FlexQoS FlexQoS 1.0 - Flexible QoS Enhancement Script for Adaptive QoS

[dave14305]

is FlexQoS broken for the 384.19A1? I saw all the post here and decided to uninstsall FlexQoS before upgrading to 384.19A1 and it seems fine. Not sure to install FlexQoS back on

It looks like ASUS QoS is possibly broken at the moment, but I'm basing it on others' feedback above, since I don't have an AX router. Are you saying vanilla Adaptive QoS seems fine on your router?

 

[mbze430]

It looks like ASUS QoS is possibly broken at the moment, but I'm basing it on others' feedback above, since I don't have an AX router. Are you saying vanilla Adaptive QoS seems fine on your router?

Yes it seems to run fine with the default Adaptive QoS

 

[dave14305]

Yes it seems to run fine with the default Adaptive QoS

Then I'd be interested in the output of these commands just as you are now without FlexQoS:

nvram get bwdpi_app_rulelist
cat /tmp/bwdpi/qosd.conf
tc class show dev br0 | grep "parent 1:1 "
tc class show dev eth0 | grep "parent 1:1 "
tc filter show dev br0 | grep -A1 "flowid 1:1"
tc filter show dev eth0 | grep -A1 "flowid 1:1"

It'll be a lot of lines in total, so a pastebin would be better than trying to paste it in a post.

Then if you're moved in any way to test FlexQoS again, run the script with the debug parameter to see what's going wrong.

 

[RMerlin]

This is because Merlin took ASUS at their word and assumes that category 4 is the default category for Unmarked/Untracked traffic. My experience says Untracked traffic goes into Work-From-Home (where category 00 Instant Messengers lives).

They said it went to the fourth prio. Work-From-Home is not a prio, it`s a category, that can be assigned to any prio.

 

[geobernd]

Upgraded form FreshQOS to FlexQOS (when switching my AC68U to 384.18) - it works great!
Thank you for all the effort to keep it alive and improve!

 

[dave14305]

They said it went to the fourth prio. Work-From-Home is not a prio, it`s a category, that can be assigned to any prio.

Sorry, I mix my terminology too much (category, class, pref, prio).

I don't know how they might be doing it, but what I observe in the tc filter list is that all traffic marked 000000 will get picked up by this filter rule (which is for Instant Messengers):

filter parent 1: protocol all pref 3 u32 fh 805::800 order 2048 key ht 805 bkt 0 flowid 1:11
  mark 0x80000000 0xc03f0000 (success 16)

And so it will show up under class 1:11 when you do the tcclass_dump.

class htb 1:11 parent 1:1 leaf 11: prio 1 rate 64512Kbit overhead 18 ceil 322560Kbit burst 79978b cburst 403200b
Sent 15620 bytes 75 pkt (dropped 0, overlimits 0 requeues 0)
rate 128bit 0pps backlog 0b 0p requeues 0

I know my ISP's speedtest (speedtest.xfinity.com) uses port 6020 and is unmarked, so I've disabled my script and using your Classification page, it shows all the data in Work-From-Home but the connections tracked in blue. But you can see no data is recorded in Web Surfing.

# tc -s class show dev br0 | grep -A2 "htb 1:11"
class htb 1:11 parent 1:1 leaf 11: prio 1 rate 64512Kbit overhead 18 ceil 322560Kbit burst 79994b cburst 403200b
Sent 547175179 bytes 364272 pkt (dropped 0, overlimits 0 requeues 0)
rate 73224bit 15pps backlog 0b 0p requeues 0

I would expect to see a tc filter at prio/pref 2 with mark 0x80000000 0xc000ffff using flowid 1:14 if they were always going to put unmarked traffic in the fourth prio (1:14). But it's not there.

 

[RMerlin]

I would expect to see a tc filter at prio/pref 2 with mark 0x80000000 0xc000ffff using flowid 1:14 if they were always going to put unmarked traffic in the fourth prio (1:14). But it's not there.

I don't know. I only looked at how tc worked a few years ago, and never looked again since then. All I know right now is that it seems completely broken on the RT-AX88U, where traffic gets classified, but nothing get queued at all - all stats remain at zero, except for 1-2 KB here and there.

I'm getting tired of dealing with this and playing the guessing game with all of this proprietary code, and miss-matched GPL releases.

 

[QuikSilver]

I don't know. I only looked at how tc worked a few years ago, and never looked again since then. All I know right now is that it seems completely broken on the RT-AX88U, where traffic gets classified, but nothing get queued at all - all stats remain at zero, except for 1-2 KB here and there.

I'm getting tired of dealing with this and playing the guessing game with all of this proprietary code, and miss-matched GPL releases.

@RMerlin is the AX88U an older unit that Asus is trying to "frankenstein" to keep it going? Curious why that unit seems to be the red headed step child from this 86U user.

 

[RMerlin]

@RMerlin is the AX88U an older unit that Asus is trying to "frankenstein" to keep it going? Curious why that unit seems to be the red headed step child from this 86U user.

No, the RT-AX88U in general is still their flagship (I don't consider ROG in this - these are a separate product line in my book). The heart of the issue here is the fact that each AX model is released with a completely different firmware, and I'm the one trying to figure out ways to make it all fit together. And it's becoming REALLY frustrating.

 

[mbze430]

Then I'd be interested in the output of these commands just as you are now without FlexQoS:

nvram get bwdpi_app_rulelist
cat /tmp/bwdpi/qosd.conf
tc class show dev br0 | grep "parent 1:1 "
tc class show dev eth0 | grep "parent 1:1 "
tc filter show dev br0 | grep -A1 "flowid 1:1"
tc filter show dev eth0 | grep -A1 "flowid 1:1"

It'll be a lot of lines in total, so a pastebin would be better than trying to paste it in a post.

Then if you're moved in any way to test FlexQoS again, run the script with the debug parameter to see what's going wrong.

I'll install FlexQoS maybe sometime this weekend

Here is the output at pastebin

https://pastebin.com/EQHvkbbW

 

[dave14305]

Version 0.8.6

• Fix rules conflict when iptables rule class is overridden by AppDB redirection
• Catch other instances of invalid AppDB entries during initialization
• Experimental backup menu functions by @maghuro

Edit: hot fix posted minutes ago...

 

[dave14305]

I'll install FlexQoS maybe sometime this weekend

Here is the output at pastebin

https://pastebin.com/EQHvkbbW

So it looks very normal, except it doesn’t have the new Asus categories. What router/firmware version is this again? Never mind, I see the title of your pastebin. If you go into QoS and customize and apply, does your nvram command return a different result?

 

[mbze430]

So it looks very normal, except it doesn’t have the new Asus categories. What router/firmware version is this again? Never mind, I see the title of your pastebin. If you go into QoS and customize and apply, does your nvram command return a different result?

I am already using "Customize"

 

[dave14305]

I am already using "Customize"

Do you see the new categories in the customize window? What if you re-save the priorities?

 

[dave14305]

@dave14305, I have a strange problem. My routed all traffic via iptables to/from my Aruba (enterprise VPN router) to the Work-from-Home category (screenshot 1) then in AppDB redirected email services and VOIP services (05**** & 06****) to Web Surfing (screenshot 2, to try to keep my Aruba traffic as isolated and high priority as possible). In the tracked connections it indicates the Aruba (192.168.1.101)(screenshot 3) IS being put into the Work-from-Home group BUT in the pie chart (screenshot 4) it seems to indicate it is going to the Web Surfing group NOT Work-from-Home. Then to throw one last curve ball at you I look at the Merlin classification page and it shows the Aruba (192.168.1.101) in the File Transferring category (screenshot 5). What am I missing here?

Today’s version 0.8.6 release will make the connections more accurately represent what is really happening underneath when there are AppDB redirections for categories used in an iptables rule. Some people may find that what is shown is not what they thought should be happening, but welcome to the party.

There is probably a better long-term fix to create special rules behind-the-scenes to keep the Classes in the intended priority level if an appdb rule would change it.

 

[andresmorago]

Today’s version 0.8.6 release will make the connections more accurately represent what is really happening underneath when there are AppDB redirections for categories used in an iptables rule. Some people may find that what is shown is not what they thought should be happening, but welcome to the party.

There is probably a better long-term fix to create special rules behind-the-scenes to keep the Classes in the intended priority level if an appdb rule would change it.

Thanks Dave. Just updated. is it just me or am i seeing more untracked connections than before ?

 

[dave14305]

Thanks Dave. Just updated. is it just me or I’m seeing more untracked connections than before ?

Probably a coincidence since I do not do anything to create Untracked connections, nor do I modify the Application name.

 

[Vexira]

Thanks Dave. Just updated. is it just me or am i seeing more untracked connections than before ?

I'd do a signature update check might help I guess.

 

[dave14305]

I'd do a signature update check might help I guess.

That will also now trigger a “flexqos check” too (restart_wrs event). Hope it works!

 

[Vexira]

That will also now trigger a “flexqos check” too (restart_wrs event). Hope it works!

Well I do have good news, fresh install works, no more rate errors, which were a left over from previous firmware versions.