You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now, all VRFs other than 0 are filtered from FPM. Yet the code already exists in the encoding to transport the table-id in the messages. As such, a surgical fix to zfpm_is_table_for_fpm is all that's needed to export VRFs.
We've done a straight patch for this in our own code, but we'd like to upstream the change. We see three solution paths:
Always enable VRFs to be sent; this changes existing behavior.
Introduce a config option for no fpm filter vrfs; this opens the question of what happens to an already-open FPM connection if this option is changed.
Extend the module init syntax, so that you specify -M fpm:netlink:1 or similar to enable unfiltered VRFs.
Option 3 seems the cleanest, but which option has the highest probability of PR acceptance?
The text was updated successfully, but these errors were encountered:
From FRR's perspective turning on the passing of table-id's is a no-brainer. Would not passing down the vrf routes be there to protect the FPM listener? The FPM listener should be paying attention to the passed table-id. So I see no reason to just pass the vrf tableid(option #1)
Right now, all VRFs other than 0 are filtered from FPM. Yet the code already exists in the encoding to transport the table-id in the messages. As such, a surgical fix to
zfpm_is_table_for_fpm
is all that's needed to export VRFs.We've done a straight patch for this in our own code, but we'd like to upstream the change. We see three solution paths:
no fpm filter vrfs
; this opens the question of what happens to an already-open FPM connection if this option is changed.-M fpm:netlink:1
or similar to enable unfiltered VRFs.Option 3 seems the cleanest, but which option has the highest probability of PR acceptance?
The text was updated successfully, but these errors were encountered: