networking in bpi-m5 with new 26.03.1 release.

[gene1934]

This system has been running for several months months as a backup server with a nobel release about a year old, maybe some older, using two usb3 7 port hubs because 1 hub could not handle the current draw of 5 TSMC 4T SSD's using startech usb3.1 cables to connect 5 if them in a /raid6 configuration for an 11.1 Tb drive.  But I decided to do an  as sudo -i, "apt update" and was advised that would update about 430 files so I figured I'd do an apt update -y, then and apt --purge autoremove".  Unforch it went into lala land while making a new initrd. After about 20 minutes I did a powerdown to be greeted by a repeating loop indicating there was no network.  So I dl'd & wrote the month old 26.03 release. Which after getting chrony to work using my own local ntpsec, finally got the correct time and I'm able to login both directions using ssh me@machine assigned address, but no outside my local net addresses are found, dhcp isn't working.  It is supposed to first query /etc/hosts wich does work, but if that doesn't resolve the name in dnsmasq, the router then query's the dns server at my ISP, which works for the rest of my 8 machines here, but does Not work for the 26.03 just released a month ago. 

 

ip a reports:

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host noprefixroute 
       valid_lft forever preferred_lft forever
2: end0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether be:63:9c:35:dd:4f brd ff:ff:ff:ff:ff:ff
    inet 192.168.71.2/24 brd 192.168.71.255 scope global noprefixroute end0
       valid_lft forever preferred_lft forever
    inet6 fe80::bc63:9cff:fe35:dd4f/64 scope link 
       valid_lft forever preferred_lft forever

 

ip r reports:

192.168.71.0/24 dev end0 proto kernel scope link src 192.168.71.2 metric 100 
gene@amanda:~$ 

 

A ping -c1 yahoo.com kills enough time its several seconds before it reports "network is unreachable"

Can you see whats wrong, or give me the newer tracing tools to find  the error?

 

Thank you.

[gene1934]

no wireless is configured except the BT mouse, which works fine.

[bedna]

What's in your /etc/resolv.conf?

 

And what do you mean with "dhcp isn't working"? Is 192.168.71.2 not the correct ip?

Where is your dhcp? Your router?

Edited Thursday at 10:48 PM by bedna

[gene1934]

the router reflashed with a recent dd-wrt runs dnsmasq, which caches the ISP.  If that fails it query's the ISP's server, which works fine for the other 7 machines here.

[bedna]

I repeat, what is in your /etc/resolv.conf?

[gene1934]

untouched and wrong as can be:

#
# This file might be symlinked as /etc/resolv.conf. If you're looking at
# /etc/resolv.conf and seeing this text, you have followed the symlink.
#
# This is a dynamic resolv.conf file for connecting local clients to the
# internal DNS stub resolver of systemd-resolved. This file lists all
# configured search domains.
#
# Run "resolvectl status" to see details about the uplink DNS servers
# currently in use.
#
# Third party programs should typically not access this file directly, but only
# through the symlink at /etc/resolv.conf. To manage man:resolv.conf(5) in a
# different way, replace this symlink by a static file or a different symlink.
#
# See man:systemd-resolved.service(8) for details about the supported modes of
# operation for /etc/resolv.conf.

nameserver 127.0.0.53
options edns0 trust-ad
search .
 

Damn, doesn't anyone use /etc/hosts for small home networks anymore? Toss in that you don't install ANY manpages so us users are running blind 99% of the time.

Thanks bedna.

[gene1934]

gene@amanda:~$ resolvectl status
Global
         Protocols: -LLMNR -mDNS -DNSOverTLS DNSSEC=no/unsupported
  resolv.conf mode: stub

Link 2 (end0)
    Current Scopes: DNS
         Protocols: +DefaultRoute -LLMNR -mDNS -DNSOverTLS DNSSEC=no/unsupported
Current DNS Server: 192.168.71.0
       DNS Servers: 192.168.71.0
 

[bedna]

42 minutes ago, gene1934 said:

nameserver 127.0.0.53

 

Either the config on the device is wrong, there is supposed to be a dns ON the device or the dhcp server is handing out wrong dns to the device.

 

42 minutes ago, gene1934 said:

Damn, doesn't anyone use /etc/hosts for small home networks anymore?

 

Not on every specific device if there is a local dns server available on the network. Then it's normally done on the local dns itself by first checking local configs (like /etc/hosts) and if not available, resolve using online dns servers and then send the response to the device.

And the dhcp server is obv configured to point to the local dns server.

That way the configs gets propagated via the dhcp lease to all devices on the network.

 

Edit

As far as I know, network manager is default on armbian (correct me if I'm wrong) and that should have edited your /etc/resolv.conf. So since it's not, you have made changes outside of default config somehow.

Could be that the specific release and device you use does not utilize network manager. But since you have not disclosed that info, hard to know.

But it looks like you use systemd-resolved so you should check how that works.

 

Edit2

I was wrong about network manager, see https://docs.armbian.com/User-Guide_Networking/

Edited Thursday at 11:49 PM by bedna

[gene1934]

NetworkManager is how I have been trying of configure end0 but it won't run by itself, ONLY when armbian-config runs it as a client.  IDK if I can copy/paste its show me output.

[bedna]

If you want to use network manager you have to make sure systemd is not overriding your configs, ie disable relevant systemd services.

 

To me, it sounds easier to just use what is installed default, in this case systemd and use networkctl to interact.

 

But it should be irrelevant. If you configure your dhcp server and dns server correctly, both systemd and network manager should pick up correct config in default config state so you should not even have to interact with either of them.

(I started assuming you are using a minimal image, therefore systemd as default. You haven't disclosed what version you use, just something 26.3)

Edited yesterday at 12:20 AM by bedna

[gene1934]

the xfce4 desktop version, specficallly:

Armbian_26.2.1_Bananapim5_noble_current_6.18.15_xfce_desktop.img

after unpacking the xz version.

 

The previous version that needed 430 some updates also had  the mdadm stuff preloaded and had no trouble recovering the /raid6 once I had created the mount point with a mkdir /raid6. this one fails on all counts so, no entries other than control in /dev/mapper.  So I'm assuming the first thing I need to install once networking is working is mdadm.  But without an external network I'm locked out of everything else.  Very frustrating.

 

gene@amanda:~$ ls /dev/sd*
/dev/sda   /dev/sdb   /dev/sdc   /dev/sdd   /dev/sdd2  /dev/sde1  /dev/sdf1  /dev/sdg1
/dev/sda1  /dev/sdb1  /dev/sdc1  /dev/sdd1  /dev/sde   /dev/sdf   /dev/sdg
So all the drives are there but not recognized as the raid6.

The odd /dev/sdd is probably the 1T 2.5" lappy drive I put in for a faster /swap and had plans for its use as a scratchpad drive for amanda's use but is essentially just 50G of swap now.

 

So what do I edit to fix the network ???

 

[bedna]

5 hours ago, gene1934 said:

the xfce4 desktop version, specficallly:

 

That is supposed to be using network manager, again, see documentation

 

 

5 hours ago, gene1934 said:

So all the drives are there but not recognized as the raid6.

 

Your question was about networking. How to configure your raid is a completely different question and has nothing to do with network.

 

5 hours ago, gene1934 said:

So what do I edit to fix the network ???

 

Since both systemd-networkd AND network manager (and you also mention dnsmasq) is enabled, you have most likely done changes to the system way outside of what you should.

If you claim the image you downloaded CAME with all that installed and enabled by default, I would suggest you provide proof of that and then report back as a bug.

 

As for your general network setup, I gave you my opinion earlier:

11 hours ago, bedna said:

11 hours ago, gene1934 said:

Damn, doesn't anyone use /etc/hosts for small home networks anymore?

 

Not on every specific device if there is a local dns server available on the network. Then it's normally done on the local dns itself by first checking local configs (like /etc/hosts) and if not available, resolve using online dns servers and then send the response to the device.

And the dhcp server is obv configured to point to the local dns server.

That way the configs gets propagated via the dhcp lease to all devices on the network.

 

[gene1934]

In this local network, there is only one name to address configured in dd-wrt.  dnsmasq The only true dns server is at my ISP and its been that way here since RH5.0 in 1998.  I have NO M$ experience in my personal history, coming into modern computing by way of a TRS-80 Color computer running os-9, to amigados, about 10 of them as I drug WDTV-5 into the computer age when I became the Chief Engineer in 1984, but I started computing in 1978 at KRCR-12 in Redding Kalipornia. Writing a utility to prepare an automatic station break machines tapes for air that ran on an RCA-1802 and 4k os static ram that cost $400 back then.

 

But lets get back on topic.

What I want for outside names NOT in /etc/hosts: the query gets passed to the router at 192.168.71,1:53 where dnsmasq checks to see if its in its cache, and failing that, fwds the query to the dns server at my ISP for resolution.  That address is not known on my side of the router.

 

This typically adds about 1.5ms to the ping time. And nowhere in that chain fails this sequence EXCEPT this new install on a bananapi-m5.

 

And I don't have to screw around maintaining a separate dns server of any kind.  That is several times the amount of monkey business of cp-ing a common hosts file around is. But you want to insist on a dns server behind every RJ-45 jack in the house.

 

I don't understand the Why?

 

This is simple, quick and bulletproof until somebody decides to fix a perceived security hole, WHICH DOES NOT EXIST IF the router is reflashed with dd-wrt.  I have not been touched since finding out about it in yr 1999-2000.   Including a decade of running my own web page right here on this machine, stopped only because the spiders refused to honor my robots.txt.

 

So I was paying bandwidth penalties of $150 or more a month on an SS income with my upload bandwidth filled by the robots from mj12.  Whoever the hell they are. M$ spiders were in there too.  People interested in my stuff couldn't get a word in edgewise.

So please tell me where in the resolv.conf  area what to edit to make it work, cuz I'll do it and a chattr +i on it to make it permanent.

 

But the fix for debian bookworm, didn't fix this.

 

 

 

[bedna]

3 hours ago, gene1934 said:

What I want for outside names NOT in /etc/hosts: the query gets passed to the router at 192.168.71,1:53

 

Does it?

 

17 hours ago, gene1934 said:

nameserver 127.0.0.53

 

Does not look like 192.168.71.1 to me...

And:

 

17 hours ago, gene1934 said:

gene@amanda:~$ resolvectl status
Global
         Protocols: -LLMNR -mDNS -DNSOverTLS DNSSEC=no/unsupported
  resolv.conf mode: stub

Link 2 (end0)
    Current Scopes: DNS
         Protocols: +DefaultRoute -LLMNR -mDNS -DNSOverTLS DNSSEC=no/unsupported
Current DNS Server: 192.168.71.0
       DNS Servers: 192.168.71.0

That is not correct, that is the global scope, no ip can end with 0.

 

3 hours ago, gene1934 said:

And I don't have to screw around maintaining a separate dns server of any kind.

 

Ehhh... What do you call dnsmasq on your router then?!?

If you don't "maintain" that, yes, things like this can happen, especially since you have set the leases to never expire.

I have no idea if this is the case, it's for YOU as sysadmin of your network to know.

 

Nobody knows what you did after install.

All that is known after all things written in this thread:

• You have a dhcp server
• you have a dns server (dnsmasq)
• you installed Armbian_26.2.1_Bananapim5_noble_current_6.18.15_xfce_desktop.img (that comes with network manager according to documentation)
• It seems you have systemd-networkd enabled

Unless you can prove the default config was with both systemd AND network manager, YOU have made changes that breaks the installation. Nobody can answer what you did, only you.

 

The normal way if having a local dns server (even if it ONLY forwards request to outside dns server) is to configure the resolving for local domains in the dns server and not touch it at device level.

For example a consumer grade router, it will set dns to router ip in the dhcp lease and  when requests come in, resolve with asking your ISP dns servers. Hence, if looking up what dns server is used on the devices, they will NOT point to your ISP dns servers (or google or cloudflare etc), it will point to your router ip.

 

Lets say you have a pihole on your network that acts as dns server, then you configure the router (dhcp server) to point to THAT ip for dns, hence, the dhcp lease will deliver THAT ip. And then you configure the pihole to resolve local ip:s and what outside dns servers to use. The DEVICES will ONLY see the local ip where the pihole lives in this situation.

 

Resolving domains to local ip:s (/etc/hosts on your device) has nothing to do with your dns detection failing.

 

This is all pretty basic networking...

 

Edit

For reference, this is what your /etc/resolve.conf should look like (or similar) if you configure your dhcp server to point at the ip where dnsmaq lives (same ip as the router itself) then propagate those settings via the dhcp lease. Then nothing has to be done on any device except for resolving local ip:s if you refuse to do that with dnsmasq. If you configure dnsmasq to do that, resolving will be done on the router instead of the devices and absolutely nothing has to be done on any device.

With systemd:

# Generated by resolvconf
domain home
nameserver 192.168.71.1

 

With network manager:

# Generated by NetworkManager
search home
nameserver 192.168.71.1

 

If you don't want to reinstall again, make sure systemd network services are disabled and network manager is enabled and then reboot. But be mindfull you might need physical access if network completely fails at boot.

IIRC relevant services are (I could be wrong here, you should read up on documentation) systemd_networkd.service, systemd_resolved.service & networkmanager.service

 

I'm in the middle of moving, so I only have an arch desktop and laptop up and running at the moment, so I can not confirm on any armbian installation that I provide 100% accurate information.

Edited yesterday at 04:49 PM by bedna

[gene1934]

Please forget I ever mentioned dnsmasq, it exists only in the router where its working perfectly  for 7 of the 8 machines here.  w/o any man pages I haven't a clue how to run any of the systemd stuff so any config changes that need to be done after 1st boot in order to enable /etc/hosts for local lookups is equ to asking 10k monkeys to retype all of Shakespear's work w/o a single typo.  We both know that aint gonna happen b4 the universe runs down in another 500 billion years,

 

So, I am going to rewrite that img to a fresh 128G u-sd, change the /etc/hostname to amanda and rewrite the the /etc/hosts file with mine. Then you tell ME what to do via systemd  to make a user "ping -c1 yahoo.com translate" to its dns address in <1.5 milliseconds.

 

That rewrite will take at least an hour. maybe into 2 because the only wat to get a boot able image out of that respin is with dd if=/dev/zero bs=512 count=20 of=/dev/sdc to wipe the u-sd card. then dd if=Armbian_26.2.1_Bananapim5_noble_current_6.18.15_xfce_desktop.img bs=512 of=/dev/sdc 

option bs=4096 is much much faster, only about 12 minutes to write but will NOT boot nor will it mount as /dev/sdc1 to /mnt/sdc1  So ATM its being written at bs=512, hoping it will work.  That takes well over an hour. 

 

You have an image writer, "Armbian.Imager_1.2.5_amd64.AppImage" which has no manpage, a useless help screen and has never written a bootable .img file to any sd card since I dl'd it in jan 25.  I haven't even bothered to try it as it this time as apparently burns up any u-sd card I've tempted it with so good but cantankerous old dd to the rescue. That, and 45 years a DM-II is slowly taking my eyesight because of macular-degeneration, not your problem of coarse, but mine at 91 yrs old.  But saying dd is slow is an understatement, glacial is a better description.

from the shell doing the writing:

 

root@coyote:/home/gene/Downloads/armbian# dd if=Armbian_26.2.1_Bananapim5_noble_current_6.18.15_xfce_desktop.img bs=512 of=/dev/sdc status=progress
6230856192 bytes (6.2 GB, 5.8 GiB) copied, 2939 s, 2.1 MB/s
 

 

 

 

[gene1934]

Because I know little to nothing about systemd, I have not touched it.  If it is enabled, and IDK how to fix it if it was enabled in the .img.  if so that is a bug & is a reason for a respin.  Its finally done but probably won't mount as dmesg says it cannot find the ext4 file system.

 

So another ugly thought crosses my curious mind, what is the DEFAULT file system, btrfs?

 

[gene1934]

your card imager needs a CLIBC_2_38, 3 versions newer than my bookworms 2-35, so I'm stuck at dd.

[gene1934]

So I'm still doing s-thing wrong with dd.  What is correct for this later spin? using 128GB cards.

[bedna]

16 hours ago, gene1934 said:

Please forget I ever mentioned dnsmasq, it exists only in the router where its working perfectly  for 7 of the 8 machines here

 

You are telling me to forget the thing I suspect is the reason for the network not working correctly...? If it is "working perfectly", why does it not resolve addresses from the bpi? Why is your bpi setting 127.0.0.53 (witch would be incorrect even if you ran a dns server ON that device, it should then be 127.0.0.1 (or the device ip), potentially at port 53 but IIRC, port 53 IS dns requests and where it will go without defining it) as dns server and not 192.168.71.1?

 

16 hours ago, gene1934 said:

option bs=4096 is much much faster, only about 12 minutes to write but will NOT boot nor will it mount as /dev/sdc1 to /mnt/sdc1  So ATM its being written at bs=512, hoping it will work.  That takes well over an hour.

 

There you go, there is another culprit. If writing 4k in each iteration fails, but 512 works, 99,99% chance the reason is the sd card is starting to completely fail. Armbian imager would have told you verification failed when verifying the image after writing it, there is zero verification when using dd.

There should be no issue writing 4096 bytes (4k) each iteration with dd on a functioning sd-card.

 

Seems you are more interested in screaming and complaining than actually getting to the bottom of it all and set it up like a NORMAL network where NOTHING has to be done on the devices because the settings will be configured correctly with the dhcp lease.

 

Btw, here is the dev documentation for armbian imager (and the source code ofc), took me 30 seconds to find... https://github.com/armbian/imager/blob/main/DEVELOPMENT.md

Not sure what useful things you expect to find in there that is not mentioned in the help. It writes an image and does verification of that image (and you can download images), that's it.

Good luck I guess, I have nothing further to add.

Edited 9 hours ago by bedna

[gene1934]

Ambian-imager requires a glibc 3 versions newer that this bookworm amd64 has, won't run at all.

restricting dd to bs=512 gets only 2something megabytes a second, glacially slow.  The more usual bs=4096 is 50x faster, niether of which reports any write errors.

And its booted and configured. time about a month off.  I run ntpsec on this machine so I'll follow the /etc/chrony/README and assign a source address by number. but ip a shows only an active ipv6 address while ip r returns nothing.  hostname is bananapi-m5 which is unk on the local net. IOW its booted but network is only ipv6 & no route of either flavor is assigned.  Is this exactly how it was spun to be?  

I've brought the card back to my reader, and will edit the /etc/hostname to match its ID for 192,168,71,2 in the /etc/hosts file I'll install.  And reboot it.  but the card won't mount despite the fact that it boots. from this bash xfc4 terminal:

root@coyote:/etc# mount -t ext4 /dev/sdc1 /mnt/sdc1
mount: /mnt/sdc1: wrong fs type, bad option, bad superblock on /dev/loop0, missing codepage or helper program, or other error.
       dmesg(1) may have more information after failed mount system call.

dmesg however has no additional info. only this:

[1618874.568139] sd 37:0:0:0: [sdc] 249737216 512-byte logical blocks: (128 GB/119 GiB)
[1618874.568944] sdc: detected capacity change from 0 to 249737216
[1618874.569933]  sdc: sdc1
[1619402.494452] loop0: detected capacity change from 0 to 12943360
[1619402.494871] EXT4-fs (loop0): VFS: Can't find ext4 filesystem
[1619458.923461] loop0: detected capacity change from 0 to 12943360
[1619458.923910] EXT4-fs (loop0): VFS: Can't find ext4 filesystem
as I made two attempts to mount it. 

 

it loads and boots normally when installed in the bpi-m5. So other than the extra server line I just put in /etc/chrony/sources.d/ which doesn't work because no ipv4 net is even available AND the nearest ipv6 enabled anything is a minimum of 125 miles from me, its about as useful as the teats on a bore hog.  Neither is the thought of editing on that tiny screen, the 40 or so lines of /etc/hosts that make up my /etc/hosts file.  Doubly frustrating is the fact that the mouse speed is measurable fraction of C speed and totally disappears when inside the xfce4 terminal borders.  But installed in the bpi-m5 it boots  normally.

back in my reader:

 

 

root@coyote:/etc# e2fsck /dev/sdc
e2fsck 1.47.0 (5-Feb-2023)
ext2fs_open2: Bad magic number in super-block
e2fsck: Superblock invalid, trying backup blocks...
e2fsck: Bad magic number in super-block while trying to open /dev/sdc

The superblock could not be read or does not describe a valid ext2/ext3/ext4
filesystem.  If the device is valid and it really contains an ext2/ext3/ext4
filesystem (and not swap or ufs or something else), then the superblock
is corrupt, and you might try running e2fsck with an alternate superblock:
    e2fsck -b 8193 <device>
 or
    e2fsck -b 32768 <device>

Found a dos partition table in /dev/sdc
???????????? wtf???????????

 

And I get the same result for /dev/sdc1

obviously something is screwed.

So I unpackedca new 200mb writeable sandisk u-sd & wrote the new image to it. fdisk then says its a dos partition table.  Can a "dos" table be used at all on a 128gb card? I've heard not but don't know where the failure is.  ipa shows only ipv6, ip r shows nothing. The only thing that could be used for net config is armbian.config which I think is NetworkManager .  but that's by now known not to work. 

 

So whats left?????

 

IOW what the hell is being used to configure network NOW with a 6.18 kernel ????   And how do I use the new tools if any, to set it up for a network address on 192.168.71.2, an ipv4 address  THAT WORKS FOR EVERYTHING???????  Or do I go back to the soon to expire

Armbian_25.5.1_Bananapim5_noble_current_6.12.28_xfce_desktop.img which I am using 2 more times here in front of klipper driven 3d printers. And which if asked to apt update, now want around 430 new files to do the update but  failing in the make a new initrd stage.

Start by putting a gpt partition table in the xz with a respin.  And stop assuming that the whole planet has ipv6 connectivity, central WV is not the only ipv4 island on this planet so blessed.