moeller0:

So a stop-gap measure would be to install sqm-scripts and luci-app-sqm and instantiate a shaper on pppoe-wan with a bandwidth slightly below the real available bandwidth. The shaper on pppoe-wan will never actually see the LCP packets and hence will not be able to drop them and by setting the bandwidth slightly below the available rate thee should always enough pppoe bandwidth t accomodate the lcp packets.

weedy:

This isn't a bug, it's "by design".

Since you're flooding your upload buffers on the router and modem the maintenance packets are taking too long and pppd thinks you have a dead link. LEDE borrowed some patches from Debian (I think) that will skip this issue for you.

You should still use QoS for interactivity reasons.

# Adaptive/Smart keepalive
option keepalive_adaptive 1 
## lcp_failure,lcp_interval
option keepalive "7,3"

nbd:

The relevant option is already set by default since some time around 2014.

cbz:

I'm not sure the option is always set - the pppd option it maps to is lcp-echo-adaptive. This isn't a default option for pppd, and its trivial to see that openwrt launches pppd without this option set (ps output or running strings on /proc/pppd-id/cmdline )

cbz:

It also seems to be the logic of this here:

https://github.com/openwrt/openwrt/blob/master/package/network/services/ppp/files/ppp.sh#L123

bill888:

fwiw, I witnessed this problem when I started using BT Home Hub 5A on a 55/10mb VDSL2 connection a year ago. Whenever the 10 mbps upstream was saturated, the DSL connection would disconnect and 'No response to 5 echo-requests' would be reported in the system log.

In LuCI->Network->Interfaces->WAN->Advanced Settings, there are two default settings:

LCP echo failure threshold 0
LCP echo interval 5

The 'default' setting LCP echo failure threshold to Zero, implies all LCP echo failures would be ignored. (see attached image) But I found this to be incorrect/misleading.

The solution was to specify a non-zero value in LuCI as in example below

eg.
LCP echo failure threshold 5
LCP echo interval 5

A non-zero value ADDS the option keepalive to /etc/config/network:

This increased the timeout to 25 seconds and resolved my DSL disconnection issues.

LuCI doesn't appear to be able to manage this keepalive_adaptive option.