hi

how does tar --exclude work?

tar vczf qwebirc.tgz qwebirc --exclude 'qwebirc/log/twistd.log.*'

^ not like this apparently

nor any of the other things I tried (adding ./, and even some weird \/ trick I found on stackoverflow)

ah

brain fart

BSD getopt stops processing options after the first non-option argument, GNU getopt doesn't

well, the synopsis is "tar {-c} [options] [files | directories]"

did you read `man tar` before asking?

yes

and then I remembered that things work differently on GNU. and that what tripped me up was the fact that tar has longopts which practically none of the other BSD tools do

but thanks for assuming I'm ignorant

goodbye

you're not passing the arguments properly, i'm not assuming anything

morning

Hi.

=)

Is OpenBSD's httpd available on FreeBSD?

antranigv: yes, in package obhttpd

kopiyka huh, it's not available on -CURRENT, weird. kopiyka thank you! <3

has anyone measured how much faster CPU equipped with AVX instruction will calculate checksums and other ZFS related stuff on FreeBSD?

in theory AVX should help with some non-multimedia stuff too

I don't think FreeBSD's binary releases use AVX at all, if I understand right

you'd need to compile it yourself

xtile, thnx for clarifying :)

by default, if I understand, most open-source OSes use only the base features of amd64 from back when it came out

Title: x86-64 - Wikipedia

the v1 baseline

though RHEL has moved to the v2 subset

Arch GNU/Linux was considering moving to v3 (with measured performance improvements) but I don't think they did

I don't keep up with GNU/Linux much

I'd be interested in a v3 binary release of FreeBSD

considered as a port/architecture like other architectures

ksx4system: AVX2 will be used for raidz2/raidz3 in sufficiently new versions of OpenZFS - but it's just about the only thing in the base system that makes use of it, because as is the case with most kernels, they operate with data at such frequency that the latency of individual assembly instrustions account for a significant amount of the total runtime.

So if you turn on SIMD instructions like MMX and SSE, it can actually make things run slower.

cgit.freebsd.org/src/tree/sys/contr…s/module/zfs/vdev_raidz_math_avx2.c this is the specific code that's responsible for the AVX2 acceleration

Title: vdev_raidz_math_avx2.c « zfs « module « openzfs « contrib « sys - src - FreeBSD source tree

agner.org/optimize/instruction_tables.pdf has a complete list of each amd64/x86_64-capable CPU, and how much latency each instruction adds - to give an idea of how big the problem can be.

aha, interesting

It's entirely possible to build FreeBSD that's optimized for the v3 and v4 sets, but the problem is that you then cannot then run it on older systems.

Currently, FreeBSD is optimized for a Pentium Pro aka the i686: cgit.freebsd.org/src/tree/contrib/l…Driver/ToolChains/Arch/X86.cpp#n115

Title: X86.cpp « Arch « ToolChains « Driver « lib « clang « llvm-project « contrib - src - FreeBSD source tree

I think the ideal would be, if FreeBSD project had the computer resources for it

to have a v1 release and v3 release

but I know that takes human and computer time

to support two architectures instead of one

It's not really a question of computer resources, though.

There is nothing in the base system that would benefit from using SIMD instructions in v2 or v3, aside from OpenZFS - and it already uses it.

The data amounts that the kernel operate on and the frequency that it operates on them at are so low and high, respectively, that it'd make things slower for the majority of people.

makes sense

But it could make a difference for packages

i.e. having not just a v3 release but all the v3 packages

since I know someone studied it for Arch GNU/Linux and did get a performance improvement

rather than slowdown

Package servers already take up +12TB.

FreeBSD isn't Linux.

exactly, it's resources :B

Also, please show the work.

Title: Uarchlevels Performance [2103142-HA-UARCHLEVE55] - OpenBenchmarking.org

Title: [arch-general] RFC: Use x86_64-v2 architecture

v3 shows improvement, v2 doesn't

Where's the student Ts, mean vs median vs average values, data confidence, and how have they accounted for the measurability bias that's mentioned in www2.ccs.neu.edu/racket/Performance/mytkowicz.pdf ?

ah :B

If you want to build third-party programs with optimizations for a narrow use-case that benefits you, go ahead.

Expecting the project to refactor the entire build infrastructure to solve that problem for a tiny portion of the userbase who have systems that're new enough to take advantage of it is an entirely different can of worms.

i wish i knew why i kept crashing

And that's assuming any of that holds up after you've controlled for the statistics principles mentioned above.

Measurability bias alone can produce 40% performance depessimization through the code being linked slightly differently.

interesting

This isn't news, Emery Berger has done two talks on it - called "Performance matters" and "Performance (really) matters"

Also, it isn't the FreeBSD project you should be asking to do this - it's the upstream projects which are responsible for their own build environments.

Packages are built from ports, but ports are just a set of makefiles instructing how s omething should be built using the software's own build tooling.

Anyone happen to have sysutils/openzfs installed? Could you pastebin etc/bash_completion.d/zfs ?

/usr/local that is.

skered: you could just extract the distfile and pick it out by hand, if that's what you're going for

Yeah, just hoping someone already had it.

I know that FreeBSD 14 now has `uuidgen -r` for real random UUIDs, but on other releases, does anyone know if there's a chance of collision?

the whole point of uuid is to avoid collision

skered, My 13.1R system uses zfs and has bash and bash-completion installed but has no zfs file. bsd.to/FZyJ/raw

Title: FZyJ

antranigv, I think the v1 UUIDs are unique enough, but they leak the generating host's primary Ethernet MAC address (a privacy concern)

we're now at v4, aren't we

jilles: if a MAC address is a privacy concern, whoever's targeting you can likely see you.

debdrup, the creator of the Melissa virus was tracked down using UUIDv1, so I don't think it's an irrelevant concern

jilles: I think maybe a bit has happened since 1999.

Also, his ISP snitched on him.

That's legitimately the only way it can be used.

fbi.gov/news/stories/melissa-virus-20th-anniversary-032519 FBI posted quite an interesting retrospective on it a few years ago.

Title: The Melissa Virus — FBI

meena, v4 (random bits) is often suitable but UUID "versions" are really different generation methods and a higher "version" isn't "better"

rwp: You need the port installed not just have zfs from base. But I've already extracted it.

Oh, okay.

Hmm... What is the executive summary of the advantage of having the port installed rather than the one from base? Wouldn't the zfs in base be preferred?

The OpenZFS port exists for people on FreeBSD 12, which has an updated and maintained version of the fork pjd@ made from Illumos back in 2006-7.

If you're on 13 or newer, there's no reason to use the port.

Awesome! Thank you for the information. That is all excellent information for me. I am on 13.1-RELEASE now. Having upgraded from 12 previously.

How much support is there to get freeBSD running on an "older-ish" laptop?

AntiThesis: if you mean something like thinkpad X220 (2011), should work fine

I've ran 12.x and 14-current on a ~2014 laptop, and am currently running 13.1 on another (refurb) that's about as old.

la_mettrie, no, Dell Inspiron 15

i7 processor, touch screen

UEFI was it's only purpose and I am repurposing it

no UEFI, straight legacy, already ripped out the UEFI partition and placed all the BIOS data back on chip

It's acting real skiddish, but I believe a good OS will whip it into shape

jilles: are you telling me you can derive my mac address from cfa343c7-bb9a-11ed-91a1-003048c4c0b6

rtprio: That suggests your box is a SuperMicro. Yes?

AntiThesis, The two areas where FreeBSD kernel is behind the Linux kernel is WiFi driver support and power management for lower power on laptops.

For WiFi there is the wifibox workaround. For older laptops with expired batteries there are power cords.

If those things are not showstoppers for you then everything else is fully functional as any other system.

I have had problems with Dell Inspiron laptops having the bad Broadcom WiFi chips without free driver support anywhere. Beware.

rwp: Does FreeBSD do tick scheduling as well?

mason, I have no idea. Sorry.

I have FreeBSD installed on my Thinkpad x201 which is rather a spare machine for me and don't yet have everything completely to my happy state yet.

WiFi in particular is a challenge. I think I have a bug found in the wpa_supplicant area. I need to find the time to debug more on it.

But on my desktop and servers FreeBSD is performing excellently.

Tick management would be more power savings than performance

Ah, so there might be variations in implementation, but it's there: en.wikipedia.org/wiki/Timer_coalescing

Title: Timer coalescing - Wikipedia

I don't have real benchmark data for battery life with FreeBSD on my x201 but... The system is always physically warmer than with a Linux kernel. And it seems like it has much less battery life. But it is also an older battery in that system.

Oh! And as of yet I don't have suspend-resume happy on the x201 either. It resumes the kernel but the display is always stuck dark.

rwp: have you disabled the TPM?

People have suggested trying a few things. I have tried a few workarounds. Haven't really spent a focused effort on it. But nothing suggested has worked yet.

I believe the TPM is disabled on that machine. I'll say yes. But honestly it's been a while and I should look before I commit.

TPM is the one thing that consistently makes suspend/resume work on all the ThinkPads I've owned.

Work by disabling it? (I am pretty sure I have it disabled. Because I just normally disable it.)

It's something to do with FreeBSD not having a driver for it, since it has no use on FreeBSD.

After a suspend-resume I can ssh into the machine. It's alive. But the graphics side of things is not happy.

Have you tried switching vtys?

Have tried switching vt's. Yes. No joy.

If I think of something else, I'll be back.

Sorry. I really wasn't wanting to debug my laptop problems at this moment. (I am trying to do taxes!)

But I was simply giving feedback to AntiThesis about FreeBSD functionality on a laptop. These are the areas that I find most troublesome.

Wish it were better. But other things can make the compromise worthwhile.

I will say, though, that I much prefer having FDE to having suspend/resume - especially when my system boots as quick as it does.

FDE and suspend/resume isn't compatible because S3 STR keeps the encryption key in memory, meaning anyone with access to DMA on the machine can read out the key in memory with very little effort.

Right. I am just doing a full shutdown and full bootup on my x201 right now and it's not seriously a problem for me.

Which is why I had almost forgotten I was having that problem. The main thing I am hitting is what seems to be a bug in wpa_supplicant.

I can't debug it at this moment but it seemed to lock dhclient out of the network device with some exclusive access or something which broke dhclient.

If I manually hacked around it then everything worked okay. But I could not get the two to work together in a native configuration (native without wifibox).

So... I have a stack of old Mac Minis in my closet that won't run any macos newer than 10.6.8. Is it reasonable to turn them back into useful workstations with a modern i386 build of FreeBSD? Trying to figure out what to do with them.

If you try one then you will know and then you can let us know! :-)

They boot up just fine, but I haven't figured out if I can update root certificates, and I can't find any modern browsers with i386 builds. Even palemoon has no i386 version. It's as if everyone things this is obsolete...

I have run into forced obsolescence of other 32-bit hardware in other systems too. :-(

Github website is marginally usable in "w3m"

That's annoying.. I have some Alix systems that absolutely do not need to be replaced, but should be kept current.

parv: thanks, but I'm hoping to make a first workstation for my 8-year-old son. I don't think w3m is in the cards quite yet... ;)

ghoti, No Doubt