Hello, hopefully quick question about mirroring. I'm on the docs page and section 3.3 says public mirrors are not allowed. I'm a new convert to BSD from Linux and I'm curious what my options are to run FreeBSD isolated from the internet? I don't exactly know the terminology to search yet so a push in the right direction would be appreciated.

what do you want to be able to do without internet access?

install packages?

Best case, install packages and update the system.

My impression is that most people who operate offline do so with a fully self-compiled system from source.

That reduces the problem to have source transferred to their system for compilation. And updating it as needed.

Since everything is in git now then one can git clone everything and then copy it to the target system.

you can build your own package repo from ports, as long as you have some way to download the required distribution files

Can compile it on a different system if the offline one is not powerful for compilation.

With a self-compiled system, will that allow me to output the binary files to a shared location for install so I don't have to compile on every server?

Aptly in Debian and RedHat Satellite has had me spoiled for too long :)

I guess the million dollar question is, what verbiage in the docs do I need to look for to set myself up for success? I don't have any special compilation requirements and essentially a rsync mirror will suffice.

In order: Yes. Yes. I am pretty sure it is covered in the FreeBSD Handbook. But there are other docs that might be more accessible.

Also see poudriere for a framework for building ports.

Ports are the parts that are not in the core base operating system.

There are a few things I use from ports currently so being able to build these and store the binary packages is something I've wanted. I'm of the mindset that you shouldn't compile in prod.

last time I looked there was some weirdness that made hosting your own freebsd-update server difficult. I don't recall the details

I don't know if it's possible to maintain a server for freebsd-update by rsyncing an official one or whatever.

mynam: poudriere is what you want for port building - it can generate a package repository for use by pkg install

(it's the same code that's used to build the official pkg repository)

Ohh nice! Definitely bookmarking that.

I'd really like my work to adopt FreeBSD. ZFS would solve a lot of problems for us and although we've got zfs on Linux, it's not supported in RHEL.

For compiling. Is "buildworld" something I'm looking for?

make buildworld is for compiling the base system (not the kernel)

make buildkernel is for the kernel (you should do buildworld first)

you may want to add a suitable -j option to that according to how many cpus you want to use

Ok Awesome!

I definitely appreciate it!

I can futz my way thru the tech bits but trying to get the right terminology is the hard part.

btw, I haven't tried it myself, but you might look into the new pkgbase system (which treats the base system as packages)

That would be nice.

I've added that to my research notes too.

I appreciate both of you for the help!

so if I were trying to test the future 14.0-release do I really need to do a source code upgrade. Is it really not possible to fast forward jails?

feurig: there's no equivalent to freebsd-update for branches which are not yet actual releases.

not last i checked, anyway

so updating via source is the way to go if you're trying to track a -stable or -current branch

Is there a decent tutorial on 13.2 -> 14?

I've read around the handbook for the specifics for a few days now.

Just weighing the update between the current work environment and what the options will be in 3-6 months.

14 is still -current, until stable/14 is branched, so you have to treat it as such

RhodiumToad: thanks thought it couldnt hurt to ask.

I would probably wait for stable/14 before doing anything too serious

That was the nature of the query in general.

the thing with working with -current is that since the abi isn't fixed, you may have to rebuild ports regularly

I just installed a pci-e ucb card I bought on amazon. `pciconv -l` tells me it is xhci0 and good things like the vendor name. And when I plug anything into the USB port, it is not noticed. No response at all

Now, the back of the card has what looks like a SATA HD power plug on it. I had an adaptor cable that converts old=style power (amphenol?) to SATA, so I hung it off one of the fan connections because my power supply has no extra SATA power. But still nothing.

Any idea what I'm missing?

does usbconfig show a root hub for it?

given that usb is expected to provide power for external devices, it wouldn't surprise me if it needed a dedicated power connection

whether what you've done is sufficient/correct ... who knows

Err, I *think* it does, but I'm not sure. The output of `dump_all_desc` has a first line of: ugen0.1: <0x1b21 XHCI root HUB> at usbus0, cfg=0 md=HOST spd=SUPER (5.0Gbps) pwr=SAVE (0mA)

I guess it's possible that the fan wiring only powers the 12v line, and perhaps it needs 5v as well for usb.

Sucks that my power supply has no sparea within reach. There are 2 SATA supply cables, each with 4 plugs spaced 2 inches apart. I need an extension, not an adaptor. Maybe.

does the motherboard have no usb ports?

is 0x1b21 the vendor id for the card?

The motherboard does, but I'm trying to get USB on a separate PCI device so I can pass it through to a bhyve instance.

vendor ASM 0x1b21 ASMedia Technology

does that look likely?

is the motherboard usb disabled or enabled?

YES, asmedia is printed on the chip.

Motherboard USB works, I have 4 ports on the back, 2 on the chassis, and 1 internal .. but I don't want to pass my keyboard through. I will track down PCI devices for them again..

The device I want to pass through is a sonoff zigbee dongle.

so you'll probably see other root hubs for the motherboard?

anyway, if you see the root hub, but not any devices, it could well be a power issue

To pass something through, I need it to be the only thing on a PCI device. None of it needs 5Gbps, so if I can find a device on a bus on an otherwise empty device, I shoud be good...

How do I list my root hubs? Is there an easy way to trace where a device is connected? I mean, easier than dmesg?

usbconfig shows all the detected usb devices

I don't know if there's a way to see the actual topology

oh, devinfo will show the topology better

wait, devinfo can show usb?

when i use some software,i got 'vm_fault: pager read error'. bsd.to/TWWn

Title: dpaste/TWWn (Plain Text)

birz what CPU?

it's most likely a mmap issue

intel i5-6300U

roughly there are 3 likely causes "non-existent physical memory" (usually a bug) "hardware error" and "out of memory"

since you get the problem with 2 browsers the first is less likely

I suspect the filesystem broken,i use fsck test. got this bsd.to/1SCb.

fsck -f /dev/ada0p1

could be that, mmap or swap causing errors

have any of you played with this kit -> ipi.wiki/pages/ampere-altra-dev-kit

Title: Ampere Altra Dev Kit – I-Pi SMARC

drobban: I know some of our developers have

that dev kit in particular, or the altera platform in general?

because there's probably some particular peripherals on the dev kit not found on other altera servers, and i think there's at least a couple of the latter in the hands of various freebsd developers

drobban: I haven't played with that one in particular, but we do run well on other Altra machines (and they're very nice)

we have a couple ampere servers at $work

i'd probably personally spring for the 64-core at a minimum

-j80 builds on an 80-core altra, though, are *chef's kiss*

kevans: trying to find a product/datasheet, the information is somewhat vague when it comes to compatible memory types.

kevans: Im thinking about getting the 80core version

does /bin/sh resolve to a POSIX compliant implementation on FreeBSD, or perhaps a csh interpreter?

pro tip: man 1 sh

and what does the SHELL (b)make macro use by default?

The

current version of sh is close to the IEEE Std 1003.1 (“POSIX.1”)

specification for the shell.

awesome

and make uses sh, not csh. right?

ah, the bmake source appears to use ksh (lol) on HP-UX, possibly sh.exe on certain Windows environments, and plain sh on everything else

Title: GitHub - arichardson/bmake: A git repository for the bmake releases

hello ?

hello

when did we add libxo support to mount?

13.1 doesn't have it

test 1 2 3

hi, I'm new to FreeBSD. Recently switched over from Linux...

Anyone here ?

omagreb: yes, there's people here

but some of them might be having dinner, or putting kids to sleep, or walking dogs, or writing code

looks like I should've shutdown my vm, before taking a snapshot… all the modifications I made to the image are lost

i wish i had ZFS on this computer lol

meena: seems to be in current and 13.2+

so I'll need a version check…

if you want libxo output from mount, I guess so?

wonder when i can remove the version check and the regex code…

when you don't need to support anything older than 13.2?

is this in a script or a program?

oh dear, gcc12, ghc, node18 and llvm13 all trying to build in parallel in poudriere

I feel a swap-space kill coming on

RhodiumToad: it's in cloud-init

bah, why does llvm13 build fortran by default, and why did I forget to disable that

i love these kinds of bugs… bugs.freebsd.org/bugzilla/show_bug.cgi?id=271081

Title: 271081 – www/firefox: crashes on arm64 with ASLR enabled

Swap: 8192M Total, 8192M Used, K Free, 100% Inuse, 7300K In, 2808K Out

" K Free"

not very good handling of 0 there

nothing's been killed yet though :-)

An is that K kelvin? It's not kilo (k) or kibi (Ki)

K as in kbytes

technically Ki, but nobody takes kibibytes seriously

that should be k as in kbytes

who cares

still it should be lower case

would be better if it showed the zero. is this freebsd , or are you showing another O/S for comparison sake?

npl, nist ...

freebsd 13-stable

as shown by top

in a VM currently thrashing itself to death trying to compile llvm13, ghc and gcc12 all at the same time

heh. stress testing, a proven way to find ze bugs

oh, and gcc48 too

I have tuned it to be somewhat less aggressive than default about actually killing processes for swap exhaustion, but I'm still slightly impressed that it kept going

(vm.pageout_oom_seq=200 vm.pfault_oom_wait=3 vm.pfault_oom_attempts=30)

the purpose of this exercise IS to find O/S bugs then? or are you looking for something else?

nope, just trying to build a bunch of packages

it's just that inevitably, sometimes poudriere schedules a lot of big stuff at the same time

it's not really very controllable about that sort of thing

it would be nice if you could add fake dependency rules, to force some large jobs to serialize

oh poudriere is awesome. yeah, I use it to, but I always short the build count by one 4 procs?, 3 threads or whatever it's called in the config.

yeah, that doesn't help much with the real memory pigs tho

my poudriere server is in a proxmox instance, has 24 cores, 24g ram, and can still be brought to it's knees

I'm sorry, I lied. it's 12 cores, 16G ram. I think I scaled it back because it was affecting the rest of the machine on big builds

mine is a much smaller VM :-)

yeah, I've never seen mine use all the swap. not sure I want to do that.

8G of swap for a 4.5G vm.

ooh, I think I might have broken it

ah, no, it was just in the process of doing an OOM kill

Apr 26 21:18:48 kraal kernel: pid 18827 (c++), jid 9, uid 65534, was killed: failed to reclaim memory

do we have an equivalent of man7.org/linux/man-pages/man1/mountpoint.1.html / /prod/self/mountinfo on FreeBSD?

Title: mountpoint(1) - Linux manual page

(in base)

in what version of base and how ugly is it allowed to be?

RhodiumToad: it'll been in put into python, so i can work around ugly

I'm trying to provide a better fix for: canonical/cloud-init #2143

Title: Fix linux-specific call on BSD. by holmanb · Pull Request #2143 · canonical/cloud-init · GitHub

can it assume root priv?

actually no need

mount -d -o current -u -- "$path" will succeed and output a mount command if it's a mountpoint, otherwise give a "not currently mounted" error

(it won't actually try and execute the command)

what's it trying to do? find which filesystem a file is on?

yes

oh, so it can look for a noexec flag on the mountpoint?

yeah

github.com/canonical/cloud-init/blob/main/cloudinit/util.py#L2606 ← this is the function

Title: cloud-init/util.py at main · canonical/cloud-init · GitHub

that is broken…

i should rewrite awful regex to at least use named groups

how exactly is it broken?

the regex looks almost ok, except for ignoring filesystems not associated with /dev/* which may or may not matter to this app

I honestly think it's silly that it's looking for that kinda thing… it should just take the first thing and be happy

which first thing?

RhodiumToad: the first "word" should be the "device"

yeah

with regard to looking for noexec flags, the right place to look would be statfs() rather than parsing mount output

dunno if python supports that

i could write an FFI…

the thing with statfs is you don't have to worry about navigating the filesystem, you give it any filename and it returns info about the filesystem that file is on

even on linux :-)

python makes baby jesus cry

It's closing it: the temp dir resolving function returns "/run/cloud-init/tmp" — if needs_exec=False; on Linux, /run is mounted as noexec.

RhodiumToad: not if the path doesn't exist tho :(

anyway, os.statvfs("/tmp") doesn't tell me about mount flags.

well, it does, i guess, but it's hard to read

statvfs is more portable than statfs but necessarily much less informative as a result

"The statvfs() and fstatvfs() functions fill the structure pointed to by buf with garbage. This garbage will occasionally bear resemblance to file system statistics, but portable applications must not depend on this."

nice…

that is definitely a documentation

"As standardized, portable applications cannot depend on these functions returning any valid information at all." is definitely a standard, if it still doesn't mandate more.

RATIONALE. \r\n\tNone.

Title: fstatvfs

i love spending 2 hours of my life on 83 characters of regex

I find regex101.com useful

Title: regex101: build, test, and debug regex

jgh: I used that, and an ipython REPL

i've now told people about possibly siwtching to freebsd

after noticing that i do really have a reason to keep using linux what with ai work and cuda ive decided that ill get another 1tb ssd soon, and switch to dual booting debian and freebsd starting there

hopefully i can time it with bookworm's release (well i technically am already using bookworm right now, its just still in testing rn)