After upgrading to 15.1 there's a message to run service local_unbound setup. Well... that deleted all my config. :( I use it internally for my local lan and I could rewrite it easily however I didn't expect to overwrite the old config.

pertho: Someone just posted a matrix of Intel support on the forums You might want to look there.

is there a list of invalid characters in network interface names?

kerneldove: ifconfig(8) man page: Name may not be longer than 15 characters. No other character restrictions listed.

even '`"? that's wild

is there anything wrong with renaming network interfaces to arbitrary strings? it would be to help me keep track of which is which. like ifconfig epair5 create; ifconfig epair5a name foo-bar

thermos: I suggest creating a dataset on the host and nullfs mounting it in the jail

Maybe you can unjail a dataset that was previously jailed, but I think they aren't meant to be unjailed

ifconfig em0 name eth0

no, no problem at all

ok cool ty!

kerneldove: i forget that bhyve adds the host to the description. does jails do not?

description: vnet0 host interface for Bastille jail dns0

bastille does

rtprio it does but i need to distinguish interfaces i give to jails. the handbook uses the 4th octet of its ipv4 ip, but not all of my jails have public ips. some only have private ips because they're 'behind' a nat jail that has the public ip and proxies for the other jails

what do you think about that?

but then it would still at least have a jail name, ya?

ya

so sounds like you're coming to my idea of using jail name as interface name?

but i still don't see the problem?

k if a jail doesn't have a public ip, what unit number to use for "epair"?

that's the problem

doesn't your jail manager assign one automatically?

in the handbook, last octet of its public ip is the {id} in jail config

i don't use a jail manager

i haven't found something like vm-bhyve but for jails

like bastille?

is it?

which i don't think is quite on par with vm-bhyve but does a lot of what i need

it does manage to pick non-conflicting epair names tho

what does it base them on?

it renames mine to ed0a_hostname

* automagically *

what's ed0?

er, e0a_ it's the epair

if you want to change a jail's hostname does it rename the if?

bastille rename dns0 dns1

did if change?

i'm sure it would, -a would restart it, which would presumably recreate the interface

ok well it's an interesting idea

i already base jail name on hostname, so if too woul dbe consistent with that

e0_ is an interesting convention too

i notice ed0a_hostname has no unit number at the end. is the 0 the unit number?

rtprio ?

kerneldove: this is how solved it to create my jails dpaste.com/EBXCQWSHJ

anybody running recent CURRENT *and* poudriere?

specifically, newer than e91e8ebefadcce9d57c8ff945ff70050cbbe1ce1

poudriere fails creating jails, with `jexec: jail_attach(6): Operation not permitted`

it seems to be specific to poudriere, normal jails can be created & attached to

nwe ty, which part should i refer to pls? it looks pretty much like the stock code in the handbook for setting up a vnet jail?

kerneldove: what are you trying to do? I skimmed history and its not clear

dch come up with a scheme for jail interfaces (epairs) that handles the case where the jail either/both 1. doesn't have a public ip address 2. has multiple epair interfaces. so the handbook code creates epairs for jails using the last octet of its ip address as the interface unit number, like epair{id} where id is the last octet. but that

convention doesn't handle case #1 or 2

fwiw i name jails based on their hostname, without anything following ".", so myjail.lan would be named myjail, so i was thinking maybe name interface {jailname}{unitnumber}, like myjail0, myjail1...

aah ok

so the script linked by nwe just shows you that you can create & rename the epairs afterwards to anything you like

e.g. `ifconfig epair12345 create`

or `ifconfig epair12345 create name kerneldove123`

note that this last one only renames the `a` end, so you still have `epair12345b` to rename afterwards

why would `ifconfig epair12345 create name kerneldove123` only rename the a end not the b end?

and where does the nwe script show renaming interface? i know the syntax but i'm not seeing it in the script

dch

kerneldove: they only alter the description. The renaming of only one end of the epair is arguably a bug

but by now, likely a bug that people depend on, so unless somebody comes up with a clever way to handle it, a bug we can't fix without breaking existing setups

man that's fkd up

and ya definitely a bug. sheesh that sux

so would you run ifconfig epair12345 name foobar \n ifconfig epair12345b foobarb?

yup

like why not just introduce fix in 16 and put in a migration note?

we can't keep cruft around indefinitely

probably it would cause more frustration and problems

e.g. everybody using a pfsense / opnsense or similar tool now needs a migration script

there's no easy way out of this\

but that's what major versions are for, an opportunity to make breaking changes that improve the OS

i respect not having churn for churn sake, but fixing a clear bug is not that

put yourself in the position of an org with 50 firewalls that are all offline because we made that change

and now you need to send out 50 engineers to fix that

its easy to say "clear bug, fix it"

its a lot harder to find a solution that doesn't break existing setups

you're welcome to file a bug for it, and maybe somebody can find a clever way to allow renaming both ends of the epair

maybe there's space for another flag

anyone upgrading major version without reading migration doc and making appropriate changes deserves that. the OS and its users shouldn't suffer due to lazy sloppy ppl

there's never enough protection to protect them anyway, so no bother

a flag could be a good compromise

hmm.. found the link: forums.freebsd.org/threads/intel-gpu-support-matrix.102846 but how do I bloody read it? Says all Alder Lake gpus are supported, but I can't get drm to enable on Intel Graphics (ADL-N) - Alder Lake-N

How do I know if it's an Xe2/Xe3 (Bxxx/Cxxx) card? Seems to not have a version like that

FreeBSDFoundation/proj-laptop #111 not much info here

ah ha.. linux-hardware.org/?id=pci:8086-46d3 Supported in Linux 6.9 and newer.. 15.0-RELEASE is 6.6 isn't it? What does 15.1-RELEASE have?

Don't suppose 15.1 will get 6.12?

i'd open a bugzilla bug with the information you collected, it'd help get that in

there's a strong push to get solid laptop support

kerneldove: drm-612-kmod was added last month but I'm not sure if it made it into 15.1

anyone know offhand what OSVERSION of 15.1-RELEASE is?

that's awesome

freebsd is so good in so many ways, that's why we need to have the courage to fix serious bugs dch!

pertho: OSVERSION=1501000

pertho: in this gist is a script called gather.sh gist.github.com/dch/68dd6f8d44dd76c558908e71dc3e1f87#file-gather-sh most of that info will be useful (pciconf, dmesg, acpi)

if you got as far as running Xorg then include /var/log/Xorg.log too but I'm guessing you don't get that far

dch: will do.. it's my wife's tablet so will have to pop the FreeBSD USB stick on it to gather the info when I get a chance :)

but it looks like graphics/drm-612-kmod just might support it

I wish I could figure out the firmware error my AX201 is getting connecting via iwlwifi(4) to my new WiFi 7 AP.

Not sure if I need newer fw or something.

Hi. Is poudriere (ZFS) broken on main (9417b115a283)? It works if I revert to f4cd99131c70.

dch recently asked about problems with poudriere on newer CURRENT too, maybe the same problem

thanks. I am running git bisect atm.

"specifically, newer than e91e8ebefadcce9d57c8ff945ff70050cbbe1ce1" "poudriere fails creating jails, with `jexec: jail_attach(6): Operation not permitted`"

Result of git bisect: e03ed9daeb49fffa1d16b8d00240c65e92650d01 is the first bad commit

Indeed, reverting this resolves the issue. des@ already requested reverting this change.

zapata: what's your test case for it?

zapata: it has been reverted

Greetings.

Welcome CuriosCat

Thanks, dnp1 .

Seems I've been slacking. Still on 14.2-RELEASE. Doing an upgrade now.

OK, 14.4-RELEASE-p6. That's better. Any thoughts or opinions on 15.x?

it's great

Good enough for me. I know I'll have to rebuild pkgs, but that's not a biggie.

I think since the upgrade to 14.4 is my ipfw nat setup broken, have I missed any interface changes?

it works on the host itself but not for the attached jails

the ABI changed between 14.x and 15.x. Did you reinstall packages?

I have updated to 14.4 (from 14.3)

so no abi change, and ipfw is in base

strange is icmp works but no tcp or udp

That makes it sould like firewall blocking.

yes, but I would like to recive mails from mailhosters which still only support legacy ip

I have mimimiced my config. this is the current output of ipfw show: paste.debian.net/hidden/22b2d6b1

Am I missing something here? Saw what I thought was a kernel panic on-screen, but the host is still running? i.imgur.com/h15rPDu.png

(15.1 on an old Dell R720)

uptime?

spork_css: that's a new one to me

this is a kernel panic, right?

3:19PM up 18:32, 2 users, load averages: 0.03, 0.08, 0.08

spork_css: It has 'EFI RT fault'. I think it's not a real panic. bugs.freebsd.org/bugzilla/show_bug.cgi?id=291193

odd, I can't really tell if this is bad or not from that.

serves me right trying EFI for the first time on real hardware. :)

oddly enough the DRAC log shows one DIMM hitting "correctable error threshold" a few days ago, and prior to that (when this host was running vmware) but no similar log at the time the fault was recorded. and having seen that DRAC message, I also had let this thing run memtest86+ for a few days last week, I think I got about 8 clean passes and 0 errors. ¯\_(ツ)_/¯

I'm trying out sylve.io as a vmware replacment, fwiw.

spork_css: I'm interested in hearing how things go. I'd like to move off of truenas here

dnp1: I walked in the room this thing's in yesterday, saw the bright white text in a familar shape on the screen, and walked out, so I haven't done anything with it yet beyond the initial install.

It looks promising, but also I see the commits have really slowed down, so I'm not sure if it's going to be around long-term. These single-dev projects sometimes kind of fizzle out.

I did raise a bug already - the whole web UI locks up (client-side) over some JS framework errors when trying to create a new zpool.

i love vm-bhyve; i don't see the appeal of the clicky ui for this

I don't run any full-time virtualization anywhere, so that's just a whole giant chunk of arcane knowledge I don't have to fill my head with.

I was using esxi before - I just wanted a simple way to supply an ISO, do an install and go about things.

I did have to start using proxmox at one job and that's... interesting, but also incredibly simple to use in most ways. But of course that also has the whole "vmotion" type thing where you can failover an entire host.

clicky ui is handy when you need fast and quick and also good for newcomers, you can always learn the internals later

Even for home use where you might have a little lab (like I do), it's still cheaper for me to grab some older hardware x2 and put proxmox or something on there and be able to test failover scenarios and such, or even use for things that I want to always be up (like Home Assistant or Plex).

i don't think any clicking is going to be faster than editing a 10 line shell file and restarting the vm, but :shrug:

I like choices!

choices exist because not only is everyone different, theyre unique

human factors in computing

But I'd already be using vm-bhyve I guess if I could migrate a win10 VM from elsewhere into it in 10 minutes. :)

when I did a few bhyve things by hand I just found myself spending hours trying to figure out things like getting a graphical UI up, how to make grub work, why there's all these loader choices, it's (for me) total throwaway knowledge that I don't want (it pushes out the good stuff when you get older, you know).

free vmware was basically feed it an ISO image (for a new VM), give it some parameters, and off you go with whatever paid work you're doing.

if I was copying a windows VM, macos VM, whatever, that was very simple - copy my vmdks and other files to another esxi host, import, answer the question about "move or copy" (ie: give it a new MAC and UUID or not) and that was it.

And free ESXi is gone now?

I believe so, I was running 6.7.

i created myself an ai that reads man pages, the handbooks, and the ports/pkg database so when i get stuck i can ask and get an answer, it helps.

I seem to recall broadcom canned it

yeah, I think they nuked not just their free "customers" but many commercial users that couldn't justify six figures or whatever noped out and that's where proxmox got so extremely popular.

another nice coup for linux - meet the users where they are with a useful replacement.

spork_css: sometimes the firmware does dumb things, so now we trap efirt calls and render that somewhat harmless

no happiness in linux though

kevans: I was looking at this, and it seems like there's some kind of parameters that would let me do something about the error (I guess tell the OS to stay out of certain memory regions EFI is using?) but I'm not clear on how that works. reviews.freebsd.org/D51146

of note sylve, like proxmox, mixes jails and full VMs in one UI - not sure about "containers" since I hate them and haven't gotten that far.

another quick one on the EFI "crash"(?) - and this is all just trying to understand the conceptual junk here, but are there processes running on the CPUs that are NOT the FreeBSD kernel and all it's child processes? Is there a little EFI shell of some sort at another layer (I know nothing about EFI except I don't use it).

spork_css: firmware still provides some interfaces that we can call into, in uefi these are the 'runtime services'

basically, you tried to do something -> we called into runtime services -> they faulted

from the context, this was probably the result of trying to set the RTC

Is true?

3.12 is the new default python?

dnp1: It's been gone for a while.

Thanks for the feedback

I think the consumerish product is free though.

vmware workstation the vbox like product. Forget the name for the macOS version.

fusion