humph! how do you split a line in ed

you can join

hey, running 15.1 w/amdgpu (RX570 iirc) after a while I'm getting corruption of Xorg video, seems to be triggered randomly by firefox playing videos? any ideas where I should start troubleshooting?

can you play videos not-in-firefox?

oh, i can play videos in firefox for a while

i'm thinking maybe it's overheating?

but it's not very stressful

that's what i was thinking

no i mean can you play videos not-in-firefox and does that corrupt your video after a while

ooh good idea to try

like using mpv or vlc or something

i was playing a video in YT Music, and had another FF window open, opened a new tab, and BAM corruption

but i'll try mpv

nyarr, it doesn't seem trivially easy to upgrade to CURRENT using pkg-base

hmm... unplugged my unused DualSense controller (PS5), no corruption yet :D

rdr: ???

I don't know if it's related

i want your problem to make more sense than that :P

haha me too

but stranger things have happened

it is true

... ok, off we go building CURRENT with -j8

rdr: how much ram?

All the RAMs.

rdr: we've fixed some problems in main with page cacheability that could be triggered by driving the system to lowmem so that it tries to reclaim pages for the various subsystems

rdr: i found a way to trigger it reliably, but I never could both trigger it and observe any foiw-up shenanigans

follow-up

i think mostly because my testing was too artificial, I had to drop usable memory pretty low but then it was tough to daily drive

anywho,

sudo kill -1 `cat /var/run/kevans.pid`

jails/capsicum folks, would appreciate if someone has the time to look at reviews.freebsd.org/D58082, thanks!

kevans: I have 16GB of old DDR3(?) RAM, 8GB VRAM in the AMD RX 570 card

hmmm, just caused it again. i'm gonna try and see if a different browser or even a change of firefox settings will help

hmmmm seems to reliably happen when opening a github link

oooh picom wasn't running :O

kevans: I switched from uefi boot to bios.

problem is one of my other servers uses seabios, which is bios-only

and I did want to replicate this on that...

anyways I can simply switch it back, and hopefully it just works

however not getting this working with bios boot is going to be problematic

worth asking on bugzilla/mailing list to see if anyone else has done it?

rdr: nice card, I have an rx 580

sitting in a box somewhere

rx 500 series were powerhouses and were so cheap!

those were the days!!!

polarian: too true! I hope having picom enabled solves the graphical corruption stuff :<

im starting to wonder if its even worth emailing the freebsd mailing list or just to put issues onto bugzilla and hope I dont make a fool of myself

I cant get IPv6 working inside vnet jails, host drops packet unless I ping the jail and then it works fine

it makes no sense ughhhhhhh

I am considering just posting it to bugzilla and if I make a cock of myself by missing some essential networking knowledge, then so be it

I honestly cant figure out _why_ its happening

polarian: Do both! Mailing list first and then a bug after a day or so if you don't get useful bites.

I hate making a fool of myself though :c

at least if its IRC most people forget when it happens :p

polarian: It's the only way to learn. And we remember. We're like Pepperidge Farm. We remember.

memory through embarrassment :p

so the host has a routable ipv6 and it's not routing to the vnet?

polarian: The shame gives you extra neural paths to the learning.

rtprio: come to think about it explaining this is gonna be hard

so I have public IPv6 addresses being routed via LLA

all the routes are there

when the jail tries to ping say google.com

or anything, even the host addr

it doesnt work

it can ping the host LLA and then it will work

for some reason a packet must be sent to the LLA first...

polarian: Have you done a packet capture?

mason: yes

echo request goes out, echo response comes back to the host via the phy, and then is dropped

oh, interesting

its not passed back to the jail

BUT

if I leave that ping going

then ping the jail

boom, it now works

messed up arp table?

this is IPv6, it would be NDP

and I dont think so... NDP looks fine

What else could it be though?

Unless it's some kind of funny stateful firewall issue.

mason: exactly, see why its hurting my head

pf is passing quick on inet6

its the first rule, no log to pflog0

polarian: Does that matter at all if it's going through NAT though?

Or is it not NAT?

mason: no NAT :)

kk

its IPv6, I am routing addresses over LLA

kk

I might write it to freebsd-questions

issue is I am currently scratching my head with trying to figure out why bhyve is being pissy

I swear I waste so much time trying to get things working on freebsd :p

LLA = local link address, like fe80:: ?

yes

is "public IPv6 addresses being routed via LLA" a common thing? because it sounds very weird, like something a lazy cloud provider would do

rtprio: How else is IPv6 routed? Honest question - I'm, again, an IPv6 noob.

the provider would give you a /48 or /56, and you'd put your jailer on :1 and route ipv6. give out addresses with rtadvd

or manually if you were so inclined

rtprio: routing IPv6 over LLA is how you are meant to do it lol

thats the joy of IPv6

unlike IPv4 where you have to keep hopping packets over a larger subnet, you can just hop packets over LLA with IPv6

I cant for the life of me to get OpenBSD to boot in bhyve either, tried 7.8 and 7.9

> wrmsr to register 0xc0011029(0x3) on vcpu 0

after this line it just exits

exit code 0

makes no sense

vortexx: iirc you use OpenBSD in bhyve, you got any idea why this could be?

(sorry for th eping)

iirc i had to set the console to the serial port

rtprio: yeah I completely forgot that

I was just thinking that haha

lemme test that real quick

I have been playing with tooo much arm64 openbsd, where tty is default

it's kinda silly it doesn't default to tty either

indeed

but freebsd doesnt either xDDDD

freebsd makes an attempt to autodetect

and the loader menu makes it less of a pain in the ass

oh it does O.O

shocking

ive had such a pain in the arse with freebsd and serial

but it works just fine when doing it via bhyve

talking about that I need to bugzilla decrypt geli with bios boot via serisal

because I will need it for another server

rtprio, In IPv6 the expected route is using an LLA route. But IPv6 1= IPv4 and so almost every we learn from IPv4 and try to apply to IPv6 has been changed. Negative transference of learning.

In IPv6 the expected route is an LLA-on-a-device. So for example on a server with an Intel NIC then in /etc/rc.conf setting ipv6_defaultrouter="fe80::1%em0" is typical, where fe80::1 is the LLA router class address any/all routers on that device segment.

It took me a while to realize and understand that in IPv6 the LL address is always associated with a device name. So it is never just fe80::1 the router LL address but fe80::1%em0 down that device segment.

you can think of LLA as /31 in IPv4

it was messy and well, doesn't work full stop on FreeBSD

LL is probably one of my most loved features of IPv6

i get LL but i guess i've never heard them just magically translate to a real public ip address

alright setting up my poudriere build server rn :)

The LL is not a public address. It does not translate to a public address. But for example fe80::1%em0 is a LL address which addresses any and all routers that are on that network segment. It does not need to be public. It's like 192.168.0.1 for a router address which is also not public.

rwp: when I compared it to /31 I assumed that you would be using RFC1918 addresses :)

But in the case of 192.168.0.1 it needs to be routed using the routing table and normally works because the device has an address on that network and so packets will route there.

the way you would route a public IPv4 over a "Link layer" without IPv6 would be a /31 range in RFC1918 and then add the IPv4 public address to lo, however on FreeBSD there is no src priority

unlike Linux

so this doesn't work full stop

was the jail polarian setting up ipv6 lla local only?

rtprio: depends what you mean

the epair host side is LLA only

OVH and some other hosting providers like to assign a gateway that is not routable, it's in another subnet, and so packets there have no way to route there, requiring a host route to be added to the host table on that network segment. IPv6 bypasses that by including the device in the route.

the epair jail side, has the IPv6 address, and a LLA (obv)

rwp: this isnt OVH, this is my own network

and was the jail supposed to be be publicly accessable?

yes

polarian, I never said your network was OVH. Why would you think that?

ohhh

you werent replying to me

or you were

idk

im so confused :p

I will write up the issue im having with IPv6 and jails to the ML shortly, because its hurting my brain to comprehend it myself, so nobody in an IRC chat will understand either

so if the jail is supposed to be be publicly accessable, and it only has a lla address, then there has to be some translation

or it wouldn't be reachable

Correct.

If the jail is supposed to be publicly accessible then I think it should have a public address. Or NAT. But a public address is just more sensical.

which is why i was wondering why use LLA in this case

rtprio: no because the packet is ROUTED over LAA

LLA*

the src header is still the IPv6 address, the route simply says send it to fe80::1%jail0 for example

the scope is important here too

You have to transport your mind back to the 1990s when IPv6 authors were developing things. They didn't like any centralization. So almost everything is decentralized and configured on the client. By using an LLA for the route the client can pretty much do it independent of any centralized DHCP daemon or whatever.

I usually use fe80::1 for host, and fe80::2 for jail/vm

well, yeah, i can not fathom how that would work without the jail having a real ipv6 address

but rwp claims it should work

rtprio: there is an IPv6 address in the vm

you have two addresses

i guess i don't see the point

rtprio, I didn't say the jail did not need an IPv6 address. I said the jail routes over an LL address. I think the jail needs a public IPv6 address. Or some complexity such as NAT.

you have the LLA address, and the IPv6 address attached to the vm, host side however, I have only a LLA address on the host epair, this is because the host already has an IPv6, why would I use more on the host?

instead I can route IPv6 to containers and vms using LLA and static routes

polarian: that might be true, if it worked

rtprio: it does work

and its how IPv6 was designed lol

polarian: this conversation started when you said it only works if you ping it first?

yes...

so it doesn't work

well not flawlessly :p

but this isn't down to the theory being wrong

your truck "Doesn't work" if you have to open the hood and start it with a screw driver

can you drive to the autoparts store? yes. but it doesn't work

alright fine, if you want to be fussy

it doesn't work

but it should, because this is expected behaviour

happy?

I am reading that your configuration is not working.

so its either A: I configured it wrong or B: theres a bug

i've done some wacky thigns with ipv6 routing, and its' worked on the first try but none of it involved lla

you dont route over LLA?

yikes.

you missed the part where i said it worked

sure it works not to use LLA

in this case, these hosts were not local

but you really should, the added benefit is you can change the IPv6 addresses and not the routes. :p

hm?

It's not required that the route be over an LL address. That's just the expected configuration. But routing over a public IPv6 address also works. And again some hosting providers only document that configuration too. (I won't mention one so that you don't think I think you are hosting there.)

it's hard for me to give a shit about ipv6 right now since my network is in shambles and my provider doesn't know what it ipv6 is

slightly offtopic but who here uses linkedin? people keep bashing me for not using it and saying its the only way to get employed these days

I feel deeply uncomfortable putting my face on a profile, with all my info fed right into a big tech slop machin e

i have an account, but linked in is insufferable to use these days

my cv gets ignored, my emails get ignored, my applications get ignored, and people tell me its because linkedin is the ONLY way these days

i can't say that's true, the market is shite

all cold applications suck

welp I guess I can make one again, leave it up 3-6 months, if I get hired then cool, if not then maybe I can consider deleting it after.

I know this is more of an openbsd thing to say, because freebsd people arent insanely paranoid, but linkedin is so good for being socially engineered

people attack your connections, get info on you through that, and then exploit that info to pwn you

if I cant keep myself secure, because $JOBS demand not to, how do I keep my services secure? is security just impossible!?!?

also for the record, asking me to use linkedin is like asking me for nudes... genuinely.

i did see a llm script to brute force apply for any and all jobs matching particular keywords, via linkedin

At the age of 21 I have turned into a fucking 80 year old I swear

I just moan about all modern stuff :p

I recommend getting offline for a while. Do your networking in real life. This is the magic sauce these days.

polarian: Nah, not 80 at all.

boru: literally just came back from a cybersec conf

People hiring (myself included) aren't going to go though the 500 resumes they get after 2 days. There's just no time for that. You just get lost in the noise if you apply the classic way.

boru: I know that, but nobody is hiring full stop in person either... unless I hit a gold mine by asking the right person at a conf

Networking is king again.

I have met many MANY people :)

idk I feel like networking in open source is a waste of time for employment

most people I bump into are either unemployed or funded through some nlnet fund

Probably. If you're looking for a paying job, you should probably look for companies that pay you to work.

boru: harder and harder to find these days

maybe this is just a UK issue?

I heard BSD folks in other countries, even younger ones have no issue finding work

No, it's not really. It's the just methods to get a job that have worked for a coupled of decades don't work anymore.

So, change tactics.

I have tried all the tactics lol

It probably feels that way, but really, networking is king these days again.

asked IRL, I have spent ages optimising my CV to get pass HR, different platforms etc

well networking is always best

Therein is your problem, you want to go direct to engineers, not through HR.

I have shot off tons of emails to firms which I match best, and none of those are hiring ever...

boru: also I find it rather humliating crying at conferences about work

Well yes, it should feel that way. No one is going to hire someone that cries about work.

but excl self employment (which isnt enough to put food on the table) I havent worked in 1.5 years now...

well I have non-profit positions but those arent hard to get

people dont say no to free labour (usually)

boru: alright I dont cry... but I feel like even bringing it up can be a big red flag

That's nice, but it's energy wasted on not finding a paying job.

my last three jobs have all been word of mouth

and I get uncomfortable when people are like "oh so what do you work as"

and I panic and try to dodge the questi0on

those resumes are on the top of the 200-500 others

welp just gotta keep trying ig... everyone I know got hired via being messaged on linkedin

and I have resisted it for 1.5 years, but at this point, it cant hurt

contracts have dried up too

Yeah, you're already fighting a losing battle there. Better off focusing your effort on what you want to do; "I hear you guys are hiring for X. I can do that with my eyes closed. Want to grab a beer?"

boru: bribe with beer you say? :p

get em drunk enough to think you are actually smart!

xD

Don't be obtuse.

(im kidding btw)

If you're good at what you do, then present that face, not "I am desperate for a job please hire me"

I dont do the latter though

I just get uncomfortable about the topic

It's your opportunity to interview _them_ to see if _you_ want to work there.

the only thing worse than having a job is looking for one...

lol

Well, the sooner you get over being uncomfortable, the better. That's a you problem.

I swear employment is like rocket science

so many ifs and buts, and conflicting info

i installed freebsd yesterday, got a job offer today 300k starting ;)

just kidding, it's hard out there

"I installed FreeBSD and my wife loves me again!"

:D

Yeah, on a serious note, the market is shite at the moment. I understand that it's demotivating, but honestly, the jobs are there, you just need to change tactics.

People hiring are also exhausted with all of these LLM applications from very specific places. A lot of jobs (IME) don't get listed anymore unless there's no choice.

Network first, go external as a last resort.

Yeah, sorry, I'm also off topic now. I will be quiet.

yeah sorry about that

Final say: don't waste you time on linkedin. It's for muppets.

oh for fuck sake, Linkedin has ID checks now?

and it requires a mobile phone? this aint going to work on a custom rom

boru: what more do you want to network, I am regularly at events and meet many MANY people, you know how many have been able to hire people? 0

not a single one

actually thats a lie

there has been, I just dont talk to them as I have no way to approach them

"Is your hiring manager here? Can you introduce them to me?"

That will be 500€. My fees are quite reasonable.

If you want to streamline things, talk to the people actually doing the hiring, not some rando.

It's good to talk to engineers to find out what it's actually like there, though.

You don't want to jump from the frying pan into the fire, and all that.

right

so hiring managers put a big sign on the head and say "HIRING MANAGER, COME HARASS ME"

in fact I lied I spoke to one on Sunday, and he said outright, no hiring manager makes themself known

Obviously. But people will introduce you if you play your cards right.

this feels like playing a game of chess

After having linkedIn being the only social media I was on for the last ~10 years or whatever it was, I deleted my account last year. Its such a waste of life and devoid of any real value.

checkmate being speaking to a hiring manager lol

got steam running yay

You're right. It's too hard. Best to give up.

oops crashed

boru: well its not like I have a choice eh? :p

rdr: you using linuxator?

yes'm

I hope you jailed steam?

it likes to shit over your system :p

it's it's own user at least

ah that works

just followed the instructions

that tends to work most of the time ;)

theres instructions?

I wish I read those for once :p

haha, combination of gemini feeding mostly true (but somewhat wrong) info and then the package when installed gave the correct info

Hi

I have little problem here, for 3 years now.. any ntfs partition I mount in my system, just get broke when copy/move files to it.

and need to reformat to get it workable again

I mostly use ntfs 3g

i didn't know ntfs-3g supported writes, i've only used it for reads

i would suggest, not writing to it anymore

I used to think the drive is died, but I tested another drive 100% health still the same

now, I will just use my winxp vm to copy files

but I need fix, to safely mount my new drive in rw mode

rtprio: ntfs-3g works fine for writes

I have written to ntfs-3g befor eafaik

unless I am misremembering

polarian: then could you help me?

mosaid: did you mount with rw as the option?

yes

and you did install this port right?

iirc the driver and the tool are in different packages or something

I believe you need ntfs and ntfs-3g

but ntfs pulls in ntfs-3g as a dep

yes installed them well

hmmm

not too sure sorry :/

I remember it working for me

im not really smart enough for this :p

it gives mostly i/o errors, which gives the feel of failing drive.. but after deep test everything is fine

whats the smart output?

can you show the output of smartctl -a

you are free to redact identifiable data like the serial number

and need reformat it to get the partition in workable state

Ok

but I used two new drives still the same

impossible to be the problem in drives not ntfs 3g

polarian: see dpaste.com/3JRVRTS8A

yeah disk looks fine

ntfs 3g is the problem

mosaid: are you dualbooting with it?

rtprio: no

those of you who use unbound as a local resolver; do you also have bind or some other primary dns or do you have a lot of local-data lines for some local names?

Look like I destroyed my drive using dd :'(

first sector now unreadable

rtprio: I used NSD for my local DNS authority and unbound as the resolver, which either forwards to NSD for local resolution, or DoT for external.

wait what were you doing with dd ?

luser: i guess i'm lazy and kind of don't want a second service?

rtprio: and that's fine.

i haven't cleaned up bind yet, but haven't fully migrated yet either

I have considered consolidating to unbound, but I have multiple NSD services for redundancy and it syncs between, so I only update the primary.

i manage my zone files with puppet, so i don't have to use primary/secondary of the nameserver

That's a good method.

rtprio: I just use bind here. Leveraged kea to keep a dynamic zone updated.

ah, i forgot all about dhcp/dns updates

I am so stupid :P saw a howto wipe disk, so I can partitioning on solid ground.. he used dd to wipe begining and end

for ntfs

but due to my low expernice in drives world, I am made wrong

uh oh, why did you need to wipe it? couldn't you just reformat the drive?

I used my winxp vm, but as I said to when I used ntfs-3g the drive completly corrupted alot of I/O errors

I tried just format didn't work, so need to wipe the disk

oh

sounds rough

drive is so healthy 100%, just ntfs-3g

Took a look to see if an arc a310 was supported in fbsd and it seems arcs still have issues

But then again it’s using the drm-kmod stuff

Wipe the beginning: dd if=/dev/zero of=/dev/da0 bs=1m count=1

The end: dd if=/dev/zero of=/dev/da0 bs=1m oseek=`diskinfo da0 | awk '{print int($3 / (1024*1024)) - 4;}'`

Is this right

see also chieflemming.wordpress.com/2020/01/…-for-ntfs-with-freebsd-and-mac-os-x

talking about some bug in ubilo