rtprio: Wireguard is on the router and is NATing, but is also NATing traffic from the server, which therefore is causing traffic going to the server pass directly out without passing through the tunnel, because the wireguard endpoint == the server IP. I can think of one solution to this problem which would be split horizon dns, the DNS server matches against the wireguard subnet and returns the

private IP of the server on the network(s)

but this obviously would be a lot of work, so I am curious if there is any "hacks" to get it functioning now, and do it correctly later on

(My new server will have a different endpoint than the router which is how you should set it up, I was lazy and I have a block of /29 IPv4 addresses doing nothing :P)

i split my dns for this.

i know we can overprovision cpu with bhyve (like run 10 vms with 4 cores given to each even though i only have 24 cores) but can we overprovision ram?

we can overprovision disk with sparse vols in zfs or sparse flat files

ram is the 1 i don't know about

l00py: it doesn't support memory overcommit in the way that something like Hyper-V does where guests can balloon into available memory. if you don't wire the guest memory (bhyve -S) i believe it can be paged out, which is a sort of overcommit, but almost certainly not what you want

ya prolly not ok ty

I hope l00py is not wanting to overcommit RAM for VMs. I have suffered through using VMs where the hoster overcommited RAM and it's really painful and causes hard to diagnose slowness in the VM.

It's hard to diagnose because if you can only see the VM then the VM always thinks everything is perfect. All good. It is not paging itself. But the host is thrashing unseen on the host side.

Has anyone else had kernel panics with the 14.1 release? I even freebsd-updated to 14.1-p6 but I still get random panics on a new Alder Lake minipc (Intel N100).

FWIW the panics are mainly page faults, plus one pte related. Can provide more info if needed. Thanks.

kpel, Unfortunately that sounds like bad RAM. I would run it overnight with a memory test in order to check for it.

rwp: memtest86 thinks ram is ok. Did 4 passes without issue.

memtest86 is pretty good.

rwp: also, smartctl doesn't show disk issues

Are you using a graphics driver from binary pkg install? Sometimes that is a source of kernel panics due to it being a separate compilation.

There are various situations where the kernel API changes in incompatible ways and the pkg binary driver needs to be recompiled to match. I have experienced that in 13.

And that is the last of my knowledge about this. I have told you more than I know and will wait for others to jump in.

rwp: I am using i915kms. The thing is, I was seeing this issue even before I got X running. I see it even when I login over ssh. Anything that is io-intensive can trigger a panic. A pkg install with lots of dependencies, an nmap run... sometimes I get a panic simply by opening and closing chromium.

How are you loading the i915kms driver? In /etc/rc.conf in kld_list="i915kms" ?? In that case it will always load even when you are logging in from ssh.

rwp: I cloned the git repo and did make buildworld. Would that build i915kms as well or do I need to build the port?

I am using the radeonkms here and the radeonkms is a port and needs to be built separately. I don't know about the i915kms as I have different hardware. I expect is a port as well.

For a test if you can do it I suggest removing kld_list="i915kms" to avoid loading it, not starting X, using it only over ssh and see if the kernel panics stop happening. That would point me that direction if the panics don't happen when not loading that module.

rwp: I might give that a go later, thanks. I checked the freebsd forums and found a couple of similar panic reports but they seem to be years old. I wasn't sure if I'm looking at the same issue.

graphics/drm-stuff has been removed from the base long time ago

mzar: thanks

Then it is definitely a port and would need to be built separately.

The kernel module in ports causing problems is one of the well known problems.

rwp: will do, thanks for the tip

if drm module is the culprit, it will crash on load, immadiately

kpel: enable kernel core dumps and debug it

no, it doesn't crash immediately. The X session can happily stay up overnight with applications open. And then I'll do something trivial like open a second browser, or close an open browser, and everything wil freeze for a few seconds, then the reboot follows.

The binary pkgs are compiled on a build server that builds for the oldest supported active FreeBSD release. So for 14.1-RELEASE-p5 that is 14.1-RELEASE as in p0 I think. If any incompatible ABI changes have crept into the kernel since then it creates this mismatch problem.

kpel: you need sth like "dumpdev="/dev/swappartion" to make it working

it should be added to /etc/rc.conf

mzar: are you referring to the cores I have in /var/crash ?

yep

mzar, Actually I know from personal experience that I could load the radeonkms module and it would not crash until I started X and then it would crash immediately.

do you have them ?

no debug symbols :(

so I rebuilt GENERIC and now I have debug info for a couple of crashes

panics could be also related to faulty hardware

rum memtest at least

mzar: did that. Passed 4 runs

I even tried running a GENERIC-KASAN kernel for a few hours. I have an output file, I just don't know where to focus. It doesn't show any out-of-bounds access if I read it correctly.

would it help if I pasted it on pastebin?

it's only 4852 lines :P

hhelo

how do I switch from my compiled kernel to default, binary kernel with default config? Because I need to do freebsd-update

kpel: good

i think you can just re-build the generic kernel

kpel: maybe you want to test 14-2-BETA ? a lot of things were fixed and changed

rtprio: if I download freebsd-update, it updates the /usr/src/sys kernel sources as well?

so I can rebuilt from my config

ScrewDriver1337: you can tell freebsd-update what FreeBSD version you are using

you can even lie

yeah sure

but I just wanna download security patches and such

mzar: I'll add 14.2-BETA to the possible things to try, thanks. Right now I'm rebooting the "patient" without i915kms.

mzar: sometimes even a "find /usr -name '*cpu*'" over ssh would cause a panic. It looks like disk io/vfs related but I don't know enough to say so.

kpel: are you using ZFS ?

mzar: nope. ufs+ffs

mzar: /dev/ada0p2 on / (ufs, local, soft-updates, journaled soft-updates)

interesting

mzar: btw thx for the tip with freebsd-update fake version. Because I got cursed VM running 14.1-STABLE

is there a way to see which version of the OS compiled without doing installworld and installkernel?

you mean like `freebsd-version `?

well, i would like to know which version was compiled without running installworld and installkernel

to see if the compiled version is actually a significant update

I'm having trouble with a portable monitor (via DisplayLink). Is there a particular channel where it'd be appropriate to inquire about that?

from what i've seen, here seems like a good place fm2279

jauntyd: what are going to compare ? git hashes ? __FreeBSD_version numbers ?

jauntyd: take a look at /usr/src/sys/conf/newvers.sh /usr/src/sys/sys/param.h

Thanks, scoobybejesus. I'm trying to connect a USB monitor (an AOC E1659FWU). Upon plugging the device in, I see it recognized as a DisplayLink USB devince (in dmesg). I have loaded the udl kernel module. But nothing happens after the USB connection. I've read conflicting suggestions in the forums. Wondering if anybody here has wrestled with a display like this.

mzar: newvrs.sh is great. ty

hii

anyone offering vps?

ash_11: forums.freebsd.org/threads/freebsd-hosting-shared-vps-etc.21472

i need free

vps

ash_11: you can sign up for an oracle cloud account and get 2 amd64 vps or 1 big arm vps

free

does it require cc

$1 verification charge

:(

don't have cards

working

it's to prevent script k1ds

say, do I need to do something special to make sure that WOL works?

rtprio can you provide a registered account

im from ukraine and i cant

sorry, i'm already using my two vps

can you arrange one?

uh no. maybe use this opportunity to install freebsd on one of your own systems