JP2014155120A - Communication device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce a waiting time to the completion of context information loading to achieve high-speed reception processing in the reception processing of a communication packet by a specific communication protocol.SOLUTION: A communication device that performs communication on the basis of a predetermined communication protocol includes: communication means for controlling communication in the communication device; control means for controlling context information for use for the processing of the predetermined communication protocol; and analysis means for analyzing data transferred through an internal bus that connects between a variety of functional units in the communication device. The control means in the communication device loads context information, corresponding to a result analyzed in the analysis means, to a cache memory accessible from the communication means.

Description

  The present invention relates to a technique for executing communication using TCP / IP at high speed.

  Among protocol groups collectively called TCP / IP (Transmission Control Protocol / Internet Protocol), TCP is a connection-type protocol that plays a central role. A TCP connection can be represented by a combination of four pieces of information of IP addresses and TCP port numbers of two network devices that perform communication. A combination of these information is called a socket pair or simply a socket. Detailed specifications of TCP are described in Non-Patent Document 1.

  In recent years, there has been a rapid increase in the number of embedded devices that are equipped with a TCP / IP protocol processing function to connect to a network and improve customer convenience. In these devices, an auxiliary device specialized for protocol processing such as TOE (TCP / IP offload engine) is added to the system to realize broadband network communication. As a conventional example of the TOE, there is a technique disclosed in Patent Document 1.

  By the way, what is important in TCP / IP protocol processing is the handling of TCB (Transmission Control Block). TCB is an abbreviation for context information prepared for each connection and its aggregate necessary for processing TCP / IP. One TCB is composed of a plurality of parameters (variables), and is a variable that is frequently accessed during TCP processing.

  Patent Document 1 discloses a technique for copying and holding a TCB necessary for performing TCP processing from a main storage unit to a high-speed primary memory such as an SRAM (Static Random Access Memory). As a result, speeding up of access to the TCB is realized. At this time, if the number of connections increases and all the TCBs do not fit in the primary memory, the replacement process is performed with the main storage unit so that only the necessary TCBs exist in the primary memory. Patent Document 2 discloses a technique for detecting in advance a timeout of a re-transfer timer that occurs when TCP / IP processing is executed, and loading a TCB into a primary memory based on the detection. This facilitates TCP / IP processing.

  In addition, there is a technique disclosed in Patent Document 3 as a technique for controlling the timing of loading the TCB into the primary memory. Patent Document 3 adopts a configuration in which protocol processing means for performing TCP / IP processing is implemented separately in the former stage and the latter stage. Then, the TCB used by the subsequent protocol processing means is loaded into the primary memory in advance by the previous protocol processing means. As a result, the subsequent protocol processing means can quickly access the TCB, thereby facilitating the processing.

Special table 2002-524005 gazette JP 2009-296546 A JP 2008-146486 A

Jon Postel, “Transmission Control Protocol”, RFC 793, http: // datatracker. ietf. org / doc / rtf793 /

  During TCP / IP communication, every time a TCP packet is received, a search is made as to which TCB is used. If the TCB is not loaded in the primary memory, it is necessary to perform a load process. In this case, since the method of Patent Document 2 uses a timer, it is not suitable for high-speed reception processing of TCP / IP packets. The method of Patent Document 3 can be applied to TCP / IP reception processing, but the protocol processing means needs to be divided into a preceding stage and a subsequent stage. Further, there is no mention regarding reduction of processing for searching which TCB is used in the protocol processing means in the previous stage. Therefore, unless the TCB search processing time and the time required for loading the TCB are waited for, the subsequent protocol processing means cannot start operation.

  The present invention has been made in view of the above-described problems, and reduces the waiting time required for completion of loading context information in a communication packet reception process using a specific communication protocol, thereby realizing a high-speed reception process. Objective.

  As a means for achieving the above object, a communication apparatus of the present invention comprises the following arrangement. That is, a communication device that performs communication based on a communication protocol, the communication unit that controls communication based on a predetermined communication protocol, the control unit that controls context information used for processing of the predetermined communication protocol, Analyzing means for analyzing data transferred by an internal bus connecting various functional units in the communication device, and the control means provides context information corresponding to a result analyzed by the analyzing means, It is characterized by loading into a cache memory accessed by the communication means.

  According to the present invention, it is possible to reduce the waiting time required for completion of loading context information in the communication packet reception process using a specific communication protocol, and to realize a high-speed reception process.

The figure which shows the structure of the communication apparatus by 1st Embodiment. The figure which shows the example of the hierarchy of the communication processing hierarchy and the hardware which performs a process. The figure which shows the structure of the primary analysis part by 1st Embodiment. The figure which shows the format example of the packet which a communication apparatus receives. The sequence diagram of operation | movement of the primary analysis part by 1st Embodiment. The figure which shows the structure of the TCB control part by 1st Embodiment. The figure which shows the example of a storage method of each element of TCB. The figure which shows the example of the management method of TCB. The flowchart figure of operation | movement of the TCB control part at the time of TCB production | generation. The flowchart figure of operation | movement of the TCB control part at the time of a TCB load request | requirement. The sequence diagram of the operation | movement of the TOE subsystem by 1st Embodiment. The figure which shows the structure of the TOE subsystem (A) by 2nd Embodiment, and a primary analysis part.

  Hereinafter, the present invention will be described in detail based on preferred embodiments with reference to the accompanying drawings. The configurations shown in the following embodiments are merely examples, and the present invention is not limited to the illustrated configurations.

[First Embodiment]
FIG. 1 is a block diagram showing a configuration of a communication apparatus according to the first embodiment. The main control unit 101 provides functions of the entire communication device and performs control for that purpose. The system bus 102 is a bus that connects the main control unit 101, the main storage control unit 103, and the TOE subsystem 105. The main memory control unit 103 controls data read / write using the main memory unit 104.

  The main storage unit 104 stores various software. Specifically, application software that implements each function of the communication device, application protocol, a device driver for controlling related hardware, and an OS (Operation System) are included. The main storage unit 104 also includes data that the main control unit 101 wants to send from the Ethernet (registered trademark) 120 via the TOE subsystem 105, and data that the TOE subsystem 105 receives from the Ethernet 120 and passes to the main control unit 101. Used as a place for storage. The main storage unit 104 also stores various communication firmware executed by the communication control unit 106. Further, the main storage unit 104 also stores a TCB necessary for performing TCP / IP communication processing. Since the number of necessary TCBs increases in accordance with the number of TCP / IP communication connections, a memory area in which the TCBs corresponding to the maximum number of connections supported by the communication apparatus can be stored in the main storage unit 104 in advance. ing.

  The TOE subsystem 105 includes a subsystem bus (internal bus) 109 that connects various functional units in the communication apparatus, and is connected to the system bus 102 via a bus bridge unit 108. The communication control unit 106 executes TCP / IP protocol processing offloaded from the main control unit 101. The communication control unit 106 loads each firmware protocol developed on the main storage unit 104 into a built-in instruction cache memory and executes the firmware program.

  Here, the role of the communication control unit 106 will be described in detail with reference to FIG. FIG. 2 schematically shows a comparison of communication processing layers and hardware for processing. The processing functions of the TOE subsystem 105 cover the layers in the range 208 shown in FIG. That is, the processing functions of the TOE subsystem 105 cover some functions of the MAC control unit driver 205, the Internet layer (IP layer) 204, the transport layer (TCP / UDP layer) 203, and the socket API 202. The MAC control unit driver 205 uses the MAC control unit 118 to exchange communication data and communication information between the Internet layer 204 and the MAC (Medium Access Control) layer 206. The Internet layer 204 processes the IP protocol. The transport layer 203 processes TCP and UDP (User Datagram Program) protocols. These functions are mainly functions for selecting a protocol and relaying data transfer and information transfer. The above is the main communication process executed by the communication control unit 106. The communication control unit 106 controls various functional units installed in the TOE subsystem 105, which are necessary for executing a series of processes. It also has a role to do.

  Returning to FIG. 1, the cryptographic processing unit 107 provides cryptographic communication protocol processing functions such as IPsec and SSL / TLS. Examples of cryptographic processing include an AES (Advanced Encryption Standard) cipher selected by the National Institute of Standards and Technology (NIST) and a SHA-1 (Secure Hash Algorithm 1) hash function unit used for authentication and digital signatures. Can be mentioned. In addition, an MD5 (Message Digest 5) hash function unit or the like standardized by IETF (Internet Engineering Task Force) as RFC1321 is also applicable. The cryptographic processing unit 107 also includes a key management function for maintaining confidentiality of cryptographic keys, random numbers, and prime numbers generated for cryptographic communication protocol processing, and a random number generator for TCP / IP protocol processing and cryptographic communication protocol processing. .

  The bus bridge unit 108 provides a bus conversion function between the subsystem bus 109 and the system bus 102 that is necessary when each functional unit in the TOE subsystem 105 except the DMA control unit 110 accesses the main storage unit 104. Have. A DMA (Direct Memory Access) control unit 110 controls data transfer between the main storage unit 104 and the MAC control unit 118. The buffer unit 111 temporarily holds the received frame processed by the MAC control unit 118. Further, the buffer unit 111 temporarily holds one or more data blocks for one packet extracted from the main storage unit 104. The timer unit 112 is used to generate a time measurement and timeout event necessary for TCP / IP protocol processing.

  The TCB control unit (context information control unit) 113 loads the TCB stored in the main storage unit 104 into the TCB cache unit 114 and returns the loaded TCB to the main storage unit 104 again (flush). Do. Details will be described later. The TCB cache unit 114 is configured by an On Chip RAM so that the communication control unit 106 can be accessed at a higher speed than the main storage unit 104, and is a memory area that stores the TCB acquired from the main storage unit 104. The TCB cache unit 114 is managed in a state where the area is logically separated into a plurality of areas. In the following, one area unit is referred to as a cache block.

  The primary analysis unit 115 is responsible for the process of snooping and analyzing the data being DMA-transferred, in this example, the data appearing on the data bus between the DMA control unit 110 and the MAC control unit 118. Further, the primary analysis unit 115 has a function of issuing a load request to the TCB control unit 113 based on the analysis result. Specifically, the primary analysis unit 115 includes an interface with a snoop bus 116 that takes in data to be analyzed, and an output unit for a control signal 117 for transmitting a TCB load command to the TCB control unit 113. ing. Further, the primary analysis unit 115 includes an interface for connecting the subsystem bus 109 for communicating with the communication control unit 106. Details will be described later with reference to FIG.

  The MAC control unit 118 is hardware that processes the protocol of the MAC layer 206 corresponding to the lower sublayer of the data link layer (second layer) of the OSI reference model. The communication device is connected to the Ethernet 120 by a PHY communication unit (Physical Layer Transceiver IC) 119. The PHY communication unit 119 is hardware that handles protocol processing and electrical signals of the PHY (physical) layer 207 located in the first layer of the OSI reference model. The above is a schematic description of each functional unit of the communication apparatus according to the present embodiment shown in FIG.

  Next, details of the primary analysis unit 115 will be described with reference to FIG. FIG. 3 is a diagram illustrating a specific configuration of the primary analysis unit 115. The register unit 301 constitutes an interface for connecting the subsystem bus 109. The register unit 301 has a register group for the communication control unit 106 and the primary analysis unit 115 to communicate, and in this example, is configured by three registers. The The instruction setting unit 302 is a register for the communication control unit 106 to set an instruction in the primary analysis unit 115. The status display unit 303 is a register for presenting the status of the primary analysis unit 115 to the subsystem bus 109, and the communication control unit 106 checks the status of the primary analysis unit 115 by monitoring this register. Can do. When the primary analysis unit 115 analyzes the data flowing through the snoop bus 116, the analysis result display unit 304 presents the analysis result to the subsystem bus 109.

  The state control unit 305 is responsible for overall control of the primary analysis unit 115. The main functions are control of various functional units provided in the primary analysis unit 115 and issue of a TCB load instruction to the TCB control unit 113. The bus decoding unit 306 has a function of decoding a data stream flowing through the snoop bus 116 according to a bus protocol. As a decoding result, a read data value 307 and a counter value 308 indicating how many bytes of the data value corresponding to the data flowing in the snoop bus 116 are output from the bus decoding unit 306.

  The bit determination unit 310 uses the read data value 307 and counter value 308 output from the bus decoding unit 306 to perform bit determination processing under the instruction of the state control unit 305. Then, the bit determination unit 310 outputs a determination result 309, which is a result of the bit determination process, to the state control unit 305. The socket information acquisition unit 312 has a function of acquiring target socket information from the read data value 307 and the counter value 308 under the instruction of the state control unit 305. In the present embodiment, socket pair information refers to a combination of four pieces of information: a destination IP address, a destination port number, a source IP address, and a source port number. The socket information is identification information that can identify the protocol communication channel by a combination of these four pieces of information. The socket information acquisition unit 312 has a function of returning acquired data (socket information) 311 to the state control unit 305.

  Next, the detailed operation of the primary analysis unit 115 will be described with reference to the sequence diagram of FIG. FIG. 5 is a sequence diagram showing the operation of the primary analysis unit 115 when storing the Ethernet frame in the buffer unit 111 from the MAC control unit 118 using the DMA control unit 110 together with the operations of various functional units. is there. The primary analysis unit 115 is intended to finally acquire socket information. In the present embodiment, a case where an Ethernet frame is received when TCP / IP (v4) shown in FIG. 4 is used will be described as an example.

  First, the communication control unit 106 performs initial setting of the primary analysis unit 115. Specifically, a process for writing an instruction necessary for initial setting to the instruction setting unit 302 of the register unit 301 of the primary analysis unit 115 is performed (501). As the initial setting value, there is offset information such as how many bytes the analysis starts for the data fetched from the snoop bus 116. For example, in one embodiment, the Ethernet frame header 430 is all processed by the MAC control unit 118, and the IP header position is different when it is not included in the data fetched by the DMA control unit 110 from the MAC control unit 118. Although the detailed description is omitted, the length of the header portion is different between the Ethernet frame and the IEEE 802.3 format. In order to be applicable even in such a case, it is desirable that the analysis position can be arbitrarily set.

  When the initialization command is written in the command set unit 302 of the register unit 301 (502), the state control unit 305 receives the written value (503), and sends it to the socket information acquisition unit 312 and the bit determination unit 310. An initialization command is transmitted (504). Receiving the initialization instruction, each processing unit and the state control unit 305 execute initialization (505, 506, 507). In this state, TCP / IP communication processing is started, and when the MAC control unit 118 receives a packet from the Ethernet 120 via the PHY communication unit 119 (508), it notifies the communication control unit 106 of reception (509). ). Receiving the reception notification, the communication control unit 106 makes an analysis start request to the primary analysis unit 115 before starting to receive the received data by DMA (510). When an analysis start request is written in the instruction set unit 302 of the register unit 301 (511), the state control unit 305 receives the analysis start command (512) and transmits a bus decode start command to the bus decode unit 306 (513). Thereby, the bus decoding unit 306 performs a decoding process on the data from the snoop bus 116 and a process for starting output of the read data value 307 and the counter value 308 (514). In addition, the state control unit 305 that has received the analysis request sends a status notification to the register unit 301 in order to notify the other function units that the primary analysis unit 115 is in the analysis process (515). Upon receiving the notification, the register unit 301 registers the state in the state display unit 303, which is a register that can be read from other functional units (516).

  The communication control unit 106 which has issued an analysis start request to the primary analysis unit 115 next transmits a DMA transfer start request to the DMA control unit 110 (517). Upon receipt of the request, the DMA control unit 110 analyzes the request (518), and if the MAC control unit 118 finds that the packet reception processing is to the buffer unit 111, the DMA transfer to the MAC control unit 118 is performed. An execution command is transmitted (519). As a result, DMA transfer is performed between the MAC control unit 118 and the DMA control unit 110 (520). The DMA transfer is repeatedly performed until the buffer in the MAC control unit 118 becomes empty.

  When the DMA transfer is started, data is also generated on the snoop bus 116. Data input from the snoop bus 116 to the bus decoding unit 306 is decoded into a read data value 307 and a counter value 308. Then, the read data value 307 and the counter value 308 are output to the bit determination unit 310 and the socket information acquisition unit 312 (521). Based on the initialization information, the bit determination unit 310 first determines that the read data value 307 input at that time is the Ethernet frame header portion when the counter value 308 reaches a predetermined value, and analyzes it. (522). Specifically, the bit determination unit 310 receives the type 403 and analyzes whether the arrival packet is an IP packet. In the IP packet, 0x0800 (IPv4) and 0x86DD (IPv6) are described in type 403. The bit determination unit 310 determines whether the type 403 has the above value. If the type 403 has the above value, that is, if it is an IP packet, the processing is continued. On the other hand, if the type 403 is not the above value, that is, it is other than an IP packet, processing such as notifying the state control unit 305 may be performed in order to interrupt the subsequent processing.

  Next, the bit determination unit 310 determines the IP header (431) (523). In other words, the primary analysis unit 115 is intended to finally acquire socket information, and performs bit determination necessary for this also on the IP header (431). Specifically, in the case of the present embodiment (IPv4), the bit determination unit 310 determines the version 404, the IHL (Internet Header Length) 405, and the protocol 412. By determining these, the location of the source IP address (414), destination IP address (415), source port number (417), and destination port number (418) required for the socket information can be specified. It becomes. Then, the bit determination unit 310 notifies the determination result to the state control unit 305 (524). The state control unit 305 that has received the determination result from the bit determination unit 310 calculates the number of bytes of the data packet in the socket information, and notifies the socket information acquisition unit 312 of the position of the socket information obtained by the calculation. (525). Upon receiving the notification, the socket information acquisition unit 312 determines the acquisition timing based on the received information, and reads the read data value 307 as socket information when the counter value 308 reaches a predetermined value ( 526). Then, the socket information acquisition unit 312 outputs the acquisition data 311 to the state control unit 305 (527).

  The state control unit 305 that has received the socket information has already completed the analysis as the primary analysis unit 115, and therefore issues a processing end instruction to the bus decoding unit 306, the bit determination unit 310, and the socket information acquisition unit 312. (528). The bus decoding unit 306, the bit determination unit 310, and the socket information acquisition unit 312 that have received the processing end command end the processing (529, 530, 531), respectively. The state control unit 305 that has received the socket information transmits a TCB load command to the TCB control unit 113 as a control signal 117 together with the received socket information (532). Since all the processes of the primary analysis unit 115 have been completed by the above processing, the state control unit 305 transmits an end state display and an analysis result display command to the register unit 301 with the analysis result (533). ). Upon receiving the instruction, the register unit 301 holds the end bit and the value of the analysis result in the analysis result display unit 304 (534). The above is a detailed description of the processing sequence of the primary analysis unit 115.

  Next, the TCB control unit 113 will be described in detail with reference to FIG. FIG. 6 is a diagram illustrating a specific configuration of the TCB control unit 113. The TCB control unit 113 has a function that allows the TCB cache unit 114 to store and manage context information used when processing each communication protocol as a TCB, and to allow the communication control unit 106 to refer to and rewrite the TCB in a short time. . Specifically, the TCB control unit 113 has the following configuration.

  The state control unit 601 analyzes various commands from each functional unit in the TOE subsystem 105 and comprehensively controls each unit in the TCB control unit 113. The main control unit 101 loads microcode into the instruction cache memory of the state control unit 601, and the state control unit 601 executes this, thereby realizing processing operations shown in FIGS. 9 and 10 described later. In the present embodiment, the state control unit 601 is a programmable sequencer, but may be configured by hardware of a finite state machine (dedicated machine) for higher speed.

  The address decoding unit 602 is connected to the subsystem bus 109 and decodes an address in response to an access request from each functional unit in the TOE subsystem 105. Based on the decoding result, whether the access request destination is the register unit 605 or the TCB cache unit 114 is distributed, and an access request is issued to the register unit 605 or the cache memory control unit 610. Further, the address decoding unit 602 returns the response result to the access request received from the register unit 605 or the cache memory control unit 610 to the subsystem bus 109. The address decoding unit 602 and the cache memory control unit 610 are connected by a data bus 604.

  The register unit 605 is used for communication between the TCB control unit 113 and each functional unit in the TOE subsystem 105. For this purpose, the register unit 605 includes a register that indicates the state of the TCB control unit 113 and a register for each function unit to set an instruction to the TCB control unit 113. Furthermore, the register unit 605 includes a control signal 117 for directly receiving a command from the primary analysis unit 115 and an interface for receiving a command from each functional unit in the TOE subsystem 105 via the subsystem bus 109. . In the present embodiment, the analysis result from the primary analysis unit 115 is received via the control signal 117, but a configuration that can be performed via the subsystem bus 109 is also possible.

  The DMA transfer unit 606 controls TCB transfer between the TCB cache unit 114 and the main storage unit 104 via the system bus 102, subsystem buses 109 and 607, and the data bus 608 in accordance with an instruction from the state control unit 601. To do. An LRU (Least Recently Used) table 609 is connected to the state control unit 601, and manages which cache block is used, which cache block is prohibited from saving a TCB, and the like. It is a table.

  The cache memory control unit 610 is connected to the address decoding unit 602 or the DMA transfer unit 606 via the data buses 604 and 608. Further, the cache memory control unit 610 controls writing and reading of various data with respect to the TCB cache unit 114 based on an instruction from the address decoding unit 602 or the DMA transfer unit 606. The TCB management unit 612 uses the TCB management table 611 to manage the socket number, such as whether or not the socket number is used. In this embodiment, socket numbers 1 to 512 are prepared, and in order to identify socket information, a number is uniquely assigned to one socket information from these numbers.

  The CAM control unit 613 writes and reads various data with respect to the first CAM 614 and the second CAM 615 in accordance with an instruction from the state control unit 601. In the first CAM 614, socket information and a socket number are registered in association with each other. The first CAM 614 is used to search for a socket number to identify a TCB using the socket information extracted from the received frame by the TCB control unit 113 as a search key. In the second CAM 615, a socket number and a cache block number are registered in association with each other. The second CAM 615 is used to determine whether or not a TCB having a desired socket number exists on the TCB cache unit 114 using the socket number as a search key. The above is the description regarding the detailed configuration of the TCB control unit 113.

  The TCB includes a window size related to transmission / reception, a sliding window control such as an Ack number, and pointer information of a buffer used for transmission / reception. Details are described in Non-Patent Document 1. In addition, if there is a parameter that can be added to the TCB in TCP / IP communication to improve the processing speed, ease, and convenience, it may be incorporated into the TCB.

  Next, an example of a TCB in which the TCB is stored in the main storage unit 104 or the TCB cache unit 114 will be described with reference to FIG. FIG. 7 is a diagram illustrating an example of a storage method of each element of the TCB. In the TCP / IP protocol processing, the TCB 701 is classified into sizes 702 to 706 for each element such as a 1-bit element 702, a 4-bit element 703, an 8-bit element 704,. Are arranged and arranged. Each element included in the TCB is normalized to fit one of these sizes. The size separation is logically set by firmware used by the communication control unit 106 and the state control unit 601, and is not physical. Accordingly, it is possible to redefine the size, the number of elements, and the order of elements by describing a header file that summarizes various parameters given when compiling the firmware.

  Next, a method for managing the TCB stored in the memory (the main storage unit 104 or the TCB cache unit 114) as described above will be described with reference to FIG. FIG. 8 is a diagram for explaining a TCB management method. As shown in FIG. 8A, the TCB groups 802 to 804 arranged in the main storage unit 104 are arranged and stored in the order of socket numbers (socket numbers 1 to 3) identifying TCP connections. . In the main storage unit 104, it is necessary to secure a TCB area 801 corresponding to the maximum number of connections guaranteed in advance in the communication device. FIG. 8A shows an example in which the maximum number of connections is 512.

  An example of the arrangement state of each TCB on the TCB cache unit 114 is shown in FIG. In this example, it is assumed that an area is secured so that a maximum of eight TCBs can be loaded. The TCBs 806, 807, and 808 to be accessed are duplicated in the TCB area 805 on the TCB cache unit 114. The TCB transfer between the TCB cache unit 114 and the main storage unit 104 is performed by the DMA transfer unit 606 according to an instruction from the state control unit 601.

  The number of TCBs that can be placed in the TCB cache unit 114 is restricted by cost or the like, but it is desirable that a plurality of TCBs can be placed. There are two reasons. The first is to make it possible to load a TCB necessary for new packet processing even when the reception processing using the TCB has not yet been completed for a packet that has arrived at the time of steep multiple packet reception. It is. Second, when the communication control unit 106 is configured with, for example, multi-threads or multi-cores and performs reception processing on a plurality of packets at the same time, it does not frequently cause TCB loading and flush processing. It is. Since loading and flushing of TCB are time consuming processes, it is desirable not to generate them as much as possible.

  In the TCB management table 611, as shown in FIG. 8C, a bitmap 809 indicating the presence / absence of a TCB in the corresponding cache block number is formed for each cache block number and is managed by the TCB management unit 612. Yes. Further, as shown in FIG. 8D, a bit map 810 indicating whether or not the socket number is used is formed in the TCB management table 611 and managed by the TCB management unit 612 in the same manner as the cache block.

  Next, the detailed operation of the TCB control unit 113 will be described with reference to the flowcharts of FIGS. FIG. 9 is a flowchart showing a process flow of the TCB control unit 113 at the time of TCB generation. The TCB is generated when a TCP connection is established. This TCP connection establishment is performed by the communication control unit 106, and the communication control unit 106 obtains socket information from information exchanged when the connection is established. Socket information is included in the TCP header of each TCP packet transmitted and received when establishing a connection. First, in step S901, the state control unit 601 is instructed by the communication control unit 106 to generate a TCB. At that time, socket information or the like is transmitted from the communication control unit 106 to the state control unit 601. The state control unit 601 starts creating the TCB based on the socket information and the like.

  In step S <b> 902, the state control unit 601 obtains an unused socket number from the TCB management unit 612. The TCB management unit 612 manages unused socket numbers, and issues unused socket numbers in response to acquisition requests from the state control unit 601. When the status control unit 601 requests an unused socket number, the TCB management unit 612 reports the unused socket number obtained by searching the bitmap 810 in advance. In addition to this report, the TCB management table 611 is updated and the next unused socket number search is naturally performed in the background operation. When the connection is released, the used socket number is returned from the communication control unit 106 to the TCB management unit 612 via the state control unit 601. When the used socket number is returned, the TCB management unit 612 updates the TCB management table 611 shown in FIG. 8D and searches for an unused socket number.

  In step S903, if the acquisition of the socket number is successful, the process proceeds to step S904, and if it is unsuccessful, the process proceeds to step S913. In step S904, socket information is registered in the first CAM 614. That is, the state control unit 601 instructs the CAM control unit 613 to associate and register the socket number obtained in step S902 and the socket information obtained in step S901. Upon receiving the instruction, the CAM control unit 613 registers the socket information in the first CAM 614 as an address. In step S <b> 905, the state control unit 601 acquires an unused cache block number from the TCB management unit 612 in preparation for generating a TCB on the TCB cache unit 114. In step S906, if there is no unused cache block and acquisition of the cache block number fails, the process proceeds to step S907, and if successful, the process proceeds to step S909.

  Steps S907 and S908 are processing for creating an empty area by saving the old TCB in the main storage unit 104 in order to generate a new TCB on the TCB cache unit 114. Specifically, in step S907, the state control unit 601 acquires a socket number corresponding to the cache block number with the oldest access from the LRU table 609. Next, in step S908, the DMA transfer unit 606 is located on the TCB cache unit 114, and the TCB of the oldest cache block number specified in step S907 is the position of the corresponding socket number on the main storage unit 104. Evacuate to. Thereby, an empty area is generated in the TCB cache unit 114.

  In step S909, the state control unit 601 instructs the CAM control unit 613 to register the socket number associated with the TCB to be created (obtained in step S902) in the free cache block. In response to this, the CAM control unit 613 registers the socket number in the address of the cache block number of the second CAM 615. In step S910, the state control unit 601 instructs the cache memory control unit 610 to initialize the cache block having the cache block number for creating a new TCB on the TCB cache unit 114. In response to this, the cache memory control unit 610 initializes the cache block. In step S911, the state control unit 601 updates the LRU table 609 according to the initialization of the cache block. In step S912, the state control unit 601 registers the cache block number in the cache memory control unit 610 together with the socket number acquired in step S902. At this time, the state control unit 601 registers the generated TCB at the address of the cache block number in the TCB cache unit 114.

  On the other hand, if the unused socket number managed by the TCB management unit 612 is exhausted in step S903 and the acquisition of the socket number fails, the process proceeds to step S913. In step S913, the TCB management unit 612 notifies the requesting communication control unit 106 to that effect, and as a result, the connection establishment is unsuccessful. The reason for the exhaustion of unused socket numbers managed by the TCB management unit 612 occurs because the total number of socket numbers is arranged in a fixedly prepared area on the main storage unit 104. This is because of the restrictions. In general, connection establishment is a negotiation and may not be established. The above is a detailed description of the operation of the TCB control unit 113 when a TCB generation request is received.

  FIG. 10 is a flowchart showing the operation of the TCB control unit 113 when a TCB load request is made to the TCB control unit 113 from the communication control unit 106 or the primary analysis unit 115. The reason why the TCB load request can be issued from the communication control unit 106 and the primary analysis unit 115 as in the present embodiment is as follows. For example, at the time of TCP / IP packet transmission processing, the communication control unit 106 loads a TCB necessary for packet generation. When a TCP / IP packet is received, the primary analysis unit 115 issues a TCB load request at a timing earlier than that performed by the communication control unit 106. As a result, the communication control unit 106 can make an arbitrary TCB load request at an arbitrary timing, and the primary analysis unit 115 can make a TCB load request at the time of packet reception. TCB loading can be realized.

  The load request is made in a state where socket information or a socket number corresponding to the TCB to be loaded is given. First, the processing sequence is divided on the condition that the socket number corresponding to the TCB to be loaded is assigned to the load request. If the socket number has been assigned, the process proceeds to step S1005. If the socket number has not been assigned, the process proceeds to step S1002. In step S <b> 1002, the state control unit 601 acquires socket information from the communication control unit 106 or the primary analysis unit 115. Next, in step S1003, the state control unit 601 obtains a socket number from the first CAM 614 using the socket information acquired in step S1002 as a search key. This process is performed via the CAM control unit 613. If the connection has been established, the target socket information in the process of generating the TCB is registered. In step S1004, the state control unit 601 determines whether the socket number has been successfully acquired. If the socket number can be acquired, the process proceeds to step S1005. If acquisition of the socket number has failed, the process proceeds to step S1015. In step S1015, the state control unit 601 notifies the communication control unit 106 of a socket number acquisition failure. The socket number acquisition failure indicates that when the received packet is a TCP packet, it is either a SYN packet exchanged at the time of establishing a connection or an illegal packet. Thereafter, the communication control unit 106 that has received this notification takes over how to process this packet.

  In step S1005, the state control unit 601 acquires a cache block number from the second CAM 615 using the socket number as a search key. This process is performed via the CAM control unit 613. In step S1006, the state control unit 601 determines whether the cache block number has been successfully acquired. If the acquisition of the cache block number is successful, the process proceeds to step S1014. If the acquisition is unsuccessful, the process proceeds to step S1007. The acquisition failure indicates that the TCB cache unit 114 has no TCB to be loaded. In step S1007, the state control unit 601 tries to acquire an unused cache block number from the TCB management unit 612. In step S1008, the state control unit 601 determines whether acquisition of an unused cache block number has succeeded. If there is no unused cache block and acquisition of the cache block number fails, the process proceeds to step S1009, and if successful, the process proceeds to step S1011. In step S1009, the state control unit 601 acquires a socket number corresponding to the oldest cache block number of the last access from the LRU table 609. In step S1010, the DMA transfer unit 606 is on the TCB cache unit 114, and saves the TCB having the cache block number specified in step S1009 to a corresponding position on the main storage unit 104.

  By the processing so far, the state control unit 601 can acquire the cache block number, and in step S1011, the state control unit 601 registers the socket number in the acquired cache block number. In step S1012, the DMA transfer unit 606 obtains the TCB of the socket number that is on the main storage unit 104 and is loaded, and reads it into the cache block of the predetermined cache block number on the TCB cache unit 114. In step S1013, the state control unit 601 updates the LRU table 609. In step S1014, the state control unit 601 registers the cache block number storing the acquired TCB together with the socket number in the cache memory control unit 610. The above is a detailed description of the operation of the TCB control unit 113 when a TCB load request is received.

  Next, how the TOE subsystem 105 including the primary analysis unit 115 and the TCB control unit 113 operates will be described with reference to the sequence diagram of FIG. FIG. 11 is an operation sequence diagram of the TOE subsystem according to the first embodiment. Note that items such as resetting and initial setting of various functional units are not necessarily related to the features in the present embodiment, and thus description thereof is omitted. First, the MAC control unit 118 receives a TCP / IP packet from the PHY communication unit 119 (1101 (same as 508)). The MAC control unit 118 notifies the communication control unit 106 that a packet has been received (similar to 1102 and 509). Upon receiving the notification, the communication control unit 106 transmits an analysis request to the primary analysis unit, and also transmits a DMA transfer start request to the DMA control unit 110 (1103 and 1104 (same as 510 and 515, respectively)). ).

  The DMA control unit 110 analyzes the request (1105 (same as 516)) and transmits a DMA transfer execution command to the MAC control unit 118 (1106 (same as 517)). DMA transfer (1107 (similar to 518)) between the MAC control unit 118 and the DMA control unit 110 is started. During DMA transfer (1107), the primary analysis unit 115 performs analysis processing of TCP / IP packets flowing through the snoop bus 116 and acquisition processing of packet socket information (1108 (corresponding to 520 to 529)). After acquiring the socket information, the socket information is added and a TCB load command is transmitted to the TCB control unit 113 (1108 (similar to 530)). Then, the primary analysis unit 115 performs an analysis end process (1109 (corresponding to 531/532)). The TCB control unit 113 that has received the TCB load command starts the TCB load process (1110, corresponding to S1001 to S1011). When the completion of the TCB loading process (1112 (corresponding to S1012)) is detected (1113), the termination process is performed (1114 (corresponding to S1013 / S1014)).

  On the other hand, the DMA control unit 110 that has detected the completion of DMA (1115) sends a DMA transfer completion notification to the communication control unit 106 (1116). Receiving the DMA completion notification, the communication control unit 106 accesses the analysis result display unit 304 of the primary analysis unit 115 and acquires the analysis result of the received packet (1117). The communication control unit 106 transmits a necessary TCB load command to the TCB control unit 113 in order to perform reception processing in the IP layer (204) and the TCP / UDP layer (203) based on the analysis result. (1118). The TCB control unit 113 that has received the TCB load instruction performs TCB load processing, but since the TCB has already been loaded in the TCB cache unit 114 (corresponding to Yes in S1006), the TCB load is completed immediately. A notification is transmitted (1119). The communication control unit 106 that has confirmed the TCB load starts reception processing in the IP layer (204) and the TCP / UDP layer (203) (1120).

  It should be noted in FIG. 11 that the primary analysis unit 115 can transmit the TCB load command to the TCB control unit 113 even during the DMA transfer (1107). As a result, there is a time required for the TCB load, and the communication control unit 106 actually reduces the time that the communication control unit 106 waits for the TCP / IP packet reception process until the load is completed. Start (1120) is possible. Further, the following effects can be given as effects associated with the present embodiment. In other words, the TCP / IP packet analysis processing that is necessary for the processing of the communication control unit 106 in the prior art is that the processing of the communication control unit 106 can be simplified by the primary analysis unit 115 taking over. This effect can also be cited as an effect of improving the reception performance of TCP / IP packets.

  In FIG. 11, the processing (1117) is started after the communication control unit 106 receives the DMA transfer completion notification (1116). In other words, if there are the IP (v4) header (431) and the TCP header (432), the communication control unit 106 determines the IP layer (204) and TCP / UDP layer (203) of the TCB load command, etc. The reception process can be started. Accordingly, at a timing earlier than the DMA transfer completion notification (1116), for example, the DMA control unit 110 transmits a reception start ready notification to the communication control unit 106, and the communication control unit 106 performs processing (1117). You may start. The above is the description of FIG.

  As described above, according to the present embodiment, it is possible to reduce the waiting time required for completion of loading from a TCB load request, and it is possible to speed up the reception process.

[Second Embodiment]
FIG. 12 is a block diagram of a communication device (FIG. 12A) and a primary analysis unit (FIG. 12B) according to the second embodiment. The difference from the first embodiment is in the function of the primary analysis unit in the TOE subsystem. The primary analysis unit 1202 includes an interface to which the second snoop bus 1204 can be connected so that the subsystem bus 109 can be snooped. The first snoop bus 1203 is the same as the snoop bus 116. Further, the primary analysis unit 1202 is provided with a bus decoding unit 1205 so that the subsystem bus 109 can be decoded. As a result, the TCP / IP packet flowing through the subsystem bus 109 can be analyzed and the socket information necessary for loading the TCB can be extracted.

  This function is suitable for receiving an encrypted packet. For example, in a packet to which IPSec is applied, socket information is encrypted, and the primary analysis unit 1202 may not be able to extract socket information from data flowing through the first snoop bus 1203. In this case, by applying this embodiment, the primary analysis unit 1202 can analyze at the timing when the encrypted decoded packet output from the encryption processing unit 107 flows through the subsystem bus 109. As a result, the TCB can be loaded at an earlier timing than the communication control unit 106 starts the reception process as TCP. Even if a special header prepared for encryption is attached to the packet, analysis can be performed by adding offset information to the analysis start point as described above.

  Note that the primary analysis unit 1202 has been described that the socket information cannot be extracted from the snoop bus 116, but it is possible to analyze whether or not the received packet is encrypted. Although detailed description is omitted, it is apparent that the configuration of the primary analysis unit 1202 according to the present embodiment can be realized by starting the analysis from the type 403 of the Ethernet frame header 430. The analysis result can be used by the communication control unit 106 to determine the next processing of the packet fetched from the MAC control unit 118, thereby reducing the processing of the communication control unit 106 and speeding up the reception process. it can. The primary analysis unit 1202 may be configured to switch and receive and analyze data flowing through the first snoop bus 1203 and the second snoop bus 1204.

  As described above, according to the present embodiment, even when an encrypted packet is received, it is possible to reduce the waiting time for load completion from a TCB load request, and to increase the speed of reception processing. Become.

[Other Embodiments]
In the above, the form which can utilize this invention was demonstrated using two embodiment. In the present embodiment, the TCP protocol has been described as a representative example of protocol processing, but other protocols can also be handled. For example, in SSL / TLS protocol communication, a plurality of sessions are identified, and a TCB for each session is required. Here, the session corresponds to a TCP communication connection. SSL / TLS is a security communication protocol on the socket API 202 and is processed in the communication control unit 106. SSL / TLS session identification is performed by the communication control unit 106, and the corresponding TCB is read into the TCB cache unit 114. In addition, an area for an SSL / TLS session is secured in the TCB cache unit 114, the TCB management table 611, and the LRU table 609. This can be realized by preparing a dedicated cache block, an LRU table, and a bitmap table for managing the cache block number.

  In the present invention, Ethernet is used as an example, but it goes without saying that the present invention can also be applied to a wireless LAN. In that case, when capturing a received packet to which a wireless MAC header or the like is added, the IP header and TCP header are analyzed in the same manner as in the above-described embodiment by appropriately applying the offset information described above. Is possible. Further, when packets are concatenated and received, the break of the packet can be detected by including the total packet length in the analysis. As a result, even in a concatenated packet, TCB loading can be performed for each packet.

  The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

Claims (10)

A communication device that performs communication based on a communication protocol,
Communication means for controlling communication based on a predetermined communication protocol;
Control means for controlling context information used for processing of the predetermined communication protocol;
Analyzing means for analyzing data transferred by an internal bus connecting various functional units in the communication device;
Comprising
The control unit loads the context information corresponding to the result analyzed by the analysis unit into a cache memory accessed by the communication unit.   2. The communication apparatus according to claim 1, wherein the control unit loads the context information into the cache memory based on a load request transmitted from the communication unit or the analysis unit.   The communication apparatus according to claim 2, wherein the analysis unit sends the load request to the control unit at a timing when the socket information is acquired by analyzing the data.   4. The communication apparatus according to claim 2, wherein the control unit creates the context information based on socket information acquired from the communication unit and a socket number managed by the control unit.   5. The communication apparatus according to claim 4, wherein the control unit manages the created context information in association with the socket information, the socket number, and an unused cache block number.   The control means, based on the load request transmitted from the communication means or the analysis means, the context information corresponding to the socket number and the cache block number related to the socket information acquired from the communication means or the analysis means, The communication device according to claim 5, wherein the communication device is loaded into a cache memory.   The communication device according to claim 1, wherein the analysis unit analyzes data received by the communication device and transferred to a main storage unit. Further comprising cryptographic processing means,
The communication apparatus according to claim 1, wherein the analysis unit analyzes the decoded data of the encryption by the encryption processing unit. A method for controlling a communication device that performs communication based on a communication protocol,
A communication process for controlling communication based on a predetermined communication protocol;
A control step of controlling context information used for processing of the predetermined communication protocol;
An analysis step of analyzing data transferred by an internal bus connecting the various functional units in the communication device;
Have
In the control step, context information corresponding to a result analyzed in the analysis step is loaded into a cache memory accessed in the communication step.   The program for making a computer perform each process of the control method described in Claim 9.

Images