hi everyone, I'm having trouble with booting 12.4 release on sparc

basically getting the same results as here: forums.freebsd.org/threads/booting-freebsd-8-1-on-sun-fire-280r.17649

Title: booting FreeBSD 8.1 on sun fire 280R | The FreeBSD Forums

my machine is sun fire v100 (also known as netra X1)

symptoms are the same: The file just loaded does not appear to be executable.

I can boot Debian and Gentoo, but not FreeBSD. NetBSD gives the same error

I've burned 4 different images and changed cdrom drive, so that's out of the question... I burn at 8x and verify iso after, no errors

it seems like some openfirmware peculiarity, but I'm really puzzled at this point

That's goingto be a tough one to solve in the 3 weeks before 12.4 goes EOL.

I don't know that we actually have anyone all tht familiar with the sparc64 boot process anymore

I used to be.

Ran NetBSD on them. But it's been a while.

(Netbooted, too. I loved that platform.)

ok I don't know what was that

but apparently :f in the end of device name is mandatory

so full command is boot /pci@1f,0/ide@d/cdrom:f

_zip100: With that, you boot?

out of curiosity, what do you guys use for basic file distribution (http) outside of a script that modifies an index.html or something?

Apache httpd's AutoIndex feature?

_zip100: that's funkadelic, but glad you got it going

you're now the channel's resident expert

:D

yes, it boots successfully and I got into installer

nice

vdamewood: thx, will check it out

I miss having all the funky hardware, PMAX the most. (DECstation 5000/200)

I miss my SparcStation IPX.

It was the first computer I ever replaced a single IC on.

I have an IPC and an IPX. Got rid of all of it when I had kids, not realizing I was a year away from a house with enough room that I'd have a dedicated machine room in the cellar.

Had* Had* sadly past-tense.

The IPC had 4.3BSD

wow nice

No. 4.1? Something old.

Got rid of it all, except for my first two computers.

In my case, I had to put it in a storage unit that was rented by my mother, and guess who didn''t pay the bill

doh

ouch =(

_zip100: Even if you default to disk on that, it'd be well worth it to get a tftp and rarpd on your network for experimental stuff.

mason: I've spent two full days trying to netboot NetBSD from Linux. Now I know that both OpenBSD and NetBSD use network bootloader which supports only NFSv2

_zip100: Oh, hm. You should be able to snag a miniroot and go to anything from that.

Same idea as a Linux initramfs.

I thought so as well, but in practice bootloader cannot speak tftp

it can be loaded via tftp, yes

but then it goes full NFSv2 mode after

hrm, hrm.

so either a dedicated machine just for netbooting, because Linux has long since remove 2nd version of NFS

or putting kernel on disk and then mounting root via NFS

third option: add NFSv3 support to the bootloader :D

which would solve all problems actually

but it's quite a task

Seems worthwhile, fun.

How do i see what fs a usb drive has ? gpart show has it as mbr gpart list shows it da0s1 what else can i find ?

file -s

try df -T

thanks richard_

Does FreeBSD support UASP?

endrift: AFAIK, no does it not.

i got a bin running with daemon and i need to see errors in /var/log/messages but they're not coming. the daemon line is daemon -P <pidfile> -f -m 3 -l local5 -s debug -T... is that not the right way?

if i run the bin's command directly it exits with an error so i know it makes them

is it possible to 'tail -f' everything going to syslog without making a new file config?

hm wow took all of those args out and just passed -S to daemon and it suppressed terminal output and instead now it's showing up in /var/log/messages like i want

is the -f the problem in your daemon command?

not sure. put it back in behavior was the same. then took -S out and behavior was the same

not sure why -f seems optional

any way to configure top's poll rate to be faster systemwide?

like in a /etc file

if a bin generates files that other system services and bins use, where should those files go?

:D hi

how do i fix "stale file handle" error when i try mounting nfs share hosted on fbsd from a linux client?

nevermind got it <3 :D

<3

welcome to /usr/home/wsky/nfs running on freebsd settled on a m2 mac running on utm :D

Title: Index of /~wsky/diabelek1

whoa

i don't know why i did that :F too much free time this morning

it fast?

what's fast?

hm makes me wonder, freebsd got drivers for AI accelerator hardware?

that using hypervisorkit?

the bsd is running on utm

what's utm

it's a vm solution for apple silicon macs

cool

using bhyve?

nope

not that i know of

UTM uses Qemu

Hi all, I'm trying to upgrade 13.2-RELEASE to 14.0-RELEASE and for some reason after installing both kernel and word (tried both via source and binary upgrade) the system still boots into 13.2 kernel. I also started this thread on FreeBSD forums so far to no avail: forums.freebsd.org/threads/upgradin…from-source-is-not-successful.91263. Any help is appreciated.

bolik: please check that you're booting into the right boot environment using bectl(8)

i want my bsd to stop getting dns from dhcp, i want static dns servers

how wouldi. do that?

mane: something like 'ignore domain-name-servers;' in /etc/dhclient.conf should work (although i haven't tested it, i just read the manpage). or you could use 'supersede domain-name-servers ...' to specify your own

see dhclient.conf(5)

thanks you! :D

I'd _completely_ forgotten about dhclient.conf and was looking at resolv.conf(5) and rc.conf(5) :)

debdrup: the latter might have something to prevent local_unbound setup haha

the resolv.conf import I mean

Remilia: only to enable it, I think.

hmm

true

Still, dhclient.conf(5) was completely paged out of my memory.

morg

morn*

Bitflip in between brain and keyboard? :D

he wanted to say "borg" but it's a secret

computer boots up faster than brain into the hivemind

:D

NVMe drives will do that :D

debdrup: evidently :)

That's assuming you reboot for anything other than kernel upgrades, of course - there's really no need for anything else.

the borg is sending instruction to retrieve fresh coffee

and upload

'morg' is transliterated Russian for morgue

could have at least made it Morgen instead

Remilia: I do feel somewhat dead atm so not far off

debdrup: I switched from a dedicated server to a VPS with guaranteed dedicated resources and post-update reboots are absolutely mind-blowing

on the old server major version upgrade meant like 20 minutes of downtime

because the BIOS took 7-8 minutes

'sea of sensors' my arse

plus it had hardware RAID1 with spinning rust on an E3-1240, while now it is 10 cores of an EPYC with NVMe SSD RAID

≈4 seconds from connection lost to SSH up

Remilia: please stop intriguing me, i need coffee.. The BORG sais so

*us

*we

the only issue with the new server is that the provider's network keeps getting blocked by MS mail relays

@debdrup: bectl boots the environment created in 2021 (it states: BE:default, Active:NR, Mointpoint:/). Can I test whether the environment is able to boot before actually defining an untested BE as the one to be booted?

uploading coffee 40% complete--we ar one

Remilia: sounds like they have to call MS and get whitelisted, unless they're known spammers? Or for some whacky political reason

MS does not do whitelisting :|

you can ask to be delisted when they block you but they'll re-block if someone *on the same network* does the thing

scoundrels!

and the problem here that my server's IP is marked 'normal' by MS and only the provider can ask for a delisting

yeah, i feel like my new Arm64 VPS @ Hetzner boot exec

*even faster than the amd64 VPS

_xor: SIGABRT is generally the result of an explicit abort() call; various libraries and runtime environments do that when unable to proceed

e.g. c++ unhandled exceptions, memory corruptions detected by malloc, assertion failure in assert builds, etc.

bolik: you can test by using the -t flag for the activate, which enables the bootonce functionality; if it doesn't come up as you expect it to, simply reboot the machine by any means expedient (even a hard reset will work, and that's how I use it along with watchdogd(8) to enable unattended upgrades, the idea for which I have to credit Allan)

for the activate subcommand*

Good afternoon

any way to configure top's poll rate to be faster systemwide? like in a /etc file.

polyex: set the TOP environment variable using login classes in login.conf(5)

It's documented in top(1) ;)

no config file route?

I'm confused, what do you mean?

@debgrup: I will try that, however, so far, when trying to change the boot environment during the startup (playing with boot options) I only see BE´s of 13.2-RELEASE and none 14.0-RELEASE, even though bectl list does show 2 BE´s 14.0-RELEASE.

like a /etc/top.conf where i can set a global poll rate

If you set the TOP envionment variable via login classes using login.conf, all users in that class will inherit that environment variable.

ya i know

You're free to modify top to be able to read /etc/top.conf, but I doubt it's going to be accepted upstream because it seems incredibly installation-specific.

Especially when you can already accomplish what you want another way.

Similarly, tools like less also support the LESS variable.

It might be useful to add the various base system utility environment variables to environ(7)..

greerings dear FreeBSD people

greetings* even

today is a day of typos, apparently

oh yes!

what's the commonly accepted method for upgrading jails after a base system upgrade to 14.0-RELEASE?

bolik: huh, I'm not sure what the difference would be..

my jails are ZFS clones of a 13.2-RELEASE snapshot template

veg: freebsd-update using -j and optionally --currently-running if that's appropriate

Neat, freshbsd.org/freebsd/src/commit/231eee17d2905 and freshbsd.org/freebsd/src/commit/e3b4fe645e50b adds software breakpoints and singlestepping to AMD CPUs for bhyve

Title: FreeBSD / src / 231eee1 / vmm: enable software breakpoints for AMD CPUs - FreshBSD

wow

if freebsd has tier 1 driver support for accelerated AI hardware maybe it takes the throne back from linux

that's the next datacenter trends

debgrup: tried to set the BE with bectl -t to 14.0-RELEASE but it still boots the old 13.2-RELEASE

All stocastic parrot acceleration hardware seems entirely proprietary, so I wouldn't hold my breath for it.

bolik: sorry, I've no clue; did you follow instructions on docs.freebsd.org/en/books/handbook/…e/#updating-upgrading-freebsdupdate ?

Title: Chapter 26. Updating and Upgrading FreeBSD | FreeBSD Documentation Portal

debdrup: Yes, i did. Thank you for your time.

does the bectl list output show the 14.0 as the currently active one?

even the proprietary AI hardware options seem limited: NVidia doesn't provide FreeBSD drivers for theirs, from what i can see

Remilia: No, it has a ¨T¨ flag next to it.

T is temporary hmm

Yes, i did bectl activate -t <name_of_BE>

so it´ll only boot once to check if itś working

bolik: I just tried -t with my poudriere VM and it worked as expected :\

specified the 14.0-RELEASE-p1 as temporary, `shutdown -r now`, booted into that, did another `shutdown -r now` and got 14.0-RELEASE-p2 again

Remilia: thank you for trying. I have tried to search what could be the issue and ran across this possibility: serverfault.com/questions/602231/co…el-version-in-freebsd/602263#602263 however, I´m not sure how to act upon it (but then again, it´s just a guess that I face similar issue)..

Title: Conflicting information about the running kernel version in FreeBSD - Server Fault

is there any way to get remaining lifetime information from an SD card? it doesn't support SMART, i already tried that. mmcsd0: 16GB <SDHC SC16G 8.0 SN 4CB2AB35 MFG 05/2022 by 3 SD> at mmc1 50.0MHz/4bit/65535-block

any Ocaml people in here?

better to just ask your question and when someone can answer they will probably answer

okey. any known issues/limitation I should be aware of while working with OCaml in FreeBSD

I run into this every so often and it still manages to both amaze me and crack me up lol...

Title: c++ - Printing 1 to 1000 without loop or conditionals - Stack Overflow

5 lines, not including the headers of course, and without conditionals.

4 lines. Apparently my brain isn't tail-call optimized heh.

is it normal for pf to keep state information for dropped packets ?

I have this IP that keeps on badgering me : it's blocked but PF now has 25000 entries in the state table for it

unixwitch: for non-volatile flash like that found in SD cards and USB flash devices, it's best to treat it as WORM media and not assume you can write to it more than once - because often they simply have not enough sectors to move things around if a cell gets damaged and you've also rewritten data

drobban: I know of people who use ocaml on freebsd, so I assume that it works

unixwitch: interestingly enough, the above position isn't even one only held by IT people; any photographer, professional or amateur, will usually tell you that they treat SD cards like they're film roll

I have a sneaking suspicion that photographers use SD cards more than just about anyone else, so it' a pretty safe bet they know what they're talking about.

debdrup: considering the SD card is this system's root filesystem, that's not really an option

unixwitch: I assume it's necessary to boot, but can you reroot (reboot -r) onto _slightly_ more trustworthy storage?

That way you can keep the SD card read-only unless you're upgrading, which saves on whatever limited amount of spare cells it has.

does anyone have the link to the source code that shows the exact changes between minor p version releases ? like p1 vs p2 vs p3, etc

the changes aren't listed on : freebsd.org/releases/12.4R/errata for example

Title: FreeBSD 12.4-RELEASE Errata | The FreeBSD Project

i'm aware of the limitations of SD cards, i'm not looking to replace the SD card right now. i'm just curious about querying its state.

unixwitch: camcontrol might be able to tell you something, I'm not sure.

I don't know that there's an official API for it; if there is, I'm not sure it's in FreeBSDs base system. Have you looked in ports?

last1: check the releng branch for the m ajor version you're interested in.

cgit.freebsd.org/src/log/?h=releng/12.4 if I remember the URI scheme right

Title: src - FreeBSD source tree

that's typed from memory, so it might not work

I found the github tree as well

Title: Commits · freebsd/freebsd-src · GitHub

cgit.freebsd.org is the repo of record, github is just a mirror

ok, I'm confused about something

I wanted to get the openzfs fix with the dirty dnode issue

same, but about a lot :D

I did freebsd-update fetch on Dec 2nd, and it upgraded me to 12.4-p6

now I see that the fix didn't happen until p8

even though it was released on Dec 1st

check freebsd-version -kru

kernel and userland can be out of sync

freebsd-version -kru 12.4-RELEASE-p6 12.4-RELEASE-p6 12.4-RELEASE-p8

what does that mean ? lol

freebsd-version without args just prints: 12.4-RELEASE-p8 and uname -a shows 12.4-RELEASE-p6 FreeBSD 12.4-RELEASE-p6

if you check the manual page for freebsd-version, you'll see that it lists the kernel, running and userland versions

uname -a derives its value from a sysctl(8) which is tied to the kernel

I'm hella confused. which version am I on ? I did fetch, install, reboot, install

you can be on multiple point versions at the same time

if freebsd-update is telling you that you're up to date, you are

it's an artifact of how updating works in freebsd, and you're not the only one that finds it confuusing :)

I was affected by that zfs bug, I updated, the bug has now gone away but when I saw that I'm on p6 it made me think I didn't update

I just want to make sure I did update - because it's got me doubting myself now

how can I get all versions to match ?

you can't

the userland can have a higher number than the kernel, without it causing issues

last1: check the modification time on /boot/kernel/kernel and /boot/kernel/zfs.ko ... it might be only the updated module is shipped, not the kernel itself

that's a good way of putting it

yeah, zfs.ko has a proper modification time of dec 2nd

also confirmed by me seeing the bug has gone away

but now I have another machine that just updated to p9, I saw this one on p6 and it confused the hell out of me

since p6 doesn't have the fix.

so p6 has the p8 fix

crazy

no, the zfs kernel module is from p8, but that doesn't affect the kernel binary

it's a bit of a trip when you learn about this, but it all _does_ actually make sense

it's not that p6 has the fix, but the kernel version that uname prints is stored in /boot/kernel/kernel. so you have the /boot/kernel/kernel from p6 and /boot/kernel/zfs.ko from p8. there is no p8 version of /boot/kernel/kernel because the kernel itself didn't require an update. (yes, this is strange and confusing and i don't really like it.)

more importantly, it has a practical upshot - if there's ever a patch for the userland (ie. utilities or libraries), which doesn't update the kernel, you can simply update freebsd without rebooting (though you'll probably need to use sysutils/checkrestart to make sure you've restarted everything that needs it

s/$/)/

shame that modules apparently don't have sccsid tags, or you could use `what /boot/kernel/zfs.ko` to confirm the fixed version is installed

yeah that'd be handy in situations like this where a module has been fixed by a point release - but those don't happen super rarely

I didn't realize we still don't have a driver for BCM4360. That's a real bummer.

rafe: Broadcom hasn't exactly been forthcoming with documentation for a BSD licensed driver

but p9 did update the kernel

debdup: Totally understand that problem, but the OpenBSD port has been around for awhile, no?

*debdrup

rafe: someone still needs to do the work, and that'd only make that one NIC work - bz@ is working on a different solution whereby the LKPI doesn't just work for GPUs, but also for WiFi devices, so that Linux drivers can be put in ports (since they're GPL, they can't be in the base system) and can be added that way.

Sounds like an even better solution; I hope it progresses soon. Is there a targeted release?

rafe, IIRC, some of it is already in 14.0. See people.freebsd.org/%7Ebz/wireless and freebsd.org/status/report-2022-10-2022-12/#_wireless_updates .

Title: Wireless Work In Progress

is perform-actual-lookups is enabled in nscd.conf, does that mean the cache is shared for all users, rather than being per-user? the behaviour for nscd -i/-I seems to imply this but doesn't actually say so

A lot of the utility of a nameserver cache would go away if it was per-user.

Being able to invalidate per-user has _some_ use, for rootcausing, but beyond that the best results are typically achieved by having everyone share the same nameserver cache.

i agree, that's why i'm wondering if this option makes the cache shared, instead of per-user like it is by default

Oh, huh, you're right - I completely missed that.

Take whatever I said and put it in the bin, please.

I guess -I is just for the convenience of not having to do -i for every user then.

this seems to suggest the cache is shared when nscd is doing lookups: cgit.freebsd.org/src/tree/usr.sbin/nscd/query.c?h=releng/14.0#n693

Title: query.c « nscd « usr.sbin - src - FreeBSD source tree

the code is rather lightly commented though...

Presumably nscd can't poison itself, which is the reason for the per-user caching - so it seems fine on first blush?

If someone can casue nscd to cache poison itself, you've got bigger problems. :D

that's why i would assume it works this way, but... i can't actually find anything that says so. maybe i should test it

There are few better way than diving in and finding out.

random weirdness of the day: LDAP's groupOfNames can't be empty (because the 'member' attribute is required)... so if you want to remove the last member from a group you have to delete the entire group, i guess? what an odd restriction.

rather annoying combined with rfc2307bis because posixGroup is an auxilliary objectClass there, so you can't (easily) have an empty POSIX group which is sometimes actually useful if the group only has primary members

Hey! ZFS-question. I put a new drive in and replaced one disk and removed two others (i only got one drive in total). After resilvering the first disk and evacutation for the other disks with all the data to the new drive can I physically remove the drives and change sata port for the new drive and boot the system? I installed boot code etc on the new drive.

gendish: only if you also remembered to put the boot sectors on the new drive

Unfortunately, ZFS can't manage boot sectors, so you have to do it via gpart - with the wrinkle, that it'll depend on your system setup.

debdrup: yeah i did that :)

gendish: moving a disk to a new port is fine since zfs doesn't use the physical port to identify the disk. removing the old disks is also fine as long as they're not listed in 'zpool status' anymore. as debdrup said check bootcode etc...

thanks. one thing i haven't understand yet is if i got legacy boot (no uefi) can i still boot a drive with zfsbootcode bigger then 2TB in freebsd?

gptzfsboot is your friend

my zroot pool is larger then 2TB

and?

hey Remilia :)

sorry I do not get it

how is that related to installing the GPT-aware ZFS boot loader in the first sector

i think the question is about issues with large disks and the BIOS interface that the CSM loader uses... i don't know the answer though

unixwitch: yeah thats correct

BIOS does not care, it will see the MBR shim

but loader uses the BIOS interface to read from the zpool, doesn't it? i mean, it doesn't contain hardware drivers for storage devices, so it must do

Remilia: ok, that was my question yesterday. thanks for answering

MBR is limited to ~2TB for disks, but GPT supports what's called a pMBR.

i f*cking love your nerds

you*

see gptzfsboot(8)

Remilia: I did but im too stupid to understand maybe. But my guess was that it would function like you said

unixwitch: the issue with 2 TB+ is that you cannot have an MBR partition that exceeds that size

boot(8) does have a note about this in the DIAGNOSTICS section which presumably also applies to gptzfsboot

NOTE: On older machines, or otherwise where EDD support (disk packet interface support) is not available, all boot-related files and structures (including the NOTE: On older machines, or otherwise where EDD support (disk packet interface support) is not available, all boot-related files and structures (including the NOTE: On older machines, or otherwise where EDD support (disk packet interface support) is not available, all boot-related files and st

the BIOS calls are not limited to 2 TB

whoops, bad paste

So when restarting my server after removing the old disks that been running for 8.5 years the server should both on one single 12TB drive with boot-codes installed!

EDD is apparently over 20 years old so this hopefully isn't an issue on any system still in use (at least one that's running ZFS)

point is, gptzfsboot does not care about MBR limits and hands off control to the loader in the pool it finds

which means 2 TB limit does not matter to it

the issue isn't MBR limits, it's that some (very old) BIOS call interfaces simply don't support large disks at all, so there's no way to read such data.

those 'very old' interfaces do not support disks larger than 4 GB

ok. I understand! Thanks for the info. It's just a waiting game now: Evacuation of /dev/ada1p3 in progress since Sun Dec 10 19:32:27 2023 231G copied out of 2.35T at 145M/s, 9.60% done, 4h15m to go

gendish: fwiw, if this is a long term change, you might be better off creating a new pool on the disk you want to use and copying the data over, instead of using zfs remove, to avoid the (admittedly fairly small) memory overhead

unixwitch: yeah, but this is the path I choosed :P

I do love my old x58 system with a intel 980x extreme edition. I want to keep it alive for as long as possible

gendish: oh, the power consumption of those older i7s…

thanks again all. is there anything to do if i want to verify that the boot-code is really there on the new drive?

130 W TDP

Remilia: got decent electricity prices

but yeah i know

gendish:

err, mistype

gendish: I find it funny that R9 7900 has 12C/24T at half the TDP of your 980

got a 3700x laying around...

Remilia: dont give me anxiety over electricity bills now (:

3700X has 65W TDP iirc

when upgrading from 12.4-p3 to p9, I did fetch / install / reboot then I issued another install

and it said: No updates are available to install. Run '/usr/sbin/freebsd-update fetch' first.

in fact, I am using a 3700X right now

is that to be expected on such minor updates ?

last1: you only need to install twice for major upgrades

13 to 14

it is related to ABI

basically read what freebsd-update install tells you

for 2-stage it will say something like 'kernel updates have been installed. reboot the system and continue with freebsd-update install'

Remilia: yeah i know. but need to buy a new mobo, memory etc

for the 3700x

last1, I have been confused by the three different versions of freebsd-version too. Let me try to help "unconfuse" things.

The kernel is a specific part and has its own kernel version. There is a version for the kernel that has been last installed on disk and a version of the kernel that is running. These will be different if a kernel has been installed but not yet rebooted to it.

Then there is a version for all of the non-kernel userland parts of the base system. freebsd-version without options reports the userland version. With -k it reports the installed kernel. With -r it reports the running kernel, the same as uname -r report. With -u it reports the userland version.

"freebsd-version -kru" seems most useful and reports all three versions. The userland version is often different from the kernel version since they are two independent parts of the system.

When freebsd-update install says "No updates are available to install. Run '/usr/sbin/freebsd-update fetch' first." it means there are no pending updates to apply. Running fetch *may* or *may not* find a new update from the distribution servers. If you are fully updated then there won't be any pending updates to apply.

The next freebsd-update fetch will say "No updates needed to update system to 13.2-RELEASE-p7." (at this moment) and no updates are pending.

On my updated system freebsd-version -kru reports 13.2-RELEASE-p4, 13.2-RELEASE-p4, 13.2-RELEASE-p7

And none of this above mentions the ports which are individually versioned and installed in /usr/local separately.

Hope this helps!

rwp: I do not think that was the question, they were just wondering why they did not need to do the 2-stage install

I think they had already done the full installation.

If I over-explained it then that's my bad. But it seemed at least some of the confusion was why freebsd-version reported something different from the kernel.

Also "freebsd-update install" always says to run "freebsd-update fetch" even though one already ran that earlier in order to get something to install. This makes sense in the grand scheme of things. Because on the first pass fetch will say there is something to install but on the second pass it will say that no updates are needed and break things out of the loop.

rwp: you would be neither the first nor the last to overexplain :3

Until I got comfortable with the design of the system I know this caught me a few times. And actually I am chagrinned to admit that just a couple of weeks ago you might remember that I hadn't had enough sleep and got snagged again just recently!

I would like a command to be executed prior nut starts. Where would it be the preferred place? I need to detach a kernel driver for a usb device -the ups- if I don't run it before I get "Can't claim USB device [0001:0000]@0/0: Other error"

is it possible for traffic to leak while pf is reloading? like service pf reload

Uh... I can't find how to install python-bcrypt on freebsd

and the pip install bcrypt requires rust compiler. I don't want a rust compiler for a damn python module just to hash things

gendish: how is the idle power draw of your cpu?

Ok I solved it