![]() > 0 times recovered from packet loss by selective acknowledgements > 132 acknowledgments not containing data payload received > 1991 packets header predicted and directly queued to user > 48684 bytes directly received in process context from prequeue > 32 bytes directly in process context from backlog > 2924 packets directly queued to recvmsg prequeue. > 0 delayed acks further delayed because of locked socket > 0 packets rejects in established connections because of timestamp > 14 TCP sockets finished time wait in fast timer > 0 packets pruned from receive queue because of socket buffer overrun > $> netstat -s eth1 > before sleep 10 netstat -s eth1 > after > Here is the netstat (from Tx point of view) : > drivers) and it works very well (adaptive-tx is off). > I have tried the same test with a debian lenny (2.6.26 kernel and HP ![]() > I suspected TX coalescing but since 3.4 I can't set this parameter > I have lots of packet loss from Tx to Rx, and I can't understand why. > ProLiant BL460c G7) which has stricly the same HW. > I am sending UDP multicast packets from one blade to another (HP > I am using a NC553i ethernet card connected on a HP 10GbE Flex-10. 6:48 ` Eric Dumazet 0 siblings, 1 reply 31+ messages in threadįrom: Jean-Michel Hautbois 6:28 UTC ( / raw) 14:46 Difficulties to get 1Gbps on be2net ethernet card Jean-Michel 6:28 ` Jean-Michel Hautbois * Re: Difficulties to get 1Gbps on be2net ethernet card $> ethtool -S eth1 > before sleep 10 ethtool -S eth1 > afterĪs you can see, there is 10 tx_stops in 10 seconds (it varies, can be 3 to 15). $> netstat -s eth1 > before sleep 10 netstat -s eth1 > afterĠ packets pruned from receive queue because of socket buffer overrunġ4 TCP sockets finished time wait in fast timerĠ packets rejects in established connections because of timestampĠ delayed acks further delayed because of locked socketĢ924 packets directly queued to recvmsg prequeue.ģ2 bytes directly in process context from backlogĤ8684 bytes directly received in process context from prequeueġ991 packets header predicted and directly queued to userġ32 acknowledgments not containing data payload receivedĠ times recovered from packet loss by selective acknowledgementsĠ congestion windows recovered without slow start after partial ackġ times receiver scheduled too late for direct processingĠ packets collapsed in receive queue due to low socket bufferĠ connections reset due to unexpected dataĠ connections reset due to early user closeĠ times unabled to send RST due to no memory Here is the netstat (from Tx point of view) : I have tried the same test with a debian lenny (2.6.26 kernel and HPĭrivers) and it works very well (adaptive-tx is off). I suspected TX coalescing but since 3.4 I can't set this parameter I have lots of packet loss from Tx to Rx, and I can't understand why. ProLiant BL460c G7) which has stricly the same HW. I am sending UDP multicast packets from one blade to another (HP I am using a NC553i ethernet card connected on a HP 10GbE Flex-10. 6:28 ` Jean-Michel Hautbois 0 siblings, 1 reply 31+ messages in threadįrom: Jean-Michel Hautbois 14:46 UTC ( / raw) Difficulties to get 1Gbps on be2net ethernet card Netdev Archive on help / color / mirror / Atom feed * Difficulties to get 1Gbps on be2net ethernet card 14:46 Jean-Michel Hautbois
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |