The method of automatic detection and driver loading of user mode network card under linux system

Abstract

The invention provides a method for automatic detection and driver loading of user state network cards under the Linux system. In the Linux kernel initialization stage, the PCI bus registers each driver and scans all devices mounted on the PCI bus; Mount the device on the PCI bus to determine whether the driver matches the device, if it matches, then continue to judge whether the class code of the device is the class code of the network card device, if not, initialize the device; if so, then Identify the device as a network card, save the network card information and give up the initialization of the network card, and obtain the network card information in user mode, and load the corresponding network card driver according to the network card information; the present invention automatically detects the network card and driver loading are placed in the user mode to eliminate the high overhead of system calls and data packet copying brought by the kernel mode driver.

Description

technical field [0001] The invention relates to the field of computer operating systems, in particular to a method for automatic detection and driver loading of a user state network card based on a kernel state under a Linux system. Background technique [0002] With the continuous development of technology, the CPU resources that were greatly relieved in the early days of the multi-core era appear to be powerless on network packet processing devices equipped with 10GE or more network cards; the reason is that a large number of interrupts are consumed due to the impact of a large number of packets Certain CPU resources (especially the X86 architecture); the performance loss of the data copy from the kernel mode to the user mode of the Linux operating system will also become a roadblock for the packet processing device to make full use of the multi-core CPU; therefore, how to eliminate the kernel mode driver belt The high overhead of incoming system calls and packet copies ha...

AI Technical Summary

Problems solved by technology

[0002] With the continuous development of technology, the CPU resources that were greatly relieved in the early days of the multi-core era appear to be powerless on network packet processing devices equipped with 10GE or more network cards; the reason is that a large number of interrupts are consumed due to the impact of a large number of packets Certain CPU resources (especially the X86 architecture); the performance loss of the data copy from the kernel mode to the user mode of the Linux operating system will also become a roadblock for the packet processing device to make full use of the multi-core CPU; therefore, how to eliminate the kernel mode driver belt The high overhead of incoming system calls and packet copying has become a major research area in the industry

Images