Conversation
Codecov Report
@@ Coverage Diff @@
## master #590 +/- ##
==========================================
+ Coverage 67.29% 67.89% +0.59%
==========================================
Files 202 205 +3
Lines 12522 12555 +33
==========================================
+ Hits 8427 8524 +97
+ Misses 4095 4031 -64
Continue to review full report at Codecov.
|
|
So with this design you will have an instance of the MPLS Push module for every unique label to be pushed. Which, in some instances, might be 1000's of labels. Does this really scale well in BESS to have 1000's modules and therefore 1000's of connections between modules? |
| pkt->prepend(4); | ||
| eth = pkt->head_data<Ethernet *>(); | ||
| eth->src_addr = src_addr; | ||
| eth->dst_addr = dst_addr; |
There was a problem hiding this comment.
To the end of keeping the scope of this module narrow, might you just copy the original ethernet header here and change the ethertype? A sequence of SetMetadata and EtherEncap modules downstream of this module can take care of changing the addresses if needed.
| Ethernet::Address src_addr = eth->src_addr; | ||
| Ethernet::Address dst_addr = eth->dst_addr; | ||
|
|
||
| pkt->prepend(4); |
There was a problem hiding this comment.
All code south of here should be guarded by a check that this call to Packet::prepend() didn't fail.
| Ethernet *eth = pkt->head_data<Ethernet *>(); | ||
|
|
||
| Ethernet::Address src_addr = eth->src_addr; | ||
| Ethernet::Address dst_addr = eth->dst_addr; |
There was a problem hiding this comment.
Doh, sorry. Ignore my comment below. I didn't notice these lines.
|
@nrdmao That's a great point. At that scale probably we would want:
(3) is what some existing modules do, such as At the moment, unfortunately, the per-packet metadata approach is not quite applicable in most cases. Existing classification modules ( |
|
Sorry, could you rebase the branch? Due to changes in #591 there are build errors in this PR... |
4 participants
No description provided.