Can Flint use SQM?

I’ve received my order, and the router is working well.

However, I have failed to install luci-app-sqm via LuCI and it shows a kmod is missing.

Has anyone successfully installed SQM? Can you share your method and experience?

Thanks

Thanks for clarifying that.

Does Flint also not support fq_codel or any other methods to control bufferbloat?

I’m don’t know yet, need ask @alzhao

fq_codel is supported. Its sqm-scripts is older version. To support cake, need to port code to tc/iproute2 and port kmod-sched-cake, and maybe more. Now kmod-sched-cake is ready. Other work is schedule later time.

Soft repo had some dependency issue for sqm-scripts, now it’s fixed. You can use sqm-scripts and luci-app-sqm (both version 1.0.7), maybe check if that mitigate bufferbloat.

I’ve successfully installed sqm and configured Luci-app-sqm

I’ve set up:

  1. interface name to eth0 (wan)
  2. ingress to 80000
  3. egress to 80000
  4. link layer to 124 (80 for Wireguard ipv6 and 44 for ethernet)
  5. enabled SQM
    And left everything else on default

Upon some testing, the ingress doesn’t seem to be affected by the settings and the egress setting works to reduce bufferbloat.

Thanks for pointing out link layer setting(124). I also found that NAT short cut driver should be turn off to make qos work.By command:

/etc/init.d/qca-nss-ecm stop

If only egress works, then you can try add another raw of settings on interface br-lan, which is downstream facing while eth0 is upstream facing.

3 Likes

Thanks, that worked for me perfectly after the command and br-lan settings

I’ll share with others my settings:

1st set of SQM:
Enable this SQM instance: tick
Interface name: br-lan
Download speed: 0
Upload speed: 70000

2nd set of SQM:
Enable this SQM instance: tick
Interface name: eth0
Download speed: 0
Upload speed: 70000

Anything above 70000 seems to have bufferbloat while running IPV6 Wireguard and nothing else heavy

2 Likes

Unfortunately, I have the problem that no matter what I adjust, nothing changes in the connection.
Both bufferbloat and QoS remain the same. When I do a speed test, the ping jumps directly from 10ms to 100ms.

How do I get QoS to work properly?

Can you share your SQM configuration?

Sorry for the late reply, but I was only able to take care of the router today.

So after the command as described above, everything works now.
/etc/init.d/qca-nss-ecm stop

What is the driver actually for that I switched off with the command?

I have a question, after the last update I noticed that with the command " /etc/init.d/qca-nss-ecm stop " the bufferbloat is only reduced during the upload, but during the download it still remains high…

is there now another command that can reduce the bufferbloat on the download?

Please try add another raw of settings on interface br-lan to see if it works.
Could you share which site do you use to test bufferbloat?

I have now reinstalled and adjusted the SQM software. + " /etc/init.d/qca-nss-ecm stop " executed. Now I don’t have a bufferbloat anymore, but now I can’t change the setting for “Download speed (kbit/s)”. no matter what value I set here, I have a download speed of max 27 mbit/s (my line is 250mbit/s).

The values for the upload speed are adopted normally.

What could be the problem?

I always use dslreports.com or Bufferbloat Test by Waveform for testing.

What I also noticed is that I can activate the SQM when I “Enable this SQM instance”, but when I remove it, it does not deactivate it until I restart the router.

My Conf:

config queue
option debug_logging ‘0’
option verbosity ‘5’
option qdisc ‘fq_codel’
option qdisc_advanced ‘0’
option linklayer ‘ethernet’
option upload ‘30000’
option script ‘nxt_routed_hfsc.qos’
option enabled ‘1’
option download ‘180000’
option interface ‘eth0.7’
option overhead ‘124’

ifstatus wan | grep device
“l3_device”: “pppoe-wan”,
“device”: “eth0.7”,

If you add setings for eth0.7 and br-lan, the download speed is controlled by setting values of br-lan.

This may be a luci bug, you can try to unclick enable and apply, not remove it directly.