so I'm back to trying to troubleshoot this issue with vt not displaying on amdgpu with drm-61-kmod

in dmesg, I am seeing: fbd0: not attached to vt(4) console; another device has precedence (err=17)

does anyone have any idea what is causing this? there's only one GPU in this laptop

it sounds like support for drm-510-kmod on the FreeBSD 14 series is going to go away sooner or later, even though that's the only driver that has properly worked so far :/

guess I could retry 515 at this point, see if these kernel panics were addressed in the last 7mo. I had meant to send in an issue regarding it, but since everything was working on 510 I just never got around to it (515 was causing page faults in the rcu subsystem, iirc)

For me 510 was causing frequent freezes. So far it has been 48 hours with 61 and I haven't noticed a freeze so looking good so far!

515 has been solid for me on 14.1

this particular laptop has like a raven 2 or picasso iGPU in it. I think like Vega 8 or something?

it seems like the "another device has precedence" error message I'm getting is from drm-kmod itself

Sorry. I said 510 but I think it was actually 515 that I had the problem with. In any case it seems it was this problem: freebsd/drm-kmod #302

rwp: hrm, with 515 I wasn't noticing performance issues, just after a couple of hours of web browsing (usually with a youtube vid playing), the system would randomly panic. it'd be locked up for like 5-10 seconds with the last chunk of the audio buffer playing repeatedly, I guess the delay was for writing out a core dump or something

but after that the system would automatically reboot

I'm wondering if it's this issue this person is having: freebsd/drm-kmod #287

they never came back with a crash log or backtrace

I will be happy then that I only had freezes/pauses because that's a lot better than having a crash. I never had a crash out of it.

if I want global shell aliases (all shells) & functions (for shells that can) is it best put them in something such as /usr/local/etc/profile.d/aliases.sh ?

(functions would be separate file)

Personally I think that if possible that instead I would install an external program in /usr/local/bin/ as more universal. As long as it isn't a chdir or similar.

If you examine /etc/profile which I assume you have then you know all that you need to know to do this. A file there seems sufficient for all shells that read /etc/profile

one PC said "'pkg: No SRV record found for the repo 'FreeBSD'. pkg: An error occured while fetching package. Unable to update repository FreeBSD. Error updating repositories!'". The other upgraded. They're on same local area network (LAN)

rwp: yeah, 'all shells' makes that a bit more complex

Yes. All shells... sh, ksh, mksh, pdksh, zsh, fish, bash, csh, tcsh, and I am sure I have offended someone for leaving out their favorite shell.

Also when that's done to me I feel it is too invasive. I then add stuff to my .profile to remove and undo what has been forced upon me.

darwin, Seems like a DNS lookup glitch of some sort. Since one worked and the other did not work. Which of course makes me suspect that your DNS server an MS-Server nameserver?

i don't have a DNS server and don't use MS

i use my ISP's settings in router and identical /etc/resolv.conf

In any case you can look up the DNS record yourself to verify the DNS operation. pkg.FreeBSD.org

Everyone has a favorite way to do a DNS lookup. Mine is "host pkg.FreeBSD.org".

It will use the first nameserver declared in the /etc/resolv.conf file.

yeah, it didn't find it on the particular one that couldn't upgrade

It would probably work if tried again I am guessing. Probably a transient glitch failure. Gremlins!

i tried many times. I'll reboot

this time worked

does qemu on freebsd support vmm for acceleration? it just lists tsc for accel

does that not answer the question?

yeah, it does not

a colleague was working on it in his spare time, but I don't know exactly how far he got / if he has any interest in continuing that

cccccchchlkghrbjhcknfkvuvgibehtfbhhhneifhhul

hello, is pam_gnome_keyring.so working as expected for you ?

it looks to me like we have regression with pam_gnome_keyring.so and virtual_oss

*regressions

rwp, jaunty, github.com/KlaraSystems/sync-be enjoy :-)

jauntyd: ^

dch: read about that tool a while ago, looks very interesting. Have you tested it?

drobban: yup, I wrote the docs for it & use it too.

how to set specific fonts with command, i have known "fc-list"

dch: nice! I have been curious to test it out. Now when I know you have been involved in the documentation I might find the courage to test it then :P

Yesterday I tried encrypted zfs root with raidz2 in a VM (14.1 disc1 install on libvirt with uefi and virtio discs). All went fine, but after a `freebsd-update fetch install` upgrade, it failed to boot. It seemingly could decrypt all disks.

Looking around, I came across issue 263171[*], and am now wondering, if I should dig into the issue I encountered any further, or if encrypted root is just not fully recommended atm. (the handbook doesn't seem to go into detail, just mentions the feature in the installer).

[*](bugs.freebsd.org/bugzilla/show_bug.cgi?id=263171)

Is there consensus on wether encrypted zfs root should work(tm), or is it even considered experimental?

doesnt the official installer provide that option?

luciusf: that totally should work, people use that since many years on their laptops.

armin: ok, thanks for the info! - I just wanted to have that question out of the way, before I invest too much time

cedb: indeed it does, quite conveniently tbh.

i mean isnt encryption at rest mostly for if someone gains physical access?

luciusf: Yeah I mean it usually does just the right thing, what's not working for your?

armin: I am already retrying the setup, to get more data and will try to reproduce the issue

luciusf: legitimate question btw, as in not rethorical

cedb: that is correct. and the fact that freebsd offers this so conveniently, was the fact to consider it for a server in a remote location where I couldn't fully control who would have physical access.

cedb: I just started with an experimental setup with VMs, when I encountered this issue

But thank you all for the quick response and pointer - that was already quite helpful!

keep in mind that disk encryption doesn't really protect the data while the system is running (I think that is pretty much impossible)

after installing zh-CJKUnifonts fonts, how to make it effect,

depends on your setup; vt? xorg? some wayland compositor? hm, I think qt and gtk do their own stuff too

nimaje: yes, data in use .. but my thread model is rather harmless. The disk encryption just serves as a precaution. The remote location is a family place. But remote nevertheless :)

my pf.rules stopped loading automatically when i've upgraded to 14.1 from 14.0 today

i need to load them with pfctl

I had this issue once, in my case it was due to an openvpn tunnel taking forever to come up

so pf_start (/etc/rc.d/pf) was called before openssl's tun0 interface was created

not sure if that can help, but probably worth checking

(and of course I had rules on tun0 in /etc/pf.conf)

I've picked up two VPS recently that use fe80::1 as the Ipv6 gateway address...anyone share ideas on getting FBSD to work?

GreenShield: something like "route -6 add fe80::1%em1 -interface em1" then "route -6 add default fe80::1%em1" ? Have auto_linklocal on interface and should work as is, at least it does on -CURRENT after some commit (was MFC'ed to 14 too I guess?)

ei: Thanks...let me try that variation.

oprs, does fe80::1 work as a default router? Can you 'ping6 trioptimum.com'?

CrtxReavr: Not currently, pings and other outbound traffic don't even appear in TCPdump...odd

I've passed the No Route error, so we're making progress

Can you pastebin your 'ifconfig' and 'netstat -rn' output?

armin: I tracked it down to the disks not being marked as boot discs in libvirt. If less than two discs aren't marked as boot disc, the system won't start

armin: Oddly, if I remove the first disc (in a four disc raidz2 setup) everything brakes. My expectation would have been, that the whole encrypted system is still resilient enough to survive two disc failures, no matter which discs those were

CrtxReavr: ei: I've gotten it to work, removed the rc.conf settings and set everything manually...seems to ping now.

Thanks for the assist

has anyone managed to get SMART disk temperatures from freebsd into prometheus, and if so, how did you do it?

GreenShield: enjoy ur modern end-to-end Internet <3

ei: HA...thanks for the tip...u'r settings worked. Now to figure out how to get them back into rc.conf

ipv6_static_routes in rc.conf(5) iirc

and for ipv6 is "ipv6_defaultrouter" for default route

Great...let me see what I can break. :)

oprs: i run openvpn, this might be it

luciusf: Oh wow, that sounds intense

armin: the weird thing is, that with bios legacy it seems to be the opposite. Whenever disc1 is one of the two marked boot discs, the decrypt and subsequent boot would fail

luciusf: Well yes THAT is something I do remember, that I had to switch between BIOS and UEFI boot somewhere in order to make it work. Quite frankly, I did *not* fully understand the root cause either.

armin: this behaviour might also be because of libvirt, but so far, encrypted zfs root doesn't seem to be as straightforward, as I expected it to be

I just fell in love with this little piece of software here: fastestcode.org/emwm.html

armin: but I gladly admit, that I always had problems understanding the boot process in it's entirety

A completely oldschool window manager that looks absolutely terrible and does not have much functionality.

Exactly what I wanted.

That thing is fantastic.

speaking of libvirt. do people actually run it? with virtbhyved?

i'm asking because it seems to not pick up the configuration changes i make, and installed documentation appears to be linux specific.

any awk/sed pros around that like to teach a newbie?

dch, Uh, okay, but I see no change in this file that I was commenting upon: github.com/KlaraSystems/sync-be/blob/main/sync-be

There's #awk & #sed specific channels.

ah neat - thanks!

Also it's always better that if you have a question you just ask the question. Then if someone knows something they might answer it.

Q: How do I print the 3rd column from a file? A: awk '{print$3}' And that type of thing.

Q: how do i implement Tower of Hanoi in sed? please keep you answer to a single line

rwp, cut(1) might be the better tool for that.

Though it depends on the file, really.

ivy, perl can do that in what looks like a few bytes of line noise.

CrtxReavr: okay, please include the complete C source code for perl in your answer

CrtxReavr, This is one of those things where there is overlap of functionality but awk is almost always the better tool than cut.

the only time i'd use cut is if i want to split at specific characters

for fielding splitting, awk is nearly always easier

It's a philosophy of design that makes them be different directions of thought though.

s/fielding/field

rwp, not when the field delimiters are set, vs. a mix of ambiguous whitespace.

We can just disagree here. But awk is designed for random whitespace. The cut, paste, join commands are designed for TAB delimited files, something rarely seen today.

rwp: don't forget col

Um. .. the -d option lets you specify the delimter.

And also col too. I had forgotten about it.

But using awk I didn't need to specify a delimiter.

wait i mean column not col. i always get those two confused

yes but column is nice to put at the end of your pipe to make it readable

There is also pr which can emit nice columns of data for human reading too.

(e.g.: column -ts: /etc/passwd)

I like perl but awk is both posix standard and also included in base whereas perl is in ports. So while I love perl for something like that I would choose awk for the portability and always available nature of it.

i try to avoid installing perl nowadays but so many things still depend on it...

awk is a crazy powerful tool. . . but its synatx looks like ass, soI avoid it when it's unnecessary.

Compared to sed's Turing complete language awk is very readable. :-)

awk is quite readable as long as you use newlines properly

more than perl, imo, although perl is certainly more powerful

Well, that's always been perl's biggest weakness, is that it's just too free-form.

I had to look up the pr options but try this for fun "echo * | tr ' ' '\n' | pr -2 -t -l1" Ha! Who needs ls? :-)

Reading other peoples' perl is just excruciating.

"Reading other peoples' /code/ is just excruciating." There! FTFY! :-)

Case in point: bpa.st/X67Q

Clearly that is intentionally designed to be the way that it is formatted. That hardly counts.

In this case, yes. . . but I've inherited perl that I was expected to make modificaitons too. . .

And it "worked," but I couldn't make heads or tails of it.

That's one thing I really like about Python. . .

YOu can use shitty variable and class names. .. but it has a minimum level of enforced syntax.

You clearly ascribe to a lot of religous dogma.

Sigh. Okay. My impression of almost all python programs is a long two page backtrace when there is a slight breeze that has blown it over. MOST python PROGRAMMERS write terrible python code.

rwp: the readme's updated github.com/KlaraSystems/sync-be if you're missing something specific from the (very short) script then add an issue

So. . . when I "code," I do bourne shell and Python. . . but I woudln't call myself a "software developer" or programmer. . . I'm a sysadmin & netadmin that automates stuff with scripts. . . and I'm good at it.

Python is one of those things that brought "programming" to the masses.

dch, I see now that you updated the project README file but of course I went directly to the script and in the script there is no comment giving any clue about what its intentions. I can read the script itself and then try to deduce its purpose but at that point I lose interest in it pretty quickly.

IT allows people to write "bad" code, where maybe no code at all would have been written.

dch, Also in order to submit a Github issue I would need to create a Github account. I have resisted doing that for so many years that I ask myself if this is the one item that would push me over the edge to do it? Not yet. Sorry.

well, I spent about 3h writing docs based on an adhoc comment in an irc repo

nice how mgmt/smartctl_exporter doesn't come with a configuration file or any pointers on how to use it

s/mgmt/net-mgmt

I always wanted to write some notes for it, just never got quite enough .. momentum

dch, I do appreciate the updated README file. That's significantly nicer than it was before.

sigh.. irc repo... I am too tired to computer

It's nearing the end of your day. It's the beginning of mine. However my days are always upside down. I wake up tire and go to bed wide awake.

Gah! *I wake up tired and go to bed wide awake. My brain is still not awake yet! Time to go make some caffeine.

rwp, BTW. . . I once tried chase down a perl linter or auto-formatter tha would help me decipher that script I pasted earlier. . . I didn't get far. Seems perl developers either don't care that other peoples' perl looks like ass or have just become used to the abuse.

hmm also it doesn't work, only picks up ada0 and ada1, none of my da* devices

oh, smartctl_exporter_devices

CrtxReavr, Well, I am one of "those emacs people" and so I just have emacs reformat code for me and that's the main tool I would use for reformatting things. (shrug)

no, that's not it. what's going on here

rwp, reformat that script for me then, please.

But obfuscated code is intentionally using obscure features in order to attain a visual effect.

I really don't want to spend more time on it. It's a clever rendering. I looked at it enough to see it. And to see that it is an obscured program. That's as much time as I want to spend on it.

oh this would explain it: root 85379 0.0 0.1 1268016 17108 - I 18:24 0:00.06 /usr/local/libexec/smartctl_exporter --web.listen-address=:9633 --smartctl.device=/dev/ada0 --smartctl.device=/dev/ada1 --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/ --smartctl.device=/dev/

hahahah

the rc script is broken because it uses 'cut', it should use 'awk' instead. yay awk

how timely

rwp, I thought you said emacs could remformat it in a canonical way?

That script does work, it's just slow AF: bpa.st/X26Q

I said I use emacs to reformat code, code that the programmer intended not to be obscured. Reformatting code that is intended to be obscured will not make the intent of the code clear.

That's a huge dodge.

Sure. I am dodging. Why shouldn't I? What's in it for me?

mystery solved!!!

hjf, What was the solution?

i was using /dev/cuaUn instead of /dev/ttyUn

I just thought you could do me the favor of auto-reformatting it for them. . . then I could actaully begin to parse the code, pieck apart its elements, give it better variable names, etc.

i could swear there was no /dev/ttyUn device

but i'm sure i'm just stupid

Some things come and go as various devices are attached, detacted, various drivers loaded, etc.

So it was a /etc/ttys selection problem?

The serial port selected wasn't listed as a secure terminal line? And switching to one that is marked as secure solved it?

I would have expected that if it was not marked as secure that there would have been an error printed to that effect rather than it simply hanging though.

rwp: wasn't the issue using a cua device instead of a tty device?

cua is for outbound connections, tty is for inbound

ivy, I thought the issue was that sh and sudo and several other commands all hung until Control-C when attempted.

right, but user just said: [#freebsd] <hjf> i was using /dev/cuaUn instead of /dev/ttyUn

so device name mismatch rather than secure/insecure issue

(as in, putting /dev/cuaX in /etc/ttys is probably going to cause issues like that)

rwp: no, it was secure too at some point during my tests

i suppose freebsd treats cua different from tty

you know what time it is?

Fri Oct 11 13:42:01 EDT 2024

le-fay.org/tmp/30d/temps.png whoo it works

hjf, What are you using to clean those keycaps? A retrobright type solution I presume?

christ that drive was hot

rtprio: the blue line is the CPU

bah

although technically, i think the datasheet on these disks says they're fine up to 60C :-P

not that i'd want to run them that hot

Are the ada0 side of the drives pre-heating the air for the da4 and da3 drives which are much warmer? :-)

The color scale is hard to match to individual drives so that was my best color matching guess there.

rwp: those are SSDs, i guess they generate less heat

rwp: yes "retrobright", just pool-grade peroxide

they actually have no cooling at all, the da* disks have two 140mm fans blowing directly onto them...

i think da3 (the hot one) is sitting at the very bottom of the case and is not quite aligned with the fans, i should probably fix that

It's still within reasonable limits okay. But cooler is always better.

yeah, i'm happy with <= 40C but finding this sort of issue is why i set up the monitoring

welp, that CH340 based serial port adapter i bought seems to have died

i mean even if it's within limits, more heat probably = lower mtbf

the UPS refused to connect to it. terminal works. i think the ups uses signaling on pins other than TX/RX

so some of the other RS232 pins may have died

on the subject of graphs i also found a nice way to graph load average: le-fay.org/tmp/30d/load.png - the red line is 15 minute average, the shaded area is 1 minute load average

(yes, the legends are unhelpful)

hjf, I doubt "RS232 pins may have died" but expect that there is a handshake protocol error from the UPS. The UPS may be expecting a 3-wire no handshake at all. Or the reverse.

Time for lunch here.

rwp: this retrobright turned amazing

hello I need some help for some vnet networking

should i go to ask here or networking

you can ask here

I have three IPs: Two for fib0(by aliasing 2nd IP), and the last IP to the fib1. I have so two nics and this second nic of the last IP to be used for my vnet jails.

For fib 1, I bridged with the external interface and a couple of epair, membered.

I see my vnet jails can pinging each other but it cannot get any internet access. I left them plain. no pf*. I believe it is due to my lack of knowledge in either such a configuration or in general networking(or both).

I enjoyed jail, enough. I need to get back to my development. It is time to get some help.

it is a vm(ec2) by the way

You don't have any pf rules for forwarding installed?

not enabled

Then no packets will be forwarded. You will need to add a forwarding rule. Most typically at NAT rule.

Example of one: nat on $ext_if inet from $int_if:network to any -> ($ext_if:0)

You will want more in the file too but I wanted to point you to where you needed to look at adding specifically.

this kind? boucek.me/blog/freebsd-jails-with-vnet-and-nat

Without looking at it longer than just that amount of time, yes.

Here is the actual file I am using at this moment. paste.debian.net/plain/1332033

And with that I am afk again for another bit. Good luck!

thx

I'm tray to track down file activity. `gstat -ac` shows me the IO per device, but I want to know which files or paths are seeing the most IO. Is this possible?

nice. i use poweredge too! :D

i wonder exactly how many they have hmm

My old poweredge 1850 with some mods runs FreeBSD: unibia.com/unibianet/freebsd/extend…ity-your-dell-poweredge-1850-part-1

what should I see in a pflog if a nat configuration works fine

goonmorning: nothing, unless you specifically added 'log' to some rules

i did

well, then the answer depends on what you added 'log' to

... and you can find the answer by running 'tcpdump -nei pflog0'

for nat

that's the question