ArmbianHelper changed the topic of #armbian to: armbian - Linux for ARM development boards | www.armbian.com | Github: github.com/armbian | Commits: #armbian-commits | Developer talk: #armbian-devel | Forum/Twitter feed: #armbian-rss | Logs: -> irc.armbian.com
xoan3 has joined #armbian
xoan has quit [Ping timeout: 276 seconds]
xoan3 is now known as xoan
califax has quit [Remote host closed the connection]
califax has joined #armbian
califax has quit [Remote host closed the connection]
califax has joined #armbian
norwich_ has joined #armbian
norwich has quit [Ping timeout: 268 seconds]
norwich_ is now known as norwich
califax has quit [Remote host closed the connection]
califax has joined #armbian
califax has quit [Remote host closed the connection]
califax has joined #armbian
crabbedhaloablut has quit [Remote host closed the connection]
crabbedhaloablut has joined #armbian
Malditron has joined #armbian
sams has quit [Remote host closed the connection]
cyph84 has joined #armbian
Malditron has quit [Quit: Konversation terminated!]
<cheakoirccloud>
I'm again having trouble with a marvell espressobin? Asking about it over in #netfilter because the leading symptom is NAT dropping packets.
FergusL has quit [Quit: Ping timeout (120 seconds)]
SallyAhaj has quit [Ping timeout: 268 seconds]
aasami_ has joined #armbian
<lanefu>
Is it overheating?
kolla has quit [Quit: %fog relay%]
kolla has joined #armbian
archetech has quit [Quit: Konversation terminated!]
archetech has joined #armbian
rtypo_bot has joined #armbian
rtypo_bot has left #armbian [#armbian]
<Armbian-Discord>
<ManoftheSea> Tell me about it in armbian-marvell
<Armbian-Discord>
<ManoftheSea> Oh wait, we don't have one of those
<Armbian-Discord>
<ManoftheSea> So ... Armbian monitor output?
TheCoffeMaker has quit [Ping timeout: 260 seconds]
TheCoffeMaker has joined #armbian
mpmc is now known as Ghoul
Toasty has joined #armbian
<stipa>
i thought that that board doesn't work, but someone told that there's some devs working on it again...
<Armbian-Discord>
<ManoftheSea> Ebin? Worked fine, mainline, no patches
<stipa>
kewl
alkisg has joined #armbian
<alkisg>
Hi all; first time here; I just installed Armbian_22.05.4_Rockpro64_jammy_current_5.15.48_xfce_desktop.img.xz but I have no HDMI sound; any help?
<alkisg>
Additionally, I'm trying to download any of the previous focal images to test in case sound works there, but they're all missing: https://www.armbian.com/rockpro64/
<Armbian-Discord>
<TheBug> Are you sure you are actually missing sound? First thing I would do is open audio util and make sure you have selected the correct output device, many images select analog audio by default and it can be clunky to select the output you want
<Armbian-Discord>
<TheBug> So if you are just expecting you boot desktop and sound magically comes out of hdmi this would be the wrong expectation, default is analog and you need to review and make sure you manually select hdmi as your output device
<stipa>
alkisg: which board ?
<alkisg>
Rockpro64; so far I tried to install a previous kernel due to something that I read in the forums,
<alkisg>
And now I'm going to test on another monitor because this one is old , although HDMI sound did work with vanilla Debian
<alkisg>
Back with new info in 10'
<stipa>
alkisg: also pay a visit to the #armbian-desktop
<alkisg>
👍
<Armbian-Discord>
<IgorPec> which vanilla debian for rockpro64?
<[TheBug]>
again did you check the audio output
<[TheBug]>
you were actually using
<[TheBug]>
'vanilla debian' may default hdmi while we often default analog
<alkisg>
[TheBug]: sorry I was typing from phone while rebooting; yes I did check the audio output; hdmi doesn't show up in pavucontrol but it shows up in pacmd; but even with aplay I wasn't hearing any sound,
<alkisg>
now that I moved to the target, newer monitor (=TV) I can hear some distorted sound; checking more...
<[TheBug]>
mm if your running desktop you should be able to open desktop tool for audio
<[TheBug]>
and just so you know
<[TheBug]>
whay x windows works you can assign output device per program also
<alkisg>
Yeh xfce uses pavucontrol, right?
<alkisg>
Btw later on we're expecting cinnamon etc, like I see in the 20.04 images, or 22.04 will be xfce-only?
<[TheBug]>
You can build cinnamon your self I believe but I am not sure if it is something we will be regularly producing or not
<[TheBug]>
using armbian builder
<[TheBug]>
I would say you could probably get away with using cinnamon image and just upgrading.. if you don't want to build your own
<alkisg>
And if I e.g. install the base image and then ubuntu-mate-desktop on top of it, the video decoding hardware acceleration will still be there due to the armbian repositories and kernel, correct?
<[TheBug]>
I am only familiar with rk3399 having acceleration in Legacy kernel...so if your using mainline images unless they recently changes something (and it is possible) I don't see how you are getting acceleration with it
<[TheBug]>
ahh maybe I am wrong there seems to be talk about it working in mainline somewhat but I haven't had a chance to test it
<alkisg>
Hmm, honestly my only way of judging was to play an 1080p youtube video and checking the nerd stats for dropped frames; in debian it was 50% dropped frames, while with that latest image I only got a few ones initially while it was still loading
<[TheBug]>
well you also will get better CPU speeds out of our overlays but I actually am not sure. I do see talk on Pine's forum but I didn't see any specific from Armbian other than the Legacy kernel with media framework
<[TheBug]>
is last I recall on acceleration and requires using legacy kernel
<cheakoirccloud>
ManoftheSea lanefu: Thanks for the replys. sshd has no problems, so I don't think thermals would explain this. I'm using thermal pads and a heat sink that comes with the case I bought, I do have paste it's just that the gap is more than a mm.
<[TheBug]>
cheakoirccloud: what board?
<cheakoirccloud>
I do have UART console, but when I say monitor I'm talking about the feature of some managed switches.
<[TheBug]>
EBin?
<cheakoirccloud>
Yes, I think 1Gb/V7 .
<[TheBug]>
okay so especially on v7 you need cooling ESPECIALLY if you are using 1200Mhz uboot
<[TheBug]>
you need to cool SoC, memory and the switch chip if you intend to consistenly have it passing traffic
<[TheBug]>
the OEM case can give you some reference to this, as they create a piece of metal
<[TheBug]>
which covers the SoC and memory at least
<[TheBug]>
that is what I would say is minimum amount of cooling for it
<[TheBug]>
when I was running them full time I had a small heatsink on each chip and had a 30x30mm 5v fan mounted pushing air across them passively
<cheakoirccloud>
The case I got came with a big chunk of metal touching 3 chips. There was a gap so I used thermal pads instead of paste.
<[TheBug]>
I was originally using them with mpcie to sata card and using them as NAS
<[TheBug]>
you should use paste if you can get some
<[TheBug]>
I am using artic 5
<[TheBug]>
those pads aren't always that good
<[TheBug]>
they are good if you have nothing better but thermal paste if available will be much better
<cheakoirccloud>
What are the symptoms to look for?
<[TheBug]>
You were basically saying you were having switch drop packets right?
<[TheBug]>
I would make sure the switch chip isn't overheating and it won't be touched by the original heatsink
<cheakoirccloud>
*Very specific packets*
<[TheBug]>
you could try for example having it on desk outside case with a fan pointing at it and see if same issue
<[TheBug]>
then you could at least rule out heat
<[TheBug]>
passive air flow should generally be enough with basic heat sink setup
<cheakoirccloud>
sshd is fine, no loss on both VLANS. It's only NAT traffic that's effected.
<[TheBug]>
hmm
<[TheBug]>
are you using it in bridges state or?
<[TheBug]>
by default armbian bridges interfaces
<[TheBug]>
into a br0
<[TheBug]>
are you using it like that?
<cheakoirccloud>
Yeah, I'm using a complicated setup. VLANs ontop of bonding.
<[TheBug]>
hmm was it working in previous versions without issue or is this only now you are setting it up?
<[TheBug]>
also does it produce any loggable information onf the lost packets?
<[TheBug]>
maybe tcpdump of interface would be worth while to see where packets end up
<cheakoirccloud>
I had problems previously with kernels refusing to create the VLAN interfaces, but this feels like a new issue with just this kernel.
<[TheBug]>
so if you step back a version this issue doesn't exist?
<[TheBug]>
could very much be a bug in the switch driver if so
<cheakoirccloud>
tcpdump shows the packets leaving onto the correct VLAN with the correct IP address.
<[TheBug]>
at that point it would be worth going to look at kernel and see if someone submitted patches related that could have changed something if this issue wasn't present before
<Armbian-Discord>
<ManoftheSea> Post nat packet leaves the soc, but doesn't arrive at the other end of the wire?
<Armbian-Discord>
<ManoftheSea> And only sometimes?
<Armbian-Discord>
<ManoftheSea> Iptables?
<cheakoirccloud>
I tested the tcpdump leaving on 5.17.5-mvebu64 and 5.15.35-mvebu64.
<cheakoirccloud>
ManoftheSea: I'm using shorwall, nothing in the default log targets.
<cheakoirccloud>
I get about %20 on a managed switch monitor port, but tcpdump on the Ebin says %100 was sent.
<cheakoirccloud>
Only routed traffic, I can see in the same tcpdump dns requests from the local dns cache being sent and answered, but a bunch of https SYN packets go unanswered.
<cheakoirccloud>
Also ping packets sent, but no pongs.
<[TheBug]>
do you have any other routers/managed switches on t he network?
<[TheBug]>
I would test also with arping and see if you get arp responses back
<cheakoirccloud>
Two modems that are doing nat.
<[TheBug]>
could have another device intercepting your traffic
<cheakoirccloud>
Can't arping(???) through a nat :( but I did test it on the Ebin side and that just works.
<[TheBug]>
you should be able to arping the destination ip and see what device responds to the requests was my thought
<[TheBug]>
one known fun with these boards is related to the mac address being duplicated
<[TheBug]>
do you have more than one running?
<[TheBug]>
I am honestly guessing outloud just things that might produce that issue but I am not actually sure how you have your network so hard to guess with accuracy
<cheakoirccloud>
The hosts having problems don't have a local route for the addresses they are having trouble connecting, arping should just exit asking if ur crazy.
<cheakoirccloud>
It wouldn't know what interface to use.
<lanefu>
Yeah shorewall doesn't support nftables I don't believe
<[TheBug]>
someone get that man a beer and tell him to take a break!
<lanefu>
Resuming break
<[TheBug]>
;)
<cheakoirccloud>
I do get timeouts as expected from arping.
<cheakoirccloud>
That just tests if the switches vlan feature workes.
<Armbian-Discord>
<ManoftheSea> That can be fixed with a recent u boot. Like, 5 years old or newer
<Armbian-Discord>
<ManoftheSea> You can set ethaddr and eth{0,1}addr
cyph84 has quit [Ping timeout: 252 seconds]
<Armbian-Discord>
<ManoftheSea> I'd also like to know if there's a difference between iptables and nftables natting
<lanefu>
cheakoirccloud: FYI there's seldom value in bonding on ebin.... only 1gigabit of bandwidth going up to SoC
<lanefu>
better off just balancing tagged vlans accross interfaces
<Armbian-Discord>
<ManoftheSea> Bonding for redundancy?
<cheakoirccloud>
Bonding to take a high road, because it's there.
<lanefu>
yeah I'm def pro-bonding in a lot of use cases.. specifically gonna be a bad time on ebin tho
<archetech>
bonding is for when ya got many users hitting a server constantly
<Armbian-Discord>
<Tenkawa> Its also best to use bonding for link aggregation
<archetech>
so lanefu is right
<Armbian-Discord>
<Tenkawa> We commonly used it in early telecommunications setups tp increase throughput
<Armbian-Discord>
<Tenkawa> back in the dark ages
<Armbian-Discord>
<Tenkawa> lol
<Armbian-Discord>
<Tenkawa> Another use we had was for load balancing as well to reduce congestion however that can be very intensive to maintain and a lot more load/resource heavy
<Armbian-Discord>
<Tenkawa> different q for any of you... Any of you currently running any of the Realtek USB3 wifi interfaces on any SoC's atm?
<lanefu>
yeah in telecom bonding T1's is quite different because you've got a common clock so it literraly gives you all the perf on a single data stream
<Armbian-Discord>
<Tenkawa> I'm trying to isolate a pstate panic
<Armbian-Discord>
<Tenkawa> indeed
<lanefu>
I bond my interfaces on my routers so I can just blindingly run all my VLANs through it and not worry about balancing. which is kinda nice
<Armbian-Discord>
<Tenkawa> I had 4 lines.. 2 voice 2 data
<lanefu>
aka the router on a stick pattern
<Armbian-Discord>
<Tenkawa> it was a nightmare
<lanefu>
lol
<Armbian-Discord>
<Tenkawa> that was just the small ISP.. that wasn't even my day job
<lanefu>
yeah bonding T1s would beef up throughput but i do remember it having some subtle latency impacts
<lanefu>
then we got to metro ethernet and life was better
<Armbian-Discord>
<Tenkawa> yeah.. this was the mid 90's though.... everyone was still on 2400-9600 baud modems mostly where this was so noone noticed heh
<lanefu>
haha nice..what did you use for a terminal server for all the dialup stuff
<clever>
this is using a special box and servers, to split your packets over 2/4/8 traditional models (like dsl)
<clever>
unlike round-robin'ing the packets over each link, this splits the packets
<lanefu>
yeah you're basically load balancing over 2 VPN tunnels to a single endpoint with that
<Armbian-Discord>
<Tenkawa> For our setup we used Livingston Portmasters connected to a router I can't even remember anymore with individual modems....
<clever>
lanefu: exactly
<Armbian-Discord>
<Tenkawa> if you can picture a room with 20+ modems
<lanefu>
USR Couriers everywhere
<lanefu>
im sure
<Armbian-Discord>
<Tenkawa> yep
<clever>
lanefu: my current ISP uses a GPON, gigabit passive optical network
<Armbian-Discord>
<Tenkawa> I was so happy the day we went to a Cisco concentrator and uplink T1's
<Armbian-Discord>
<Tenkawa> I learned IOS asap
<clever>
where a single fiber and upstream modem, goes thru a passive optical splitter, and connects to 8 customer modems
<clever>
download is tagged by destination, and upstream decides the balancing
<clever>
upload is timeslots, so different customers dont collide
<clever>
and the size of the timeslot determines how much upload you can do
<clever>
i just recently realized, my chromecast is busted
<clever>
i had to move the wifi access point, and i think its just out of range now
<clever>
so the chromecast just says connecting, and then nada
<lanefu>
Wifi is pain lol
<clever>
thats why everything i can hard-wire, is hard-wired
<clever>
the laptop is an edge case
<lanefu>
I just spent half a week in NYC compensating for bad commercial wifi for a client demo
<clever>
in the past, i setup link failover on my laptop
<clever>
with ethernet as the primary, wifi as the failover, and both on the same ip
<clever>
so tcp connections seamlessly handover when i rip the cord out
<clever>
ive had very bad luck with getting good wifi devices
<clever>
originally, i had a dlink router with the dhcp disabled, acting as an AP, but it died one day
<clever>
the next router i tried doing that with, refused to stop sending ipv6 RA packets, so it null-routed all v6 traffic
<clever>
the next one, a wifi extender, had weird interfactions with my android tablet
<clever>
if the tablet was at the edge of range, the 2 devices would conspire to totally JAM the wifi
<clever>
causing all other wifi devices to just drop off
<clever>
until i turn the wifi off then on, on the tablet
<stipa>
the weakest links in my wifi network are shitty
<stipa>
APs
<stipa>
actually just one
<clever>
also, half the AP's, have 100mbit ethernet ports
<stipa>
everything else runs linux
<clever>
so even if the wifi can run at 500mbit, the ethernet port caps it to 100
<stipa>
100 Mbs is enough fro internet
<clever>
but its also accessing my nas over the lan
<clever>
which reminds me, the isp supplied router, bans access between wifi and lan
<clever>
so you cant even ping the nas
<stipa>
yeah, it sux
<clever>
the website for the router, claims it doesnt even have that feature
<stipa>
even if you get good stable network connection to the ISP might suck
<clever>
in the past, i had a nasty bug on my linux router
<clever>
i think its a motherboard flaw in the pcix controller
<Armbian-Discord>
<ManoftheSea> Did anyone explain why it's no good to bond on the ebin? The Topaz switch chip has 5 1gbe ports, only one of which goes to the SoC, and 3 go to the ports on the v5/v7 (and I assume the v4). So bonding can't improve the performance to the SOC.
<clever>
randomly, the tx rings in the WAN NIC would just timeout, and the driver would have to reset the NIC
<Armbian-Discord>
<ManoftheSea> But, if you're doing a failover with just redundant ports, sure.
<clever>
stipa: as a result, you just randomly loose internet for seconds to minutes
<stipa>
yeah, that's what i get with 4G connection
<stipa>
i can feel cells behaving differently, they probably run different hardware brands
<stipa>
but i can't see them since the modem is locked
<clever>
ive got an android app that can display the current cellid
<clever>
and its uploading it somewhere, which should be making cell maps
<stipa>
so, that hardware on the cells is probably crashing non stop as well
<Armbian-Discord>
<brammie15> hey im using a orange pi 3 lts and keep getting this error i dont even have to touch it to do that as far as i know it only happens on the CLI release and desktop been working fine