how do i prevent x from loading at start

i made a mistake with my configuration and im getting a black screen

i am unable to change to another tty

ctl f1 or ctl alt f1 does nothing

i’ve tried all the function keys

Boot Environments. FTW! Boot to a previous boot environment. Mount the default one. Fix it. Boot back to the default one.

how do i do that?

Title: BootEnvironments - FreeBSD Wiki

At the boot screen there will be a menu of actions. Instead of booting directly select the boot environments entry (I forget the wording) and select a previous boot environment.

Boot. Then use "bectl mount default /mnt" to mount the default boot environment on /mnt.

Then edit files in /mnt/boot/loader.conf or /mnt/etc/rc.conf as needed to repair the problem.

"bectl umount default" to unmount it. Then "shutdown -r now" to reboot normally back to your default environment. FTW!

This does require ZFS. If you installed using UFS instead then this won't work. In that case boot the install disk, select a shell, mount the drive, and do basically the same thing that way.

i guess they left, but an easier way might be to just boot -s and disable whichever rc script starts the X dm

rwp: btw, since you seemed interested in my ZFS problem the other day... after the system rebooted the problem vanished. which is nice, i guess, but makes me suspicious there's some other problem that will make it recur later.

i did run a memtest86+ and it found no problems

hi, my desktop fan goes on for 1 seconds and then goes off, and it is like that all time

any ideas on why that could be happening?

uskerine_: since FreeBSD doesn't usually control fan speed by default, my first thought would be incorrect fan speed control settings in firmware. e.g. if the temperature fluctuates around 50C, and the fans are configured to go from 10% to 50% at 50C

(temperature goes to 50C, fan goes on, increase airflow reduces temperature to 49C, fan goes off again)

ok thanks

how can I change that?

BIOS?

BIOs/UEFI, yes, at least on most systems i've used

next time I shutdown I will check

thanks

it is really annoying

in the mean time, you could use something like 'sysctl dev.cpu.0.temperature' or htop (which can display CPU temperature in real time) to see if there is a particular temperature that triggers the fan to come on

enabling powerd (if you aren't already using it) might also help to scale down CPU frequency when idle

or it might be something completely unrelated... but this is at least the first thing i'd check in this situation

ooi, does anyone run -STABLE in production nowadays? i remember back in 3.x/4.x that was the normal way to do it, but it seems like the releng branches are preferred now

ISTR one regular mentioned using -STABLE as their daily driver.

(I prefer -RELEASE because of freebsd-update. Back in the 4.x days there was no way other than building from sources.)

I definitely daily drove -STABLE back when I last used freebsd on desktop

the only systems I have now are -RELEASE, but I still build from source, because sooner or later I'll have to backport a patch (most recently: D41330)

which also lets me enable extra hardening (full relro, kernel retpoline), disable stuff I don't need (e.g. krb5), and compile for newer CPUs (because why not?)

uskerine_, (unixwitch), +1 for powerdxx (plus plus) but -1 for powerd which does not function sufficiently on my system. But powerdxx works great.

VVD: on the off-chance you remember my 10GbE link issue from the other day, i moved the system closer to the switch and connected it with a Cat6 patch cable, and now the 10G link comes up fine. so i guess the problem is with the cable, but it seems odd for a cable to work at 1G and not 10GE... i don't suppose you have any ideas what might be wrong with the other cable?

unixwitch, check standard for cables different cats

supported frequency

it's Cat6A STP, should be fine for 10GE

the working cable isn't even 6A, just 6 (but it's also a bit shorter)

unixwitch, I could see a shorter better cable working for 10 GHz and a longer not so good cable not working at 10 but working okay at 1 GHz.

rwp: yeah, that's reasonable, but the 'long' cable is only ~20 metres, it should be well within spec (10GBASE-T over Cat6A is rated for 100m)

It's amazing to me that things work in the Giga Hz range at all. It's all electro-magic up in those frequencies.

maybe that cable is just bad, i could climb into the attic and replace it...

Title: Twisted pair - Wikipedia

> Building infrastructure

Cat 6 - 250 MHz

It might also depend upon the port on the switch. If those drivers have been damaged with an ESD in the past. If the connector on the RJ-45 socket was perfect or if it had some lint caught in it.

VVD: yeah, that page says Cat6A STP is rated for 100m at 10GBASE-T

so the cable is fine by spec

Cat 6A - 500 MHz

"Improved standards, tested to 500 MHz. Full 100 m distance at 10GBASE-T ISO/IEC 11801 2nd Ed. Am. 2. (2008), ANSI/TIA-568-C.1 (2009)"

unixwitch, Regarding your zfs problem that disappeared after a reboot, well, I guess I would still take it as a win. What controller are you using? I noticed the /dev/da0 name so should I assume a serious controller like an LSI2008 or something?

rwp: pretty sure it's not the switch, i tried two different copper ports and also a 10GBASE-T SFP+ i had spare which had the same behaviour

the working cable is fine in the same port

It could also be the end on the cable too. Just a not-perfect crimp might cause it not to be great at that point. Or it might have a damaged section in the cable at just a bad harmonic distance from an end which affects things badly on one of the twisted pairs. It only takes one bad twisted pair to prevent the full speed from working.

yeah, i already recrimped the socket end, i guess i'll try redoing the patch panel end as well

if i do have to climb into the attic i might just replace the entire cable with 10GBASE-LR, i hate this copper 10GE anyway

or maybe 10GBASE-SR, idk if anyone makes PCIe NICs with LR SFPs in

rwp: yeah, it's a bog-standard 2-port LSI: mps0: <Avago Technologies (LSI) SAS2008> port 0xf000-0xf0ff mem 0xfbbc0000-0xfbbc3fff,0xfbb80000-0xfbbbffff irq 24 at device 0.0 on pci1

although i don't have a proper backplane so it's using SFF-8087 to 4x SFF-8482 breakout cables

Sounds good to me though. Those work really well. I basically have three of those myself.

mpsutils says: Board Name: SAS9211-8i

yeah, i've been using mpt/mps controllers since the days of the old Sun X4xxx systems and they've always been great

except for one weird bug we had in the Solaris driver back in the day... but i had a chat with the driver developer on IRC and got a fixed version the same week, which was pretty neat. i miss opensolaris :-(

I'll just note that they get quite hot. I have made some fan ducting to make sure the heat sink on mine gets airflow. Otherwise they get so hot I worry they will cook off.

the PCIe slots here are packed pretty tight, there's an Intel X540-AT2 (10GbE copper NIC), this card, and a graphics cards (just a basic PCIe x1 VGA card) all next to each other... but the case does have pretty decent airflow and all the reported temps seem okay

I noticed it early on setting them up. My finger felt like it got burned when I touched it! After that I always improved the airflow on it. I added a fan on it for one and I made a ducted air tube from a fan on the other one.

With a little airflow directly on it though the temperature noted by my finger seems to be just fine.

actually i found an older photo of the layout and it's not as bad as i thought i.imgur.com/Nc30765.jpeg

i really hate how PCIe-starved these AM4 boards are

i should replace it with a proper threadripper/epyc/something one day

Looks like the cpu cooler fan will swirl air around it.

I need to shop for two motherboards to replace two aging and partially failing motherboards that I am using. One has a failed dimm socket reducing memory by half.

yeah, the CPU cooler is oriented very weirdly, but it flows bottom to top so it should pull air from the PCIe cards through the top of the case

rather than pushing the CPU's hot air onto the cards which wouldn't be ideal

i wanted to put the VGA card in the bottom x1 slot but it didn't work there, it only works in this slot. why? who knows.

If you reach in there and put your finger on the SAS/SATA controller hit sink and find you can't keep it there for it burning your skin then you might consider adding a fan on it. But if it feels cool enough to you then it is good enough as it is and don't worry about it.

i might check that tomorrow

I don't care about looks but only functionality and no one can see in my case anyway so I have literally taken a bit of string and tied a fan into place on systems as needed. Around the disk controller as we are discussing. Around disk drives that needed more airflow. I mostly use 120mm fans that are pretty quiet so the extra fan does not add appreciably to the noise level but still moves a lot of air.

this case does have a clear panel on the side but only because it was the only full tower case i could find at a reasonable price with a decent number of 3.5" slots (9!). i've been thinking i should at least add another fan at the back at some point

the airflow from the front goes straight over the disks so their temps seem okay

I wish I could find a decent deskside case with that many 3.5 inch slots! Four is about the normal and it is really very difficult to locate one with more. There are two home NAS cases that keep popping up for reviews but only one is available and the other is $300.

# smartctl -x /dev/da0|grep Temp

Current Drive Temperature: 36 C

Drive Trip Temperature: 40 C

I eventually gave up and bought rack mount cases. I have been using them ever since.

rwp: this is a Corsair Obsidian 750D (corsair.com/uk/en/p/pc-cases/cc-901…n-full-tower-atx-case-cc-9011078-ww) - if you can live with the clear side panel it's actually not a bad case for the price. but you will need to buy some extra 3.5" cages because it only comes with 2 by default (3 disks per cage, so you need 3 for 9 disks)

also the plastic disk brackets are super cheap and break if you look at them wrong, but you can buy more of those

That appears to me to have three front side accessible 3.5 inch bays that someone could install up to 3 hot swap cages.

Oh! I guess there are 6 more inside mounted sideways. That would be okay.

it looks like this: i.imgur.com/wlcROjD.jpeg - except imagine the third disk bay is on top of the other day, that's how i have it configured

s/other day/other two/ (wtf)

according to the manual that's not a supported configuration but it's obviously meant to work this way, there are slots at the top to hold all 3 cages securely

(when i had it configured like in that photo, the disks in the third cage got very hot, since as you can see there's not much airflow over them down there)

and yes, re: front bays, i have an Icy cage with 4x 2.5" bays that i use for my SSDs. you could fit up to 3 of those if you wanted

That looks pretty good. Looks like all of the drives would have solid power and connections. Looks like reasonable airflow, though the drives are planar in the wrong direction for great air flow across them. But your data shows they are good enough.

I would definitely tie fans on with twine if needed! :-)

rwp: disk data+power looks like this: i.imgur.com/iN2dIYB.jpeg (except i have 8 disks now, and they're all in a single vertical column). lots of places to tie down cables at the back side which is nice.

those blue cables are the SFF-8087 breakouts

I think it looks good. Dust and spider webs aside. I can't throw stones on that either though. Looks like as good of a case as you could find for it today. Lots of space and lots of airflow. Good!

i did ask Noctua support about having two fans blowing over 8 disks in that configuration and they said it would be fine, fwiw

although there is some variation in disk temperatures now that i look at it

# smartctl -x /dev/da7|grep Temp

Current Drive Temperature: 29 C

Drive Trip Temperature: 85 C

# smartctl -x /dev/da6|grep Temp

Current Drive Temperature: 36 C

Drive Trip Temperature: 65 C

i'm all about doing things as cheaply as possible though so not too concerned about that :D

true/false: shielded twisted pair cable (STP) has to be grounded at both ends, or else it's worse than UTP because the shielding acts like an antenna to pick up interference?

Don't quote me on this, look it up, but IIRC shielded twisted pair is grounded on only one end.

But it's not due to becoming an antenna. It's to avoid creating a ground current loop.

hmm. interesting

You want zero current flow in the shield. So if it is not connected through there can't be current flow in the shield.

so does this mean NICs don't have the shield connected to ground, while switches do? or the other way around? or are you meant to use unshielded patches to connect to hosts?

I am embarrassed I don't immediately know the answer to that question. But I don't know what the strategy is there. I mean I have a BSEE degree but I have slept many times since then and most of those brain cells have expired.

But remember that most networking equipment is expected to be indoors and unshielded.

When dealing with shielded equipment that's unusual and so I think you would be expected to Do The Right Thing at that point as required.

community.spiceworks.com/topic/4330…g-for-shielding-on-a-cat-6a-network agrees with you and says to only earth at one end, but it doesn't say how actual behave here

Title: How to set up grounding for shielding on a cat 6a network

i suppose the obvious answer is to use UTP for the last-metre connection to the host, so the shield isn't connected at all at that end

(... which is actually what i'm already doing, i just thought this was wrong)

About 15 years ago I had a project installing weather cameras up towers and that used outdoor rated shielded twisted pair due to the hazards of lightning and trying to protect against it.

interestingly one post there says _not_ to run the shield to the switch, but to terminate it at the patch panel. ground the patch panel, and use UTP to the switch

my understanding (could be wrong) is that standard networking equipment - switches and nics alike - won't ground a shielded cable through its connector alone

We had lightning arrestors up at the top of the tower. Another at the base of the tower. And every couple of years one would be blown out by lightning either a strike or just close and I woudl have to replace the arrestor.

... which does make sense, and our patch panels have an earth terminal, so maybe i should do it like that

For that job crimping on the shieled connectors was truly a painful experience. Just awful.

rwp: we use Keystone connectors which are quite nice to terminate. the shielding is just a bit of copper tape over the wire, and it snaps in like a Krone terminal but without needing a tool

We did not have a patch panel available at those sites. But as I understand it things would have been just as you mentioned that the patch panel would be used to either ground or not ground that end and the final cable from the patch panel to the switch would be unshielded.

i rewired one the other day (because of my 10GE issue) and it only took 10 minutes or so

I could have used those nicer connectors. But that's water long gone under that bridge now.

On nice weather days it really wasn't bad to be working up a tower but when it was cold with a wind that would freeze my fingers off! That was not the normal plan. But I remember needing to do a repair after a lightning strike and needing to run a new cable and install new boxes and had to crimp an end on the tower and that was truly miserable.

what in god's name is this: amazon.co.uk/Grounding-Connection-E…ding-Pillowcase-White/dp/B0CMT5BYY5 - it claims to be an "earthing cord" but it converts a 2-pin USA plug to a 3-pin UK plug... but a 2-pin US plug doesn't have an earth pin?!

Title: Grounding Cable Connection with UK Plug, 4.6 m Earthing Cord for Grouding Pillowcase, Grounding Sheet and Grounding Mat: Amazon.co.uk: DIY & Tools

We are talking like 30 feet up and not like really tall. These were airport beacon towers. Low to the ground in the grand scheme of things.

That is just what it says it is but I think it is pretty funny. I believe the idea is that in the winter when it is dry and you walk across carpet you can build up quite the static ESD zap which then arcs when you touch something. That mat is meant to discharge that harmlessly into the ground. Er, Earth.

Americans say "ground" and the British say "earth" and the two are the same. We are a people divided by a common language.

but how can it discharge anything to earth (ground) when it only has a 2-pin plug?

So that cord has only the ground/earth terminal on the power plug. The other plug pins are plastic, insulating, and there just to hold the plug in place.

the 2-pin plug is live+neutral, right? there's no ground?

It has a 3-prong plug for the AC mains socket. Only the ground/earth terminal is metal and the other two for the AC mains is plastic insulating and just mechanical support.

click the second picture

On the other end it has one snap connector. It's a snap like a coat snap. That connects to the blanket.

it's a US 2-pin plug with a US-to-UK adapter

if that's a normal adapter of that type, only live+neutral are connected and ground is not connected anywhere

Amazon might have redirected me to a different image. I am in the US and that is a UK site.

I am seeing only British style plugs.

Oh! I see the photo you mean. It just has one prong.

this is the imagine i find concerning i.imgur.com/5SS1kTg.jpeg

That is just the ground/earth prong. Just push it into the ground/earth hole of any power plug anywhere in the world that has ground plugs.

oh, i thought it was a 2-pin plug viewed from the side so it looks like a single pin

British ground holes are rectangular. US are round. US sockets: en.wikipedia.org/wiki/AC_power_plug…/File:NEMA_5-15_Outlet_120V-15A.jpg

Title: AC power plugs and sockets - Wikipedia

And that amazon mat allows you to stand on it to discharge built up static before sparking something. That's the plan.

But for years we would have said attach a wire to the house water pipes. That's been the usual standard.

Except now all new construction uses plastic water pipes! Making hooking to a convenient water pipe worthless now.

Which is probably why they are plugging into the ground/earth hold of an AC mains power socket.

Meanwhile... Those type of anti-ESD mats are very familiar to anyone who has ever worked in an electronics manufacturing site.

We would have wrist straps, or shoe straps (I preferred the shoe straps), and everything would have an anti-ESD mat to avoid zapping things.

well, in any case, i went with a standard UK re-wireable plug and some green/yellow earth wire, which i hope is BS-compliant :-)

there's no water pipes or other earth sources near the patch panel so i'll just have to plug it into a socket

That actually means not having a direct wire connection but always using something like a 1 Mega Ohm resister so that there is no spark zap but it slowly drains the charge away without a spark.

In the US almost all power wiring is in metal conduit with metal junction boxes. Though those are also turning into plastic in new construction. But can almost always get a good ground/earth connection.

Unless we are dealing with amateur ham radio transmitters in which case it's not sufficient there and you should run a separate heavy ground/earth but that's really getting off topic.

I am getting called IRL and must go...

But ground wires are just wires. They don't need to be special wires. Any wire will do. Bare wire without insulation is generally used.

rwp: yeah i know, but you're suppose to use green/yellow wire for earth here, so people know what it is

a blue wire going from a 240V mains plug to a patch panel might raise some eyebrows :-)

Is there any possibility to fix the error I caused?

Here is what happened:

I have a FreeBSD installation with root-on-zfs (dual boot with Windows), the pool was called "root". I was trying to rename it to zroot. For that I booted from 14.0 memstick and opened a shell. There I did as follows:

zpool import root zroot

zpool export zroot

now I cannot boot anymore and it says "insufficient replicas" and below "invalid lable" when trying to import that pool from usb stick

well i managed to import the pool from memstick boot. It says ONLINE and a scrub revealed no errors.

nevertheless I cannot boot it, it says "Invalid file system specification"

I set vfs.root.mountfrom to zroot in loader.conf

somehow it seems to find 2 pools named zroot and the conflict is causing a püroblem

'zpool import zroot' says 'cannot import zroot, more than one matching pool', 'zpool import -a' import the pool, i can set mountpoint to /mnt and it is mounted and looks good.

does FreeBSD have a driver for Mellanox ConnectX-2 cards?

oh, mlx4en(4) says yes, the summary is just incomplete

hello great master's (nice UGT) I need some help here, I'm building one debian ltsp server to two raspberries pi4, but have something wrong, kkk...

I Connected my laptop to my tv via hdmi. used arandr and it was fine. problem was that the sound was coming from my laptop not the tv. any ideas ? running 14.0-RELEASE

jb1277976: hw.snd.default_unit is probably what you're looking for; see: docs.freebsd.org/en/books/handbook/multimedia/#sound-setup

Title: Chapter 9. Multimedia | FreeBSD Documentation Portal

how do I specify which minimum version of Go a port wants

encountered Ruby-like behaviour with Go 1.21

Remilia: did you check the porters handbook?

I actually didn't, thanks for reminding me (half asleep)

hmm, apparently I cannot specify '1.21+' like illustrated for python but it works with 1.21, which is why I asked that question above haha

I guess I should have tried without + from the start

wonder if you cannot specify a minimum version for Go, only exact

I'm trying to get mysql57-server running in a jail. I'm getting an error that libpcap libraries are missing, preventing the service from starting up. However, I do see them in my corresponding Poudriere build jail.

Do I need to manually install net/libpcap?

The answer to that is no. I built and installed libpcap, and the error persists.

wait, why does mysql server require pcap?

and libpcap is part of base system anyway

Schamschula: is there a reason you need that EOL'd version btw?

That's what I've found out, but both ping and mysql-server throw the error

your jail has Issues then

with its distribution set

check `freebsd-update IDS`

Yes, I need to get back up and the update to mysql 8.0 did not go smoothly. I've got duplicate .ibd files.

The jail was a fresh install last night (13.2 RELEASE)

can't you just restore from the dump you made before upgrading

hmm

you installed it with bsdinstall in jail mode I presume?

Nope. I can't even get it to give me a clean fresh empty structure.

what

sorry I do not understand

Schamschula: did you use `bsdinstall jail /path/to/jail`?

or something else?

I used pkg with the jail

s/with/within/

once again sorry I do not understand, how did you use pkg to create a new jail?

are you using pkgbase?

Remilia: for years now, yes

lol, sorry…

I have PkgBase in my hilights…

meena: oh

if they used pkgbase I have no idea so will withdraw

I had no idea that with pkgbase you can use pkg to create jails w

No, I followed docs.freebsd.org/en/books/handbook/jails/#classic-jail

Title: Chapter 17. Jails and Containers | FreeBSD Documentation Portal

Remilia: yeah, my jails are tailor made for the software they run and only have very minimal stuff installed

Schamschula: I would try to create a new jail using `bsdinstall jail` and seeing if stuff works there

also check if you have /usr/lib/libpcap.so in your jail at all

and /lib/libpcap.so.8

OK. I'll need to figure out how to get the right OS version under bsdinstall

it will use host version though

if I recall right

Yes t's there.

The hoist version won't work.

Here't the actual error I'm seeing: ld-elf.so.1: Shared object "libcap_fileargs.so.1" not found, required by "wc"

Schamschula: in your jail, do `ldd /usr/local/sbin/mysqld` and check what it is linked against

oh

libcap

you should have started with that

sorry, I thought you meant libpcap

based on this [15:11:59] <Schamschula> I'm trying to get mysql57-server running in a jail. I'm getting an error that libpcap libraries are missing,

libcap is /lib/libcap_fileargs.so.1 though

in this case

and if you do not have that in place your base system is ???

In the meantime, I do see that I have a mysql skeleton on the host OS.

I may try restoring there.

libcap_fileargs is not in base.txz

It's installed on the host system and the Poudriere jails.

yes, it comes from something else

OH

yeah it makes sense

Indeed, I did a pkg list pcap, not there.

why are you looking for pcap?

I am just curious

are you sure libcap and libpcap are the same?

Because I misread the library name...

I've used libpcap for so long, I ended up assuming something.

this is actually a bug, 13.2-RELEASE base.txz is missing libcap

meena: ^ wow

whew. wow. that sounds bad

meena: i.koumakan.jp/2023-12-18/1702914489.png

All of them are missing here's ping: ld-elf.so.1: Shared object "libcap_dns.so.2" not found, required by "ping"

fetching the 13.1 one

Not there, either. Only my 14.0 build jail provides libcap_*

I wonder if this is just new in 14

it's not in my install of 13.2 either. couldn't that mysql build be from a 14 system?

yeah

that is noticeably more likely

Could be, but the executables in the jail seem to be linked to it.

Schamschula: `freebsd-version -u` in the jail

BTW: mysqld is in libexec

meena: disregard that

they are in /lib/casper in 13.

I was blind \o/

Ahh: It shows to be 14.0-RELEASE-p3. Not good.

well there you go

you have a mix of 13.2 and 14.0

or

you are not inside the jail

Time to detlete it and start over.

tbh I am curious as to how you ended up with this arrangement

you probably have /lib/casper/ in there

have you tried to mysqldump and then restore into the new version? most likely you already have the dumps from your backups?

Definately in the jail via jexec myssql /bin/sh

ridcully: yes, but the nearly week old scheduler dump.

And, I can't log into mysql -u root.

i doubt there where new mysql 5.7 releases in the last week, that would give you a different result. so did that work?

No, 5.7 is pretty much done.

5.7 is utterly EOL

Indeed.

It is, but I had previously updated to 8.0 twice, and gone back to 5.7 twice. Unusably slow.

the problem with the current situation is that if they ran 8.x with 5.7's data directory and that failed, there is a high chance the files are already incompatible

On linux, the scheduler can make a difference. I don't know what the BSD equivalent is.

I backed up the file structure before updating. So I have a copy for 5.7.

why not use MariaDB instead btw?

do your tables use some mysql-only storage engine?

I've also tried that over the summer. I had issues getting it up and running.

then again did 5.7 have any

huh

If you migrate to MariaDB, you have to do a logical dump + restore.

There are other steps needed, as well.

yeah but you must always do a dump before switching major versions anyway

Progress. Reset mysql server root passwd. Next: I'll see if I can restore the dump.

exciting.

evening guys, wonder if anyone could help me a bit, messed up with R13->R14 upgrade and zpool update - forgot to update bootloader aaand electricity went down, so I can no longer boot the devil

boot from CD or usb?

its basically home NAS with two pool and m2/ssd acting as system drive

paulf: its basically home NAS with two pool and m2/ssd acting as system drive, i have shell from usb now, but i messed up bootloader trying to recover the system probably

*two pools, zroot and zdata, let's say

Title: dpaste/JLKc (Plain Text)

nimmen_dimmur: As paulf noted, if you can boot from rescue media, just copy in the new bootloader.

nimmen_dimmur: For an EFI system, copy loader.efi from /boot (as mounted in your rescue system) to whichever drives and names are appropriate.

A local example, not taking into account where you've got stuff mounted in the rescue environment: cp /boot/loader.efi /boot/efi/efi/boot/bootx64.efi

Note the doubled EFI, as it's common to mount these on /boot/efi, and therein is the EFI directory. Variations depend on whether you're using EFI boot variables or just the fallback naming, bootx64.efi as noted.

nimmen_dimmur: If you have more detailed questions, go ahead and ask them.

yeaaah, well thats the part i messed up... I dont remember how I did setup system so I went with some forum post about bootloader update, and obviously I did guess wrong

i did some gpart bootcode with wrong stuff

Title: dpaste/riuG (Plain Text)

nimmen_dimmur: Is your system UEFI?

zfs

That's unrelated to the question.

sorry, a bit tired, not sure to be honest

i've set it up years ago and it was running with no problems

I ask because "gpart bootcode" is only something you'd need for a legacy-boot system.

And your first paste shows EFI System Partitions.

But assuming it is UEFI, you seem not to have your ESP mounted in that second paste.

ESP? with second paste i'm showing that I thought i'll just mount zpool, copy over /etc and reinstall everything, but i don't see any zroot / data

nimmen_dimmur: 0x0.st/HY77.png

yeah, it was different before i messed up that part, I think second item was - maybe third was uefi? (definitely *efi*)

You've also got that freebsd-boot partition there, so I guess it could be either.

aha, found the wrong command - gpart bootcode -p /boot/boot1.efi -i1 nda0

so now what i'm pasting is after it - which messed up the bootloader to non boot state

command was issues after i booted from usb image, trying to fix the old bootloader problem (zpool upgrade after R13->R14 upgrade, without bootloader upgrade)

*issued

nimmen_dimmur: If you "man gpart" it talks about the different boot-related files.

Title: gpart(8)

Specifically look at the "BOOTSTRAPPING" section.

Not an exhaustive list sadly.

The EXAMPLES section might interest you too.

do i assume right if its zfs base, than pretty much gptzfsboot is the only option to use?

*zfs root

Since we see EFI stuff, GPT is a safe bet.

That said, you don't need gpart bootcode at all if it's UEFI.

Do need it if it's a legacy boot.

should be uefi, i did fresh 13 install there with mostly close to defaults (if i remember right)

so gpart bootcode -p /boot/gptzfsboot -i1 nda0 does sound right? since i did pretty much mess up with similar command

not what does not mean i'll solve the issue, but i'll be back to where i was?

*not that it does mean

>> That said, you don't need gpart bootcode at all if it's UEFI.

Sorry, a bit busy with work and concurrently sick, so my bandwidth is a bit lower than normal.

In short, find and mount your EFI System Partition. See what's there. Copy the correct boot loader into place.

It's probably going to be /boot/loader.efi (relative to where you have your /boot mounted)

mason: no worries, thanks for help anyway, well thats the problem i would have done that but i'm not able to mount it or i'm too dumb today to figure it out, bsd.to/Jsk2

Title: dpaste/Jsk2 (Plain Text)

obviously i tried to mount it before i tried the gpart bootcode stuff

so lets keep this simple, since the zroot seems to be ok, but why i can't see the zroot/ data? how can I mount it, i'll just copy it over and do a reinstall, since I think it will take more time for me to understand the bootloader things than going for this workaround =]

I see nothing immediately wrong with your mount_msdosfs syntax.

Your paste bsd.to/riuG suggests that your ZFS stuff is accessible...?

Title: dpaste/riuG (Plain Text)

mason: it is, but not the / part???

i did zpool import -R /mnt/zfs zroot

obviously /mnt is from usb drive and I did mount -uw /

Title: dpaste/Bjug (Plain Text)

you can ignore the 'unionfs' bit as i've used it to boot sshd so could connect it in more convenient way from another computer

I do the same.

nimmen_dimmur: Not sure what you mean by, "but why i can't see the zroot/ data?"

Is there stuff that you expect to see in /mnt/zfs/zroot?

The typical layout uses something like zroot/ROOT/default or zroot/ROOT/some-snapshot

mason: i did expect to see the / data - lets say /etc so in this case -> /mnt/zfs/zroot/etc ?

You probably shouldn't expect to see anything in the base zroot dataset.

i see

nimmen_dimmur: Have a look at 'zfs list' as well

so my main concern is how do i recover /etc ?

sorry, /usr/local/etc ideally

nimmen_dimmur: If you paste "zfs list" output we can likely tell you where to look.

mason: formatting is pretty bad so i'll send you this one then: paste.ofcode.org/37SNrrzaw4pcFfJTq9JnPLf

Title: Paste ofCode

So, yeah, one of those zroot/ROOT options needs to be mounted up as the root of your hierarchy.

Apologies, meeting, but I'll check in after.

Almost certainly easier to reinstall bootcode than reinstall EVERYTHING. I can poke about some more later.

mason: no worries, you already helped a lot

well ok, i'll ready more on bootcode then if you think its easier

but I may already messed it up one more time with - gpart bootcode -b /boot/pmbr nda0

nimmen_dimmur: That shouldn't harm anything. The real thing to figure out is if it's UEFI or legacy. I suspect it's UEFI, so you need to figure out what that mount_msdos is failing. If you figure that out, then you can mount it up and copy in the correct bootloader.

mason: so would dd /boot/loader.efi /dev/nda0 not mess things up?

nimmen_dimmur: dd is commonly used for block devices. You want to copy a file to a filesystem.

You want to mount that first partition somewhere, and look at what's in it, and then copy loader.efi to the right places based on what you see.

Hi, I'm experiencing freezes on my APU2 based router since upgrading to 14.0-Release, never had any such issues with 13.2-Release. Any known issues? It's this board: pcengines.ch/apu6b4.htm

Title: PC Engines apu6b4 product file

espen: Regardless of what you learn here, it'd be a great idea to file a bug about it.

Thanks, but I haven't been able to get any more debug information. So the bug report would be pretty empty. I tried to enable dumpdev=AUTO, but there's nothing in /var/crash after the freeze.

Even that's a start.

/boot/loader.efi[23~~

sry

bugs.freebsd.org/bugzilla/show_bug.cgi?id=275830 filed a bug 🤞

Title: 275830 – APU2 based router freezes after upgrade to 14.0-RELEASE

well at least i got my zroot/ but now i remembered that mason is probably bot? hah

Most people can't tell.

I can pass the Turing test, though.

I'm pretty sure mason isn't a bot.

At least not tained on social media comments, he seems nice ;)

*trained

"not tainted" also works. :-)

for a reason freebsd wont apply my pf rules on boot

espen: =]

i was able to mount /boot but surely this did not help as there is no /boot/loader.efi so then i remembered from older times that there was a nice bot in here

I added mason on linkedin, so this will be most embarassing if he's not real

mane, 2 possible reasons: not having pf_enable="YES" in rc.conf, and syntax error in your rules file.

well ok, but surely there used to be a bot here

what was the name?

kevans: LinkedIn is not real

VimDiesel is a bot, but he mostly scans for URLs

mason is integer. At least in Fortran.

heheh

To be fair, I'm sick today (not covid thankfully) so I'm at my most botlike.

well even if mason is bot im doing a copy of / and reinstall since bootloader was too tough for my determination today

nimmen_dimmur: There are more smart people on and paying attention now.

mane, pfctl -vnf /etc/pf.conf

sadly i'll have to get to this anyway, as i have 2 nas servers and they both have same problem

If we can figure out your mounting issue, I bet you'll be mostly done.

I think I've accused meena of being a bot before

except i did not mess up the bootloader of the second :/

nimmen_dimmur: The thing that comes to mind is that if you somehow corrupted your ESP, that might be why you can't mount it. Not sure how reasonably to verify this.

V_PauAmma_V i have pf_anble="YES" in my rc.conf

espen that just returns all the rules for my pf

mane: try pf_enable="YES" ?

nimmen_dimmur i have that

ah

mane, it's verifying it, so if there's something wrong it should tell you

if i run service pf start manually all works fine

pf_flags?

mane: are configuring things, that are not there yet/when it runs while booting? e.g. vpn, tunnels, things from a local rc-file, ...

*you

everything works fine but pf

is order of the services in rc.conf relevant/

?

mason: no clue why it doesnt mount, thats how i went to radical uknown solutions and messed up my bootloader

mane: no, rc.conf is basically just env-vars read before running the actual scripts

ok, i tried working around it and tried to start to service in /etc.local but rc.local seems to not be executed

i thought maybe it doesnt mount as it it zfs on root, as in, its not msdosfs and needs something else, i'm as lame in boot stuff as i am with 8086 assembler(or any other, for that matter)

i've noticed that pkg update/upgrade has picked up nothing for a while (a week or two?)...checked today and $ABI as used by the update script is an empty variable. what *should* it be for 14.0-RELEASE on amd64, and why might it be empty please?

ps. FreeBSD.conf is set to use 'latest'

cant tell as my all bsd boxes are out of order :D noticed that R13 didnt have much updates, thats how i spotted there is R14 and did upgrades...

nimmen_dimmur: fair play! i'd have expected something, however minor though...

FlibberTGibbet: quick question, you did upgrade or fresh install? I remember some minor issue with pkg

nimmen_dimmur: upgrade from 13.${whateveritwas}. all smooth otherwise

you would double check from R14 git /usr/local/etc/pkg.conf and compare

if no diff, then there could be more things involved

could be. i'm guessing (heh) that the absence of the environment variable is kicking pkg into touch, as that's what the conf file seems to use to select the repo. but why that variable is empty could be down to something else, like you say

FreeBSD does not need to have a pkg.conf file. Many installations will work well with no pkg.conf at all or with an empty pkg.conf (other than comment lines).

my bad

nifty

you could try this then, no/empy pkg.conf

well, i'll back it up first :D

smart choice

/etc/rc.local i mean

thanks for help mason, i'll have to recover other server tomorrow with bootloader, this one will do the crude copy/reinstall way

Is there another way other than parsing "zpool status" output to get a list of devices that part of a pool?

all things being equal the timestaps at pkg.freebsd.org/FreeBSD:14:amd64/latest/Latest are 8 December, so there might not be any updates anyway!

Title: Index of /FreeBSD:14:amd64/latest/Latest/

unixwitch: you know, I should test my ConnectX-2s on FreeBSD... I've generally had very bad experience trying to get them to work

only really got anywhere on RHEL 6, newer Linux just wouldn't work

how do i change password with pw?

ok got it

espen: I wonder what is the major difference between apu2 and apu4 because I upgraded my apu4d4 to 14.0-RELEASE without any issues

a good while ago

dh: zpool list -v as well

(that one can be made "parsable")