Large receive offload for virtual machines

• US 9,384,033 B2
• Filed: 03/11/2014
• Issued: 07/05/2016
• Est. Priority Date: 03/11/2014
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Jain, Jayant, Sengupta, Anirban, Wu, Weiqing
• Primary Examiner(s): Lee, Adam
• Application Number:
  US14/205,143
• Publication Number:
  US 20150261556A1
• Time in Patent Office: 847 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 2009/45595 Network integration; Enabling network access in virtual machine instances
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: H04L 49/70 Virtual switches

Abstract

A network interface controller (NIC) that includes a set of receive NIC queues capable of performing large receive offload (LRO) operations by aggregating incoming receive packets is provided. Each NIC queue turns on or off its LRO operation based a set of LRO enabling rules or parameters, whereby only packets that meet the set of rules or parameters will be aggregated in the NIC queue. Each NIC queue is controlled by its own set of LRO enabling rules such that the LRO operations of the different NIC queues can be individually controlled.

Claims

1. A method for operating a host machine in a virtualized network environment, the host machine hosting a plurality of virtual machines, the method comprising:

specifying a set of rules for determining whether or not to aggregate packets at a plurality of incoming packet buffers, each rule in the set of rules controlling a packet aggregation operation in a corresponding incoming packet buffer of the plurality of incoming packet buffers, wherein a first rule specifies not aggregating packets for a first set of packets at a first incoming packet buffer and a second rule specifies aggregating packets for a second set of packets at a second incoming packet buffer, wherein packet header contents distinguish the first set of packets from the second set of packets;

forwarding a non-aggregated packet from the first incoming packet buffer to a first virtual machine of the plurality of virtual machines, wherein the first virtual machine forwards the non-aggregated packet; and

forwarding an aggregated packet from the second incoming packet buffer to a second virtual machine of the plurality of virtual machines, wherein the second virtual machine terminates the aggregated packet.

2. The method of claim 1 further comprising specifying a five-tuple identification for the particular transport connection.

3. The method of claim 1 further comprising specifying a protocol identifier, a source IP address, a destination IP address, a source transport identifier, and a destination transport identifier for the particular transport connection.

4. The method of claim 1 wherein the first rule specifies not aggregating packets associated with a particular transport connection at the first incoming buffer.

5. The method of claim 4, wherein the first rule further specifies packets that belong to a different transport connection will be aggregated.

6. The method of claim 4, wherein the first rule specifies a five-tuple identification with at least one wild card, the particular transport connection being associated with the five-tuple identification.

7. The method of claim 1, wherein the plurality of incoming packet buffers are queues in a network interface controller (NIC).

8. The method of claim 1, wherein the second rule specifies aggregating packets having a particular destination MAC address, the destination MAC address being an address of the second virtual machine.

9. A computing device serving as a host machine in a virtualized network environment, the computing device comprising a set of processing units executing a computer program comprising sets of instructions for:

executing a plurality of virtual machines;

specifying a set of rules for determining whether or not to aggregate packets at a plurality of queues, each rule in the set of rules controlling a packet aggregation operation in a corresponding queue of the plurality of queues, wherein a first rule specifies not aggregating packets for a first set of packets at a first queue and a second rule specifies aggregating packets for a second set of packets at a second queue, wherein packet header contents distinguish the first set of packets from the second set of packets;

forwarding a non-aggregated packet from the first queue to a first virtual machine of the plurality of virtual machines, wherein the first virtual machine forwards the non-aggregated packet; and

forwarding an aggregated packet from the second queue to a second virtual machine of the plurality of virtual machines, wherein the second virtual machine terminates the aggregated packet.

10. The computing device of claim 9, wherein the first rule specifies that packets belonging to a particular transport connection are not aggregated at the first queue.

11. The computing device of claim 10, wherein the particular transport connection is a first transport connection, the computer program further comprising sets of instructions for aggregating packets belonging to a second transport connection into an aggregated packet at a third queue.

12. The computing device of claim 9, wherein determining whether to aggregate packets for the particular transport connection further comprises analyzing a five-tuple identification for the particular transport connection.

13. The computing device of claim 12, wherein the five-tuple identification comprises at least one wild card.

14. The computing device of claim 12, wherein the five tuple identification comprises a protocol identifier, a source IP address, a destination IP address, a source transport identifier, and a destination transport identifier for the particular transport connection.

15. The computing device of claim 9 further comprising sets of program instructions for directing an assignment of packets received from a physical network to one of the first and second queues, wherein a packet assigned to the first queue is for one or more first virtual machines and a packet assigned to the second queue is for one or more second virtual machines.

16. A method for controlling large receive offload (LRO) operations at a network interface controller (NIC), the method comprising:

receiving, at a plurality of NIC queues, a plurality of incoming packets from a network, wherein each of the incoming packets contain transport connection information;

specifying a set of filters, wherein each of the filters comprises a five-tuple specifying particular transport connection information and a rule determining whether or not to aggregate an incoming packet containing the particular transport connection information at a NIC queue;

for each of the received incoming packets;

matching the transport connection information of the received incoming packet with the corresponding transport connection information specified by the five-tuple in a filter of the plurality of filters;

aggregating, at the NIC queue, the incoming packet according to the rule of the filter; and

retrieving the packets from the NIC queues.

17. The method of claim 16, wherein the five-tuple comprises a protocol identifier, a source IP address, a destination IP address, a source transport identifier, and a destination transport identifier.

18. The method of claim 16 further comprising operating a plurality of virtual machines and forwarding the retrieved packets to the virtual machines.

19. The method of claim 16, wherein a five-tuple includes at least one wild card.