Maintaining packet order in offload of packet processing functions

Network interface apparatus, comprising:

• a host interface for connection to a host processor having a memory;
• a network interface, which is configured to receive over a network data packets in multiple packet flows destined for one or more virtual machines running on the host processor;
• and
• packet processing circuitry, which is coupled between the network interface and the host interface and is configured to pass the data packets to a virtual machine monitor (VMM) running on the host processor for preprocessing of the packets by the VMM, which delivers the preprocessed packets to the one or more virtual machines,
• wherein the packet processing circuitry is configured to receive a first instruction to offload from the VMM preprocessing of the data packets in a specified flow in accordance with a specified rule, and responsively to the first instruction to initiate preprocessing the data packets in the specified flow by the packet processing circuitry in accordance with the specified rule while writing one or more initial data packets from the specified flow to a temporary buffer, and upon subsequently receiving a second instruction to enable the specified rule, to deliver the initial data packets from the temporary buffer, after preprocessing by the packet processing circuitry, directly to a virtual machine to which the specified flow is destined, and after delivering the preprocessed initial data packets, to continue preprocessing and delivering subsequent data packets in the specified flow to the virtual machine,
• wherein the packet processing circuitry comprises a transmit pipe, for processing outgoing packets for transmission to the network, and a receive pipe, for processing incoming data packets received from the network, and
• wherein the packet processing circuitry is configured to deliver the initial preprocessed data packets from the temporary buffer to the virtual machine by loopback from the temporary buffer through the transmit pipe to the receive pipe, which writes the preprocessed data packets to another buffer in the memory that is assigned to the virtual machine.