386BSD (and not 4.4BSD) so it's technically not wrong if you consider FreeBSD a fork of 386BSD, i suppose

this reminds me that i had a vague desire to port 2BSD to i386 at some point... although nowadays it would be more trendy to port it to ARM, i suppose

unixwitch: quick, port it to armv6 before it dies too

i'm sure NetBSD will support ARM for the next 30 years for the 3 people still using it :-d

s/ARM/armv6

true, true.

awhhh, buildworld failed 0x0.st/H324.png

can freebsd see inside of a gpt label that was created by openbsd ?

jb1277976: it should be able to do that, yes

mtrkss: CURRENT? or a release?

meena there a doc showing all the fstypes i can use to mount it ?

jb1277976: lsvfs will show all supported filesystem types, if that's what you mean. if you want to mount an OpenBSD partition, it will be formatted with FFS (UFS) but i'm not sure if FreeBSD supports OpenBSD UFS off hand...

ok

i *strongly* suggest mounting it read-only if you try that, so it doesn't damage the filesystem

ok

The BSDs do not support each other's FFS/UFS partition. I've been through this exercise several years ago. The common file system between the BSDs ? ext2/ext3 :-)

unixwitch: stable/14

sorry for late response

s/do not support/do not have inter-operability/

did a git clone again, hope it works now

mtrkss: curious. i just built stable/14 yesterday (with the p2 commits) and it was fine

wait no. that was releng/14.0

here's the /etc/src-env.conf i'm using 0x0.st/H323.conf

right now it's building fine

oh god i'm getting a lot of warnings on openssl

shouldn't SSE2 be default by now?

it's been around for about… as long as i have… used computers.

is there a base utility that will generate crypt(3)-format password hashes?

or a ports utility would be acceptable too...

unixwitch: dovecot

it has `doveadm pw`

(as an example)

also there's Perl

Remilia: feels a bit heavy to install all of dovecot for it's password utility

would rather not install dovecot (thanks though) but it turns out slappasswd can do it: slappasswd -c '$5$%s'

true

i wonder if it's worth submitting a PR with such a utility... seems like it might be generally useful

wait, can't openssl do this

hmm apparently not

Ugh...

Copied around 700GB using rsync from a linux (skill issue, I couldn't use luks drive on FreeBSD)

Now

30GB RAM in use

ZFS caching is fun (the copy is already completed)

did they remove -crypt from `openssl passwd`

Remilia: ah, `openssl passwd -6` does do what i wanted

unixwitch: there's pwgen in ports

unixwitch: but that's SHA512, not crypt

meena: i used pwgen all the time but i don't think it can generate crypted hashes?

oh, right, missed that bit

Remilia: it generates SHA512 hashes in crypt(3) format: $6$LwohPYPLMYR9X/jY$cAiGEZwKwkwJCmC0M7dYwMND1bpXRWTIP8so1y1GTgslWvOOTBOE3.bdcIwteyRWEUTjYbet6yhHBG1o7p0d00

I thought you wanted the crypt algorithm

well, i want something that crypt(3) will accept, and it understands $6$ hashes

you never said what algorithm you wanted so

Zyxer: it's normal for your filesystem to use all available memory; it will free memory as needed. the alternative would be having memory sitting around not used by anything, which is a waste of money

Zyxer: all filesystems (more or less) work this way

unixwitch: i think OpenBSD has something

I know ZFS does that

And nothing weird, I was just suprised that copying files and dirs would make it cache everything

At boot only 1337MB RAM is eaten for cache

Or, around 1GB, 1337 is total RAM

But I haven't heard of any other filesystem caching that much on unused RAM

all filesystems work like this. including UFS on FreeBSD, and ext4 on Linux. it has to cache everything read from disk because that's how the cache works

Oh, yea

right, OpenBSD has encrypt(1).

but ZFS takes it further afaik and caches more than "necessary". Like no other setup using dwm eats over around 300mb RAM at boot

It starts caching before you use the files or anything

Zyxer: it really doesn't... if booting requires loading 300MB of files from disk (like /bin, /usr/bin) then you'll have 300MB cached after boot.

it won't just randomly cache things that aren't read from disk, because that would generate i/o for no reason

ZFS does tend to be a little more insistent on keeping things in the cache (since it has its own ARC cache instead of using the normal filesystem cache) but that just makes it more efficient

I know, it literally is a feature. "Unused RAM is wasted RAM" actually applies in ZFS cache

I'm not worried. I just never seen cached RAM not getting freed up when the files copied to the system using scp or rsync

My OpenBSD server has like, 50MB more due to cache after huge files copied to

Hmm... On _other_ systems I am used to using the Apache utils htpasswd command to manage those hashed passwords. I don't see it available on FreeBSD. Please say it isn't so!

so apparently `make installworld DESTDIR=/jail/whatever` doesn't produce a functional jail (due to missing files like master.passwd)... what else do i need to make it work?

unixwitch: I think you can use distributeworld to generate distribution sets

but not sure if they include that stuff

distributeworld + packageworld would give you stuff like base.txz kernel.txz etc.

Ok so, I am sorry if this is a dumb question, but why doesn't du -a list all files in specified directory? Could the issue be that the folder is a symbolic link?

du -aL

to traverse symbolic links

Thanks

Remilia: i was hoping i wouldn't need to use bsdinstall since i already have a built world in /usr/obj...

I got 13239 memes locally. Although that includes directories since I didn't filter for only files.

rwp, "pkg provides" gives several possibilities other than just installing Apache.

V_PauAmma_V, Are you saying that htpasswd is included in the main Apache package?

Time to tag them all (already tagged maybe 20% or somethign)

I really expected it would be in a separate package of utilities. (Like on, ahem, other systems. Me glances shyly away.)

Technically, I'm not. But "pkg provides" definitely is. Along with a Python htpasswd implementation.

Zyxer, If you want to push _through_ a symlink then put a trailing slash on the end. For example "du foo/" with a trailing slash. "foo/" is the same as "foo/." and dot is the current directory.

V_PauAmma_V, It does seem to be bundled in the www/apache24 package. cgit.freebsd.org/ports/tree/www/apache24/pkg-plist#n4

Title: pkg-plist « apache24 « www - ports - FreeBSD ports tree

Is this a ramification of the problem that one can't make multiple binary port packages from one port source? Do I have that right? And so at the moment it's all together even if splitting them out would make sense?

rwp: Thanks

unixwitch: Well after boot and me check RAM, fresh boot, it is 1337MB

sometimes 1338MB

rwp, that's been possible for a while, using flavors. See eg, packages workrave and workrave-xfce, both from x11/workrave.

Then this would be a perfect example. Those are useful utilities to have without needing the entire server installed.

I should install that server package and verify. I am one of those people on the Nginx side of things and so avoid Apache all other things being equal.

V_PauAmma_V: no, that isn't subpackage stuff there and flavours shouldn't be used to emulate subpackages

I don't understand the distinction you're trying to make.

rwp: Indeed. I've been building apache24, even though it's broken as a server (at least when using the php module), just to get a couple utilities, e.g. logresolve.

Or perhaps I don't understand what subpackages are.

The other utility I use is "ab" apache-benchmark which is good for stress testing http servers.

workrave and workrave-xfce are two build of the same software with differnt options set (flavours); the wanted behaviour is to build the software once and split specific files (htpasswd) in its own package instead of creating only one package from that build (subpackages)

Well something like libnghttp2 and nghttp2. Both are built from the same source code, one is the library and the other are the utilities.

Are they built in one build or built in two builds? I think that is the question being pondered.

Two builds.

JFTR but ab and htpasswd are both in the apache24 package as expected.

Two builds. Which is more build work than if it were one build that then produced two packages. I think that is the point under discussion.

It's still on my learning path to learn how to build packages. I haven't ever done it and don't yet know the details.

Life and time are what keep everything from happening all at once.

I just had a look at the apache configure --help info. I don't see how you can not build the server binary. So you'll end up building it, just not packaging it.

Are subpackages already available? I don't remember seeing them mentioned in the Porter's Handbook. Did I miss it?

yeah, seems like libnghttp2 / nghttp2 is a good candidate for subpackages once that lands (no idea were informations on that are, afaik there is support in poudriere for it and now it needs testing or something)

*nod* Thanks. I can see the point of wanting to use less build cluster resources.

It's also the huge packages which can benefit the most and the huge packages are going to be bigger build resource hogs.

The lack of subpackages for FreeBSD has puzzled me. I'm used to MacPorts which has had subports for a long time. I maintain the nghttpd2 over there github.com/macports/macports-ports/…b52cd6b7ed26b2/www/nghttp2/Portfile

Title: macports-ports/www/nghttp2/Portfile at 54f3a3b23e8d07573e6c8b1783b52cd6b7ed26b2 · macports/macports-ports · GitHub

Note: there are some dead subports there, as python support has been removed upstream.

hm, that subport stuff seems more like flavours, as it changes configure args

Yes an no. The python subports are, but nghttp2 = libnghttpd2 and nghttp2-tools = nhttpd2 over here.

and I don't find where anything says which file should be put in which package

nimaje, anyway, thanks for the explanation.