KR20020039492A - Packet processing method - Google Patents

Abstract

PURPOSE: A method for processing a packet is provided to decrease a load caused by a frequent switching between a kernel mode and a user mode and decrease a processing time. CONSTITUTION: In a method for processing a packet from a kernel mode to a user mode, a shared memory being used in the kernel mode and a user mode is allocated and informed to a kernel mode driver and the shared memory address into a system address(S100). The shared memory area is divided by a packet structure unit, and creates the first linked list(S110). The second linked list is created for storing a packet(S120). If a packet is entered from a network interface card device driver, one space list item is obtained from the first linked list and the packet is copied in the memory area, and the copied packet is inserted in the second linked list(S130a). If packets are accumulated in the second linked list by a repetitious process, the packets are transmitted to a user mode program(S140a). The second linked list is initialized(S150).

Description

Packet processing method

The present invention relates to a packet processing method, and relates to the implementation of a protocol driver in charge of packet processing between a kernel mode and a user mode in a network. .

1 is a hierarchical diagram of a protocol.

The bottom layer in the figure serves to make fragment data into bit data or combine bit data into fragment data for data transmission to the link / network access layer. .

The network interface card device driver (NIC) controls the NIC, a card that inserts into an expansion slot on a PC or workstation to allow a computer terminal to connect to a local area network (LAN).

In the figure, the middle layer is a network layer that processes communication using frames. The middle layer examines all frames sent and received by each node.

The Network Driver Interface Specification (NDIS) provides a standard interface for Network Interface Card device drivers.

NDIS polls the stack one by one until the stack requests the packet to match the packet coming from the NIC driver with the appropriate protocol stack.

In the figure, the upper layer is the transport layer, which controls the flow of networked data.

The packet processing method of the present invention relates to a method of implementing a protocol driver of this layer.

FIG. 2 is a schematic diagram of a conventional packet processing method, and shows an embodiment of a conventional conventional method of processing a packet from a kernel mode to a user mode in a network. It is.

This method waits, and if there is a read call from the user mode program, the kernel mode program sends one packet input through the network interface card device driver. Is copied to the user space buffer, and the user mode program processes the data and reads it again, repeating the above operation to transfer the packet from kernel mode to user mode. Will be processed.

However, in the conventional packet processing method, since the number of asynchronous procedure calls are made to copy packets, the overhead due to frequent switching between kernel mode and user mode is high and repeated. There was a problem that there is a high probability of packet loss during packet transmission.

The present invention has been invented to solve the above problems, and an object of the present invention is to provide a packet processing method which can reduce the processing time due to the frequent switching between the kernel mode and the user mode.

Still another object of the present invention is to provide a packet processing method that can reduce packet loss during packet transmission.

1 is a hierarchical diagram of a protocol.

2 is a schematic diagram of a conventional packet processing method.

3 is a schematic diagram of a packet processing method of the present invention;

4 is a flowchart of an embodiment of processing a packet from kernel mode to user mode in the packet processing method of the present invention.

5 is a flowchart of an embodiment of processing a packet from user mode to kernel mode in the packet processing method of the present invention.

According to an aspect of the present invention for achieving the above object, the packet processing method of the present invention is coded by using a shared memory (Shared memory) used in the kernel mode and the user mode, the packet is shared It is characterized by creating a linked list by dividing it into structure units to process packets.

According to an additional aspect of the present invention, there is a feature of interlocking so as not to simultaneously use the same memory area of the shared memory in kernel mode and user mode.

Before describing the present invention in detail, the computer and communication related terms used in the present invention will be described.

Kernel is the most basic system function of the operating system. It is responsible for core tasks such as process management, CPU scheduling, I / O control, and storage management.

Kernel mode program is a program that operates as part of the kernel that processes a user mode program request using a kernel service and a network interface card driver.

A user mode program is a program that receives or sends a packet using a kernel mode program, or controls a NIC driver. The user mode program operates under an operating system, and actually looks at the contents of a packet and performs various tasks. To deal with.

A packet is a bundle of data used in data transmission. Each packet contains not only a certain amount of data but also control information such as a data destination, an address, or a control code.

A linked list is a data structure that corresponds to a linear list. Each node in the list consists of a data part and a link part. The link is the next node in the list. It has a pointer value.

Address translation refers to the translation of the address of an instruction or data into the address of main memory where it is loaded or relocated when the program is loaded into memory.

That is, the virtual memory address of a data item or command is converted into the actual address of main memory.

Shared memory refers to memory shared by multiple processors.

Interlocking refers to preventing another processor or user from touching one processor or hardware or software being used by the user.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily understand and reproduce the present invention.

FIG. 3 is a schematic diagram of a packet processing method of the present invention and illustrates an overview of processing a packet from a kernel mode to a user mode in a network.

The method of the present invention waits and if a read call is received from a user mode program, the kernel mode program receives a packet inputted through a NIC (Network Interface Card device driver). Copy certain packets from the stored linked list to the user space buffer, and the user mode program processes them and then reads them again and repeats the above operation from kernel mode. The packet is processed in the user mode.

That is, the present invention does not process and repeat one packet during packet transmission, but processes and repeats a plurality of packets in predetermined packet units.

4 is a flowchart of an embodiment of processing a packet from kernel mode to user mode in the packet processing method according to the present invention.

A packet processing method according to an aspect of the present invention is a method of processing a packet from a kernel mode to a user mode, in the kernel mode and the user mode to be used as a packet buffer. A memory allocation step (S100) of allocating a shared memory to be used simultaneously and informing a kernel mode driver, and converting the shared memory address into a system address;

Generating a first linked list by dividing the shared memory area by a packet structure unit (S110);

A second connection list generation step (S120) of generating a second connection list for packet storage;

When a packet is received from a NIC (Network Interface Card device driver), a packet storing step of obtaining an empty list item from the first connection list, copying the packet to the memory area, and inserting the packet into the second connection list. (S130a);

A packet forwarding step (S140a) of forwarding the packets to a user mode program when predetermined packets are accumulated in the second connection list by repeating the packet storing step (S130a);

And an initialization step S150 of initializing the second connection list.

This embodiment allocates shared memory used simultaneously in the kernel mode and the user mode to be used as a packet buffer, informs the kernel mode driver of the shared memory allocation, and Convert shared memory address to system address.

Meanwhile, the shared memory area is divided into packet structure units to generate a first linked list, and a second linked list for packet storage.

When a packet comes in from a NIC (Network Interface Card device driver), one empty list item is obtained from the first connection list, the packet is copied to the memory area, and the packet is repeatedly inserted into the second connection list. When a predetermined packet is accumulated in the second connection list, the packets are delivered to a user mode program, and the second connection list is initialized to terminate one cycle of packet transmission.

5 is a flowchart of an embodiment of processing a packet from a user mode to a kernel mode in the packet processing method according to the present invention.

A packet processing method according to an aspect of the present invention is a method of processing a packet from a user mode to a kernel mode, in the kernel mode and a user mode to be used as a packet buffer. A memory allocation step (S100) of allocating a shared memory to be used simultaneously and informing a kernel mode driver, and converting the shared memory address into a system address;

Generating a first linked list by dividing the shared memory area by a packet structure unit (S110);

A second connection list generation step (S120) of generating a second connection list for packet storage;

When a packet is received from a user mode program, a packet storing step of obtaining one empty list item from the first connection list, copying the packet to the memory area, and inserting the packet into the second connection list (S130b) )Wow;

A packet forwarding step (S140b) of forwarding the packets to a kernel mode program when predetermined packets are accumulated in the second connection list by repeating the packet storing step (S130b);

And an initialization step S150 of initializing the second connection list.

This embodiment allocates shared memory used simultaneously in the kernel mode and the user mode to be used as a packet buffer, informs the kernel mode driver of the shared memory allocation, and Convert shared memory address to system address.

The shared memory area is divided into packet structure units to generate a first linked list and a second linked list for packet storage.

When a packet comes in from a user mode program, one empty list item is obtained from the first linked list, the packet is copied to the memory area, and the packet is repeatedly inserted into the second linked list. When a predetermined packet is accumulated in the second connection list, the packets are delivered to a kernel mode program, and the second connection list is initialized to end packet transmission in one cycle.

That is, according to the method of the present invention, the address translation between the kernel mode and the user mode is performed only once when the shared memory is configured by using the shared memory that is used together in the kernel mode and the user mode. It can be accessed from both sides without additional address translation.

Conventional methods that do not use shared memory require address translation between kernel mode and user mode each time a packet is processed.

In addition, the packet processing method of the present invention may interlock to prevent the connection list from being mixed when the packet is inserted into the second connection list in the packet storing step (S1340a) (S130b), and the packet forwarding step (S140a). In operation S140b, since the second connection list is in the form of an array using shared memory, indexes of the beginning and the end of the array may be transferred without having to transfer a memory address.

The most problem in using the shared memory in the present invention is the problem of interlocking to prevent simultaneous use of the same shared memory area in kernel mode and user mode. Because locking is very difficult, we solved this by using a linked list to use the index of the array instead of the memory address.

In addition, the packet processing method of the present invention is characterized in that the number of delivery times of the packets inserted in the second connection list in the packet forwarding step S140a or S140b is less than the number of packets to be inserted.

In other words, if the packets of the second connection list are frequently delivered in the user mode, a lot of overhead is applied as in the conventional method, so the number of delivery of the packets inserted in the second connection list should be less than the number of packets to be inserted.

However, if the number of packet delivery becomes too small, the delay of the user mode program becomes large, and the number of packet delivery is increased because the demand for shared memory increases because a large number of packets must be maintained in the connection list. Needs to be set appropriately.

If the goal is to handle 200Mbps traffic, the number of packets is as follows.

Assuming that 300 bytes of packets flow on average,

(200 × 1024 × 1024) (bps) / 8 (bit / byte) / 300 (byte / packet)

= 87380 packet second

By the way, the maximum size of the packet is 1514 bytes in the case of Ethernet. Based on the maximum size, about 126MB of shared memory is required. In fact, it's reasonable to see twice that, since there's the memory needed to process it.

Based on the current computer environment, it is ideal to deliver the user mode at a rate of about 100 times / second.

In addition, the packet processing method of the present invention adjusts the packet delivery period in the following manner in order to appropriately adjust the number of packet delivery times.

First, in the packet forwarding step (S140a) (S140b), by using a clock provided by the kernel when delivering a packet of the second connection list, the number of packets delivered when the previous packet is forwarded from the last delivery when the previous packet is forwarded. After the passage of time, the delivery cycle is controlled by transmitting packets of the second connection list.

In other words, this method allows a packet to be delivered every predetermined time.

Second, in the packet delivery step (S140a) (S140b), if the length of the second connection list reaches the value of the packet transmitted to the kernel mode program, the packet is the user mode, and the length of the second connection list is transmitted to the user mode program. In the case of reaching the value of, the packet transmission of the second connection list in the kernel mode is controlled.

In other words, this method is to deliver packets every certain number.

On the other hand, as a method for adjusting the packet delivery period may be a method of transmitting a time period or the number of packets to a kernel mode program or a user mode program.

This method adjusts the packet delivery cycle by passing arguments through a special program function to the protocol driver. This method can be implemented in various ways through language program means. Detailed description will be omitted.

Therefore, a protocol driver for controlling a network interface card (NIC) by sending and receiving a packet through an interaction with a network interface card driver (NIC) through the packet processing method as described above. driver).

In addition, it is very useful in terms of security if the encryption algorithm is applied to transmit the packet by encrypting the packet during transmission.

Therefore, the above object to achieve the object of the present invention.

As described above, the present invention allows address translation between kernel mode and user mode only once when configuring shared memory so that both addresses can be accessed without additional address translation by using the address. This can reduce the burden of frequent switching between modes, reduce processing time, and reduce packet loss during packet transmission.

Although the present invention has been described with reference to the accompanying drawings, it will be apparent to those skilled in the art that various modifications may be made therein without departing from the scope of the invention, which is covered by the following claims.

Claims (7)

In the method of processing a packet from a kernel mode (Kernel mode) to a user mode (User mode), Allocate shared memory used simultaneously in the kernel mode and the user mode to be used as a packet buffer to inform a kernel mode driver, and allocate a memory to convert the shared memory address into a system address. Steps; Generating a first linked list by dividing the shared memory area by a packet structure unit; Generating a second linked list for generating a second linked list for storing packets; When a packet is received from a NIC (Network Interface Card device driver), a packet storing step of obtaining an empty list item from the first connection list, copying the packet to the memory area, and inserting the packet into the second connection list. Wow; A packet forwarding step of forwarding the packets to a user mode program when predetermined packets are accumulated in the second connection list by repeating the packet storing step; And an initialization step of initializing the second connection list. In the method for processing a packet from the user mode (User mode) to kernel mode (Kernel mode), Allocate shared memory used simultaneously in the kernel mode and the user mode to be used as a packet buffer to inform a kernel mode driver, and allocate a memory to convert the shared memory address into a system address. Steps; Generating a first linked list by dividing the shared memory area by a packet structure unit; Generating a second linked list for generating a second linked list for storing packets; A packet storing step of, when a packet comes from a user mode program, obtains one empty list item from the first linked list, copies the packet to the memory area, and inserts the packet into the second linked list; A packet forwarding step of forwarding the packets to a kernel mode program when predetermined packets are accumulated in the second connection list by repeating the packet storing step; And an initialization step of initializing the second connection list. The method according to claim 1 or 2, And interlocking the insertion list so that the connection list is not mixed when the packet is inserted into the second connection list in the packet storing step. The method according to claim 1 or 2, In the packet forwarding step, since the second connection list is in the form of an array using a shared memory, the first and the end indexes of the array are transmitted without transmitting a memory address. The method according to claim 1 or 2, And a number of delivery times of packets inserted in the second connection list in the packet delivery step is smaller than the number of packets to be inserted. The method according to claim 1 or 2, In the packet forwarding step, a packet of the second connection list is delivered after a time equal to the packet number delivered in the previous packet delivery from the last delivery in the previous packet delivery using the clock provided by the kernel when delivering the packet in the second connection list. Packet processing method, characterized in that for adjusting the delivery period. The method according to claim 1 or 2, In the packet forwarding step, if the length of the second connection list reaches the number of packets delivered to the kernel mode program, the user mode; if the length of the second connection list reaches the number of packets delivered to the user mode program, the kernel mode; Adjusting the delivery period to deliver the packets of the second connection list packet processing method.