ant-x: whats wrong with twm?

rtprio, it behaves wrong with pop-up menus of e.g. FireFox. When I press the sandwich button, a pop-up menu is shown, but it disappears as soon as I shift the mouse cursor, unless I keep holding the LMB down.

^ This does not happen on Windows and I am sure is not meant to happen anywhere.

hello. i have an old laptop with intel ivybridge and HD4000 graphics. it was working well up until i updated to freebsd 15. now, it doesn't load i915kms on boot; i have to load it manually using kldload

i have kld_list="i915kms" in /etc/rc.conf, i also tried full path, kld_list="/boot/modules/i915kms.ko" (file does exist) and in both cases, i see a warning during boot that says: /etc/rc WARNING: can't load i915kms: No such file or directory

but then logging in and running sudo kldload i915kms does load it. any thoughts on why rc can't find the file? zfs filesystem if it matters...

i've been using the laptop since freebsd 11 and it was working fine until freebsd 14.3 and this issue only crept up after updating to freebsd 15...there weren't any obvious errors/failures during update to freebsd 15, followed the usual process as documented in handbook

using drm-kmod package

but you say `# kldload i915kms` works just fine

from cli, yes, it does

but not when rc tries to load it during boot

rc says file not found...even if i give it full path

you shouldn't have to give it a full path

right, it doesn't find it with or without

how did you upgrade to 15? freebsd-update?

yes

after booting, try running `service kld start`

we can try and diagnose that

ok

will disconnect to reboot, brb

ok, running 'service kld start' in cli loads it fine

and at boot there's no related errors?

during boot got that same error that it couldn't find file

i ran service kld start after boot in cli and that loaded it

should i run service kld enable or something so it runs during boot?

i need to step out to walk the dog, will be back in 20 min or so...

CorvetteZR1: it should always run

i don't know, sorry, maybe someone else knows

yea, this is weird...ok, i'll keep digging. thank you!

howdy

EHLO

anybody know if there is some issue with freebsd samba when using it to backup proxmox containers?

i keep getting error 70 bad address from proxmox and am not sure if that's fbsd samba or something

no issue AFAIK

i don't see these issues using nfs

NFS is better choice

i'm actually copying all my backups to another dataset that i'm exporting with nfs

just because of this weird bad address thing with samba

add certs for auth and you will be lucky

it's so random

it's to a local nas so i'm not too concerned about certs so much as i'm at a loss with this bad address thing from samba heh

INFO: zstd: error 70 : Write error : cannot write block : Bad address

i'm not sure if that is telling me it's a bad network address or bad block address

like maybe the zfs backend on the fbsd side is messing with how it's supposed to write to zfs?

NFS is perfect choice for FreeBSD, it's completely fine to share ZFS over NFS

i'm sure .. but i'd still like to sort out why the error is happenign with smb

*happening

are you piping samba through zstd ?

no. proxmox is piping tar to zstd to backup to smb

OK, so proxmox have problems with samba

yes. but i'm not sure which side is having the issue here. i'm guessing it's the proxmox side. i don't see these sorts of issues elsewhere

try NFS then, use nfsv4 and delegate the whole share to the client

but again. i don't know if it's telling me the bad address is a network hangup (which i'm not seeing at all) or if it's because when it tries to write to the fbsd samba share it's not writing correctly

yeah i already did that

Macer: you have to troubleshoot it

i'm moving the backups to a new dataset that i'm exporting with nfs.. but again... i'd like to sort out where the issue is with smb :)

because i've seen this on more than one proxmox server connecting to more than one fbsd server

sniff on the inrerfaces, take a deeper look what is going on filesystems etc

or better, workaround it by changing to NFS

yeah. i've been trying to find it.. it's very random

that's what i'm doing for now. moving it to nfs

just to make sure the backups work

good choice

but i'd still... like to find out what is going on with samba and determine if it is a proxmox problem, fbsd samba problem, freebsd zfs problem, or something else

i've been quadruple checking settings all over and monitoring logs on both sides to sort out where / why it's happening and so far the only thing i can see is the bad address error from proxmox

i don't see nic hangups etc

maybe i'll monitor the fbsd debug log

go ahead, troubleshoot it, increase debugging in samba and dive into the problem

but if you want reliable solution, use NFS

i'm wondering if maybe mount.cifs on the proxmox side is too aggressive on minor hangups lol

you can tune samba server, you can make mount.cifs less agressive

/192.168.1.13/proxmox on /mnt/pve/devil type cifs (rw,relatime,vers=3.1.1,cache=strict,upcall_target=app,username=backup,domain=DOMAIN.ORG,uid=0,noforceuid,gid=0,noforcegid,addr=192.168.1.13,file_mode=0755,dir_mode=0755,soft,nounix,serverino,mapposix,reparse=nfs,nativesocket,symlink=native,rsize=4194304,wsize=4194304,bsize=1048576,retrans=1,echo_interval=60,actimeo=1,closetimeo=1)

wow proxmox does a lot just to mount cifs

i'm looking at reparse=nfs and retrans=1

i'm wondering if zfs is losing the block during the backup from another write process

Macer: what is happening, you're backing up from proxmox to freebsd

?

Hi

since my upgrade to FreeBSD 15.0, my host regularly breaks with jail: nginx: getpwnam: Too many open files in system

i was thinking that bastille is the problem or a jail, but the whole system is affected

a reboot solve it

i just notice something strange

on Linux, syncthing only eats 63 FDs

i'm out of what i understand, so i asked chatgpt which tells me that freebsd kernel design is different and uses kqueue instead of inotify. kqueue needs one FD per watched file

is that true ?

i will wait for it to break again to verify the number of FDs

my question is : setting higher kern.maxfilesperproc is a workaround or a fix ?

rtprio: yes. from proxmox using smb to fbsd using samba

INFO: zstd: error 70 : Write error : cannot write block : Bad address

and this happens randomly

*fbsd using samba and zfs

eoli3n: inotify is 15.0, it is one of the new features man.freebsd.org/cgi/man.cgi?inotify(2)

mzar: hm, let's check if the dev knows it

eoli3n: Too many open files in system <- kern.maxfiles can be set only at boottime

nice

ok, so it's a proper fix before syncthing move to inotify ?

to increase it, i mean

I don't know, I am not using syncthing, but kern.maxfiles can be set from loader.conf, it's lodader tunable

eoli3n: you can try to patch the code, beging with allowing to build this part on FreeBSD github.com/fsnotify/fsnotify/blob/main/backend_inotify_test.go

i will let this to anybody more competent than me :)

but thanks for the flag

oh.. it's test, not right file

this one

eoli3n: you can learn by fixint it

yes i could, but i miss time for now

mzar: how do you propose 'fixing' the module that is specifically Linux only?

to fix*

considering fsnotify has backend_kqueue that is specifically Free/NetBSD + Mac OS only

Remilia: I'd begin by adding FreeBSD to the build

when making BE how do i include my pkg repository ( built by poudriere ) as well /usr/ports tree? would bectl create -r work? it is that /usr is not incuded in BE due to canmount=off for /usr

mzar: I'm sorry, what?

Remilia: mzars suggestion was to use inotify on FreeBSD instead of kqueue, not that that inotify backend is somehow broken

nimaje: I thought inotify is Linux only?

or is it like, via the emulator

read the man page mzar linked above, FreeBSD 15 has inotify and with it you can monitor a directory without needing an fd for every object in the directory (no idea why kqueue didn't get some handle for that use case or maybe it did and I just missed it)

oh

if the suggestion was instead to make it build on FreeBSD using devel/libinotify then it would have been useless, as devel/libinotify is just a wrapper around kqueue to have a inotify API to make porting easier, but it would have the same problem of needing many fds to monitor directories

dch: is oracle free tier now better with subscription? and are there more up-to-date guide for setting up freebsd image

angry_vincent: I recommend people to subscribe & upgrade to credit card, but then stay below the $$ monthly usage limit

setting up freebsd should be super easy, just choose the 15.0-RELEASE UFS or ZFS image from the web ui

if you're using command line I have tools to help

I think Installer is broken in 15.0-RELEASE

i just dd'ed the freebsd image over the oracle one from linux and restarted

checkpoint: what's wrong with it?

Remilia: recently I was trying to install 15.0 on Horizon N5 mini-PC, it could not connnect to network - neither wifi nor ethernet could connect and/or obtain IP from local DHCP server.

Remilia: same time 14.3-RELEASE worked like a charm on same mini-PC.

curiouser and curiouser

Has anyone noticed significant degradation in system responsiveness under heavy disk IO under 15-RELEASE? I'm noticing this in two wildly different systems: an Oracle Cloud arm64 instance using zfs, and a VM on Linux KVM using UFS. When disks start to saturate, ssh responsiveness suffers greatly.

This is utterly subjective, but more than on 14.

(KVM vm is amd64)

amigan: do you have ZFS?

oh

amigan: check `top -SH` and see if you have arc_prune at 100%

it's what completely devastates my server under tons of MediaWiki request

s

Remilia: nope on the ZFS VM. Most CPU is idle tasks, which makes no sense why it's such a dog