i'm feeling pretty cool right now. writing you from a system i archived and restored to a new OS drive. i'm shocked it just worked. :)

nice

isn't that a neat feeling when something that "prolly won't work as planned" actually does?

totally. what's crazy is how simple it is to do.

yeah I have a server on VULTR and they gave me more HD space, it blew me away how easy it was to expand the filesystem out to fill it out

i noticed when i booted the installer on a usb drive, i was asked for the geli passphrase on the OS device. is that expected?

yes

and there's no good way around it, even if you don't need it

If it finds old metadata, you're entering it or suffering through a bunch of prompts.

markmcb: asked by loader or by the kernel?

kevans: Is it ever by the kernel any more? I thought that default swapped some years ago.

It used to bite me fiercely because I was using USB keyboards.

kevans, i assume the loader. it was the usual prompt you get before anything else loads.

markmcb: bios or uefi boot?

mason: it's usually the loader, but it can be the kernel in some circumstances

IIRC, geom_eli_passphrase_prompt="NO" would force it back to the kernel.

uefi boot of a usb drive with the bootonly installer image

e.g., uefi loader /shouldn't/ prompt unless there's a geli partition on the same disk it was loaded from, iirc, these days

FreeBSD-13.2-RELEASE-amd64-bootonly.iso to be exact

flow was: i installed to a nvme ssd, rebooted, selected to boot again to the installer, was prompted for the passphrase, then booted into the live usb env.

you'd think chsh would check if the shell exists and/or is in /etc/shells before letting you do that, unix.stackexchange.com/questions/54…-file-or-directory-in-dragonfly-bsd

Title: shell - su: /bin/bash: No such file or directory in Dragonfly BSD - Unix & Linux Stack Exchange

meena: Things are getting complicated again...

I have got kde plasma installed on my older asus laptop. This laptop has got intel and nvidia hybrid gpu. So drm-kmod i915kms module and nvidia-driver390 needed to run kde. all was working fine. i have a bootable backup copy about this system on external SSD also but the main system is on 2.5 inch SSD in the laptop.

But that laptop suddenly died. Motherboard issue.RIP. 😔 So i have got a newer laptop Lenovo T480 with Intel Graphics only. No Nvidia at all. Just for curiosity i was placing the ssd inside from my older laptop and was trying to boot. And was booting into with KDE Plasma flawelessly. 😆 No issue at all! So i was trying to boot the backup OS also with the same configuration and also was booting

with no issue.

And after that was starting the problems...

the problem is i was installing two packages from ports collection. x11-fonts/mkfontscales and textproc/markdown. one of this two package fucked up xorg and kde....🙄 So after giving my password on sddm loginscreen the KDE logo was popping up but after i got black screen only. it was not not loading the desktop and taskbar. So i was switching to tty but also failed. Only hard reset did work

using the power button but that was causing serverauth.xxxx file does not exists error later on... I was doing pkg delete the mkfontscale and markdown packages and do pkg autoremove but that removing kde5 also.... and xorg was not working anymore. So i was reinstall xorg and kde5. And now xorg complains about serverauth.xxxx file. if i delete .Xauthority file then complains about that one

also...🤬😡 if i change enable_xauth=0 in /usr/local/bin/startx then not complains about serverauth anymore but has got a new problem: Segmentation fault for gpu driver 0 1 So i was removing nvidia-driver390 as i dont need anymore... maybe will help. Also I was removing the nvidia-modeset from rc.conf. But still having the Segmentation fault error....Any idea? 🤔🧐

The external SSD system does still boot on the newer laptop with nvidia config even if i no more have nvidia gpu....

Do you not have a ZFS "snapshot" to "rollback" to?

Hello

do you know how can I use xmodmap to switch two key of a keyboard?

tyler82: you could just paste a link to your forums post :D

parv: no i dont have...

yuripv: yes but i did post here first and the forum post had created just after this. :) So i could link this to the forum. :))

i see :)

tyler82, Segmentation fault could be due to missing library that was originally used; try rebuilding the driver if that is possible. In any case, I would start from scratch: remove all X11 things; then install the window manager, GPU driver, etc

parv: this is what i did.

reinstall xorg and kde5

So that link?

but what driver u mean? i dont need anymore nvidia...

sudo pkg delete xorg kde5 then pkg autoremove

this is what i did so far

shall i pkg delete drm-kmod & pkg autoremove also??

You need to stop guessing, build the software with debug symbols as described on wiki.freebsd.org/DebuggingPorts and then find out what's causing the issues with lldb.

Title: DebuggingPorts - FreeBSD Wiki

hi. I am writing my installation script for freebds. It fails on 3/4 so I decided to test it on VM and here is my probolem.

i have no idea how can I get access to my script from freebsd that is running as quest on vm.

i have tried share folder in virtualbox - it isn't work

that would be great to share as standard ext4 device.

any idea how to get access to this script?

using ssh / scp ?

no internet

or meybe you can tell me how can I turn internet on

I boot *.img and i use "shell" option

maybe another question

"gpart show" tells me there is usb disk under da1

once in the shell what prevent to set up the network interface manually with ifconfig / route ?

but i can't mount /dev/da1 - no such file or directory

there is no da* in /dev

checked because i even tried /dev/da1s1.

ls /dev says there is no /da and even ada

ok i am sorry I did something wrong

there is da1s1

gpart: No partitioning scheme found on geom /dev/da1. Create one first using 'gpart create'.

i destroyed /dev/da1 and trying to create portable partition on usb disk. I don't understand this error.

google isn't helpful

wait, if da1s1 existed then there must have been a partitioning scheme

you destroyed that one?

i created freebsd sheme

ow i can't add partition

gpart add -t freebsd -s 1G /dev/da1

gpart: invalid argument

i am googeling and cant find anything helpful

what are the exact gpart commands you ran

parv: if i remove xorg and all xorg related files and folders then im theory should not get authserver.xxxx error to get. So why am i having it? Probably because not all files and/or folders removed....But how could i purge it properly?? i was using autoremove also after pkg delete...

1) gpart destroy -F /dev/da1

2) gpart create -s bsd /dev/da1

3) gpart add -t freebsd -s 1G /dev/da1

gpart: invalid argument

you don't want the bsd scheme.

how caqn I create poirtable usb stick?

if you want the drive to be recognized by anything else, you want either the MBR or GPT schemes

i want temporary usb drive to share it with guest system on virtual box

then if you want the data to be readable on things other than freebsd, you have to decide on a filesystem type to use

tyler82, I do not know. If an X11 session is not being used, do not see any issue with ruthlessly removing various related files in home or in /tmp/. All I can suggest from here is to check logs; trace the processes that are started with "startx" or however you do that. When all else fails, also re-login. After that more logging & process tracing

(msdos being the most widely recognized but the least useful)

parv: does not sounds good... but okay...

thanks, mbr works

gpart is extremely primitive :|

It's probably more that you're simply not used to using it.

true

but for example 'gpart show' is one huge puke

very hard to read

gpart show da1 to show a specific disk

Any known concurrency bugs with tmpfs? I have a sporadic test failure in the mariadb testsuite that occurs frequently on freebsd 13 with `mount -t tmpfs tmpfs /tmp`, never anywhere else (including freebsd 12.3/tmpfs and freebsd13/notmpfs)

Looks like a writer thread writes 44 bytes and the end of a file, and occasionally reader thread reads 44 bytes of zero instead of correct data

Current repro requires running a complex test with multiple database processes until it fails, so not ideal for a bug report...

pretty sure the write takes an exclusive lock on the vnode.

can it be reproduced with ktrace active?

RhodiumToad: I can try, any pointers to how to use ktrace (I'm not that familiar with freebsd)

Currently trying to reproduce with truss, but not successful so far

ktrace -i somecommand will trace (output goes to ktrace.out in binary form) the specified command and all processes it spawns

ktrace -i -d -p somepid will trace an existing process and its current and future descendents

kdump will interpret the trace output

I'll try, thanks for the suggestion

Doesn't seem to reproduce when run with ktrace -tc :-(

I'll let it run for a bit more to see. Otherwise maybe I can at least record the IO pattern done to the file and see if I can make a small test program with similar access pattern to reproduce the issue...

unfortunately ktrace will inevitably slow it down a bit

-tci would have been better, wanted to see the actual data

Ah, yes we need the data too...

most of the other default trace options aren't much overhead, so better to leave them all on

tyler82 what 'authsever error' ?

rtprio: i have fixed it already. the problem was that i did not delete properly xorg

the correct command was pkg delete -xi xorg. Not just pkg delete xorg.

that deletes everything with xorg in the name

the package named "xorg" is just a metaport

"That's your problem, right there"

lots of xorg-related ports don't have xorg in the name, though

that command was deleting a bunch of additional files. after do autoremove and rebooting i was reinstall everything from scratch. Now all works fine.

RhodiumToad: Ok. i did not know that.

It's nice that that worked out, but it means you won't know what caused the issue, so you'll be hard-pressed to identify and fix it next time, without resorting to the same.

debdrup: xorg was causing the issue but i donno which package exactly responsible for the authserver.xxxx error.

reinstall seems a little excessive for a few files, but sure

pkg query -a "%n-%v %C" | awk '$2=="x11"' # lists every package in category "x11"

or awk '$2 ~ /^x11/' to also get packages in the various x11-foo categories like x11-fonts, x11-toolkits etc

however, if you just installed xorg to pull everything in, they'll all be marked auto-installed so autoremove will remove them once the xorg port is gone

i was expecting that also. but as u can see not happend

i was installing xorg via pkg not from ports

besides, the cool kids use wayland these days

rtprio: does KDE support wayland?

tyler82: KDE supports wayland, FreeBSD however doesn't support KDE Wayland yet.

kenrap: i mean that one. :)

anybody has tried restic backup manager?

(btw, I recommend you pop in #freebsd-desktop and chat with the folks there regarding KDE)

kenrap: noice

Some really enthusiastic devs there :)

kenrap: aham....

RhodiumToad: Ah, with some tweaks and patience I managed to reproduce it with ktrace default options, thanks for the help! Will analyze the result and see what's really going on

Indeed, it looks like the tmpfs is returning incorrect data in this case. A write(343 bytes), then a read(387 bytes) then a write(44 bytes). The read sees the length of the file including the second write, but the data from the second write is missing (it just reads zeros).

hm.

can you show the actual part of the dump - use kdump -E to get timestamps

oh and also -H

(3 different threads; and the read() returns before the second write())

also what fbsd version

(kernel version, specifically)

RhodiumToad: yes, I'll collect the information, and I can also make the full dump file available (60MB)

`freebsd-version -k` returns: 13.0-RELEASE

`uname -a`: FreeBSD freebsd-130-amd64 13.0-RELEASE FreeBSD 13.0-RELEASE #0 releng/13.0-n244733-ea31abc261f: Fri Apr 9 04:24:09 UTC 2021 root⊙rnfo:/usr/obj/usr/src/amd64.amd64/sys/GENERIC amd64

13.0-release with no patches?

RhodiumToad: Might be (or how can I check). This is a qemu/kvm VM setup for continuous integration, it might have been setup originally and never modified

ea31abc261f is definitely the original 13.0-release kernel.

I don't know if any of the 13.0 patches change the kernel but I bet they do

I believe 13.0 is out of support, too

so the question would be, can you reproduce on 13.2

Right, I'd need to install a new VM for that...

I searched for bugs mentioning tmpfs on bugs.freebsd.org, but didn't spot anything relevant

I don't recall any tmpfs fixes between 13.0 and 13.2, but then again, my memory is shite. otoh, i do remember quite a few vfs fixes (again, my memory shouldn't be trusted)

RhodiumToad: Thanks for the help

Here's the extract of the kdump output, showing how the read sees wrong data from a concurrent write to tmpfs file: bsd.to/D94o

Title: dpaste/D94o (Plain Text)

Is there a deterministic order to kernel modules loaded via /boot/loader.conf vs /etc/rc.conf?

_xor: rc.conf should come after loader.conf, and iirc, should load modules in order of appearance in kld_list

I would need to look at loader's code to say anything about that order

but there's basically just two possibilities: alphabetic and declaration order

rc might have other dependancies which could alter the order; but loader i would expect declaration order

RhodiumToad, meena: This C program reproduces the problem for me when run on tmpfs in FreeBSD 13.0 (in a VM) : bsd.to/3r5L

Title: dpaste/3r5L (C)

After typically a couple hundred iterations, it reads wrong data