can anyone get github.com/open-telemetry/opentelemetry-collector to build on freebsd?

Title: GitHub - open-telemetry/opentelemetry-collector: OpenTelemetry Collector

Hey if I have the latest binary pkg of ifstat, but I want SNMP support, which for some reason was disabled in the latest, what is the cleanest way to update just that with a version supporting SNMP?

Not sure who thought it would be a good dieal to change ifstat pkg to no onger include snmp, unelss it was a security thing.

good idea... (not one)

Hmm, are you talking about net/ifstat? SNMP support in that has been optional since 2004, per cgit.freebsd.org/ports/commit/net/i…524983b12a7223d8564ce24e60f3400ed9b .

Title: ports - FreeBSD ports tree

s/optional/a not-default option/

can anyone get fluent bit with the http or opentelemetry outputs working? they both break for different reasons for me

V_PauAmma_V: Hmmm. Well I had it, and then after a pkg upgrade the snmp suport in ifstat went away.

Ah, look slike I had ocmpield it myself before, and somehow package upgraded it. I probably overwrote the old package with the custom compile, and then pkg didn't know that and upgrade it to the new version, which has no snmp. Thx

Damn my typing.. I had compiled it before...

I use it to display BW usage in my screen status line

Updaded from github and recompiled... Deleted the pkg and reinstalled manually. All good.

is there something like github.com/clopez/dellfan to disable BIOS fan control in FreeBSD?

Title: GitHub - clopez/dellfan: User space utility to control the fan speed on Dell Laptops

ah it has nothing to do with my lid. If I switch between ttys and back to X it also swaps out the video driver for llvmpipe

(ctrl-alt-f3 then back with ctrl-alt-f9)

Erhard: maybe you can ask to have it converted to a FLAVOR option so you get both

o/

I just installed FreeBSD on my AMD/AMD desktop, and I can't for the life of me get into X, even after following the relevant chapter of the documentation. What, if anything, can I do?

Halian|MBA, did you install X system ?

madmax: Of course

Halian|MBA, and driver for amd

madmax: AFAICT yes

I'm very n00b at BSD so

Halian|MBA,did you do --> pw groupmod video -m your_user_name

madmax: Yes

Halian|MBA,ok did you login to your user

Yes

startx

Did you logout and login _after_ adding your user to the video group?

lts: yes; rebooted, even

Sorry, gotta BRB

what happen when you run startx

madmax: hokay so. sprunge.us/5nrEAS

termbin.com/0xni also here's my Xorg.0.,og

.log *

where are the list of names for tcpdump's 'icmp6[icmp6type]' documented? e.g. 'icmp6-echo'?

Halian|MBA, did you add the module to /etc/rc.conf file, executing the following command sysrc kld_list+=amdgpu

madmax: not that exact command; I added kld_list="andgpu" to /etc/rc.conf manually

Is "andgpu" instead of "amdgpu" a typo?

V_PauAmma_V: yes. :x It's spelled "amdgpu" in /etc/rc.conf

*nod*

(I've wasted hours due to typos of that ilk.)

(Felt af)

Halian|MBA,and now is working ?

madmax: not at all

Halian|MBA,after you install x org did you install any DE

madmax: yes, Xfce

OK. That X log reminds me vaguely of my first try at starting X, when I didn't actually have the module for my GPU loaded. So I think yours doesn't get loaded, despite the kld_list. So I'd try using the full path with the .ko extension, like so: kld_list="/boot/modules/amdgpu.ko"

OK

(and reboot)

OK

Nope, still no GUI :(

Halian|MBA, you follow instructions from xfce

termbin.com/z3xn here's the output of startx

madmax: Yes.

Halian|MBA,did you load dbus_enable="YES"

madmax: yes

What does "kldstat -n amdgpu.ko" display?

V_PauAmma_V: termbin.com/2g6q

OK, it's loaded.

YEAH

Yeah*

So next we need to see what your X log looks now.

V_PauAmma_V: termbin.com/ks81

Halian|MBA,try '% startxfce4'

madmax: that just spits out the same kind of X error as above

Halian|MBA,when you make your user did you add to wheel group

OK, so based on your X log, you now have a different problem. amdgpu doesn't recognize your GPU. What GPU do you have, and is it integrated or external?

V_PauAmma_V: XFX Radeon RX 6700, discrete

Hmm. Have a look at wiki.freebsd.org/Graphics#NVIDIA_Proprietary_Graphics_Driver and see if that helps. I have no experience with those, but I strongly suspect amdgpu is the wrong module for your graphics card.

Title: Graphics - FreeBSD Wiki

NVIDIA driver is definitely not going to be the correct driver for an RX 6700

I stand corrected.

i have an RX 6800 XT and amdgpu works fine with this card, which is the same generation as 6700

I use amdgpu for it on Linux

(however, i don't use X, so i'm not really sure how to configure that)

Halian|MBA,you trying to set MidnightBSD

madmax: yes; unfortunately, their IRC channel is deader than disco

Halian|MBA: can you install a Wayland compositor such as sway or hikari and try to run that? that should not require any configuration just to start and will tell you if the driver is at least working

lw: *sigh* I guess

i'm not saying you have to use wayland, i'm just trying to narrow down the problem

Of course

I'm just not a fan of Wayland :þ

i missed the start of this conversation, but when the amdgpu kernel module loads on boot, does it set the monitor resolution etc. correctly on the text console?

TBH I haven't checked closely

lw: [backend/backend.c:371] Failed to open any DRM device

can you paste the output of dmesg somewhere? or relink it if you already did

lw: termbin.com/0xzo

what version of drm-kmod are you using?

I don't know how to check that

in freebsd, "pkg info | grep drm" should show it... i don't know anything about MidnightBSD or what that even is

lw: drm-fbsd12.0-kmod-4.16.g20220

ouch, that's a bit old

quite possible too old to support RX 6xxx series

Hrm

lw: i think in the pcap-filter manpage

The 6700 launched in 2021

yes, but we only just got support for RX 7xxx like... a couple of months ago?

Ahh

first it has to be supported in Linux, then the Linux driver has to be ported to FreeBSD, then you need to be usable a recent enough version of FreeBSD to use the newest drm-kmod

…yikes, 12.0 is from 2018

s/usable/using

OK MidnightBSD is not the way forward lmao

yeah

BBL installing proper FreeBSD ><;

i'm pretty sure that card will be supported in 14.0

because i remember using it there before i upgraded to 15.0

*nod*

jmnbtslsQE: aha yes. thanks

Halian|MBA: if it helps, the "4.16" in your drm-kmod version is the Linux kernel version it's based on, so you can check when Linux first supported this card

in 15.0 we have drm-61-kmod which is Linux 6.1... i think 14.0 has Linux 5.15

hmm

16:14:26.136629 rule 161/0(match): block in on ix0.103: 10.1.3.130 > 10.1.3.1: ICMP echo request, id 3447, seq 0, length 1480

16:14:26.136646 rule 161/0(match): block in on ix0.103: 10.1.3.130 > 10.1.3.1: ip-proto-1

"ip-proto-1" is ICMP... i guess that means tcpdump can't parse the packet somehow?

and why is one of my wireless clients sending a 1480-byte ICMP packet

I GOT X WORKING (in FreeBSD)

OMG!! gz

TY ^_^

W00t!

lw: might be a fragment..should be able to confirm if you do -v or -vv to see the IP packet

yeah, i wondered if it was a fragment but i thought tcpdump would mark that somehow... maybe not without -v

but even then, it shouldn't be fragmented to begin with... and why is a wifi client sending a huge echo request to its router?

well if it's that large, that is the reason for the fragmentation i assume. but i don't know why it would send that

ok yes it's a fragment

16:34:53.817441 rule 161/0(match): block in on ix0.103: (tos 0x0, ttl 64, id 11783, offset 0, flags [+], proto ICMP (1), length 1500)

10.1.3.130 > 10.1.3.1: ICMP echo request, id 3447, seq 0, length 1480

16:34:53.817457 rule 161/0(match): block in on ix0.103: (tos 0x0, ttl 64, id 11783, offset 1480, flags [none], proto ICMP (1), length 52)

10.1.3.130 > 10.1.3.1: ip-proto-1

i'm actually not even sure why pf is blocking this, i thought it reassembled fragments by default, and icmp echo request should be passed

rule 161 is: "block return in log quick from any to self" but before then: pass quick inet proto icmp from any to any icmp-type echoreq

firewalls have mysterious ways

yeah. like my fun recent issue: scrub (reassemble tcp) breaks connections to Amazon/AWS... which reminds me, i need to file a bug about that

very mysterious. i have reassembly using ipfw on some hosts that don't have any issue with AWS (as far as i know)

how does it break your connection?

the connection is established, client sends HTTP request, Amazon sends HTTP response, router forwards the response to the client, the client ignores it, Amazon continues to retransmit the response

client is separate from the pf machine just on the same network?

yes

client <-> router <--pppoe--> Internet

(not an MTU issue, MTU is 1500 all the way)

i guess i wonder if something in the tcp payload was corrupted. but you said your scrub is only fragment reassembly?

i mean, the ip payload that includes the tcp header and its own payload

match in all scrub (reassemble tcp)

pf.conf man seems to say that reassemble tcp just processes TCP and doesn't do IP reassembly (not sure why it's called reassemble). but yeah i guess it sounds like some bug where the tcp data is corrupted by that processing

hmm, i would think if the client just ignores it then maybe the checksum is wrong. not sure

i might misunderstand the pf.conf man page though

yeah, i thought that as well - from what i can tell it *always* reassembles fragments and 'reassemble tcp' is the other stuff like sequence renumbering

... which is odd

do you have checksum offload on that interface? not sure if that is relevant in this case

actually yeah that can't be relevant

on the internet interface no (it's pppoe) but on the internal interface yes: options=4600703<RXCSUM,TXCSUM,TSO4,TSO6,LRO,RXCSUM_IPV6,TXCSUM_IPV6,MEXTPG>

hmm, not sure

the packets involved in this issue weren't fragmented anyway though (it was only about 400 bytes) so not related to that

once my users go to bed i'll re-enable (reassemble tcp) and take some packet captures for a bug report

heh

... wonder if i could fix this by turning off ipv4 and forcing them to use NAT64

fixed as in the machine will be set aflame by an angry mob

heh

i actually did that for a while and surprisingly few complaints

but i turned it off due to issues i wasn't sure were related, once this new freebsd router has been stable for a bit i might turn it on again

i think i should publish more of my tools eh

save layout

i shall not

save ur own layout

jauntyd:

Password:

funnily i've not pasted anything sensitive in my >23y of ircing

more like 7wg

:)

excuse me for repeating myself, but is there any fan control interface available in freebsd? seems my dell laptop has bios control enabled and fan does not work by some reason, i want to gain manual control on it.

i have a reproducible fluent bit seg fault bug i can't find a solution to. any help greatly appreciated: termbin.com/ntly

it's using the fluent-bit directly from pkg

Does lightdm have anything like SDDM's Xsetup file? I need to run some xrandr commands when X sets up

Halian: I think you might want to look at /usr/local/etc/lightdm/Xsession

skered: will do

Looks like that's running stuff in /usr/local/etc/X11/xinit/xinitrc.d/

how do I use WITHOUT_LLVM_TARGET_ALL from src.conf(5)?

concussious: put it in /etc/src.conf and enjoy ?

i tried "options WITHOUT_LLVM_TARGET_ALL" in /etc/src.conf

oh no that is not the right way to do it

WITHOUT_LLVM_TARGET_ALL=yes

aha

although instead of 'yes' you can set it to literally any value, such as 'no' or 'false' and it still behaves the same way, what matters is the variable being set

can it go in the kernconf?

uh, maybe but that doesn't seem right

either you want to build all of src that way (including src/sys) or you don't

that makes it not build anything for cross compiling, right?

based on my brief scan of src.conf(5), yes, but i wouldn't swear to it

okay, thank you so much

Is OpenJDK the only java compiler in packages?

last time i checked oracle license on java, that you have to agree to their license to use the "offical" jdk. where openjdk can be reisitrubuted in other means

Thank you

hrm.. they do have a license Oracle No-Fee Terms and Conditions, but the binaries are here: oracle.com/java/technologies/downloads

i have a reproducible fluent bit seg fault bug i can't find a solution to. any help greatly appreciated: termbin.com/ntly

freebsd have any api to make tailing 1000 files efficient?

what's everyone use to collect logs on hosts and ship them out?

regarding tailing files: select, or (better) kevent