daemon: theissues arose after i could feel the heat. And it gets worsewhen i start doing simple tasks, etc

but no load showed via top

daemon: yes, right now.That's the test

ACPI / power profile mishandling?

what i amatalking about is experience and the need for why we started looking at this inthe first place

its certainly weird

i gave up once a month ago, tryingto figure out abnormal heating after freebsd install,i amgoing it another go,and this timei hopingto find the cause and RhodiumToad is helping me

what asbout netbsd and openbsd

I think if you can demonstrate a significant difference between 12.4 and 13.2 then it definitely merits a bug report

works wellinlinux, even with fairly hard daily use,i don'tfeel the heat on my laptop

just out of curiosity

a bug in what though

don't complicate things yet :-)

RhodiumToad: booting up 12.4,i think its goingto be the same as 13.2,regardless of earlier short term result

its got to be some weird shared resource acpi shinnangans

if you fance boot 5.x :D

fancy a party*

don't be silly :-)

there is a 5.0?

RhodiumToad, fan/cooling controlled via OS is my bet :P

sozuba, yes there is, that is where we all had sound blasters that required MAKEDEV :)

Hahaha

I miss the old days; dialing 'into' the internet via a diamond 56k on com3 ahhhhh rose tinted glasses

anyhow your overheating problem; you need to isolate the actual problem

if it is ONLY freebsd

then you have a starting point

the question then becomes what does FreeBSD do that the other OS's do not

does it apply to other BSD's?

we swap alot of code

so netbsd and openbsd do they have the same issue?

if they do not you narrow down the issue further and ffurther

because heck you are an edgecase either way

its going to be some power control thing

yeah, looking into acpi_thermal is probably going to have to be the next stop

Just booted and its already interesting, you are going ot love it

do tell

this is 12.4?

yes

that is still bloody hoy

hot*

from a cold boot

Powerd is not running

look I do not want to be rude here but you are <SURE> you have decent thermal compound and a decent cooling setup on this thing right

ok. so a main difference here is that it's not going below C1 state

can you do sysctl hw.acpi ?

cpu 100% idle

sure

-1 must be infinity (I really do not read the acpi sysctls often :D)

hrm. so ACPI isn't really giving any control here

maybe that is the issue

the system may have been designed with the thought the OS will control via ACPI

hm. but this is a lenovo, aka ibm?

ibm sold it long back didn't it?

was'nt there a kernel meta module you could load to try load a bunch of proprietary ACPI controls?

kldload acpi_ibm

try that

and if it loads, try the sysctl hw.acpi again

ok

it looks like acpi_ibm has its own thermal subsystem separate from acpi_thermal

loaded

hw.acpi.thermal.tz0._CRT: 127.1C

-- CRT?

critical

ah

that's the temperature at which it forces a shutdown

should i give 5 mins before doing dev.cpu, for the changes to settle?

nah, just do it now

ok

do not hold us in suspense :P better?

oh, also, sysctl dev.acpi_ibm

thqat is a LOT more lines of text

so it did something

more than what? that's looking at a different sysctl node

I was looking at dev.cpu >.>

its late

hm

sysctl dev.acpi_ibm.0.fan=1

try that

report if it starts hovering of course :P

ƒout of sysctl dev.acpi_ibm.0.fan=1 is -> dev.acpi_ibm.0.fan: 0 -> 1

ouput*

did the fans spin up?

ok, now do sysctl dev.acpi_ibm again, and also sysctl dev.cpu and sysctl hw.acpi

that "fan" setting is supposedly for automatic vs manual mode, with 0 being manual

daemon: i just checked, it looks like it is.

when you run linux, is the fan usually on?

just as an aside RhodiumToad is way smarter than me :P I was just curious but it seems we have the solution to the problem

ƒtermbin.com/eizq

I thin k modifications to /etc/sysctl.conf are nexty

well, none of the sensors ever showed the fan was running, so i got a bit worried and checked it, it always runs, very silent, I have to placemy ear very close to the vent

Also revently i wanted to try controlling the speed of the fan, so i installed think pad specific fan control tool, it messed up, so uninstalled and let the kernel handle it

and the everythign was back to norma

well

01:31, <RhodiumToad> that "fan" setting is supposedly for automatic vs manual mode, with 0 being manual

would suggest it wants manual control via software

so ... yeah acpi shinangans

sozuba: what do we see for sysctl dev.cpu and sysctl hw.acpi now?

dev.cpu -> termbin.com/kfrl

is the fan still running?

hw.acpi -> termbin.com/t0q6

yes, just checked again, it is running

ok. powerd is not running?

no it is not running

can you do service powerd onestart wait a few seconds then service powerd onestop

sure

done waiting few seconds

ideally we want to generate some cpu load. what's a good cpu-heavy program in the base system?

compile opneoffice

stopped

that or do a quick perl -e 'while(1) { int(rand(9999))*int(rand(9999)) }'

but that is a single core

would it be better to do the load test in an installed system, or it doesn;t matter if it is a live medium

?*

does not matter

okay :)

dd if=/dev/zero bs=4k count=10000000 | xz -T0 -c >/dev/null &

larry approves the perl method ;o

then look at top while that is running

ƒok

perl is not in base

I know :(

xz is a convenient cpu-heavy thread-enabled program

even though they are gpl perl and zsh should be 'options' on the installer they are both so handy

top shows 96% usage for user and 4% usage for system ~

ok, and sysctl dev.cpu shows what now?

tem is close to 79,81

still below tjmax

fan is still running?

let me check the fan

its a lenovo I havea 6th gen

you will feel that shit if its fan is on

put your hand behind it or under it depending on model

yes fan running, but suprisingly still just less fan sound, usually i can hear the sound

let emcheck again

remember that the tjmax is 90c

the fan should scale with load

yes can confirm fan is running

i.e. it should get louder with the higher the load

it sounds like it is not operating correctly by all accounts

yup, that's what it has been doing so far, but here i don't know

I mean its not hitting thermal cut out

that is the point you have a problem

teem is 92, should i stop the stress test?

temp*

no

yeah, kill %1 should stop it

the cpu will auto thermal throttle if it is at risk

no need to push it that far

DONE

we not going to get de-lidding? :D

sorry for the caps

no worries

im just glad its not only me that somehow manages to hit caps lock accidentaly

i have butter fingers, i do all sorts of things with my keyboard :P

kinky bugger ;)

has the temperature gone back down?

come on man, i didn't mean it in that sense

:)

I know I just could not help my self lol; but indeed with RhodiumToad on this one; has the thing cooled?

yup temp is 58,58,61,61

so ... sysctl.conf? or rc.conf and kld_list

if you do a proper install, then the thing to do is probably this:

acpi_ibm_load="YES" in /boot/loader.conf

dev.acpi_ibm.0.fan=1 in /etc/sysctl.conf

okay will do, noted down

and probably a good idea to do service powerd enable

ƒokay

so the issu is the fan and power are notproperly managed?

I have this fun upcoming; that 6th gen X1 I mentioned earlier is going to be BSD in a month or two -_-

there controlled in software

BSD does not control anything unless you tell it to

daemon: ah okay

really kinda crap to be honest for fundamental hardware like cpu's and the ilk

I would expect that to have a fallback

daemon, Yeah I agree :D

well I will get the same fun as you in a month of so; so I guess at least I know now :D

installing freebsd on a system that is XHD this will be fun

haha, i think yo wont have much problwm with yours.

RhodiumToad: i'll have to go out now toget somethings done,are we done for now? If so, i'll finish my stuff then cmeback and install anddo what you suggested, then check everything we checkednow and the, keep you posted whenyou are available

if not, i can continue testing now and postpone my suffs for later :)

no worries, just let meknow

I mean ~ if your problem is solved then surely you should offer thanks?

I have other stuff to do today

you can catch me around tomorrow or so, most likely

RhodiumToad: when ever you are free and available, i very mcuh understand you have a life beyond this

daemon: well, we don't know if it's really solved

daemon: always, i amnot sure of the issues is sorted, RhodiumToad hasn;tsaid anythingof that sort. But i amvery thnakful tohimand you for spending your time and effort tohelp me

RhodiumToad, everything is grey -_- be british about it; we had sun today

Thank you guys :)

sozuba, and we hope it is solved :)

sozuba, have a good evening

i hope so too,i can'twait tofinally play around with freebsd and alsoget some work done with it

daemon: you too

~ I use windows bsd and linux, use the right tool for the job :P

have a lovely eveningyou two daemon and RhodiumToad

thanks again

np always welcome :)

daemon: I rarely use windows, even then on a vM

And members of family who still have windows

linux has been what i used

I amstillfiguring out the right tool for me

well,honestly I amstill figuringout a lot in life.

sometime i wonder where I am, haha

I need to bail its 2am in the uk :P so after this message I am gone, but if you like freebsd and its ports system you should also checkout gentoo; bsd for servers, linux for docker, windows for ui :P bye bye o/

anyway,that's life.

i used gentoo for 7 months

loved it

unfortunately my laptopcoundlt handlethe compiling

understand have a good ngiht then

bye guys

take care

catch you later

gentoo + laptop = great in the winter time, lol

#freebsd is probably not the right place for that.

right place for what?

markmcb: waxing poetic about gentoo

lol, ok, sorry. it's not an exaggerated comment. compilation = heat. laptops don't tend to handle heat well. i wasn't aware humor was not allowed.

i think it was _mostly_ sozuba

ah, ok. i thought you all just really hated my joke. :)

anyone know if it's possible to setup geli such that it looks for a keyfile, and if not found prompts for a passphrase?

have you tried the example in the man page?

yeah, unless i'm doing it wrong, it seems to want both the key and the passphrase vs. key or passphrase

the one with the "two User Keys: one for your employee and one for you as the company's security officer"

yes. it's more of a loader.conf questioni guess. i can get it to work with just a key, if the provider has been setup like that, but i can't seem to get it to not prompt me for a passphrase if i've provided a key, but fallback to passphrase if the key isn't present

it will have to remain an unsolved mystery for tonight. good night all.

Anyone know where I could search to find out info about what broke when updating py-cryptography past 3.4.8? I see that it got reverted in ed4bec1e01390965296c2f510e87de9ccf8cef63 but I can't find info about why

maybe in bugs.freebsd.org or the mailing lists?

Could try Ports Git repository cgit.freebsd.org/ports ; the "review" section reviews.freebsd.org; & Bugzilla bugs.freebsd.org/bugzilla

Title: ports - FreeBSD ports tree

I mean, in SW development I am used to have the "why" referenced in the commit message of a specific change

haven't had a look on the commits of freebsd - is it any different here?

See the reason in src/; in ports/ sometimes not

well, in that case I'd directly ask tcberner or leave a comment at reviews.freebsd.org/R11:ed4bec1e01390965296c2f510e87de9ccf8cef63

Title: R11:ed4bec1e0139

or the maintainer of that port

Thanks to RhDoc's URL for "reviews", reason is "Revert due to multiple breakage reports."

yeah, but endrift was curious about what exactly broke. "multiple breakage reports" without mentioning the reports themselves wouldn't help in that case

In that case need would need to trawl through mailing list, bugzilla

unfortunate

it's fairly simple, more or less

I'm now facing a separate issue that py39-boost-libs is shipping a static lib that was built without -fPIC, but I'm trying to build something that builds a shared lib against it

because....there is no py-libtorrent-rasterbar2 port yet

the yak shaving continues

endrift: are you compiling your own ports?

wasn't trying to, but the port I was using before an update seems to be currently broken

ultimately I just want to use deluge, but installing that via pkg shows nothing, installing via pip tries to pull a newer twisted[tls] that requires a newer cryptography

found I could install deluge by forcing it to use the existing twisted install, or at least an older one, but then it complains that libtorrent is missing

so I'm trying to install that, without much luck

libtorrent-rasterbar2 itself installs, but it doesn't include the python libs, and there's no py-libtorrent-rasterbar2

and trying to build it manually incurs the linker error mentioned above

Title: pipx

I feel like you've missed the problem

install the c and c++ deps, and try pipx for the python stuff

those are already installed, but compiled without -fPIC

by the installed pkg

endrift, You may not like this plan: remove the existing related packages; compile however you want & what you possibly can (use packages for the rest)

aye, i was indeed missing the gym context

parv: I was planning to see if I could just rebuild the boost python libs at this point

I think everything else should just work after that, though I know mixing ports and packages is a bad plan

endrift, Right, right

unsure I have anything else that depends on this

it's a static lib anyway so...wait

I can just build a local copy and link against that

let's back up for a second: why are the… things compiled without -fPIC?

Dunno, that's just how the pkg builds it

It might be an issue with boost's bespoke build system

it wouldn't surprise me, it is boost

that said, I think there may be ports currently marked as broken as a direct result of this

freshports.org/devel/boost-python-libs this is the port that ships a static lib built without -fPIC

Title: FreshPorts -- devel/boost-python-libs: Framework for interfacing Python and C++

ah I double checked, the broken port was marked broken for a different issue

Title: FreshPorts -- devel/boost-python-libs: Framework for interfacing Python and C++

actually, hm, why is my local build of libtorrent-rasterbar trying to link against the static lib...? I can probably get it to link against the shared lib instead

Installing freebsd, and if i am making a uefi setup gpt is the prefered partition right? or bsd would suffice?

i am already reading the documents, just asking for peopels opinion and experience :)

going ahead with uefi

i mean gpt

sozuba: good choice

imm_: thanks and reason?

now days rust port is become outdated wile latest version is still compiling by the poudriere ...

readung this -> forums.freebsd.org/threads/wireless…rface-not-listed-but-it-works.59150 it says wireless interface wont be visible under ifconfig and we just have to accept it and work with that

Title: Solved - Wireless interface not listed (but it works) | The FreeBSD Forums

is that the case? if so why?

sysctl -n net.wlan.devices shows iwm0 exists, so why is not listed under ifconfig

sozuba: you need to map it to a WLAN(4) device

meena0: ah okay, will do thank you :)

sozuba: this feels like tmi docs.freebsd.org/en/books/handbook/…vanced-networking/#network-wireless

Title: Chapter 33. Advanced Networking | FreeBSD Documentation Portal

meena0: yes read and configured

thank you, i though tere was something missing, some module or stuff

:)

what is tmi?

okay too muc information, urban dictionary:p

guys, does anyone know if my syntax is correct?

dev.acpi_ibm.0.fan=1

should there be space between charcters and = sign?

in sysctl.conf, no space needed

Hi RhodiumToad thanks :)

installation complete, butt the tempreature is still high with nothing running and cpu 100% idle.

55,55,53,53 C

you changed /boot/loader.conf and /etc/sysctl.conf? and rebooted after?

yup, as you said

and powerd is enabled

ok, let's see the sysctl dev.cpu and sysctl hw.acpi

sure

dev.cpu termbin.com/d824

also, is the fan running?

yes it is running, i can hear it clearly

clearly meaning, movig my ear close to the wentt

vent*

hw.acpi termbin.com/mny5

hmm.

can you try sysctl dev.acpi_ibm.0.fan_level=7

sure

ƒdone, the temperate is still high 57,57,54,55

should i post dev.cpu again?

does the fan sound the same? and did the output say 0 -> 7

the fan still sounds the same

and yes, dev.acpi_ibm.0.fan_level: 0 -> 7

ok, now do sysctl dev.acpi_ibm.0.fan=0

ok

and when you do that, does the fan go off, speed up, or stay the same?

dev.acpi_ibm.0.fan_level: 7 -> 0

stays the same

but will try again and see

checked again, it feels like it has no impact on the fan

it stays the same

no, set fan_level to 7 and .fan to 0

oh sorry

setting .fan to 0 should put the fan on manual mode, and fan_level=7 should be the fastest controlled speed

but if it doesn't actually change, that suggests that none of these settings have any effect

ƒi dodn't see any change

in the way fan functions

sound, inyernsity/umpact remains same

temps are still high?

dev.acpi_ibm.0.fan: 1 -> 0 and dev.acpi_ibm.0.fan_level: 0 -> 7

yes, temp is high 57,57,53,53

ok. I think I'm out of ideas; this should probably be a bug report.

you'll need to collect this info to attach to the report:

both dev.cpu and hw.api?

the exact model name and age of the laptop, the contents of /var/run/dmesg.boot, the output of sysctl dev.cpu; sysctl hw.acpi; sysctl dev.acpi_ibm

okay, will do. Thank you very much RhodiumToad

mention specifically that changing the fan= and fan_level= settings did not do any good, and that powerd isn't helping

okay will do, will post a link here after,to keep you updated

also, point out specifically that it's running pretty warm even when the system is idle and the clock speeds stepped all the way down

and stress that this is OS-dependent and that it's noticable cooler in linux

i.e. not just different reported temps, but you can actually feel a difference (assuming that is the case)

Got it, will mention that.

also that it was running hot both with and without acpi_ibm loaded

Bug component should be kernel right?

while filing

yes

sure, will mention that too

ok

Does this title sound apt? "Excesive Laptop Heating with 99-100% CPU idle"

mention the type of laptop in the title

ok

sounds good? Excessive heating on Lenovo Thinkpad E450, even at 99-100% of CPU idle

good enough

awesome :)

might be worth posting to the forums too

(with the same info)

Yeah was thingking the same. Will do

RhodiumToad: i was going through the output of sysinfo -a, and i noticed this-> termbin.com/g1w8 , does this have ny significance to my issue?

mainly, The acpi_wmi.ko module is loaded, however it is not being loaded upon the system boot time from /boot/loader.conf.

did you load coretemp yourself manually? also what is kld_list set to in /etc/rc.conf, if anything?

RhodiumToad: yes i loaded it manually

there is no kld_list in my rc.conf

it's fine for modules to be brought in as dependencies of other modules or via rc.conf rather than having to put everything in loader.conf

understood

hm

pchtherm is in there?

what does sysctl dev.pchtherm show?

I'm guessing that was loaded by devmatch

ƒoutput of dev.pctherm -> termbin.com/bl1ys

rc.conf termbin.com/wl2uu

loader.conf termbin.com/b1ap

ok. so the pchtherm output is consistent with the rest

that also tells us that 93C is this system's limit for normal operation, and 115C is when it will shutdown

ah okay cool :)

pchtherm just lets us read info from the device, it doesn't seem to have any tunable knobs that might be relevant to the problem

i thought it was something i forgot to set and may be is what causing the issue :D. I also got a bit hopeful :P

ƒƒdev.pctehrm, dev.cpu hw.acpi, should those be attached as files or pasted in the summary?

probably best as files

okay

every morning i wake up to sendmail driving a load of 5+ on 8 cpus and doubling my systems wattage. how do i completely disable it? i have sendmail_enable="NONE" set

what's it doing?

smmsp is trying sendmail -i root

(you could do cp /usr/share/examples/dma/mailer.conf /etc/mail/mailer.conf to ensure that anything trying to send mail uses DMA instead)

you also want sendmail_submit_enable="NO" in rc.conf

thanks, i'll give those a shot.

hm, sendmail_submit_enable="NO" should not be needed as long as sendmail_enable="NONE"

but the mailer.conf change is needed to prevent programs directly invoking sendmail

dma will handle local mail delivery only

(or you can configure it to punt to a smarthost)

so there's no way to globally disable email? redirecting to other things is the only option?

well, you could point all the mailer.conf entries to /usr/bin/false

you'd then want to go through the periodic.conf files and make sure every periodic thing was configured not to send mail

also, cron

the ability to send mail is assumed by a lot of things

ouch. ok, i'll try that. three weeks on freebsd now. maybe i don't understand enough yet, but this is a weird design element of the OS. regardless, thanks for the help!

all Unixen

I have a CUPS related question (FreeBSD on both ends). I have a printer that prints with 'lpr file.txt' no problem. I want to print from a remote computer. Thus, I thought 'cat file.txt | ssh me⊙tc lpr' would work. However, it reports that the printer is not found. Sshing in first, and then using lpr works as expected. What would cause lpr to see different printers, for the same user

account, based on if it was run directly with ssh or after loging in with ssh?

afaik I don't have anything weird in .bashrc or .bash-profile that would be causing this.

no tty?

I suppose that is true, although I am not sure why lpr would care

try with ssh -t

"lpstat -a" lists the printer in both cases

ok will do thanks

my guess is the PATH is different

echo $PATH when logged in normally, and ssh you@where env | grep PATH

"pseudo-terminal will not be allocated because stdin is not a terminal" which I guess makes sense due to cat

ok lets see what PATH says

note that /usr/bin/lpr is the old-school BSD print subsystem, while /usr/local/bin/lpr is the CUPS one

indeed

if you're relying on a .profile to set $PATH (or provide an alias for lpr), non-interactive ssh won't run that

ok, indeed, `which lpr` is different, I will investigate that; you guys are quick

(best place to set such things is in ~/.login_conf)

should ahve thought of that

also, lingo wise, what would you call "ssh somewhere; command" vs "ssh somewhere command"

e.g. "running over ssh" vs "sshing and then running a command", is there a smarter sounding way to say that

so, I had added /usr/local/bin to PATH in my .zshrc, hence the difference. I will man about .login_conf

man login.conf

note that the per-user file is ~/.login_conf with an underscore not a . which is slightly confusing

right gotcha

stuff set in ~/.login_conf should apply both to normal logins and ssh, and in recent fbsd versions also to things run from cron

side note, and thanks for pointing me in the right direction with login.conf, why would man -k or apropos login_conf return nothing appropriate? I find that often man -k doesn't return hits for exact text matches in man pages, or maybe i misunderstand what apropos does

it's just searching the manpage titles, and the manpage for login.conf doesn't have login_conf in its title

i see, thankyou

I tweaked all the mail stuff ... is there a way to trigger whatever is happening in periodic nightlies so I can test the config?

you can run 'periodic daily' as root

thanks. assuming that was the source of the issue, it seems to be resolved.

looking through the main pages for mailwrapper and mailer.conf and not seeing anything, but if i configed a smtp mailer, is there a way to substitute my actual email for anything mailed to "root"?

which smtp mailer?

msmtp is the one i usually use

presumably that will have an aliases or forwarding mechanism

ah i see, so handle it with the mailer (vs. substitute before then)

sendmail and dma both respect /etc/aliases by default, I believe

most other mailers should too

so just root: my⊙ra in there should help :-)

see the default /etc/aliases for more commentary

perfect, looking in the docs and it has a flag to point to a standard alias file, so i should be good

behold. my first perodic email! :) thanks for that help.

hi

is there some sort of package that needs to be installed to get $XDG_RUNTIME_DIR defined?

defining it manually in profile feels like a bit of a hack

i manually define XDG stuff in my .zshenv ... i've found this to be the most portable way to go and have it work across OSes

i never found a better way, so i have my own "setupconfigs.sh" i run on any new install and this is part of it

hi all

RhodiumToad: Looks like i can attach only one file, should i upload a zip file? or is it fround at for security reasons?

frowned*

one at a time

meena: how? there is no attach button, just submit button.

sozuba: it just means you have to do it for every single file

meena0: I don't understand what you are saying, sorry.

i can attach files one by one, if there is an option. The only option i have is submit the PR

if i press the submit button with one file attached, will i get an option to submit more after the submission?

yes

but also only one at a time

ah okay, that's fine, as along as i can attach all the files.

Thank you

Title: amdtemp: Fix missing 49 degree offset on current EPYC CPUs by valpackett · Pull Request #754 · freebsd/freebsd-src · GitHub

I wonder if this could be a similar issue

RhodiumToad: Posted to PR -> bugs.freebsd.org/bugzilla/show_bug.cgi?id=271938

Title: 271938 – Excessive heating on Lenovo Thinkpad E450 even with CPU @ 100% idle

meena0: looked at that Link, I am not savy enough to say if that's realted to mine, but looks related :D

i'll dig through

okay I gota go now. Bye people

yuripv: do you see 'Spam' in the header at Bugzilla?

grahamperrin: do you have a newer version of bugzilla somehow?

meena: no, the same version.

If I recall correctly, all committers automatically gain some privileges (I didn't realise this until some time after I got my commit bit … don't know whether it's documented).

grahamperrin: it looks like you just set a different style

Oh, right, the blue. I find the Classic skin easier on the eye. <bugs.freebsd.org/bugzilla/userprefs.cgi>

Title: Log in to FreeBSD Bugzilla

Nothing exciting for me in the Administration tab, incidentally. I'm quite certain that I don't want to "Set up whining".

am i correct to say there is nothing like linux's udev unique names for devices, e.g., /dev/disk/by-uuid/ ? seems gpt partition names are the closest thing.

markmcb: not a direct answer, but as far as I know it's commonplace for loader.conf(5) to include these two lines:

kern.geom.label.disk_ident.enable="0"

kern.geom.label.gptid.enable="0"

this looks promising. i'm finding blog mentions, but can't seem to find the relevant docs. regardless, thanks for the insight!

the installer puts those lines in loader.conf on zfs systems

is there a reason they should be off?

or is it simply people want to see their labels and disk ids override that?

yeah, zfs insists on tasting every device, so if they're on, it's inconsistent what device names show up as members of a given zpool

i.e. sometimes the members will show up as /dev/ada0p2 or whatever and other times the same disk will show as /dev/gptid/xxx...

turning them off was seen as less confusing

interesting, so a file system tool's output drove disabling an OS feature? is this viewed as a ZFS bug?

it's not the tool output, it's that zfs would actually attach the /dev/gptid/... device

right, but i assume that's ok if they point to the same thing

seems like it's just the human side of it that's confusing, or am i misunderstanding?

yeah, but it causes issues

in particular if you detach the device under one name, it can immediately reattach it via the other

ouch. and it's not considered a zfs bug?

(gmirror can have the same issue, incidentally)

zfs is a prickly pear

it's not clear how to fix it

> … finding blog mentions, but can't seem to find the relevant docs. …

I think I encountered the same (lack of documentation) a few weeks ago …

yeah, i feel like a big fat warning should be a comment above those lines lines in loader.conf

part of the problem is that glabel's labels disappear and reappear according to whether the provider is open on another path

so when you close a device, the labels reappear, and that provokes other geom classes (including zfs) to taste them

so e.g. zfs or gmirror sees them as a newly-arrived device whose metadata shows them to be part of the zpool or mirror, so it attaches them

and when they say taste, they mean "influencer in a supermarket licking ice cream and putting it back" taste.

lol

not quite :-)

geom is very flexible, but sometimes so flexible that it ties itself into knots

I seem to be having issues with ctld

I have a setup to export a handful of targets, which used to work until I rebooted today

Now when the initiator on the other end attempts to connect, daemon.log reports the target can't be found, and the initiator errors out

however, ctlstat clearly shows the target is present

restarting ctld does not help

when was it last rebooted before today?

it's all going over the correct subnet too

yesterday

I'm wondering if you did some manual configuration that was lost on the reboot

I upgraded from 13.1-STABLE (because oops I forgot to rebuild -RELEASE) to 13.2-RELEASE last night, but it worked fine after rebooting into it

Via <lists.freebsd.org/archives/freebsd-questions/2022-January/000540.html>: <forums.freebsd.org/posts/550463>

Title: Re: kern.geom.label.disk_ident.enable="0" and kern.geom.label.gptid.enable="0"

and I haven't really poked at it since

I definitely didn't change anything with ctld in the meantime either

I also tried re-rebooting and that didn't help

hm

actually...I'm unsure it worked last night. I'm using it to export zvols for VMs, but the VM on this target didn't get booted due to IOMMU issues

which I've since fixed

so I'm not sure where in that process it failed

so it might have been the update to 13.2 that broke it?

I suppose it's possible

not sure where I'd go to start fixing that though

hm

I'm unsure exactly where in 13.1-STABLE I was

bunch of changes in ctld at least between 13.1 and 13.2

I didn't really update /etc when upgrading, which might be a problem

I'm still not sure how to do that properly when doing a source build

man etcupdate

ah thanks

if that doesn't help, we need to fix it

my own builds are WITHOUT_CTL so I can't easily test it

er, WITHOUT_ISCSI

ok, so it looks like the major change to ctld was factoring a bunch of code out into a library

fair few changes in sys/cam/ctl but they mostly look like error checks

I don't see anything regarding changes in /etc that spring out as relevant to this

have you tried increasing the debug levels for ctld and cfiscsi?

RhodiumToad: how would I do that?

also wow I had a bunch of conflicted cruft in /etc, that's all resolved now, thanks

cfiscsi has sysctls to set debug level

ctld has a -d option to run in foreground and spew debug output to stdout

after re-merging /etc (and accidentally messing a few things up) and rebooting, it seems to be working again

thanks for the man page RhodiumToad

It might be a coincidence but...I'll take it

hm, interesting

I have a 13-STABLE jail with syslogd and apache24 running, and nothing logged

truss -p `pgrep syslogd` # doesn't show anything either

fun!

well, almost nothing…

tried running 'logger foo' in the jail?

erk, why is my /home full

is there any resource that shows what's planned to be in future releases? like a road map?

only looking at what's going on in -current, afaik

The steady state of disks is full.

markmcb: 14.0: hackmd.io/JczFDHtiQYSeEyeK9182jw?view 15.0: hackmd.io/@jhb/ByWrxQmr2

Title: FreeBSD 14.0 Planning - HackMD

note that as OpenSSL and LLVM 15.0 integration delays 14.0 branching, more stuff from that 15.0 list might make it into 14.0

ooh, nice. this is great.

grahamperrin: yep, i just didn't want to mark entire mail account as spam, and just one post

i asked this before, but to upgrade my existing freebsd 13.1 to 14… do i just untar the kernel.txz and base.txz over my running system?

phryk: that'll possibly wipe a good deal of your /etc

meena0: yeah. i vaguely remember there being some snapshot mechanism for ufs. if i make a snapshot of the current step i should be able to easily roll back if something important gets nuked, right?

not easily.

phryk: rsync -a /etc /var/backup/$(date)

ufs snapshots are read-only, and in 13.1 you can't do snapshots on journalled ufs afaik

probably easier to upgrade from sources

Haven't built kernel/base from sources in ages. Essentially just check out the right branch for both, do make && make install, reboot, done?

make -jN buildworld buildkernel

then make installkernel and reboot

then make installworld, and then do etcupdate (make sure you did the preparation steps for etcupdate before installing anything)

then reboot, and then there's a make delete-old-libs or something like that to remove old files

this is all documented, so read the docs

ye, just found it. 25.6 in the handbook

the various reboots are because a 14 kernel can run 13's binaries (assuming you didn't remove COMPAT_FREEBSD13 from the kernel build), but not vice-versa

RhodiumToad: logger works, and pkg also logs, so something's up with apache24

the git branch for 14 is "main"?

Yep.

good. :)

o/~ where has all my diskspace gone o/~

mhh. my /usr/src is currently empty and "etcupdate extract" fails. if i understand the etcupdate docs correctly, i should first check out the code for the version i'm currently running, do "etcupdate extract && etcupdate diff" and only then update the code in /usr/src to 14?

something like that