We've got a strange problem with lagg(4) interfaces built on Intel's 82580 chipset igb (dual port 1Gb card). We cannot send over lagg interface more than ~1Gb/s, yet receiving ~2GB over it. Looks like something indicating the problem can be seen here: anri@host:[8:13]~#ifstat -i lagg0 -i igb1 -i igb3 1 lagg0 igb1 igb3 KB/s in KB/s out KB/s in KB/s out KB/s in KB/s out 9116.50 26515.16 4147.70 28871.43 5004.86 23683.31 8423.08 26544.62 3853.22 28980.60 4594.21 23520.97 8796.48 26395.28 4248.46 28344.00 4567.42 23978.17 Note (IN) traffic - everything's ok, about 50% on each igb interfaces and total summary on lagg0. But (OUT) traffic looks weird - there is more traffic on the single igb1 than on lagg0! Tried with default driver came with the system, also the new one igb-2.3.10 from Intel site - no luck. Set net.link.lagg.0.use_flowid=0 not solve the problem. Testing lagg interface built on other cards (em, bce) on the same machine shows expected normal behavior for both directions. Fix: My system is freebsd 8.4, use the e1000 driver from freebsd 8.3 , the result is OK .
With outbound bytes double the expected value on an igb interface, you're probably running into http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/182828 Sincerely, Andrew Kinney
Responsible Changed From-To: freebsd-bugs->freebsd-net Over to maintainer(s).
This seems related to the other issues where folks disable LRO. Does this help with your test?
batch change: For bugs that match the following - Status Is In progress AND - Untouched since 2018-01-01. AND - Affects Base System OR Documentation DO: Reset to open status. Note: I did a quick pass but if you are getting this email it might be worthwhile to double check to see if this bug ought to be closed.
Feedback timeout (over 2 years). Also, this is believed to work just fine for all supported branches.