mwan3: refresh source ip in mwan3track #27745
Conversation
It is possible for the source IP to change without an ifup event. As the USR1 and USR2 signal are already used for ifup and ifdown, there is no possibility to add another trigger for refreshing the source IP only. When receiving an ifup, then the source IP is refreshed. However, in addition to that, the current interface state is set to the assumed initial interface state which is an unwanted side effect in this case. This patch refreshes the source IP during every iteration of the main loop, that is all sleep_time seconds where sleep_time is by default ten seconds. There is one mwan3track process per interface. Thus, I make the assumption that this does not cause a noticable amount of extra system load. Signed-off-by: Jonas Lochmann <[email protected]>
| } | ||
| sleep_time=$interval | ||
| network_flush_cache | ||
| mwan3_get_src_ip SRC_IP $INTERFACE |
There was a problem hiding this comment.
I think we should move the call of network_flush_cache into the function mwan3_get_src_ip. If some in the future needs the function mwan3_get_src_ip anywere else in mwan3. The developer needs to know that the function network_flush_cache needs to be called first to get the latest SRPC_IP value.
We also call this function there
packages/net/mwan3/files/usr/sbin/mwan3
Line 229 in 1a46649
There was a problem hiding this comment.
I disagree with that. The other place where this is used in mwan3track (in firstconnect), there is also the manual call to network_flush_cache. This only refers to a cache within the bash script.
mwan3 use is short running/only calls the function once. Thus, there is no moment where this could be refreshed.
2 participants
📦 Package Details
Maintainer: @feckert
Description:
It is possible for the source IP to change without an ifup event. As the USR1 and USR2 signal are already used for ifup and ifdown, there is no possibility to add another trigger for refreshing the source IP only.
When receiving an ifup, then the source IP is refreshed. However, in addition to that, the current interface state is set to the assumed initial interface state which is an unwanted side effect in this case.
This patch refreshes the source IP during every iteration of the main loop, that is all sleep_time seconds where sleep_time is by default ten seconds. There is one mwan3track process per interface. Thus, I make the assumption that this does not cause a noticable amount of extra system load.
🧪 Run Testing Details
✅ Formalities
If your PR contains a patch:
git am(e.g., subject line, commit description, etc.)
We must try to upstream patches to reduce maintenance burden.