Edits to the VLAN Trunk Wiki

@howl - I have observed that you've been making Wiki edits to the 'Wi-Fi configurationMesh Wi-Fi802.11s - VLAN Trunk Limitations' page, after you discovered Undocumented UCI Configurations to setup Unintended Behavior of certain Wireless chipsets. You've now resorted to reverting those Wiki corrections and adding additional text - accusing a software contributor of unilaterally maintaining the Wiki as justification for your Wiki comments. You conclusions seem to be quite false and your actions are damaging to the Wiki and those who rely on the accurate, clear and concise information provided there.

It's been brought to the attention of a few users with long standing Wiki editing privileges that you've been making series of edits. Personally, I ask that you rationally discuss your concerns here - instead of altering the Wiki for the global public to consume without community collaboration or consensus.

I ask that you please refrain from the changes, and I encourage/invite you to discuss the matter in this thread. It's becoming quite tedious task over the past few weeks/months for users to keep watch and tidy up Wiki pages such as this.

Thanks for your consideration, and I look forward to the discussion.

That particular wiki page have resulted in just being a clone of https://github.com/openNDS/mesh11sd/blob/master/docs/vlan-trunk-over-80211s.dokuwiki.

The documentation has several issues and also uses things that were negated by the author without checking them.

Let's say for example, when treating a mesh network as just pure layer 2 it shouldn't be added to a network. When a VIF is added to a interface even if that interface is set as unmanaged it get a layer 3 overlay. When is managed i easy to see, when it's unmanaged, OpenWrt changes the default ipv6 enable behaviour of devices so just by adding the vif it get IPv6 LL Addresses just because it's just added to the unmanaged interface. This is just one of the edits reverted by the author of that clone page.

Another one is about the nodes being aware of VLANs, is very easy to test that not all traffic is replicated over all nodes, have to test the multicast traffic. But, using this as a limitation is a bias, exactly the same is true with any other trunking setup, even more biased in the scenario about compromised node that the page states.

One more, ath10k-ct has troubles with mesh wifi setups, even with simpler ones. Also, the information about mt76 since OpenWrt 23 is info stated in threads I have opened as tests done by @LongQT-sea, but that doesn't mean that even previous versions are supported. Changes about this has also been reverted.

That's just some of the edits that have been reverted by @bluewavenet and are mistakes that indeed there in the page cloned from his own project, this is a community driven wiki should not endorse unilateral edits that are hidden advertisement of a project in the way of inventing openwrt's limitations that doesn't really exists to promote its own project scripting that breaks openwrt's own configuration system.

Let's try another approach.

You chose to make extended remarks in the thread regarding the incident. I responded to your comments there. To be clear, as a fellow Wiki editor (and in discussion with others). I opened this thread on my own accord to:

  • To understand if it's actually your ultimate desire, intention or goal to discuss and add technical information to the Wiki
  • If so, understand a written and coherent description of the "Undefined Functionality" and Undocumented UCI configuration that you hint at in Warnings placed on the Wiki (and since removed in agreement with other editors - for others, examples linked above)
  • Maybe offer help in editing it, testing what you describe, etc.

Many years ago, bad documentation regarding wireless bridging and Layer 2 wireless configurations caused me to delay use of OpenWrt for years. This particular topic is particularly important and attention-grabbing to me.



If you intend to add technical information:

  1. I did not see the reverted edit you describe. Can you explain in UCI what "alternative" you are describing, and also the documented approach, recommendation, suggestion?
  2. I don't understand what you mean, or perhaps I misunderstand the functionality you discuss. My understanding is you setup VLAN tagging over an Mesh.
  3. What troubles? Can you describe here, please? :smiley:

BTW, I wanted to add in all fairness - I did see this edit, but don't understand your context on the Wiki, nor here. I hope that you don't mean there would be different instructions for one's setup, depending on the wireless chipset used?

Interesting, actually years ago I used mesh for just one network, about 2021, for trunking GRE was the best option then but finally I didn't even need a mesh because the area I finally needed to cover. When I started few months ago again with mesh, the first and almost the only option for trunking was a solution that messes the whole openwrt's own system. Seeing that solution is presented as the main one was surprising and also was a very surprising thing that finally researching a little I get to the place that @bluewavenet describes as something that I could achieve somehow, but till some more people stated is possible and a discussion with no arguments just fanaticism about doing a impossible things, then a unreliable things, and finally a just a lucky thing was possible. The reality behind is different.

Is not an alternative, option network must not be set for mesh vif to be used as a pure layer 2. If you intend to overlay for example a VXLAN the is needed, even more I have documented this in the V"X"LAN thread in the forum time ago.

They point there is about nodes being aware of VLAN indicating is a limitation, the reality is that if you insert a VXLAN nodes are also aware of the VXLAN, and even the VLANs you want to insert inside the VXLAN if you want a more complex setup. Conclusion, unreal invented limitation.

This one is funny, I can quote mesh11sd main README, https://github.com/openNDS/mesh11sd#dependencies-full-function

If the image uses ath10k-ct, switch to non-ct ath10k packages; otherwise the daemon logs an error and exits.

ath10k-ct has, or at least, had (but taking into account the own @bluewavenet note should be yet problematic) issues to work with 802.11s.

Offtopic about ath10k, I don't know why actually ath10k-ct has not been reverted to be used pure ath10k nowadays and for devices with 128 MiB or less to use the smallbuffers kmod Archer C7 radio seems to restart by itself - #2 by howl

I will try to advance the documentation these days, at least the non LuCI setup parts, but wanted to do it at a whole to check the images positioning for the beginning, that's what I asked for image upload permission.

  1. My apologies if I wasn't clear, could you explain in UCI. Again, only if you intend to provide usable [configuration] information to the wiki. I want to test the underlying functionality on various chipsets.
  2. I still don't understand (with a UCI example). Are you suggesting I can bridge VLANs to Wireless? You mention VXLAN and VLAN in the same paragraph. I'm trying to understand an actionable configuration I can replicate, but your commentary is not specific.
  3. In a simple sentence, without trying to be funny, without discussing bluewavenet, describe what you are trying to say about various chipsets. (I think you're saying there's outdated information about receiving errors when using ath10k-ct?) I should add, I don't have access to a device with the ath10k chipset, so I can only understand your point of view from your discussion.

You explained so well, perhaps was me that not, here https://openwrt.org/docs/guide-user/network/wifi/mesh/vlan#resulting_configuration_files you have the configuration resulting files for this kind of setup. I have ath9k and ath10k over non-DSA architecture but can't test them in the near time.

The configuration as stated above. You don't knowing why VXLAN things is on the thread gives hints that all this mess about mesh11sd being presented as the main solution because is the only simplier way to achieve that mean that it seems this is not being seeing what is happening more than the people are directly involved.

Let's summarize:

1.- I asking about if VLAN over mesh is possible (there where hints that is).

2.- The forum consensus from some people that only repeated @bluewavenet ideas was that is impossible and I must use mesh11sd, or batman-adv.

3.- Anyway I tested and it worked even passing the supposed @bluewavenet limitations said.

4.- One concern was that a solution is doing the mesh11sd thing about using a VXLAN overlay in the mesh but as stated by @bluewavenet that is impossible with pure OpenWrt and he was doing some scripting and installing full tools versions like ip-full to achieve it and use mesh11sd to avoid the issues of VLAN.

5.- I started a thread about it and finally achieved with just few lines in pure uci and just installing luci-proto-vxlan to bring up a VXLANv6 with Multicast IPv6 Local Link VTEPs. The config to achieve that is in the thread.

6.- Again @bluewavenet stated that mine was a complex workaround. That's what I have stated several times a highly hilarious, a solution installing only just one simple package using few uci lines is complex but the mesh11sd solution using scripting and needing ip-full among other packages is the solution.

The result has been that every step is negated to be possible or being complex by @bluewavenet. What about if someone shares a way to dynamically set up multi wan nodes? Because is just the only step left to be using simple and own OpenWrt configuration to achieve what mesh11sd do and stated that is the only solution because can be done in pure OpenWrt and that's why that software messes up all the uci system.

I know the threads, even more with the distracting interactions (I include myself in this), could have diluted this fact. But is very dangerous, not that mesh11sd exists, options are always welcome, but minimizing and negating OpenWrt pure options just because promoting the use of that software that biased stated is to solved OpenWrt's limitations is what is dangerous.

There is just only two things that prevent me now from trying to set up multi wan mesh nodes in the most pure OpenWrt way possible (is very probable that this needs at force wan interface up down hotplug scripting), I don't have multiple WAN connections despite I could simulate them, and is not a use case for my own normal use. But if anyone tries and does it, and @bluewavenet appears negating it, your alarms should arise, is the only thing actually not documented to cover the important features in pure and simple OpenWrt uci system.

I have also pending to set up another wiki page how to do the VXLAN thing, I'm in doubt if it can be unified with the pure VLAN, but there are details in the mesh config section, interface section, that probably makes it better to set another separate one.

No, this part only says that someone lost the context window and forget that ath10k-ct is not suitable for mesh networking. The same person advise against it and at the same time says is ok to use it. The reality is that for sure ath10k-ct has been giving troubles with mesh but at the moment I think no body has stated that the issues has been solved and reported a stable mesh network with it. About the thread I linked, actually I think a revision about using -ct versions over normal ones should be consider and also 128 MiB RAM devices and less should use smallbuffers kmods because out of ram issues. I could reply a solution in that thread because I achieved years ago to bring out of memory the three Archer C7 I have solving it with the smallbuffers kmod. Using a 128 MiB device with the full buffers kmods even as a Dumb AP only gives you more time till the RAM gets out of memory.

  • It appears you're adding wireless interfaces to the network configuration?
  • Is this the configuration you want to be discussed and shared with the community?!?!

:man_facepalming:

I see this Wiki entry was created by you. I guess when folks warned you about that configuration, you ignored them and instead got mad at bluewavenet and blamed them of some conspiracy?

Understood.

To be clear, are you suggesting that Wiki page remain?

:spiral_notepad: I've seen no such consensus.

There is a consensus that you keep adding irrelevant warnings to another Wiki page - and they were removed. I've asked numerous times now that you not purposely attempt to conflate that with my topic here.

I think this might work as well. Now your configs present many other issues - and frankly require warnings. I fear repeating those warnings will cause you to react negatively again.

This is unrelated to VXLAN, or any other technology you keep mentioning.

Link, please?

Frankly, I'm getting tired of hearing about him. Every time you mention him, I get really confused if your issue is with him, or your [messed] up configuration, or if you're trying to harm other Wiki readers out of spite for him.

Is this "someone" supposed to be bluewavenet?

Again, you totally lost me.


From you're UCI example, it's clear that you somehow rigged OpenWrt to bridge wireless and Ethernet in a manner akin to how an unmanaged switch [possibly] handles traffic trunked to it.

:warning: This is dangerous - advising users to configure undefined functionality and behavior is unwise, and should be done at their own risk.

You seem to be misunderstanding warnings about this ill-advised configuration. My suggestion is that you feel free to edit your Wiki page. I'll be sure (and I'm certain others too) to begin adding technical warnings you choose to ignore (e.g., OpenWrt official documentation does not advised adding wireless PHYs to the /etc/config/wireless in the manner you show in the UCI.

That configuration can be exactly achieved also by using LuCI that is a more limited environment not letting you to do non intended things.

Is completely supported and yes, it works because is correctly translated to the underlying tools. Can you explain more about why is advisable when the generated interfaces are as they should? And the exact people that say that? Bring up real arguments about it, not just your word that is not advisable, tell why exactly is not, and why openwrt scripts achieve to do a working underlying linux network setup.

That configuration has been exposed in the forums threads, are you really knowing all the underlying issue about this or are you being aware just now?

You are mixing things, if you want to treat this fair you should not do that.

The subject in my sentence is a forum thread.

The topic is about uniperson documentation and biases in the wiki, all I have exposed shows that is happening.

Mine has to be tested, and it really works, others might, maybe... Interesting, so much interesting.

Then you should provide why is undefined functionality, I provided argumentation why adding a network in a mesh vif must not be done for pure layer 2 as port mesh.

Technical warnings? The same warnings that the same person warned about it is contradicting himself. Why you don't read all the contra argumentation about mesh with VLANs. The most important VLAN tags inside payload. The warning you say look logical somehow and coherent structured language, but actually are fallacies as has been probed.

VLANs over VXLAN with IPv6 Link-Local VTEPs over 802.11s Mesh here you have, even more simplier if the VVXLANv4 parts are omitted, that is stated as a complex workaround. This confirms that you are being aware of this now, if you are trying to judge something as it seems, you should read the whole thing before, not in the process.

Edit:

Assistant seems that get out of tokens and lost the context. By the way something like ath10k-ct yes and ath10k-ct no somewhere else is not the only contradiction about all this topic.

If you say so.

Surprise - I've been aware over 20 years, actually.

My understanding of this is spans to the beginning of 802.11b and the consumer release of the original WRT54G hardware (nearly 22 years). I'm curious if you know the underlying history. I first observed this behavior in DD-WRT. You can see a discussion on this, circa 2019.

What you discovered has nothing to do with mesh.

OK.

I see your edit, but I really need to know the contradiction you speak of.

I think your're intentionally overlooking the issue with trying to send Layer 2 frames (Ethernet) - that must be translated to a different Layer 2 medium where the behavior is unspecified (wireless), or this may be the communication issue - regarding why you don't understand the issue with bridging wireless interfaces.

Given you're using a "known-bad" config - I think prefaces/warnings on your wiki page would be a good idea.

I invite others to chime in.

Edit:

I recall other uses mentioning security concerns, these must be what they're refernecing.

But have to do in a proper mesh VLAN, actually setting a network to a vif as I has previously said, this happens:

OpenWrt changes the mesh device default from ipv6 off to on, so the default now makes it to get IPv6 Link Local Addresses even if the interface is set as unmanaged makes to overlay a unneded layer3 on top of the mes that is going to be used as a pure layer 2. So the configuration in the clone page is wrong.

Actually you are not saying anything about what is exactly not expected to be supported by doing that.

Still you have to specify that know-bad.

And still, if the issue is just because the desire of getting a VXLAN or batman-adv layer into to avoid the biased issues, the VXLAN with IPv6 LL VTEPS is still probing the fierce need of categoric negation about this issue in all aspects.

The edit just say that one person with some supposedly contradicts itself just to argument against and the ath10k-ct yes/no is just an example of that contradictions and appreciation because the arguments about the yes option seems to arise from being AI assisted.

About the other things you say quoting that part, nothing to do is not about that.

From your response:

  • I take it you have an issue with bluewavenet, or
  • You genuinely don't understand this issue is unrelated to mesh

Feel free to read the post I linked. (This is really making me think you may not understand the problem. Your responses here and in other threads seem to support this.)

I'll wait for others to respond. Unfortunately after more than three times asking you to provide some description without discussing bluewavenet, I've concluded that your issue is with him, and you truly have no clue about the issue with simply bridging Ethernet to wireless.

:white_check_mark: It works and it's good enough for you. I understand.

:warning: I would suggest to others that your posts be prefaced with warnings, and if the previous behavior of attacking bluewavenet's page persists, possibly banned from editing the Wiki. As others mentioned, folks have been banned for less.

As always, I invite others to reply, comment, etc.

As I stated, I don't have any problems with people using what they want, but, that only makes to know that I don't have issues with the project.

Also I have stated clearly and exposed categorical negation about this person about anything that goes against his interests (hence the VXLAN topic because while you are telling me with mesh VLAN I'm not seeing some potential issues, is not the case with mesh VXLAN but still negated by this person). So the issue is not with the person, is with the categorical negation about something that conflicts with his own. A normal person would try to understand how could be possible that indeed a way to with simple UCI can be achieved something he is trying to do with scripting and full-fledged tool versions, instead of negating what exposes the main idea behind what he develops is wrong. No, is not difficult to make a VXLAN mesh by just using OpenWrt.

Is so different person that a particular attitude of that person, don't confuse that please.

Great, please address the concerns with this wireless bridging - please don't discuss VXLAN or mesh. For example, let's say I add the following to bridge:

eth0.10 (or whatever on your equipment)
eth0.20  (or whatever on your equipment)
phy-ap0 
  • Can you provide an example of the STA
  • and Client
  • both wireless and network config

Also:

:thinking: What about devices where this "trick" doesnt work - what about the security of the VLANs?

I think someone asked you about this. Perhaps your response will clarify things.

The configuration you are asking for is exactly that same as a normal one from stock as from the configuration in the wiki. You have a interface for lan and another interface for guest.

What I'm using is even more "complex" (understanding complex as more VLAN tags) 10 lan, 11 media, 12 iot, 13 not, 20 guest.

I have replied about this multiple times.

First of all is not trick, your prejudice just reared its head.

Second, one of the limitations is the frames that can be managed by the radio hw/sw. This issues is even more present in other solutions as for VLAN mesh only 4 byte overhead that are even not needed to be defined because the kernel take care itself of the VLAN MTU, that's why VXLAN needs to set its own plus the VLAN overhead and hence VXLANv4 1554 MTU and VXLAN 1574 MTU.

Third, the pending tests with ath9k, ath10k has been tested by @LongQT-sea, is just to ensure, but basically chipsets that are not encourage at all to be used for 802.11s mesh like broadcom and marvell ones are what is not advisable to use for this.

Edit: About the edits, I thought I was the only needing to edit posts after sending them because English limitations, but I could see at least two times, with this last post from you being the second, edit fluctuations in the time window the forum doesn't notify the edits :laughing:

I remembered later, what could be interesting to provide is examples about using config wifi-station and wifi-vlan, I have used it to use only one VIF AP for all the VLAN (works quite well with WPA2 and is a pain in the ass to maintain with WPA3 because the need to add MACs).

That can be interesting not just a simple AP VIF for a interfaces already exposed.

Not sure if you're joking.

If what you claim is true, please provide the make and model of the device you're referencing, so I can download it from the downloads.openwrt.org website (or compare, etc.) to examine that myself.

OK, please use 10, 11, 12, 13 and 20 instead in the example.

If you mean Electrical Engineering standards being my bias, sure. You can also call it a bias we call this a "discovery", that you are attacking an author and developer of software that was designed to actually overcome the "limitation", etc.

This is why your Wiki entires need a warning.

Yea, it would prove what you're saying about it "working" for you and LongQT-sea. So, feel free to identify when this began to work - and please, I'd like a sample config, so I can begin testing sending my VLAN traffic to another floor of my house without use of an an Ethernet cable.

Not be harsh, but if you'd stop wasting time bashing bluewavenet, you'd realize I don't give a damn about your discussion or contentions with him, but I'm quite offend it's taking such a long time to handle something as simple as this - which has been known for over 20 years.

If there's a new advent in 2 decades, please provide a simple example - I know my family would like me to get rid of this ugly network cable.

The fastest way to generate uci configs to test this is to use the 5 GHz 802.11s mesh backhaul toggle on wrtnova.com

It's basically a firmware-selector, but with a GUI for the uci-defaults script, the core logic is at:
https://raw.githubusercontent.com/LongQT-sea/WrtNova/main/wrtnova.sh

wrtnova.sh includes an important detail mesh*) sleep 4; /etc/init.d/network reload in the hotplug script that works around a race condition where the mesh interface comes up late after a /etc/init.d/network restart

OK, cool. I'm not the best at reading thru other's scripts, but thanks.

Yea, there's a history that explains that. We're touching on it here.


There are only 2 or 3 other users that I would inquire about this. One of them I cannot consult (bluewavenet). It's been made clear that his involvement would automatically create some kind of bias.

Frankly, the other is probably reading already, the third probably wouldn't entertain this.

  1. I'm understanding that your script, an LLM produced:
config bridge-vlan 'vlan_1'
        option device 'br-vlan'
        option vlan '1'
        option local '0'
        list ports 'mesh0:t'

It added:

The actual ports list may contain additional Ethernet ports and other interfaces.

The important distinction: Without batman-adv, the mesh interface is directly attached to the VLAN-aware bridge and carries the configured VLANs.

  • And the remote end will be the same identical configuration?
  • :warning: What is the remote end does not have a VLAN-aware bridge, does not translate these frames as intended, etc.?!?! Is this wnat @howl has been discussing when I inquired about various chipsets?
    • I understand "it works" - that doesn't meant it's safe
  • So you propose the following works?
config bridge-vlan
        option device 'br-vlan'
        option vlan '1'
        option local '0' #?
        list ports 'phy1-ap-0:t'
        list ports 'lan1:t'
  • What is local 0 (I can't find that in the official Wiki)?

Also, from experience, there's history about bridging Ethernet to wireless. Since this "working" configuration/rig/trick is used, you mentioned you had to perform a network reload.

This is after years of the developers working out various issues. One result of that was that it's a clear indication that some manual altering of the network configurations has been done, because official OpenWrt scripts, etc. never add wireless to bridges inside the network UCI configurations. Unfortunately, howl was alarmed that evidenced some bias in me. It actually is the result of 15+ years of interest in Ethernet to Wireless bridging.

I'm interested in your reply.