FlexQoS FlexQoS issues with 388.4 HND5.04 models

[Kingp1n]

Can someone test this setup? I think I might found a bug but I can't troubleshoot it.

I have the default rules setup plus 3 additional iptables rules (2 streams & 1 gaming rule) added with FlexQoS on my new GT-AX11000 PRO (testing out new router), however, once FlexQoS is installed, the router keeps restarting every 30 secons/1 minute even after rebooting. I uninstalled all other scripts Diversion, Skynet, YazDHCP and it keeps rebooting.

Is there a way I can troubleshoot this? It does the same thing with fq_codel and stock "Asus".

Once I un-install Flex, the router stops rebooting constantly. I have no other script running, just a swap file and entware. I setup from scratch after multiple resets and it's the same setup from my old GT-AX11000 router.

Should reinstall FlexQoS without those 3 iptables rules and test again for any restarts?

My 3 rules are attached:

 

[dave14305]

Once I un-install Flex, the router stops rebooting constantly.

There is your solution. FlexQoS is dead. We’ve been swimming upstream ever since @FreshJR came up with his first ideas. We fought the good fight but in the end, closed-source crap killed it.

 

[Kingp1n]

There is your solution. FlexQoS is dead. We’ve been swimming upstream ever since @FreshJR came up with his first ideas. We fought the good fight but in the end, closed-source crap killed it.

Thanks for the quick response. It might not be fully dead, I noticed by just installing FlexQoS (w/fq_codel) with the gaming rule only it does NOT cause out any reboots. So it must be something in the Streaming rule (!04***) that must be throwing something off. I'll keep testing with this rule only for now.

 

[dave14305]

So it must be something in the Streaming rule (!04***) that must be throwing something off.

Yes, I can imagine what’s probably happening. The temp fix for the crashes was to check for a valid upload or download mark before changing the mark.

When you use ! in the Mark field, it simply says “not that mark”, which allows the “bad” marked packets to enter the chain and cause a crash.

I made a fix that might help. Please update and let me know. I did this on my iPad while on the sofa, so it’s on-the-fly…

Check mark earlier in chain · dave14305/FlexQoS@39c467d

When using negation in an iptables Mark field, the oddly marked packets causing crashes were allowed to be re-marked and cause a crash. Put the mark check on the jump to the up or down chains.

github.com

After this, then it’s dead…

 

[Kingp1n]

Yes, I can imagine what’s probably happening. The temp fix for the crashes was to check for a valid upload or download mark before changing the mark.

When you use ! in the Mark field, it simply says “not that mark”, which allows the “bad” marked packets to enter the chain and cause a crash.

I made a fix that might help. Please update and let me know. I did this on my iPad while on the sofa, so it’s on-the-fly…

Check mark earlier in chain · dave14305/FlexQoS@39c467d

When using negation in an iptables Mark field, the oddly marked packets causing crashes were allowed to be re-marked and cause a crash. Put the mark check on the jump to the up or down chains.

github.com

After this, then it’s dead…

Just updated to FlexQoS v1.3.5, added the Streaming rules and I'm not seeing any crashes....yet (fingers crossed). I appreciate it @dave14305 for the quick fix!!!!

 

[dave14305]

Just updated to FlexQoS v1.3.5, added the Streaming rules and I'm not seeing any crashes....yet (fingers crossed). I appreciate it @dave14305 for the quick fix!!!!

You can delete your earlier screenshot. There is some personal information included.

 

[Kingp1n]

Anyone testing the latest 388.5 alpha and currently using latest FlexQoS? I know it's early but im just wondering if these QoS changes in the new 388.5 alpha builds will need to get Flex updated or it should not affect anything at all? Thanks in advance!

Newer builds (latest 388.5 alpha)

FIXED: Various issues with the QOS Classification page

- CHANGED: Display tracked connections on the QoS/Classification
page even if QoS isn't set to Adaptive QoS.
- CHANGED: QOS/Classification page can now resolve local IPv6
addresses.

 

[nzwayne]

Anyone testing the latest 388.5 alpha and currently using latest FlexQoS? I know it's early but im just wondering if these QoS changes in the new 388.5 alpha builds will need to get Flex updated or it should not affect anything at all? Thanks in advance!

Newer builds (latest 388.5 alpha)

FIXED: Various issues with the QOS Classification page

- CHANGED: Display tracked connections on the QoS/Classification
page even if QoS isn't set to Adaptive QoS.
- CHANGED: QOS/Classification page can now resolve local IPv6
addresses.

I've been running on the first 388.5 Alpha for a week with no issues on an AXE16000. Just converted to the 2nd 388.5 Alpha. Running Add-ons as per my Signature detail.

 

[raion969]

I also think that still many people use Flexqos sadly there are no more updates.

Is the Stable or the develop channel version the latest of flexqos?

 

[SomeWhereOverTheRainBow]

I also think that still many people use Flexqos sadly there are no more updates.

Is the Stable or the develop channel version the latest of flexqos?

Compare this https://github.com/dave14305/FlexQoS/blob/master/flexqos.sh#L14-L15

with

this https://github.com/dave14305/FlexQoS/blob/develop/flexqos.sh#L14-L15

and tell us.

I was a bit confused myself... From what I gather from @dave14305 prior posts though, there is not much else that can be done with this script due to the current state of Asuswrt's closed source QoS code. It appears he modified the master directly probably because it was just a minor change from the develop version.

Its a shame the script has been forced to this direction because @dave14305 had such a great run with it picking up where @FreshJR left off at.

 

[raion969]

so you would not use flexqos anymore on routers like rt/ac-5300 ?

I think it still works, havent looked into the categorys placement but bufferbloat looks good.

 

[dave14305]

so you would not use flexqos anymore on routers like rt/ac-5300 ?

It should work fine on AC routers. Newer AX models pose problems.

 

[Kingp1n]

I saw a minor update to FlexQoS this morning...anyone know what was updated? Thanks

 

[dave14305]

I saw a minor update to FlexQoS this morning...anyone know what was updated? Thanks

Just cleaning up the develop branch and pushing some pending IPv6 GUI changes to master.

Still frustrated by the flow cache behavior. iptables rules don’t do a damn thing anymore on HND5.04.

So now I only run FlexQoS for the benefit of the AppDB rules.

 

[Kingp1n]

Just cleaning up the develop branch and pushing some pending IPv6 GUI changes to master.

Still frustrated by the flow cache behavior. iptables rules don’t do a damn thing anymore on HND5.04.

So now I only run FlexQoS for the benefit of the AppDB rules.

Thanks for the info...

Dumb question but is there any way to bypass the iptables rules so they still work for HND5.04 models or this all falls under the dreaded closed-source stuff? I guess none of the "well known" iptables rules no longer work with newer AX models correct?

I wonder if Asus try to implement some changes but ended making Adaptive QOS worst instead of fixing it.

I'm still hoping they will be a fix for this but doubt it, if it's closed-source!

 

[Kingp1n]

@dave14305,

I'm not sure if this part of the update, but I noticed the list while looking at the FlexQoS tab under the "tracked connections" only list my local IP number now (see screenshot) vs the Source IP (Client Name). Is there a way to switch back to reflect the Client Name under the Flex tab?

I tried hitting the Reset button under the filter connection and name appeared for a few seconds but quickly switched back to IP number:

Back it goes back to showing IP number:

 

[dave14305]

I'm not sure if this part of the update, but I noticed the list while looking at the FlexQoS tab under the "tracked connections" only list my local IP number now (see screenshot) vs the Source IP (Client Name). Is there a way to switch back to reflect the Client Name under the Flex tab?

I don’t think anything changed in FlexQoS in this regard. I use the same functions that Merlin uses. It gets data from networkmap, so make sure it’s updating properly on the client list page.

is there any way to bypass the iptables rules so they still work for HND5.04 models or this all falls under the dreaded closed-source stuff? I guess none of the "well known" iptables rules no longer work with newer AX models correct?

Delete all the rules but one that you know won’t actually match any traffic (I keep the Usenet rule). If you delete all the rules, it will revert to the 4 original default rules.

I even experimented with creating tc filter rules to mimic iptables rules, but they still didn’t impact the traffic as expected. I don’t yet understand if Adaptive QoS even works correctly with flow cache on stock firmware.

I try not to think about it too much. I’ll be looking for another abandoned Addon to adopt and resuscitate this year.

 

[Kingp1n]

I try not to think about it too much. I’ll be looking for another abandoned Addon to adopt and resuscitate this year.

Thanks again.

This sounds like great news!!! Is the plan maybe to continue working on a hybrid Flex/Cake QoS thing or just CAKE only?

Either way, looking forward to...

 

[dave14305]

Is the plan maybe to continue working on a hybrid Flex/Cake QoS thing or just CAKE only?

No, hopefully something completely unrelated to QoS.

 

[langesgnu]

Not sure if this is related to the above post, but with the latest minor update it seems that the columns for remote and local IP are mixed up now.