Method and system for forwarding Ethernet packets between multiple network ports

Abstract

The invention discloses an Ethernet packet forwarding method and system among multiple Ethernet ports. The method comprises the following steps: A, presetting the number of buffer areas where each Ethernet port receives the packets as a first predefined value, wherein each buffer area for receiving the packets is allocated with a memory of 1600 bytes in advance, presetting the number of buffer areas where each Ethernet port sends the packets as a second predefined value, wherein each buffer area for sending the packets is not allocated with a memory in advance; B, establishing a Ethernet packet network data package receiving assignment through each Ethernet port, and adopting a polling mode to receive the package; C, when the Ethernet port receives the Ethernet packets, judging whether the Ethernet packets need to be forwarded; D, when the received Ethernet packets is judged as need to be forwarded, sending the packets from the receiving packet buffer areas by invoking the package sending function which needs to be forwarded, and setting an identifier which indicates that memory does not need to be released on current packet sending buffer area of the Ethernet port which needs to be forwarded. By means of the Ethernet packet forwarding method and the system thereof, zero memory application and zero-copying forwarding of the Ethernet packets among multiple Ethernet ports is realized, and the system performance is obviously improved.

Description

technical field [0001] The invention relates to the technical field of network data forwarding, in particular to a method and system for forwarding Ethernet messages between multiple network ports. Background technique [0002] In the field of data communication, there is often a need to forward large-traffic Ethernet packets. How to improve the forwarding capability of the processor and the data throughput of the system is a problem that system engineers need to think about. [0003] Traditional design mode: After the network port A receives the report, after the actual processing, the message pointer is passed to the network port B to send the function to forward it, and at the same time, the memory is re-applied for the network port A and mounted on the current BD. After the network port B is sent, it is responsible for releasing the memory. In addition, the traditional network port drivers all use the interrupt method. Although the real-time performance is very strong, ...

AI Technical Summary

Problems solved by technology

In addition, the traditional network port drivers all use the interrupt method. Although the real-time performance is very strong, very frequent interrupts under heavy traffic will lead to increased system scheduling, and the kernel will also consume CPU resources.

[0004] The above methods can be handled calmly under normal traffic, but when the large traffic receives more than 50,000 packets per second, a large number of packets will be lost, the CPU usage will be extremely high, and the system will become very unstable.

The system consumes too many CPU resources in memory application and release operations, at least 50,000 times per second.

Images