CN107528923B - A kind of data transmission method of network adapter and network adapter - Google Patents

Abstract

The invention discloses a novel network adapter supporting a TCP/IP protocol cluster and a data transmission method thereof. Both sent and received by the host application layer are metadata. The novel network adapter can independently complete some protocol negotiations, such as functions of establishing data connection, encapsulating/de-encapsulating network data, encrypting/decrypting data, compressing/decompressing network data, a hardware firewall, responding network detection messages, various routing and link detection mechanisms, even PPP/DHCP user dialing and the like by realizing a TCP/IP protocol stack. It provides the possibility for the operating system and existing hardware to process only the metadata. The data throughput of the core hardware is greatly improved. The invention can form a novel host framework which is as small as a PC and as large as various servers, and is suitable for wide application.

Description

A kind of data transmission method of network adapter and network adapter

technical field

The present invention relates to computer network technology, in particular to data transmission technology in computer network, in particular to a novel network adapter supporting TCP/IP protocol cluster and a data transmission method thereof.

Background technique

At this stage, because the TCP/IP protocol is widely used, using the TCP/IP protocol for data communication is the mainstream configuration of the host. In the OSI seven-layer model of the network, a common network adapter is a network component working at the link layer, which covers the physical layer and the data link layer, and is the interface connecting the computer and the transmission medium in the local area network. Physical connection and electrical signal matching, as well as data frame transmission and reception, frame encapsulation and decapsulation, media access control, data encoding and decoding, and data caching functions. With the widespread application of cloud computing and cloud storage, the amount of network data increases dramatically, and the relationship between network data and users is getting closer and closer. In the future, most of the data of personal PCs and cloud servers will come from the Internet.

In today's increasingly mature electronic manufacturing technology, hardware products are updated faster and faster, but hardware manufacturing costs are getting lower and lower. All of this stems from the growing maturity of electronic manufacturing technology. Nowadays, the hardware configuration of many home PCs can reach the configuration standard of a small and medium-sized server. From the side, it also shows that the processing speed of the host is more and more related to the user's experience.

In this context, the desire of high-performance CPU + large memory to improve the processing speed of the host will be dragged down by the operating system's mode of carrying a simple network adapter. Figure 1 shows the working mode of the traditional network adapter on the host. In the current mode, the network adapter works at the data link layer, data packet parsing/checking, data packet encapsulation/checking, user data Packet encryption/decryption (user IPSec VPN access scenarios), user packet compression/decompression, in-depth analysis of network data and virus identification, etc., consume a large part of the host's software and hardware resources. At the same time, in this "sick" mode, the user experience can only be maintained by continuously upgrading the CPU, memory and even the motherboard. This would be a lot of overhead worldwide.

At the same time, because the operating system has task scheduling in operation, it is difficult to respond quickly to the link detection packet like hardware. Therefore, once the link is interrupted or the traffic is congested, the host cannot converge on the data processing in time, resulting in data loss. It is often necessary to ensure that communication data is not lost by means of connection-oriented communication protocols that are relatively expensive in communication. If the host has a hardware that can independently support the detection protocol, the link convergence speed of the host will be greatly improved.

If the above overhead is removed from the existing operating system and hardware resources, it is handed over to an independent hardware to complete it alone. Operating systems and applications are only responsible for processing metadata, which will greatly improve user experience. Cloud hosts and distributed servers can also focus more software and hardware resources on data business processing.

Therefore, it is urgent to have a new type of network adapter that supports the TCP/IP protocol suite.

SUMMARY OF THE INVENTION

Purpose of the invention: In view of the deficiencies of the prior art, the present invention provides a data transmission method for a network adapter and a network adapter, which can realize network data processing based on the TCP/IP protocol cluster.

Technical solution: The data transmission method of a network adapter according to the present invention includes the following steps:

The application constructs a unique identifier of the application in the local machine according to the module ID or task identifier;

The application notifies the information of the new network adapter peer through the destination IP address and destination port, and then initiates the session;

The application registers a message receiving and processing interface with the new network adapter for processing return data;

The return data is delivered to the application program through the message receiving and processing interface corresponding to the session, and the data reception is completed.

Wherein, the data sending method includes the following steps:

Construct the information required by the application layer to send data to the network adapter, including application module ID or task identifier, destination IP address or destination IP address plus destination port, and packet receiving and processing interface;

The application layer calls the interface provided by the network adapter and passes in the constructed data;

The network adapter searches for the corresponding session information according to the incoming data information, encapsulates the metadata of the application layer by layer, and initiates a session request to the remote end;

Create a session and build a session table.

The data receiving method includes the following steps:

Obtain data from the link layer and decapsulate the data layer by layer;

Query the session table according to the current session information;

If the session table is valid, call the corresponding message reception processing function, and hand over the metadata to the application for processing; if the session table does not exist or is invalid, the data is discarded.

A network adapter that includes:

Hardware processor, which can independently complete common protocol interaction and data processing;

Independent hardware memory or host memory that can be used independently to cache messages and data for protocol interaction;

An independent non-volatile storage unit that saves the software, user configuration and operation logs required for its operation;

Multiple link transceiver module slots to support the drive and automatic identification of multiple physical layer transceiver modules or flexible plug-in cards.

Beneficial effects: The present invention discloses a novel network adapter supporting TCP/IP protocol cluster and a data transmission method thereof. When using the network adapter for data transmission, the host application layer does not need to create a session, maintain session state, and do not need to Participate in the encapsulation and decapsulation of data. The host application layer sends and receives metadata. By implementing the TCP/IP protocol stack, the new network adapter can independently complete some protocol negotiation, such as establishing data connection, encapsulating/decapsulating network data, data encryption/decryption, compression/decompression of network data, hardware firewall, network detection Message response, various routing and link detection mechanisms, and even PPP/DHCP user dialing and other functions. Provides the possibility for operating systems and existing hardware to process only metadata. The data throughput of the core hardware is greatly improved. Using the present invention, a new type of host architecture can be formed, which can be as small as a PC and as large as various types of servers, which is suitable for wide application.

Description of drawings

Fig. 1 is the mainframe working mode based on the traditional network adapter.

FIG. 2 shows the host working mode of the new network adapter supporting the TCP/IP protocol cluster based on the present invention.

FIG. 3 is a flow chart of data transmission performed by the novel network adapter according to the present invention.

FIG. 4 is a flow chart of receiving a message by the new network adapter according to the present invention.

FIG. 5 is a system structure diagram of the novel network adapter according to the present invention.

Detailed ways

The technical solutions of the present invention will be further described below with reference to the accompanying drawings.

The invention discloses a novel network adapter supporting TCP/IP protocol cluster. As shown in Figure 2, the new network adapter covers the complete TCP/IP protocol stack, and can independently complete the encapsulation/decapsulation of network data, data encryption/decryption, network data compression/decompression, firewall, and network detection packets. Answers, various routing and link detection mechanisms, and even PPP/DHCP/L2TP user dial-up functions. It reduces the processing of packets by the operating system protocol stack as much as possible, improves the processing speed of network data through hardware, and saves core resources such as CPU and memory of the host.

The main functions of the new network adapter are as follows:

1) Independent data transmission and reception

The application constructs a unique identifier of the application in the local machine according to the module ID or task identifier;

The application notifies the information of the new network adapter peer through the destination IP address and destination port, and then the session can be initiated;

At the same time, the application needs to register a message receiving and processing interface with the new network adapter to process the return data;

The return data will be handed over to the application program through the message receiving and processing interface corresponding to the session, and the data reception will be completed;

In the above process, the application layer does not need to create a session, maintain session state, or participate in data encapsulation and decapsulation. The host application layer sends and receives metadata.

2) In the network layer processing of the new network adapter, the processing of firewall and deep virus analysis is added to effectively protect the security of the host.

3) The new network adapter can complete the PPP/DHCP/L2TP dial-up independently through the dial-up mode, user name, password and other information configured by the user, and shield the network maintenance for the user.

4) In the IPSec VPN scenario, the new network adapter will encrypt or decrypt the data sent and received at the network layer.

5) Support common BFD, VRRP, STP and other network link detection and maintenance protocols, and independently complete protocol negotiation and response. However, the user needs to hook up the corresponding processing in the protocol state change event to notify the application.

Specifically, taking the following data transmission and reception as an example, the working process of the network adapter of the present invention will be described in detail.

As shown in Figure 3, sending data through the new network adapter includes the following steps:

Step 301: Construct the information required by the application layer to send data to the new network adapter once, including but not limited to: application module ID or task identifier, destination IP address or destination IP address plus destination port, and packet receiving and processing interface.

Among them, the application module ID or task identifier is used to distinguish different local applications within the new network adapter;

The destination IP address or the destination IP address plus the destination port is used for the network adapter to establish a session with the remote end;

The message receiving and processing interface is responsible for handing the return data to the corresponding application layer program for processing;

Step 302, the application layer invokes the interface provided by the new network adapter, imports the constructed data, and prepares for data transmission;

Step 303, the network adapter queries corresponding session information according to the incoming application ID or task identification, destination IP address or destination IP address plus destination port as the Key value;

Step 304, determine whether there is session information, if the session information does not exist, go to 312 for processing, if the session exists, go to 305 for processing;

Step 305, determine whether the queried session table is valid, if the session table is valid, go to 306 for processing, otherwise go to 310 for processing;

Step 306: Obtain the socket information corresponding to the session, and refresh the last used timestamp of the session table. The timestamp indicates the last used time of the entry, which can be compared with the current time by the entry aging mechanism to determine whether the entry has exceeded the aging time. cycle to be deleted and recycled;

Step 307, performing TCP/UDP encapsulation on the metadata of the application;

Step 308, perform IP layer encapsulation on the application metadata or the data frame that has been encapsulated by TCP/UDP;

Step 309, complete the link layer encapsulation of the data frame;

Step 310: If the time when the session was initiated last time has expired, go to 312 for processing, otherwise go to 311 for processing; the connection timeout time is used to suppress the frequent sending of data requests to trigger connection establishment requests and reduce network congestion. This timeout time is: User-configurable;

Step 311, returning an error value to the application, informing the application that there is currently no available session;

Step 312, initiate a session establishment request to the remote end according to the destination IP address or destination IP address+destination port;

Step 313, if the new network adapter successfully initiates the session according to the information specified by the application, go to 315 for processing, otherwise go to 314 for processing;

Step 314, according to module ID or task identification, IP address or destination IP address plus destination port group Key, establish a pseudo session table, save information such as session failure sign, the time stamp of this session initiation time;

Compared with the real session table, the pseudo-session table is characterized in that the failure flag is adaptive, and its function is to record the timestamp of the first session initiation, to ensure that sessions are not frequently initiated within a certain period of time after the timestamp, and to reduce unnecessary system consumption. , after the specified time, you can try to establish a session with the remote controller again;

Step 315, according to the module ID or task identifier, query the message receiving and processing interface corresponding to the task;

Step 316: After the session is successfully created, establish a session table according to the module ID or task identifier, IP address or destination IP address plus the destination port group key, and associate the packet receiving and processing interface. Overrides if a pseudo-session table exists. And perform the operations of steps 306 to 309 to complete the packet encapsulation.

After the return data reaches the new network adapter at the local end, the data will be decapsulated layer by layer. Then, according to the session information, the session table and the corresponding packet receiving and processing interface are queried, and the data is handed over to the application program for processing. As shown in Figure 4, the specific steps for receiving data are as follows:

Step 401, obtain data from the link layer, and decapsulate the link layer;

Step 402, parse and check the IP layer encapsulation, and then peel off the IP layer encapsulation;

Step 403, parse and check the TCP/UDP encapsulation, and then strip off the TCP/UDP layer encapsulation;

Step 404, obtaining corresponding session information according to the current session information;

Step 405, constructing a Key according to the current session information, and querying the session table;

Step 406, determine whether the session table exists and is valid, if the session table exists and is valid, go to 408 for processing, otherwise go to 407 for processing;

Step 407, if the current session table does not exist or the session table is invalid, discard the current data;

Step 408: If the current session table exists and is valid, call the corresponding message reception processing function, hand over the metadata to the application for processing, and then update the timestamp used by the session table.

In addition, there is also a set of session table/pseudo-session table aging mechanism for network adapters. Ensure that the session table and pseudo-session table whose aging time has reached are deleted and recycled.

Compared with traditional network adapters, the new network adapters supporting the TCP/IP protocol cluster need to be extended in the hardware architecture to achieve the above functions. As shown in Figure 5, the new network adapters include but are not limited to the following components:

1) Hardware processor, which can independently complete common protocol interaction and data processing. In addition, it can also include a co-processor, for example, a co-processor for TCP/UDP/IP header checksum calculation or other hardware units with this function; network adapters for security application scenarios preferably have co-processors that support common encryption/decryption algorithms A processor or other hardware unit capable of this function.

3) Independent hardware memory or host memory that can be used independently, for buffering messages and data of protocol interaction.

4) An independent non-volatile storage unit, which saves the software, user configuration and operation log required for its operation. Make sure that after the host is powered on or restarted, the network adapter can automatically return to a normal working state and the running log is not lost.

5) Certain business data processing capabilities and business configuration interfaces to meet the business needs of users such as firewall and virus defense.

6) Multiple link transceiver module slots to support the drive and automatic identification of multiple physical layer transceiver modules or flexible plug-in cards. The working mode of the network adapter is similar to an independent network terminal, and its physical layer link connection module can support flexible card insertion and automatic identification, such as ETH link transceiver module, wireless link transceiver module, PPP link transceiver module, ATM link transceiver module It can greatly enhance its practicability in different link scenarios without paying the cost of multiple complete network adapters. For example, in the ETH link scenario, the ETH transceiver card or module is used together with the network adapter. When switching to the wireless scenario, a wireless transceiver card or module can be replaced or added, but the main part of the network adapter does not need to be replaced. This hardware architecture reduces the cost of network adapters to a certain extent.

In terms of maintainability, the new network adapter provides users with a simple management interface or interface, which is convenient for users to configure or upgrade, such as the network access method and IP/DNS configuration of the network adapter.

Claims (1)

1. a data transmission method of a network adapter, is characterized in that, the data sending process comprises the following steps: Step 301, constructing the information required by the application layer to send data to the network adapter once, including: application module ID or task identification, destination IP address or destination IP address plus destination port, message receiving and processing interface; Step 302, the application layer invokes the interface provided by the network adapter, passes in the constructed data, and prepares for data transmission; Step 303, the network adapter queries corresponding session information according to the incoming application ID or task identification, destination IP address or destination IP address plus destination port as the Key value; Step 304, determine whether there is session information, if the session information does not exist, go to 312 for processing, if the session exists, go to 305 for processing; Step 305, determine whether the queried session table is valid, if the session table is valid, go to 306 for processing, otherwise go to 310 for processing; Step 306: Obtain the socket information corresponding to the session, and refresh the last used timestamp of the session table. The timestamp indicates the last used time of the entry for comparison between the aging mechanism of the entry and the current time to determine whether the entry has exceeded the aging period. but to be deleted and recycled; Step 307, performing TCP/UDP encapsulation on the metadata of the application; Step 308, perform IP layer encapsulation on the application metadata or the data frame that has been encapsulated by TCP/UDP; Step 309, complete the link layer encapsulation of the data frame; Step 310, if the time when the session was initiated last time has expired, go to 312 to process, otherwise go to 311 to process; Step 311, returning an error value to the application, informing the application that there is currently no available session; Step 312, initiate a session establishment request to the remote end according to the destination IP address or the destination IP address plus the destination port; Step 313, if the network adapter successfully initiates the session according to the information specified by the application, go to 315 for processing, otherwise go to 314 for processing; Step 314, according to module ID or task identification, destination IP address or destination IP address plus destination port as Key value, establish a pseudo session table, save session failure sign, the timestamp of this session initiation time; Step 315, according to the module ID or task identifier, query the message receiving and processing interface corresponding to the task; Step 316: After the session is successfully created, a session table is established according to the module ID or task identifier, the destination IP address or the destination IP address plus the destination port as the key value, and the packet receiving and processing interface is associated, and the operations of steps 306 to 309 are performed to complete packet encapsulation; The data reception process includes the following steps: Step 401, obtain data from the link layer, and decapsulate the link layer; Step 402, parse and check the IP layer encapsulation, and then peel off the IP layer encapsulation; Step 403, parse and check the TCP/UDP encapsulation, and then strip off the TCP/UDP layer encapsulation; Step 404, obtaining corresponding session information according to the current session information; Step 405, construct the Key value according to the current session information, and query the session table; Step 406, determine whether the session table exists and is valid, if the session table exists and is valid, go to 408 for processing, otherwise go to 407 for processing; Step 407, if the current session table does not exist or the session table is invalid, discard the current data; Step 408: If the current session table exists and is valid, call the corresponding message reception processing function, hand over the metadata to the application for processing, and then update the timestamp used by the session table.

Images