so I think daemon is botching the error message

Oh?

CALL execve(0x239240ddeb30,0x239240dde780,0x239244217140) / NAMI "/usr/local/bin/consul" / RET execve -1 errno 13 Permission denied

Hmm, for the life of me, I can't see why it's getting a permission denied on that though? It's 755, as are the parent dirs.

I even tried 555 in case it was exiting early since it doesn't need to be wirtable.

it's getting the same error from fstatat. that implies it's a directory not the file

I mean 555 on /usr/local/bin/consul

Hmm

FreeBSD's make implementation still has BSDmakefile as the highest precedence search path, correct?

Strangely, Homebrew's man page for the bsdmake package documents this filename, but FreeBSD's man page does not: man.freebsd.org/cgi/man.cgi?make(1)

Title: make(1)

_xor: any mountpoints involved?

this could tell you which part is failing

/cluster, /cluster/consul, & /cluster/consul/data are all zfs datasets.

I did briefly wonder if that could be part of the issue as well earlier, but haven't had the chance to poke around yet.

Other than that though, /usr/local/bin/consul is just on zroot/usr

Let me try it with non-zfs dirs to see if that makes a difference, quick and easy to test.

you checked the perms on all of / /usr /usr/local /usr/local/bin yes?

Yeah, just tried it with `-u consul` and all instances of /cluster changed to /t (and rsynced /cluster/consul/ /t/ with --chown=consul:consul). Same error, no such file or directory.

Yup, will re-check just to make sure.

WAIT A SECOND...

What should the perms be on /?

755

It's 700. If this turns out to be the problem, I'm going to burn down an orphanage or something.

yup, that'll do it

Heh, when I checked earlier I ls'ed /, but somehow didn't look at . (was confused for a split-second there as to why I wasn't getting POSIX perms for /, so I used getfacl and just looked at that really quick)

preyalone: I believe the search path is actually set in sys.mk

Wait, so would this mean that anything that tried to run as non-root on this box was probably failing or something?

preyalone: which does seem to have BSDmakefile makefile Makefile as the order

_xor: yup

All day yesterday and half of the day today...spent on this.

I'm still trying to figure out how / got set to 700.

there's a reason I ask people to check on all the directory prefixes including / itself :-)

i see

can we update the man page to reflect this name?

I usually ask for a pastebin of the output of ls -ld / /usr /usr/local /usr/local/bin but I was sloppy today

RhodiumToad: Yeah, I somehow missed that earlier. Took a triple-check to catch it.

Doh!

what is the relationship between bmake and FreeBSD make?

I just checked a few other boxes and it looks like it was just this one, so that's good.

Aren't they both the same file?

preyalone: freebsd make is bmake, but I'd have to check the history of contrib/bmake (where it lives) to see versions, etc

Well, I'm chalking this up as a positive. At least I learned some more details about rc.subr, limits, su, and ktrace/kdump.

what's goofy, is that the Homebrew bmake and Homebrew bsdmake packages behave differently regarding BSDmakefile. The former treats it as a missing makefile error. The latter loads it just fine.

I just used truss, dtrace, or previously strace before.

it's a bit of a pain that daemon is botching the error. would be good to track that down, since it's a bug

all of those would probably work, I just happen to prefer ktrace

Yes, that was the annoying part. I was really hoping there was an argument to daemon that allowed more verbosity or something.

How much of base is dtrace instrumented? Most of it?

Is it standard practice for FreeBSD to provide both a make man page and a bmake man page? I guess, since one binary would be symlinked to the other

They're not symlinked, but they're the exact same size and file $FILE returns the exact same info.

not aliased to each other, either?

are they hardlinked?

I was just wondering that.

does `bmake` load the same sys.mk as `make`?

Btw, when was strace replaced with truss? Not that I mind, just curious because I'm pretty sure I used strace from base before years ago.

does it install a bmake link by default? I don't recall that

"/usr/bin/bmake was installed by package FreeBSD-utilities-13.snap20230316060122" (using PkgBase)

oh, pkgbase

is it just a hardlink to make?

Might be, let me check

It would appear so: `159595068566595282 263496 -r-xr-xr-x 2 root wheel 18446744073709551615 208528 "Dec 31 19:00:00 1969" "Mar 16 07:02:37 2023" "Mar 18 03:24:37 2023" "Mar 16 07:02:37 2023" 131072 233 0x800 bmake`

bah. it's execvp() that botches the error

When it tries to run /usr/local/bin/consul?

if (eacces) errno = EACCES; else errno = ENOENT; -- very informative, not

lol

it's... technically right

it's supposed to be doing a $PATH search, and if it finds that a directory in $PATH isn't accessible it skips it rather than treating it as an error

Can daemon just be revised to differentiate errno and dump a more specific message? Or would execvp need to be revised to set errno more specifically? If it's technically right, then wouldn't want to break it I imagine.

daemon can't do any better since it's just calling execvp

execvp could, arguably, do better if the path name has a / in

Oh is the error message being output directly by execvp? It's not daemon checking errno and outputting it?

execvp sets errno, and daemon just looks up the text message corresponding to errno

and it's execvp that sets errno to ENOENT rather than EACCES

ah that's what I was wondering.

it's in the execvp manpage, fwiw

Yeah will check it in a bit. Going to take a quick break after this stupid ordeal had me obsessing. I didn't want to backlog it and come back to it later.

Thanks for the help.

anyway, lesson is: always check perms on / and other key dirs if stuff works for root but not other users.

Yup, going to make sure that's checked from here out. Though I'm wondering how that got set in the first place. (not a big deal as it's isolated to this specific box)

Are there any good resources to see what's planned for future releases? I've found pages on the wiki in the past but the ones I saw didn't have a whole lot of detail; just wondering if I've managed to miss the right resources

there's usually an at-least-yearly have/need/want session at one of the devsummits

e.g., hackmd.io/JczFDHtiQYSeEyeK9182jw?view

Title: FreeBSD 14.0 Planning - HackMD

they're fairly raw, but maybe paint a good picture

Thanks,, I'll take a look

Checking in case I'm misreading pkg-clean(1) or it's incomplete: is there a way to delete cached package files for packages that were current but that I uninstalled meanwhile, other than "pkg clean -a" which deletes even those for still-installed packages?

(I know rm is an option, but I'm reluctant to rm them by hand.)

Is there a way to put the primary display to sleep when connected via HDMI? I have a screen which I have my laptop plugged into, and my FreeBSD box, and because the console is always on, whenever my laptop sleeps or disconnects, the screen switches to the FreeBSD box and stays on. Haven't been able to find the right search term to find anything yet

kevans: I didn't realise PkgBase Falls under need

Has anyone been able to run a FreeBSD desktop in Proxmox with a resolution greater than 1280x720?

havent tried a desktop on my proxmox fbsd instances, hmm maybe i sohuld on e day

rustyaxe, CLI works fine, but the resoluiton on the desktop caps out at 1280x720. Debian guests and window guests are able to go to 4k on my system. Obviosuly a setting is missing for the FreeBSD guest

i tried to remotely update my backup box from 13.1 to 13.2 yesterday. got to "new kernel has been installed, please reboot and run freebsd-update again", so i rebooted but the box never came back. had a look through IPMI, "Mounting from zfs:zroot/ROOT/default failed with error 6."

google had some forum threads from people with very old zpool versions, but my zroot pool was created during a fresh 13.1 installation so i don't think that's it

from the bootloader i can start a 13.1 boot environment that's probably been created by freebsd-update, and it even boots through to the login prompt. so the zroot zpool itself should be fine. (but i can't get ssh access to the box in this case because of a mismatch between the running kernel and the kernel modules inside the rootfs)

any ideas on what to do?

Grabunhold_: did you run `freebsd-update` again once you booted in 13.1?

trev: you mean when i booted into the alternate boot environment? no, i just noted that the filesystem seems to be fine and shut it down again

yeah

i would've aggressively done that, but it's just me. i am a noob

i don't think running freebsd-update on the alternate boot environment would do me any good. i need to figure out why mounting zfs:zroot/ROOT/default fails with error 6 when doing a normal boot

6 is ENXIO, and hints about openzfs.ko module is not available during mountroot phase.

either it is not loaded by loader or failed otherwise.

i haven't changed the loader config, it was working fine with 13.1 and the zroot pool. maybe the openzfs.ko module problem has something to do with the update. I'll try to read the loader config.

(if it even is a module problem, not sure. might be.)

escaping to the loader prompt and typing "load openzfs.ko" doesn't work, not sure if it's supposed to

"lszfs zroot/ROOT" from the loader prompt lists "default" and the boot environments, so it seems to be working from the loader itself

 

Pretty scary story :/ *double-thinks before proceeding from 13.1 to 13.2

i'm still wondering if this is a one-off error or if it will reproduce with my other 13.1 machines

Grabunhold_: let's first get it fixed before you go break any more machines

so, first off: why does the 13.1 boot environment not work?

second, what does your loader.conf(s) look like?

Could absence of "/etc/rc.d/zpool" cause that, lists.freebsd.org/archives/freebsd-questions/2023-April/003220.html ?

Title: freebsd-update(8) upgrade peculiarities (was: FreeBSD 13.2-RELEASE Now Available)

By "that" I meant not being able to boot 13.2

As for the new boot environment for 13.2, "freebsd-update" has started using ZFS snapshots (when available) for some time during 13.1 period without fucking any mention in its manual page

s/fucking any mention/any fucking mention/ # sorry for the misplaced cursing

meena: the 13.1 boot env boots, but version mismatches between the kernel and the ethernet driver modules on the rootfs prevent network access. also, the login prompt doesn't come up. it prints the current time but no login: line.

regarding loader.conf: is there a way to print it from the loader prompt? i can't boot a live environment right now, the virtual media functionality of the IPMI device is dysfunctional and I am waiting for my coworker to put a boot stick into the machine

parv: i don't think it's the absence of /etc/rc.d/zpool, in my (limited) understanding of the FreeBSD boot process that would only be touched once the root filesystem has been mounted, which is what we fail to do

Grabunhold_, Right, right

meena: telling the bootloader to load "kernel.old" instead of "kernel" makes the ethernet modules load, but still no login prompt

and ssh?

i'll try that right away

(and when i speak of loading kernel.old i mean with the 13.1 boot environment, of course)

meena: no network despite interfaces showing up. "devd: cant open devctl device /dev/devctl: no such file", devd fails to start and that seems to prevent networking from starting up

and still no login prompt

oookay, got my coworker to plug in a stick, let's see...

i wonder why on earth would the kernel (no matter of it's kernel.old or just kernel) be able to mount zfs:zroot/ROOT/(name of boot environment snapshot) but not zfs:zroot/ROOT/default

yeah, got the livecd booted

ctrl-w is no good when running the shell through an ipmi interface in your browser -.-*

I would try to check if strange stuff happened with /boot/

gotta love ipmi, set scroll lock but can't un-set it

is it safe to import the zroot pool or will the loader fail to import it because it's been used on "another machine"?

Force import via -f option

easy, but is it safe? will the loader stumble afterwards?

*bootloader

I use it in single user mode. Cannot say about your situation

interesting. where might this second "zroot" pool come from?

maybe the usb device?

(the original zpool, the one with the 118 ID, is a single-device pool)

importing the 118-id zroot pool via it's ID gave me no errors, but "zfs list" or "zpool status" just says "no such file or directory"

what the f is going on here

oh, yeah, importing the zroot pool mounted zroot/ROOT/default on /

the live environment didn't seem to like that, should probably have provided some sort of alternate root to the import command

but we can get at loader.conf now

meena: imgur.com/eYT10OI.png this is the loader.conf from zroot/ROOT/default

I'm tempted to copy my homedir, nuke zroot from space and do a fresh 13.2 install

okay, got the zpool mounted using "zpool import -R /mnt", now the live environment is working. zpool status reports the pool to be in fine condition and zpool list has all filesystems including the boot environments

Grabunhold_: But what was the problem?

'now the live environment is working.' seems to me like still in the process of finding the problem

08:28 <tsoome> 6 is ENXIO, and hints about openzfs.ko module is not available during mountroot phase.

it seems that kernel and modules were out of sync, but it's not clear why

Hello, everyone. Can anyone help please? I got this: dtrace: failed to compile script "/usr/ports/lang/tcl86/work/tcl8.6.13/generic/tclDtrace.d: "/usr/lib/dtract/ipfw.d", line 1: syntax error near "in_addr_t"

I'm not sure if it's correct channel to ask, sorry if no.

the problem with the live env was getting my coworker to stick in a boot medium and working around IPMI bugs.

the original problem is this: <Grabunhold_> i tried to remotely update my backup box from 13.1 to 13.2 yesterday. got to "new kernel has been installed, please reboot and run freebsd-update again", so i rebooted but the box never came back. had a look through IPMI, "Mounting from zfs:zroot/ROOT/default failed with error 6."

i've got the live environment going, got ssh access in there but still unsure what the problem is.

booting the 13.2 kernel with snapshots of 13.1 works (but has the aforementioned module mismatches, so no ethernet because modules don't load)

<Grabunhold_> i wonder why on earth would the kernel (no matter of it's kernel.old or just kernel) be able to mount zfs:zroot/ROOT/(name of boot environment snapshot) but not zfs:zroot/ROOT/default

meena: any ideas with the loader.conf?

Sounds complicated.

Strange, installing tcl86 using package manager, solved problem with building it?..

glenda: fun!

Grabunhold_: loader.conf looks sensible

Anyone else using FreeBSD on server?

Grabunhold_: this is very strange because why would freebsd-update install a kernel without the right modules

meena: Do they can install modules manually?

glenda: /usr/lib/dtract/ipfw.d looks strange, is it a typo by you?

no

Anyway, it is solved somehow.

And that file exists.

dtract and not dtrace? and installing tcl86 fixed building tcl86 from ports?

That's weird.

Wait.

It is a typo by me, when I was writing error message, yes.

Sorry.

meena: i guess the right modules are there, but they are on zroot/ROOT/default, and that can't be mounted because of error 6

at least during boot

can be mounted just fine from the live environment

glenda: Yes, I do use FreeBSD for too many web servers. Can help you if you wish.

and why can the 13.2 kernel mount the 13.1 snapshot but not the zroot/ROOT/default, that irritates me the most

Registration for BSDCan 2023 is now open.

I got another problem. I am running iRedMail-1.6.2, and getting this: Port directory /usr/ports/lang/php74 doesn't exists. What to do?

I know this port was deleted.

It is required by RoundCube I think.

tercaL : Nice. The obly problem I have now is what I wrote above.

you try to *run* something and it complains about missing something in */usr/ports/*? nothing should require /usr/ports/ to exist; it shouldn't matter if you have a copy of the ports tree and if you have one it shouldn't matter where it is or what you decide to do with it

It is a script that should build/install this all.

Grabunhold_: did you update loader in efi?

And port of php version that it requires is missing.

no?

meena: not manually, no. haven't really "touched" anything so far aside from backing up my homedir

still poking around the system

I don't think freebsd-update… updates that

but either way, it shouldn't need updating if you haven't upgraded the zpool (not sure if there's other criteria)

i haven't upgraded the zpool, it's been created by the 13.1 installer and i haven't touched the pool in any way, just booted the OS off of it and that's it

i need this machine functioning again by the end of the day. i'm close to nuking it.

hm, that iRedMail installer doesn't seem like something I would want to run, maybe create a propper port for it instead of running that shitty installer?

Sounds like a good idea

iRedMail is just a script, that setup mail server, etc.

I have not found any information on setting up mail server manually.

Title: How To Run Your Own Mail Server

Even despite search engines' best attempt at becoming worse, it's still possible to search using keywords - it's just a question of using the right ones.

Getting an IP address that's not already in a blacklist somewhere, or getting an IP off blacklists, is usually a much more difficult problem than the software stack.

glenda: You'll need to use php 8.x because 7.x is no longer supported

vkarslen: Yeah I see.

Okay, no shitty scripts ever again.

debdrup : Thanks.

Time to purge jail, and make new.

(the keywords for the above results would've probably been something like postfix dovecot rspamd dkim sieve and maybe freebsd

And "coffee"!

hey folks

gzar: nope

sorry for that hilight!

hi gzar

hello

glenda: nope, the URI has nothing to do with keywords

ooh, both names start with g :P

debdrup : Doesn't matters, you can't setup mail server without coffee.

Title: postfix dovecot rspamd dkim sieve freebsd - Google Search

It is hard for me to browse web.

I cannot open most of sites, sadly.

looks like there's at least two other guides, one of which also specifically targets freebsd.

i don't know what that means

Nevermind.

Can you remind me please what package command "dig" from?

glenda: Use /usr/bin/drill

You can use drill, it's a drop-in replacement.

I think dig comes from bind-tools, but you can use sysutils/pkg-provides to find out.

Thank you

Oh, rDNS? Sounds complicated. Knocked my ISP.

Mail won't work without it.

As a bonus, it gets you a neat hostname on IRC. :P

mail.example.com? :p

It will work but messages will get into spam

I checked by dig, got: my-ip.my-isp.com

So I mailed them, I guess it's what I supposed to do?

They need to set up the rDNS, yes.

Hope they will not tell me to go away.

Does anyone have experience with hosting mail server at home?

if there's something specific you want to know, it's better to ask that, instead of asking for people

No, nothing specific.

in the +25 years i've been on irc, i don't think i've ever seen someone get help by asking for people.

Oh no, I wasn't asking for help.

I am just wondering

i've never lived somewhere with an ISP that didn't block port 25 :x

My ISP didn't

nice

My ISP is like: do anything as long as it's not commercial

Can I buy public ip?

- nah, you got it for free.

glenda: I used to, I no longer do because there were more interesting things to try to get working.

(hosting a mail server)

But what prevented you from keeping hosting mail server and trying interesting things?

Because the universe is vast and computers are varied, mail is interesting to learn but a pain to maintain

So I let others maintain it and just use a custom domain.

Why it is pain to maintain?

No way sadly, no mail providers.

Google? It's a shit.

glenda: because scammers have ruined the open public commons of email, and so the filtering/antispam wars make server upkeep high maintenance

I suppose it depends how complex or not you want to make it, but when you include spam checking, blacklist updates/removals/checks, updates of software, occasional bounces from people who don't trust you, etc. it gets tiring

Yeah, what Demosthenex said

Yeah, what dubiousness said that Demosthenex said.

I wouldn't dissuade you, especially if it's your only option for whatever reason, but I would also suggest that it's not an "easy" thing

Just a simple mail server, just for me.

Yeah I have heard it's very complicated.

glenda: if it's not internet facing, it's easy

I ran a mailing list once, and it was easy.

It's not just spam (though correctly dealing with it by doing spam filtering before accepting the mail isn't always easy to configure) - it's also that the biggest providers cf. netmeister.org/blog/mx-diversity.html are either only interested in being able to send mail to everyone (in the case of Google) or they're only interested in being able to send and receive mail from other companies

Title: Who reads your email?

using their service (in the case of Microsoft).

It's especially a problem with Microsoft, because they don't publish their blacklist, and getting anything off it is an actual multi-month herculean effort.

Mail deliverability also isn't really promised by the protocols used, so it's not a reliable protocol for things that people think it's reliable for (ie. alerting and notification systems).

The only thing in the RFCs that's a requirement is that mail servers don't throw away email once it's accepted it, and until the user tells the mail server to get rid of it.

..and most of that is handled on the filesystem level, not by the mail server.

Interesting.

By the way, is there a specific reason for ISPs to block port 25?

Blocking spam

They don't want people sending email from their home connections.

That's weird

A lot of ISPs used to offer mail hosting as an extra service, so permitting people to host their own mail meant they'd lose out on that extra money.

Most of ISPs charge a lot, and they not permitting you to host things.

the reasons for this are political, alas

ISP charges depend on where you are, they can be dirt cheap in most of Europe.

Why permit you to host things yourself when you can force people to pay?

My network costs about 5 usdper month.

lol

yeah debdrup is right here

debdrup: Yes you are right.

It also doesn't help that at one point before they started closing port 25, there _were_ worms which would distribute themselves via mail on port 25.

I will not be surprised if my ISP will forbid using rDNS

But that hasn't really stopped even after ISPs started blocking port 25, so I'm not convinced it's a valid reason in and of itself.

I'd be surprised if the support personelle you get in contact with know what rDNS is.

oh yes

my current ISP is using CGNAT on ipv4 and doesn't support ipv6. so there isn't ANY public IP on my WAN line, not even a dynamic one.

and that's the only ISP that's able to supply more than 12mbit/s here

awful

And what is the max speed and price?

yeah, my current ISP doesn't have IPv6 either, it sucks, I got so used to haveing IPv6 before I moved

60mbit/s, around 45 euro / month

wow.

1000 mbit/s, around 5 usd / month..

no public IP is even worse...

yeah, i'd trade a few mbit for that public IP

there should really be a definition what legally constitutes "internet" in the context of shitty ISPs

next year they'll get away with giving you a http proxy, it's a race to the bottom

agreed. i'm not a fan of the gov, but just as food labels aren't supposed to explicitly LIE about the ingredients, the same should be true of ISPs

if they block ports, or don't provide ipv6, etc, they shouldn't be allowed to call themselves ISPs, but have some alternative name

I didn't liked my ISP, but after reading about your ISPs. I love my ISP.

mobile ISPs here used to inject lower-res versions of jpgs using transparent proxies and block TLS/SSL connections some 15 years back.

Comcast would inject ads in sites accessed via http:// rather than https:// back when I used it.

at least that stupid crap came to an end

Also I don't use ipv6. Should I?

glenda: I find it convenient to be able to use both ipv4 and ipv6

ipv6 makes it so much easier to run multiple servers from home

without NAT

if you *can* enable ipv6 in your router, doing so is nice

I run multiple, also every Jail has it's IP, and every VM (local).

as recently as a few years ago, they'd redirect you to their ad-laden "search engine" instead of sending NXDOMAIN if you tried to resolve something that didn't exist

I have only local ipv6

that DNS requests don't always come from a browser didn't seem to faze them

oh yeah, i remember when that was a thing for phones

Reverse hostname have to be "mail.example.com", right?

And I can use mailboxes: user⊙ec

I understood it right?

Hi, anyone has solution equivalent to shadowsocks+ck-cloak on freebsd? Kinda have to use it on my laptop but cannot seem to find a way to get ck-cloak to work...

glenda: Different things. The term "mail.*" is not mandatory. It can be "aristo.example.com" as well. Just depends on what you set for MX record (Your domain DNS panel).

it's been a few years since i ran a mailserver. if you had an A record and PTR record for mail.example.com and your IP, you could then use an MX record to map mail.example.com and example.com together

rDNS should match the MX record of your domain.

Let's say the MX record of your domain states a hostname like: peril.yourdomain.com, you create an A record: peril.yourdomain.com with its IP address (IP addr. of your mailserver) and that IP address should reply host queries the same hostname; peril.yourdomain.com - that's all, imho.

When it comes to mail server configuration, mail-tester.com helped me a lot, back then.

Title: Newsletters spam test by mail-tester.com

Very simple and much detailed online check.

for mail servers.

tercaL: Thank you

tercaL: I don't really understand other thing. "myhostname = something" is the same as mx record, right? (eg mx.moo.org). What is "myorigin"?

glenda: I don't even use "myorigin", well if you refer to Postfix configuration. I only have myhostname = mail.mywebsite.com

And btw, there're really good people with great experience, in #postfix if you wish to ask about further details.

Oh, thanks, I will join.

But as I asked already, do you have mydestination = or no?

I got this there -> mydestination = localhost.$mydomain, localhost, mail.mywebsite.com

thanks!

`uname -a` tells me: FreeBSD frog 13.1-RELEASE-p6 FreeBSD 13.1-RELEASE-p6 GENERIC amd64

Now, freebsd-update says: I've installed `world/lib32`, among other things.

And then asks me whether that looks reasonable.

Well, does it?

The full line is: kernel/generic src/src world/base world/lib32

Harmless, it should be considered reasonable.

Can you copy and paste the entire output of the freebsd-update command to pad.bsd.to ?

okay, all data is safe plus another just-in-case backup of zroot/ROOT/default. i'll nuke the 13.1 install and do a fresh 13.2 one now.

RoyalYork_: Here's the entire output: paste.debian.net/plainh/2d8f64f9

It seems that lib32 is for also being able to run 32-bit applications.

So, it should indeed be fine.

You should be safe to proceed

msiism: I believe you remember what you checked at install time?

I don't, actually.

But it was probably reasonable.

msiism, Chapter 25. Updating and Upgrading FreeBSD -

25.1. Synopsis of the Handbook has an example similar to yours.

It happens :) no worries

By continuing (yes), freebsd-update will attempt to download all files required for the upgrade. In some cases, the user may be prompted with questions regarding what to install or how to proceed.

Okay, thanks. I'm now upgrading.

Good luck, it should go smoothly

I have yet to upgrade my production server.... I should get to it

I hope this will bring the Falkon browser back to a working state, maybe. The last pkg upgrade kind of broke it.

Maybe because it wasn't tested on 13.1 anymore.

Browsers, the source of all evil

Well, w3m usually works.

Except for 50% of the Web, that is.

But that's not w3m's fault.

Agreed

50%?? try 90% these days to garbage like cloudflail

thanks to

No wonder, these days even toilet paper runs on Javascript

debdrup jail review?

antranigv: didn't you have a review involving some jail stuff?

Do I need /dev/fd/ if I don't use linux compat?

linux compat?

fdescfs(5)

/dev/fd is floppy

Not on FreeBSD.

ah

fdescfs(5) isn't the same as fd(4) either.

Sorry then

fdescfs /dev/fd fdescfs rw 0 0

Oh, yeah I'm also getting confused.

fdc(4) creates /dev/fd%d devices, whereas /dev/fd/ is a filesystem created by devfs(5) for fdescfs(5).

So far as I know, the only thing that needs fdescfs(4) outside of things in Linuxulator is bash.

fdescfs(5)*

Linuxulator lol

What a name

That's the name of the Linux syscall translator.

Does it emulates kernel?

It doesn't emulate, it simply translates syscalls.

from linux to freebsd?

kernels

src/sys/<arch>/linux/syscalls.master iirc.

Ah, thanks

linux(4) has more information.

i think java also requires it

Huh.

I don't have fdescfs mounted on /dev/fd and use some java programs sometimes, so it does not

it might be like bash, where a small subset of the functionality depends on fdescfs(5), but it can mostly run without.

cgit.freebsd.org/src/commit/?id=8935a3993219 neat!

Title: src - FreeBSD source tree

fdescfs has been a source of a "few" bash related problems in the past

fstab: that i usually use on systems that mainly get affected by those things...

procfs /proc procfs rw 0 0

fdescfs /dev/fd fdescfs rw 0 0

linprocfs /compat/linux/proc linprocfs rw,late 0 0

linsysfs /compat/linux/sys linsysfs rw,late 0 0

Use a pastebin!

tmpfs /compat/linux/dev/shm tmpfs rw,mode=1777,late 0 0

yeah i should of

appologies

other notes ... ports emulators/* may need to have something to so with that

at least production systems on my part are using linux_base-c7-7.9.2009_1

solved the majority of linux/bsd integration problems ive hadddd

Sysadmin by cargo-cult doesn't seem advisable.

CrtxReavr: IIRC theres python stuff works better with fdescfs (no details sorry)

Well, for now I've unmounted it and removed it from fstab. . . it can go back if needed.

I think though, it was just mounted by default by this VPS provider and I've left it alone for muiltiple upgrades.

tercaL: sorry for delayed response, yes. enable zfs compression everywhere, all the time, and don't worry about it. lz4 on any h/w more recent than last 10 years is not going to make any negative effects. prior to that, maybe, you can test easily enough.

yeah, I'd bet java's exactly like bash in that way

I did double-check to make sure I'm not crazy (just now); the install message in the pkg recommends both procfs and fdescfs: freshports.org/java/openjdk18

Title: FreshPorts -- java/openjdk18: Java Development Kit 18

but yeah, bash only uses it for <(cmd substitution) iirc, which is also relatively minor; I wouldn't be surprised if it's only one or two special java things

anybody interested in having an invidious port?

"relatively" carrying a lot of weight, it can have security implications

I have a "works on my amd|arm machine" thing, and it could do with some help turning it into a usable thing.

I hope to get the rest of the port on my trip next week, and then a hand making useful "first run setup" notes would be good

that was that seflhostable alternative youtube fontend? while I currently won't use/run it myself, it seems like a good idea to have as port

dch, ? have a link somewhere to follow ?

even a depot

sure CmdLnKid nimaje when I've done the rc.d bits I will share the link

exploring hardware with arm64 that could be a cheap and capable server able to run freebsd, any recommendations.

sata and pci-e is needed =D

i enjoying my proliant, but it seems to make a bit too much RF noise for my use ;(

proliant?

:D

why arm64 specifically

thx

exploring new tech

energy effiency is a plus

i guess it depends on what you intend to do with it

throwing around some data. not computation heavy

if it were me and i wanted to play with a different server arch, i'd go with riscv

not because i think riscv is good, but it's just more "interesting" at the moment

moonshine: are you able to present a cheap capable alternative with sata and pci-e with riscv?

that is able to run freebsd, in that case im interested

can bsd run on a energy starved cpu...

probably not, but let me do a quick search

drobban: define 'cheap'

lol

well. not sure what USD is at from SEK. but perhaps around 1.5k?

no, there probably won't be any cheap riscv boards anytime soon

what did you find? and price range?

assuming google's not lying to me, 1500 SEK ~ 145 USD. hmm

Dollar to Swedish Krona Exchange Rate Today, Live 1 USD to SEK = 10.243

yea, around 1500USD

oh, that was 1.5k USD

depending on how much power you need, I've heard great things about the Windows devkit that clocks in at ~$600USD

windows devkit :-O

wtf is happening.. Im in the freebsd channel right?

:-)

Aye

it runs freebsd

I mean, naturally it runs windows by default, but there's nothing stopping you from wiping that and installing freebsd instead

forgot an s

there are some small riscv sbcs but they're not exactly what i'd call a server

also they're overpriced for what they actually do

are we talking about the Volterra machine?

compared to other archs, that is

yeah

volt errabn\y\\\

sorry, toddler attack. volterra

missing pci-e?

i started testing arm64 servers in 2013 or so, just getting things bootstrapped... riscv today is kind of where arm64 was in 2013

in my estimation, anyway

oh, yeah, it probably is, sorry

moonshine: what kind of gear did you use back then?

i can't even remember. we (where i was working at the time) one that was loaned to us to test

all those companies are gone now, i think :)

i have completely wiped the 13.1 installation, did a fresh 13.2 install, rolled my ansible playbook and we're back in business.

unfortunately all of the arm64 stuff I use is way over the suggested budget

let's hope the next 13.1 -> 13.2 upgrade goes more smoothely...

except for dell, maybe

dell tried to get into arm early, iirc

Even after upgrading to 13.2, Falkon gives me errors: paste.debian.net/plainh/1ff5ff94

The effect of that is that the window area for Falokon is occupied, but nothing is every drawn.

s/Falo/Fal/

Now I'm looking for someone also using Falkon to confirm that the problem doesn't just occur for me.

s/every/ever/

at dells site it seems to only offer intel/amd

msiism, Wish I could assist with your Falkon issue, however this one goes beyond my limited experience

I mean, I can use Firefox for now. So, it's not all that urgent.

But I'd file a PR after some more investigation, I think.

i use this self powered usb hub with my machine, every device on it works fine except my trackball which keeps getting removed (by udev i think?), i dont really see how to track this down. There is no such problem if i plus the mouse directly in the mobo

drobban: HP proliant server.. 24 cores, 256gb ram. bunch 15k SAS drives

10gbit connection from it to LAN and another 10gbit to my radio, which tends to throw a 1-2gbit/sec firehose at the poor computer, depending what im playing with. ;)

seems to be amd64 not arm64

rustyaxe: poor computer. :D

most arm64 servers you would probably need to talk to sales to get pricing

maybe there's stuff you can pull off ebay

I don't have lang/rust installed, but something I'm upgrading on my system is trying to bring it in as a depedency - how do I tell what?

how do you upgrade? pkg upgrade?

Was going a ports upgrade with portmaster.

hello, when using lsof I e.g. get "lsof 1173 vagrant txt VREG 0,92 135120 28461 /usr/local/sbin/lsof (/dev/gpt/rootfs)" in the output. Is there a flag or an option to remove the (/dev/gpt/rootfs) part? Or should I pipe to sed/perl/awk?

What does the lsof manpage say?

I suspect is has an option to omit itself, but I don't have it installed and don't plan to for your question.

I'm sure you could also: lsof | grep -v lsof

iirc portmaster shows a dependency path when building stuff and note that portmaster builds and installs on your running system, so you build in an unclean environment (shouldn't lead to problems, but sometimes does) and has your system in a partial updated state while building, which can be very bad, I suggest using poudriere or synth to build ports

Well, sadly more & more ports don't build well outside of pouderie or whateverthefsck they call it.

really?

Don't act like you don't know that.

what is the performance of Windows OS's virtualised with freeBSD as a host?

I need to use some Adobe software sometimes, I have a thinkpad x270. Will it be enough?

drobban: I was reminded of these: ipi.wiki/products/com-hpc-ampere-altra?variant=43142801129634 just a little higher than $1.5k @ $2k for a 32-core beast

Title: Ampere Altra Dev Kit – I-Pi SMARC

kevans: that's for zero 🐏

meena: zero... sheep? o.o

ram

oh, I guess I need a magnifying glass

yeah that's a bit lame

morpho: try it and see

I need something to run as a daemon to listen on ipv6 port and send stuff to ipv4 port

I guess this is not something pf can do

nope that will need a couple of chained netcats; you could do it with perl to but eh ~ I mean you could do it with any programming language

unless you can find a userland tool to listen on ipv6 and send ipv4 .... which you MIGHT be able to do actually with ncat

nmap's version of netcat

super annoying, its actually this java issue Graylog2/graylog2-server #15177

Title: unable to start GELF input on IPv6 address · Issue #15177 · Graylog2/graylog2-server · GitHub

15177 – not able to set CC in port, graphics/mpeg2codec bugs.freebsd.org/bugzilla/show_bug.cgi?id=15177

fennel-lang.org/tutorial works in Flua

Title: Fennel – Getting Started with Fennelthe Fennel programming language

it used to work just fine in previous graylog/jvm combo and in the last edition it broke

meena: that is pretty neat

ffennel

:D

^^^ kevans you now have a lisp in base, what have you done

fennel is a lispy flavoured lua

Fenne is very good, because Lua is very good.

eh

Hmmm... Lua is good because of what? What makes it so good?

its good because its light

that is why its used for the AI in NPC's in games so much

idk.. syntax is (puke) ...

but every language is good

every tool has a job

True, it is light. What else?

no idea I just know its used a lot in games for NPC's for its lightness

im a perl/c#/rust guy :P

and this is a freebsd channel lol

Probably is fast, as well, right?

speed does not matter

Freaking interpreted

geting the job done matters, you use what you like

Ok

threading and forking will do the rest

worry about optimizations after you deliver

before that write in what you like

or what community you like to deal with

I am not dealing with communities, mate :)

you are dealing with one right now

I was just curious about how good, fast, light and shiny Lua can be

as good, fast, light and shiny as any other language :)

for that I could not answer and to be honest; I do not think most of the lua community could either

Ok, sorry for disturbing the community with my ignorant questions :)

Peace out

Homework question on IRC brings back memories.

dclau, its good to chat it was no problem :) but yes indeed any language is what you make it

Lua is better than a lot of languages and worse than some

I'm a bit late for a homework question :)

as long as you're not a byte late ... it's okay :)

Now I see the community. Relax, guys.

now I'm trying to remember if there's been CVEs with 1 bit leaks

lua isn't as fast as some interpreted languages (unless you go for luajit, which has its own issues) but it's pretty nicely embeddable

is there a native equivalent of lsof?

fstat

tried that, it won't tell me what's open on my zfs dataset

which I can trivially unmount -f

what exactly did you try?

Title: Snippet | IRCCloud

RhodiumToad: that, `sudo fstat -v`

sudo fstat -f /jails/h2o

thats blank

hmm

fbsd version?

1400082 approximately 2 weeks old

main-n261593-491263d7a640 so actually 17 March almost a month old

ha got it

its a socket

syslog-ng 97187 root 23u unix 0xffff000174425900 0t0 /jails/h2o/var/run/log

thats from lsof

Title: Snippet | IRCCloud

sockstat -u should also show that

obvious in hindsight

fstat only shows `root syslog-ng 97187 25`

that suggests that the fstat program, the sysctl interface to the files table, and the files table itself are somehow out of sync on what type values exist

it seems a bit unfortunate that fstat doesn't recognise those

this isn't the first time I've seen this, but its the first time I've had the time to dig around instead of `umount -f` and doing whatever it was I needed to do

RhodiumToad: ok, looking on a nice 13.2-RELEASE instead, I see the same thing - sockets aren't recognised

uh, they are, but maybe not on zfs?

I see that libprocstat has some zfs-specific code, possibly that hasn't been kept properly updated?

hm, fstat -f ... doesn't list sockets on the specified FS even on ufs

RhodiumToad: I think you're right, only zfs. I wonder if I have any UFS atm.

try `fstat -s` maybe?

and paste output please, I'm curious to know what I should see :D

Title: ⚙ D21880 fstat(1) -s option

Title: dpaste/hsbAE (Plain Code)

RhodiumToad: this is great. I want `-s` on all the time, everywhere.

note that the first command didn't show the sockets though

cool, I can close 3 old BZ with this now, and then feel good about asking for a zfs specific fix :D

did I read this correctly, it only shows up into when the pid is given?

it shows up in fstat -s without other args

but if you use -f /path to examine just one filesystem, they don't show at all

and you don't see it for a single fs, aah.

hmmm

see dpaste.org/SiWBN

Title: dpaste/SiWBN (Plain Code)

root syslog-ng 84577 16* local dgram fffff80023b52d00 /var/run/log

it does show up in zfs /

but not in child datasets

hm, on the system I tested that on, /var is actually just a subdir of / (ufs)

on the ones I have, /var/run is just a subdir of / zfs, but there are a bunch of jails with syslog-ng providing the sockets

cgit.freebsd.org/src/tree/usr.bin/fstat/fstat.c#n247 probably explains why sockets aren't printed if -f is specified

Title: fstat.c « fstat « usr.bin - src - FreeBSD source tree

Can't remember, but HEAD on -STABLE branches (e.g. stable/13) is usually buildable, no? (or is it expected to break sometimes?)

(and why it only prints for the same FS as specified path and not child datasets)

_xor: stable branches should pretty much always build

That's what I thought.

_xor: it should be buildable, yes, it's the reverse (something older being built on main) that isn't guaranteed and not really supported

Hmm, probably my config/environment then. Though the build did error in the same place on two separate machines, so it made me wonder.

what was the error?

/usr/src/crypto/heimdal/lib/hdb/hdb-ldap.c:819:17: warning: cast to 'unsigned char *' from smaller integer type 'unsigned char' [-Wint-to-pointer-cast]

I guess I can remove WITH_OPENLDAP=yes from src.conf.

It's not mandatory, but would have been helpful.

are you expecting to build without warnings?

Oh wait sorry, that was the warning line. Let me get the error line.

No, mis-paste. One sec...

erm

I was about to go for the easy route first and try diabling ccache, until I saw that it's a source compile error and not a linker error.

yeah, that's wrong in the source

hrm, but this code is like 11 years old???

Hmm, how can I see the default value for a make macro? I wanted to make sure WITH_OPENLDAP=no by default and that I explicitly changed it to yes.

Tried `make -V WITH_OPENLDAP` but that didn't dump anything.

I guess not many people compile it, heh.

yuripv: so are you saying with a small logic change, fstat could show sockets in child filesystems?

s/WITH/MK/ ?

yuripv: Just tried that, no dice.

I read the code you linked but can't spot the logic yet

No biggie, going to try to with that set to no to see if that allows it to skip past that part of the source.

the code has an obvious glaring bug

dch: nope, i'm saying that if -f is specified, sockets (or anything else that isn't fifo/vnode) are not printed

maybe it compiled with warnings before but compiler is stricter now? would have to check

aaah yes I see now

RhodiumToad: Yeah, that's what made me wonder if I was wrong about -STABLE always being buildable.

hm

my build logs from way back show it compiling without error

oh wait, wrong file

I guess it could have compiled with a warning before, but a pointer where it expected a char? That seems odd to be intentional. Like truncating a wider type to a narrower one where you know the extra bits aren't used...ok, I can maybe see that.

but a char vs word-size address? (then again, been a long time doing systems stuff, so I'm very out of touch)

what WITH/WITHOUT options are you using?

WITH_OPENLDAP=yes

I just commented that out and am recompiling now.

Oh optionS...one sec

Only other thing I can think of is one of the other options causing the error, but not sure which one.

That's the current /etc/src.conf I'm compiling with now, the error'ed one had WITH_OPENLDAP=yes enabled.

WITH_OPENLDAP doesn't look like it's supposed to be enabling compilation of heimdal/**/hdb-ldap.c

Well that sucks, heh. Guess this one is probably going to bomb too then.

hm

maybe it is

MK_OPENLDAP adds -DOPENLDAP=1 to heimdal cflags

If it makes a difference...

There's src-env.conf

and the whole file is #ifdef OPENLDAP it seems

I guessed it was WITH_OPENLDAP due to Heimdal.

so this option may never have worked?

Guess not?

Last commit was actually 11 years ago?

I mean that touched those files.

the broken source line is blamed on "Import Heimdal 1.5 distribution -- 12 years ago"

Oct 5 2011 to be precise

Lulz

looks like it was fixed in heimdal 1.6, but apparently 1.5 was the last version imported

the idea of running crypto code that has not been updated since 2011 and contains obvious this-won't-compile bugs seems a bit iffy

Yes. Yes it does. I should check stuff like that before I enable it. Had I seen it was 11 years old, I'd have probably not enabled it, not at least without looking a bit more into its origin/status/etc.

the description sounds like it shouldn't be compiling this at all

someone with more familiarity with the makefile mess would have to comment

If it's not being used, maybe it should be deprecated?

ah