Macer: that's quite a few drives in that pool

hye, I'm trying to install freebsd in a vm on linux, and am getting this error: CAM status: CCB request completed with an error

at what point do you see this error. are you installing under kvm or what

yes, qemu/kvm. here's a screenshot: 0x0.st/HwyK.png

I can also upload the libvirt xml, if needed

no ideas?

olk, The read-only file system status looks not-good. Since you are installing it will need to be be writing data to there. So the first task will be to figure out why it is read-only.

Did the host system run out of disk space?

nope. I even turned of the read-only flag on the emulated usb install media in virt-manager

*off

But you are using the USB disk to install from, so that one won't be the issue. It will be the VirtIO disk that you are installing to that will be of interest.

Also, why are you using a USB disk mount? Why is it not a cdrom mount?

because cd-rom mount with memsick image causes entirely different issue: 0x0.st/Hwyc.png

olk, Here is a working FreeBSD 13.1 libvirt xml file from here for you to compare. termbin.com/6y2i

Why are you building a VM using UEFI? It's more trouble. Easier to use Legacy BIOS for VMs. However they do work if you have the ovmf package installed.

here's the disk I'm installing to: 0x0.st/HwyT.png

I suggest building a VM using Legacy BIOS boot first because that is simple. And then after that is working then build one booting UEFI.

That last paste looks reasonable, though mine says "Hypervisor default" for both of those.

I need UEFI for a feature I want to test before going bare-metal. and here's my xml, it shows that I have ovmf installed and needed file chosen: bsd.to/wgYL

Title: dpaste/wgYL (Plain Text)

here's the feature: github.com/sadaszewski/freebsd-patch-geli-password-from-tpm2

Title: GitHub - sadaszewski/freebsd-patch-geli-password-from-tpm2: A patch for the FreeBSD source tree which enables fetching of GELI password from TPM2 and booting a trusted root filesystem

rwp: I have the same issue with the second option set to hypervisor default

I think the option-discard-mode:unmap is actually the default. But it made me wonder if when creating the VM if FreeBSD was selected in order to set most of the configuration to template values for it or not.

I selected freebsd13.1 template, yes

and edited it

Editing it is the only way to select UEFI. I usually let it default to the other values. I note yours is changed to Q35 but the default is i440FX. I don't know if that is important for this or not.

I suggest creating a VM using all of the defaults and a default installation to keep things simple. That works for me. Then repeat the installation changing one thing at a time. That should isolate where things become trouble.

uefi doesn't work with non q35 models

and the bios installation is worthless to me :)

I guess freebsd wasn't tested on kvm+uefi

I am sure that kvm+qemu+uefi was not tested on FreeBSD!

I'll try a VM creation here and see what problems I hit.

thanks a lot

UEFI for all systems has been nothing but a long path of thorns.

quick question: is secure boot a thing on freebsd?

I don't know. I am just a simple user. But I don't use it anywhere.

I have a great secureboot+fde+tpm setup on linux, but want to try freebsd

I am basically also finding it impossible to install 14.0R this way too. I am thinking of dropping back to installing 13 and then doing an in-vm upgrade to 14.

so it's a bug

UEFI? Yes. I would definitely agree with that assessment. But UEFI is being forced upon everyone regardless.

rwp: Allan Jude said he made a faster way to rename them :)

i'm just waiting for it to resilver and i already added all the labels.. i just have to do the thing

rtprio: it has 12 disks raidz2

- 1 hot spare

plus... damn matrix turning + into a bullet point

Macer, It looks like you are in good shape, with just a long time ahead of you spinning disks sync'ing data. But you can get there one disk at a time okay. I have done that before.

sure. the only time it was a problem was when it was resilvering the first disk failure with the hot spare and the 2nd disk died lol

that was a little nail biting

now that the hot spare is fully resilvered i'm not really on pins and needles. i have the 2 new drives resilvering at the same time now so i'll just wait it out for 5 more hours or so

but at least this made me work on the gpt labeling of the partitions so i can keep better track of what breaks

Macer, Just for comparison, this is how I labeled my disks. I used somewhat shorter names here. bsd.to/iz9Q/raw

Title: iz9Q

ah. i put coordinates of my nas in too heh.. plus the brand and serial

at least then i quickly say 'oh... another barracuda died'

*can

I am using the bay number plus the driver serial number.

olk left but I have tried a bunch of combinations and I cannot get FreeBSD to boot UEFI in a VM. I can boot Devuan (Debian fork) UEFI in a VM okay though.

do you need efi for fbsd in a vm?

i thought it does legacy booting just fine

Macer, olk above just before you came back says it is needed to test this feature patch: github.com/sadaszewski/freebsd-patch-geli-password-from-tpm2

Title: GitHub - sadaszewski/freebsd-patch-geli-password-from-tpm2: A patch for the FreeBSD source tree which enables fetching of GELI password from TPM2 and booting a trusted root filesystem

oh

I have booted Devuan on UEFI in libvirt VMs often so I assumed it would work in FreeBSD too. But responding to olk I have to eat my words and say that I can't get FreeBSD to boot UEFI in KVM+QEMU either.

oddly enough in proxmox i was able to get macos to boot in a vm with efi and opencore

heh

Previously on bare metal I have had mixed results with FreeBSD UEFI depending upon the specific system. I always blamed the UEFI firmware because my UEFI booting systems all have buggy firmware anyway. I have one workstation that UEFI seems to work okay and one laptop and the rest all seem to be trouble. I have frustrated myself wrestling with UEFI on them and given up more than succeeded.

bbiab

i am going to throw it out there.. i am setting up netatalk for time machine backsup, i can get the "homes" share to work but not the timemachine one.. has anyone ever worked with netatalk recently? bsd.to/eat7 <---configuration file as it stands now

Title: dpaste/eat7 (Plain Text)

the [homes] one shows up.. but NEVER the second one

rwp: i moved over to openzfs with those drives :)

Title: powerd not working - Pastebin.com

so i'm trying to get powerd to throttle my cpus and they don't seem to be throttling

not really sure what i'm doing wrong here

ah so powerdxx seems to have slowed it down

wow so i've been running the cpus maxed for years and never even noticed

well having variables in path variable is NOT good..

Hi

hi all

hi

Hi, I have now upgraded to 14-RELEASE and the upgrade went smooth. I will not update ZFS for now, that will be done in a week or two from now.

After the update to 14 I also upgraded my jails with ezjail-admin. The upgrade procedure went through without errors. Thereafter I rebooted the machine.

Now I can not use "su" after getting a root shell one of the jails with "sudo jexec <id> <shell>" so "sudo jexec 1 sh". The error I get is "su: pam_start: System error". I have compared pam.d/su on the host and the jail and they look identical to me.

I have also done the following:

sudo pkg --jail=testingjail upgrade -f

sudo pkg --jail=testingjail check -Ba

sudo pkg --jail=testingjail check -da

"sudo ezjail-admin console testingjail" returns error 1

"jls" reports all the jails as running with correct path, ip, etc

found the following in the jails auth.log "in openpam_load_module(): no pam_opie.so found" that should narrow it down a bit

fgrep opie /etc/pam.d/*

(inside each jail)

 your.ls/amkqa

Title: Paste.to

seems to be present

you need to get rid of those

Is opie a one time password module for pam?

was

I wonder where that came from. I have no recollection of installing that but these jails have been living with me for a long time. They are like pets :-)

My jails are coming up one by one now. Removing "opie" from pam.d/* in the jails seems to do the trick.

what was the reasoning for not enabling block-cloning by default?

are there any problems with the feature?

(I mean with zfs 2.2 in 14.0-RELEASE)

oooh, checked the release notes for zfs 2.2, and

... a block cloning bug (#15526) that can result in data corruption ...

I guess that answers my question

this sucks, I was really looking forward to that feature :(

which public pgp key server has the freebsd key to verify the checksums for 14.0?

as usual this feature will be better in future

can someone sign the checksums with a valid pgp key?

voy4g3r2: have you figured it out and do you have a specific howto/tutorial to recommend? I'm about to embark in the netatalk/tm journey myself

clemens3: The release announcement contains information about the signatures. freebsd.org/releases/14.0R/announce

Title: FreeBSD 14.0-RELEASE Announcement | The FreeBSD Project

Title: OpenPGP Keys | FreeBSD Documentation Portal

Does core FreeBSD not support wifi roaming?

clemens3: They public keys you are searching for should be available via the link above

I've never found hashes of files to be useful

I guess it has a place if you're using a mirror

veg: I have this running locally since a while, but dan.langille.org/2023/09/24/moving-…ecting-my-macbook-to-zfs-on-freebsd and his earlier dan.langille.org/2017/04/19/creatin…an-apple-time-capsule-using-freebsd covers it

Title: Moving time capsule from host to jail and connecting my MacBook to ZFS on FreeBSD – Dan Langille's Other Diary

dch: And Mine hasn't been working lately, I have not figured out why.

dvl I should check ours and see how its going. I tell the kids, the mac has no backups, keep your work on the server where it belongs...

thanks, I love the 2023 time tag :)

thorre: thanks, will look into the pgpkeys page!

Remilia: Have you figured it out? (Periodic issue) - I wonder.

thorre: well, it has a lot of data, importing. let's see...

tercaL: no

veg: yes i was, i used this as a baseline blog.khubla.com/freebsd/time-machine-backups-using-freebsd-zfs and here is my conf i got to.. not 100% aligned to the entry but gives an idea: bsd.to/Hnqj

it is even more shocking today

Title: dpaste/Hnqj (Plain Text)

i may end up doing a little more playing.. but first i have to move the 230 gig of stuff :)

but I found what is using 100% of a core, it is the kernel

at 03:00 today load avg jumped by 1.0 like every night since the 14.0 upgrade but this time it never dropped back down

veg: i ony have 1 user, using time machine, and the CNID database and other stuff seemed excessive..

plus i was having headaches with getting it to work.. so i stripped it down till it worked :)

Remilia: Is that a VPS? An instance of a vmware or such kind?

it is a guaranteed-cores VPS from netcup

Remilia: Could you please try adding this into rc.conf file and check one night;

devmatch_blacklist="virtio_random.ko"

and ofc, reboot the system after adding it

the module is not loaded

it does not show up in kldstat at the moment

kldstat or kldlist wouldn't list it.

as it's embedded - I think.

once, I solved my VPS, kernel using 100% of a core, with that line.

Still having it.

the hypervisor is KVM hmm

weird that I had zero issues with 13.2

thorre: ok, it works! thanks

clemens3: glad I could help you.

tercaL: the man page for virtio_random says you need to explicitly include it in your configuration for it to be built in

Remilia: bugs.freebsd.org/bugzilla//show_bug.cgi?id=254513

Title: 254513 – virtio_random: random_harvestq spinning on a CPU with Q35 virtio random device

tercaL: I think this is great because this bug was fixed and because I used 13.x without issues before and because my kern.conftxt does not include any mentions of virtio_random

and because CPU usage is not in rand_harvestq

thorre: if I may do an addendum:). the first link about the announcement, I don't see a link to the second keyring page?! Maybe I overlooked it. If it is missing, you have to go to the public key servers, which are not up to date. While with the information you gave me it is all fine.. let me look again.

tercaL: i think we should close that. it's fixed in 14.0. although maybe we should backport it to 13.x

I found the page via a webserarch. Usually I am satisfied with checking the MD5 sum after a iso download.

Remilia: try the hotkernel script (docs.freebsd.org/en/books/handbook/dtrace/#dtrace-using)

And yes, "just because you think that they are not out to get you does not mean that they are not out to get you".

Title: Chapter 27. DTrace | FreeBSD Documentation Portal

bugs.freebsd.org/bugzilla/show_bug.cgi?id=269823#c13 i already said that here

Title: 269823 – rand_harvest produces 100%CPU on 1 CPU with virtio_random.ko loaded

yuripv: I get kernel`acpi_cpu_c1 254458 82.5%

and if you use -m for modules?

the first or the second link?

sorry, redundant line

yuripv: if you mean hotkernel -m: kernel 197287 99.8%

thorre: ok, yeah, so you knew this keyring existed, I search for key server and couldn't find anything

so imho this info is missing on the announcement page

yuripv: to rewind back a bit, this started at 03:00 (periodic run) on the night after I upgraded to 14.0-RELEASE, and until today, the spike usually lasted no more than 2.5 hours

Remilia: that really looks familiar, I don't know if I've seen this before myself or just in bugzilla

yuripv: upgrade was on the evening of the 18th of November: i.koumakan.jp/2023-11-25/1700919264.png

you can see that before that, periodic tasks did spike the usage but not for long

this is a netcup 'root server' VPS with guaranteed cores, seems to be under KVM

I feel responsible cuz i touched some of that code

(no, wait, that was something else)

well, acpi_cpu_c1() is "idle"

top shows 'swapin' as the state but that probably means nothing

clemens3: Maybe it's time to submit a problem report? docs.freebsd.org/en/articles/problem-reports

Title: Writing FreeBSD Problem Reports | FreeBSD Documentation Portal

yuripv: the other high % are kernel`lock_delay 1147 4.1% and kernel`vnlru_free_impl 3002 10.7%

Remilia: not running out of memory and heavily swapping?

Mem: 7578M Active, 16G Inact, 3140M Laundry, 3701M Wired, 764K Buf, 695M Free / Swap: 32G Total, 633M Used, 31G Free, 1% Inuse

with 16G inactive it probably should not swap much?

yuripv: systat -vmstat shows no swap paging but lots under 'VN PAGER' which I am not sure what is

What is the default/standard/shipping version of python in freebsd 14?

it is not tied to freebsd version

There's no python in FreeBSDs base system.

thorre: it is 3.9 so far

Ok, thank you angry_vincent. It has not changed since 13.2 then.

I am not sure it is 3.9

The meta-port called lang/python is 3.9, but DEFAULT_VERSION in make.conf(5) can change that if you build stuff.

Remilia: it is.

Remilia: i guess it has something to do with vnodes (see also the vnlru_free_impl above)

oh

thorre: again, it has nothing to do with freebsd version

debdrup: That was the info I was searching for. Thank you.

are you running out kern.maxvnodes may be?

I was looking at the wrong package

yuripv: I am currently looking at it and I guess I might be

thorre: you can also check freshports.org/lang/python#packages as an example to see what version are available on what specific ABI tuples.

Title: FreshPorts -- lang/python: "meta-port" for the default version of Python interpreter

but I have no idea *why*

Remilia: cgit.freebsd.org/ports/tree/Mk/bsd.default-versions.mk#n142

Title: bsd.default-versions.mk « Mk - ports - FreeBSD ports tree

debdrup: Teach a man to fish ....

angry_vincent: that's where a ports committer would change it, yes - but DEFAULT_VERSION in make.conf(5) is where a user should change it.

yuripv: bumped maxvnodes three times but it did not change the picture

Oh, it's DEFAULT_VERSIONS

think I will try stopping some jails and see which one does this

Title: Ports/DEFAULT_VERSIONS - FreeBSD Wiki

It's rather handy to be able to set versions like that for an entire tree of managed software if you're custom-building ports.

thorre: Yeah, that's always been my basic philosophy when helping people :)

vfs.freevnodes: 364540 / vfs.numvnodes: 599039 / vfs.wantfreevnodes: 375000

kern.maxvnodes is 1500000

If people are receptive to it, it makes it less likely that they need help in the future, and more likely they might be able to help someone else.

thorre: ok, thanks!

will look into it

veg: heads up the first backup takes FOREVER..

yuripv: I am starting to suspect that something fails to release vnodes

debdrup: Thank you, I appreciate it.

debdrup: i know

actually it may of been he OneDrive location that got in there that as causing headaches.. removed that directory and he backups are MUCH faster

yuripv: post-reboot, vfs.numvnodes is 92473 and CPU usage is fine

but stopping jails and all services prior to reboot did not help

veg: man.freebsd.org/cgi/man.cgi?query=afp.conf <-this is also very helpful to read through, some tweaking was possible with stuff.. like i never under mimic model option until that conf explained it more

Title: afp.conf

Title: some copied files are corrupted (chunks replaced by zeros) · Issue #15526 · openzfs/zfs · GitHub

15526 – [NEW PORT] security/pgpgpg: a wrapper for GnuPG to emulate PGP 2.6 bugs.freebsd.org/bugzilla/show_bug.cgi?id=15526

looks like a zfs corruption bug

looks like freebsd has vfs.zfs.bclone_enabled = 0 by default. it's not clear if that avoids it though. i set vfs.zfs.dmu_offset_next_sync = 0 (from default 1).

oh, this was posted yesterday too I think

I didn't upgrade my pool yet, so, apparently, I still have feature@block_cloning disabled.

I am in FreeBSD 14.0 in qemu, and my native resolution is 1280x1024. When starting xorg, it is 1920x1024 or something like this. I can set it with xrandr to 1280x768, but when i try 1280x1024 it says no such resolution.

any hint how I can fix that?

Probably a missing driver ? Do you have the correct driver installed and running ? Also is your xorg.conf setup correctly ?

didn't setup any xorg.conf, using qemu on linux

ahh you're running on FreeBSD which is running in qemu which is running on Linux ?

yes

and i searched, zero xorg.conf exist

but using openbox startx worked fine

just too high resolution

i think i used -vga std option for qemu

yeah that probably inherited from the system

had similar problem on older netbsd installation

but it worked fine on openbsd

all in qemu on linux

using xrandr --size 1280x1024

there was some trickery adding the resolution via 2-3 command line commands

Not sure. Its been over a decade since I've used X11

but last time that didn't not succeed, so i was asking again today if someone has a pointer

ok, wayland user, too bad:)

:-) have never used wayland in my life.

simple, straight forward command line

If I had to use a windowing system, would be X11 + (vtwm || stumpwm )

oh, cool, tty/tmux heavy user myself

i switch between 12 consoles, two of which run X11

home PC has to be Windows for work reasons

I said goodbye to Windows when 95 was still state of the art

thought liked the user interface (except the command line)

yeah same kind of reasoning for me as well Remilia, though for me its Mac at home and at work.

when 95 was state of the art we had no Intel-compatible systems at home

SPARCstations and SGI stuff only

the love for motif

correction

I had an Intel-compatible PC I made myself because my uncle told me I am not getting a real PC until I solder one

heh

but it would definitely not run 95, it was a Z80A

CP/M 2.2 power

greybeard

ha ha

do not have a beard but yes I was born in 1986

yuripv: are vnode counts supposed to just rise?

vfs.numvnodes was around 50k shortly after reboot, it is 258k now

wish I had stats from before the upgrade

Remilia: I think they should be increasing

endlessly?

not sure about endlessly

vnodes as I recall is just a struct, similar to an inode but at vfs layer. as long as files are being created it will grow. is there something that's writting to the file system ?

mns: not really

I thought vfs.numvnodes is the number of active vnodes in use

and consequently that if I stop most processes, the number should drop

not active processes, no

Remilia: not sure how accurate this is: wiki.freebsd.org/BasicVfsConcepts?h…ghlight=%28%5CbCategoryStale%5Cb%29

Title: BasicVfsConcepts - FreeBSD Wiki

I don't have an UEFI system but trying to help a friend of mine who has FreeBSD with UEFI enabled system, for EFI system update after updating from 13.2 to 14.0. Wanted to ask his ada0p1 has efi partition, is it possible to mount that partition while the system is already running?

Like, to do mount_msdosfs /dev/ada0p1 /boot/efi and cp /boot/loader.efi /boot/efi/EFI/BOOT/Bootx64.efi and the rest..

or if "/boot/efi" cannot be written, perhaps mkdir /tmp/efi and then mount_msdosfs /dev/ada0p1 /tmp/efi

Or perhaps, in order to update EFI system partitions, he should boot into single-user mode or directly by a usb stick and do the mounting there?

why would you not be able to mount the EFI partition?

on my ex-13.x system it is in fstab and I do not recall adding it there tbh

was a fresh install somewhere in June this year

'/dev/gpt/efiboot0 /boot/efi msdosfs rw 2 2'

still stuck on the fact that for 5+ years i never noticed that powerd wasn't throttling the cpus

Macer: it doesn't ?

mns: i had to disable powerd and install poewrdxx

which seems to work fine. but i never noticed the cpus were maxed out at 3GHz.. now they're staying around 1.6GHz.. that's huge

that is huge

what CPUs ? which system ?

I'm running powerd so curious

some ancient xeons.. i think those are 5560s

though it seems to be working in my case as far as I can tell

Macer: IIRC, powerd tells you on startup if it'll be able to do anything for your system

tercaL: Did you see "Upgrading from Previous Releases of FreeBSD" in the release notes?: freebsd.org/releases/14.0R/relnotes The instructions for updating the EFI loader worked fine for me.

Title: FreeBSD 14.0-RELEASE Release Notes | The FreeBSD Project

X5670

meena: i didn't see that.. all i saw was it starting up when the service was started

either way. powerdxx works just fine.. 1.6Ghz is the minimum the CPU can go and they're pretty much stuck there since i only use it as a nas

but yah for like 5 years they were pegged @ 3GHz lol

i only happened to enable the htop frequency stuff just because.. on a whim.. and noticed they were running around 3GHz and weren't coming off of it even though the cores were mostly idle

how do you enable the "frequency stuff" in htop ?

it's in setup

display options has an option to show the cpu frequency

AHH OK

oops caps lock

the fbsd manual has a command for it too

but i wouldn't have even bothered.. i just assumed powerd would handle it and enabled it and forgot about it

and just by chance happened to notice it wasn't throttling heh

i was also curious about swap. i guess the standard is to segregate a 2GB partition per disk for swap but how would you actually set that up? i guess most people mirror it?

oh nice, it has temperature there as well, though doesn't show me anything for temp.

yah i didn't even try with temp heh

but i was wondering what the freqs were looking like and when i turned it on everything was maxed out lol

i probably lost like $500 over that span of time because of it

you can create a stripe zpool over spare space on multiple drives and create a raw dataset for swap in that pool

that would likely be more effective than swapon'ing multiple separate partitions

or you could do a raidz

Title: 275308 – EN tracking issue for potential ZFS data corruption

hide your data too

Macer, Thank you for the mention of powerdxx as that will help me hugely on the laptop. I have listed power management on the laptop as a problem and I think you just solved it for me.

torpedo tubes 2 and 4 opening

rwp: that's good. like i said though it was just on a whim. so i guess we both got lucky lol

Macer: yeah so seems like powerd works for me as far as I can tell, in that my CPU is running on the lower end of its frequency range, especially when idle

I am not currently using FreeBSD on my laptop and power management and battery life is one of the reasons. This will fix one of them.

mns: yah i guess it's hit or miss with either or

in my case it wasn't working and i just expected it to and never checked because i suck

now to figure out why temperature doesn't show up in htop

torpedo tubes 2 and 4 closing

mns: let me know how that works out. i can't even get that work in debian :)

Macer: in all honesty, there are just too many things to keep track of

(work properly)

My desktop also has the same problem but it's less of a problem here. My desktop has been running at 3.3GHz with powerd running. Disabled powerd, installed enabled started powerdxx, now running at 1.6GHz.

lm-sensors sees it but htop doesn't

rwp: exactly the same as my isilon :)

and that's a dual xeon so running all the cores at 3GHz ... the money lost there .. it's just a nas. most of its workload is really light weight

I'll put a Kill-a-Watt meter on a box and measure the power difference between the two configurations. But fortunately for me here the desktop is a lower power machine at the base.

rwp: i've been running it like that for like 5 years

that may not be true though. i was running freenas for a while so maybe it was working for that.. that was at least 2 of the years before i decided to go vanilla freebsd

but i guess i'll never know.

mns is correct. There is too much to keep track of! Impossible to do it all. We all just do the best that we can.

well i mean with fbsd it's usually just set it up and forget about it until an update / upgrade

and even then the releases have such a long lifespan that you can probably forget about it for years if it's set up properly

Since mns says powerd is frequency scaling there it means that it is hardware dependent. And other than battery life on a laptop it's not immediately obvious elsewhere.

Also looking at the various powerd operating modes it might be that because my system is plugged in on AC mains that it was intentionally running higher performance. I will need to try some of the other modes and see if perhaps configuring it differently would work better here.

mns, powerd++ references the temperature sensor as dev.cpu.%d.coretemp.tjmax and on my system "sysctl dev.cpu" does not report any temperature sensors. Probably needs a module that is not loaded.

I had to "kldload coretemp" in order to get temperature sensors online.

rwp: In my /etc/rc.conf: powerd_flags="-a adaptive" is all I have.

let me try the coretemp

On my desktop upon which I am working now the battery power status is "unknown" so that probably affects something. I need to boot up the laptop and test there.

Running "powerd++ --verbose -f" and then interrupting it after it prints out the initial configuration is informative.

It tells me that hw.acpi.acline is not available. And that explains why I see "uknown" in the display of that field.

Otherwise letting it run and then stressing up the machine with a handful of "yes | sha256 &" jobs and watching the result is informative. Then killing those off (fg, Control-C, repeat) returns things to normal.

rwp: I loaded coretemp, but no temp sensors listed for 'sysctl dev.cpu', maybe its not just coretemp that's needed. Will need to go through htop manual

It's almost certainly hardware dependent. I have an Intel cpu system here and "coretemp" makes me think it requires an Intel.

ok so according to the htop(1) man page, need to have libsensor installed besides coretemp. I can't find the libsensor package in ports though

I have AMD CPU

amdtemp then?

wait, "core"temp as in Intel Core ? I was thinking "core" as in core of the kernel or something along those lines

I just have this vague memory that "coretemp" is an Intel specific module, but my memory is often vague so don't put much weight to it.

coretemp – device driver for Intel Core on-die digital thermal sensor

amdtemp – device driver for AMD processor on-die digital thermal sensor

(from man pages :)

bingo !!

it works! I've got temp

yuripv: thanks. I was interpreting "core" differently

now I got to make sure I load it in all the time

Thank you everyone! Today has been a good learning and system improvement day. At least three things are better here now. Time to take a break before I really break something! BBIAB.

now to figure out what AMD family the CPU falls under

should be in dmesg (or /var/run/dmesg.boot)

There is always the "lscpu" port too.

Try "sysctl hw.model".

yeah found it with dmesg, Family: 0x15

rwp: that's different from AMD Family apparently

hmm wonder what other AMD related modules there are

apropos amd

should nearly do it

(or apropos -s4 amd if you are looking for drivers only)

yeah that gives me enough yuripv

5 hours of uptime and at 317k vnodes, unsure how bad these stats are: paste.ee/p/uEB2g

Title: Paste.ee - View paste uEB2g

though vnlru did not wake up yet

at least not in a noticeable way

Remilia: how did you collect these stats ? I dont get anything for 'sysctl vfs.vnode' and for 'sysctl vfs' I don't get any vnode listing

mns: stats might live in a different name space. you might wanna check sysctl -a

mns: `sysctl -a |egrep "^vfs.vnode"`

sysctl does not accept 'prefixes' or wildcards

you must list every variable or use -a

yeah it does accept prefixes

I never got it to output anything matching a prefix before

did it change in the past 15 years?

oh you're right

I dunno, I've only been around for 13-6 years

wild! Zooey learns something new every day, to quote Granblue Fantasy

meena: last time I tried this it was 4.3-STABLE

maybe I did something wrong at the time

Remilia: you need to become more daring, or forgetful

mns: btw which version are you on?

here's what I get paste.ee/p/TquCb

Title: Paste.ee - View paste TquCb

Remilia: I'm on 13.2-RELEASE-p4

14 here

ahhhh ok

mns: do you get (legacy) if you `sysctl -d kern.maxvnodes`?

kern.maxvnodes: Target for maximum number of vnodes (legacy)

yes that's the output I get, just without the '(legacy)' part

mns: seems like this is a very fresh change then

seems like it

Remilia: reading freebsd.org/releases/4.3R/announce and it feels like a very different time

Title: FreeBSD 4.3 Announcement | The FreeBSD Project

It's time to fax my order for the 14.0 cd set

vkarlsen: i think alpha and i386 are both bad candidates to ask for

meena :)

vkarlsen: Brother and Panasonic fax machines can use SIP \o/

NetBSD still got you covered tho

Remilia: Really? :D

the current models support PSTN, ISDN, SIP, and X.25

Remilia: That sounds like a fantastic way to send a letter

also I think there's a proprietary Brother extension that lets you send faxes in colour

How does SIP help with faxes ?

I just did an update to 14.0-RELEASE on a server with a zroot on a ZFS mirror and updated the first efi partion with the new loader.efi just fine, but I can't mount the second. mount_msdosfs throws me a "invalid argument". what could I be missing?

mns: in some very rare cases you might not have access to a landline, and SIP solves that issue

oh derp. it's already mirrored....

Remilia: so that's how all those online fax places do it, they're using SIP

some use fax modem pools

there's currently a petition going on here in the country side of Ireland to please keep landlines

why?

because they are accessible

and always available

because they don't have internet connections ?

mns: no, because cellular networks are not as reliable or accessible

you got me before I could correct myself.

a landline phone needs no power supply and it is trivial to backup-power a PBX

mns: when there's storm here, our mobile Internet goes to shit

mns: as an illustration, when I lived on the outskirts of StPetersburg in a relatively small settlement (5k pop) a storm tore down the 6 MV power line and the repairs took 3 days, cell towers went offline after 40 hours, the landlines were working throughout

(ADSL over those landlines was up as well)

6 MW, I presume

flatdog: megavolt

last I checked volt was V

true, like on the good-old tram lines. Got it

trams usually run on 580 V

some, maybe

in most of eastern Europe it used to be 580 and I think a few Japanese networks are 580 too

no idea way though \o/

is it still needed to add the partitions yo use as swap to /etc/fstab? it seems it mounts them auomatically?

weust: I have it in my /etc/fstab

I had the old NVME disk naming in there, but the the /dev/nda was was swap enabled automatically.

I don't know much about NVME

weust: the installer adds swap partitions to fstab

so I am not touching that

Remilia: For non-NVME drives that won't be a problem. But 14 changed the naming of those disks

'/dev/nda' sent me tbh

From the relnotes:"NVMe disks are now nda devices by default, for example nda0; see nda(4)."

non-disclosure agreements in your devices

haha yeah

We don't talk about those here

(and also not in #freebsd-social, in case you're wondering)

meena: you mean we don't disclose information about those

so i changed gpt labels to zfs partitions after the fact.. and i want to use them vs using device names (da12p2 ... etc).. but the pool is in use and the gpt labels don't appear in /dev/gpt after a reboot

is there some sort of magic i need to perform here in order to use the gpt labels?

to which extend, current FreeBSD socket C programming is similar to BSD4.2?

very

even to the point of old examples working fine ?

-out of the box I mean-

simple ones, yes

I read a column about how BSD4.2 has recently introduced its network IPC mechanisms, namely sockets. Mentions that Xerox Network Systems and IBM SNA might be alternative domains to UNIX system domains and Internet (i think it refers to the socket type AF_UNIX and AF_INET) -the year is 1985-

yes, that's about when I was using 4.2

and yes, the socket paradigm has basically not changed in 40 years

I am testing the code that is included in the article, even the header files seem to be in place

the code is pre ANSI

they never quite did sort out any distinction between AF and PF, mind

what is AF and PF suppoused to be?

address family, protocol family

it seems like protocol and addressing were mean to be different things, the column talks about "The arrchitecture of 4.2 provides an extensible set of communication domains or standardized address formats. The two curerntly implemented domains are the UNIX system [AF_UNIX] and the internet [AF_INET"

"Sockets using the Xerox Network systems (XNS) protocol would likely be more efficient and easier to use than those using TCP, in fact, a socket type (SOCK_SEQPACKET) has already been allocated for the XND protocol"

It is my understanding that back then it was not clear that the ARPA protocols were mean to be the de facto standards one

there were multiple camps. Exciting times

I am finding the UNIX magazines of those times pretty interesting. Reading what was being published and discussed back then is different from the background noise that remained afterwards. Also I am surprised of the so many UNIX vendors that were back then. I am also surprised on the amount of commercial software that has been lost, there is no trace of all those productivity, office software,

databases, etc.

Macer, A detail that you might be missing is that a device might initially be exposed by various paths but as soon as any of those are used then the other alternative paths are immediately removed.

So if you attach to /dev/da0p2 for example then any associated GPT label disappears immediately. Enforcing a singleton use policy at the kernel level.

In order to use the gpt labels one needs to import the array using those labels. You can't boot off of them and then change.

So I think the only way is either one disk at a time as a "replacement" or booting the installer media and doing an "import -d /dev/gpt" on the array to force those names to be used.

rwp: yah it seems like fbsd 'hides' them

but zfs (newer zfs) allows you to rename devices on the fly .. so they pop up in /dev/gpt after reboot

uskerine, jgh, If you were ever using HP-UX then a tidbit lost in time is that I was told when I talked with the kernel guys there while I was debugging a tty problem that HP-UX was based off the BSD4.2 kernel not the 4.3 and that's why the tty driver was different there. So I guess I have used the BSD4.2 based kernel a lot by using HP-UX.

i tested it on a slop pool

and yah.. seems like the gpt devices are showing up after a zfs rename of the device

Macer, I am still using zfs-2.1.9 so I guess I will need to wait for 2.2.1 for that feature.

yah. guess it's a 2.2 feature lol

i just changed a couple more devices to the gpt labels in a pool and reboot just to make sure it doesn't come off the rails

Macer, Did you see that earlier mns and I learned about coretemp & amdtemp after you dropped off? I now have cpu temperature sensors enabled now! Cool beans!

rwp: no .. missed that heh

i was too busy with this zfs mess :)

i'm in the home sretch now though

renaming the devices in zfs and reboot makes them show up so i guess i can do the rest of them now

rwp from the last commercial UNIXes I only used Sun/Solaris as a user around mid 90s (it was still widely used as workstation for EDA software). Linux was spreading quickly all around the place though.

But I do remember seeing the HP-UX workstations, they were used for microwave/radio simulation

PA-RISC had a pretty good level of floating point "grunt" performance during that time. My perspective was cranking through electrical SPICE simulations.

But the tty driver was the funky 4.2 one. select(2) operates differently there on non-blocking file descriptors. I needed "expect" to work and ported it.

I loved the sparcs, not so much the PA-RISC, and also liked the POWER IBM ones, along with Alpha of course :-)

SPARC, Alpha, POWER and then PA-RISC would be my order of things

Honestly I will be the feelings was really for the OS on the software side of things. HP-UX out of the box was pretty bare. I spent a lot of time porting software to it. But SunOS/Solaris had nicer software with more features, and a lot more bugs.

SPARC was a nice idea initially, but fell behind chronically as time went on. IMHO the boss of Sun Microelectronics should have been fired about three times

What happened to all software written for System V/BSD4.2? Did each one of the HW vendors require recompiling or at least when underlying CPU was the same it was not needed?

I am curious about the commercial business software, apparently there were a vast amount of commercial productivity and database software

Cheyenne: what's the failure? floating is always a pain in the ass .... debug and optimized builds often give different results

rwp: SunOS and Ultrix usually had better open source software, as they were more accessible on campuses. Everything got ported to the other unix vendors after the fact. That was true most of the time

Yup!

uskerine: most of those got ported to Windows and Linux and then discontinued. At least non-Solaris and non-AIX OSes.

I once had a similar case, I was basically hitting some boundary condition, but not reproducible in debugging, because while data lives in the FPU it's stored with higher precision (80 bits IIRC), and when store occurs its rounded to 64

I wonder if single stepping causes that as well

mns one might think that the internet is an infinite repository of legacy software, but actually all that software seems to have been lost in the night of time.

uskerine: there's probably still some copy on a floppy in some basement/attic/storage room ... but fading fast! :D

dtomato I have had issues with floating point too, do you know that it is because of the FPU? I ended up removing the optimisation flags

uskerine: not everything needs to be kept around. Just the stuff I'm interested in, that's all :-)

uskerine: it depends on what the problem actually is, and some weird issues can crop up from language and compiler flags, but in this case it sounds like manually written assembly, which is why I enquired about the exact nature, because as you know, floating point, there be dragons :D

I do not recall the exact issue, but back then it was extremelly tricky to debug, we ended up thinking that it was an intermitent bug in gcc (we found something documented). But I was never sure about it. What I remember is that removing the optimisation made the issue disappear

I'm also not going to claim deep expertise, i've just had my hair singed by them :D

uskerine: if I remember correctly -fast-math is often cheeky because it allows the compiler to be quite loose with order of floating operations (which matter) and code optimizations/foldings

mns sure thing, but I have recently being reading these old UNIX magazines and it is remarkable how much business software was available for System V and BSD back then. The use case of a few employees sharing the same server for accounting, databases, etc was apparently pretty widespread (of maybe that is what the ads tried to convey)

but none of that remains available anywhere, not even a single screenshot

dtomato that is good to know, you always assume things are going to behave normally

uskerine: the one we were relying on was `-ffloat-store`, which forces that reading out of FPU causing loss of precision, we needed it because client insisted new code give exactly bit for bit same values as old code (and we are talking generations difference) ... that seemd to mostly work, but even adding printf()s or basically the way you break

down your computations in code might influence the actual results :-/

so it wasn't realiable, you just have to accept that floating-point just breaks your maths and accept that it's chaos :D

uskerine: yeah I remember those magazines and what they showed. There was a use case for what you're referring to. A lot of the software just is no longer around, sitting in the vaults of the corporations. Like Lotus 1-2-3 which had a Unix version, but is now most likely in some vault at IBM. For the most part though, the Unix desktop were acting as front ends for mainframe based software back then.

You have examples of what you're finding in those Unix magazines and which magazines ?

Probably Byte for a lot of it.

There were many word processors and spreadsheets. Many databases too. Those seems like the most common ones. Then several compilers too. It seems that depending on the year the focus was on one thing. That is interesting to see as you can see now the whole sequence.

UNIX Review and Unix World more I'm guessing.

Yes I have been reading UNIX Review

I think it was better than UNIX World

yeah I still have some of my copies around too

they are now in archive.org. There are extremely good articles, for example one number covers the early history of UNIX, both inside AT&T and commercially

I had one where they had an article about how to test your new Unix system, by using things like grep on the dictionary, and things like that lol

there is also another article about networking that was remarkable insightful (back at that time), it is written by the owner of 3com