Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberCN1698337 B
Publication typeGrant
Application numberCN 200480000656
PCT numberPCT/US2004/017655
Publication date30 Oct 2013
Filing date4 Jun 2004
Priority date5 Jun 2003
Also published asCN1698337A, DE602004030562D1, EP1629656A1, EP1629656B1, US7363572, US7412488, US7609696, US7613109, US7991918, US8417852, US20040246974, US20040249881, US20040249998, US20040257986, US20040258075, US20040258076, WO2004110025A1
Publication number200480000656.8, CN 1698337 B, CN 1698337B, CN 200480000656, CN-B-1698337, CN1698337 B, CN1698337B, CN200480000656, CN200480000656.8, PCT/2004/17655, PCT/US/2004/017655, PCT/US/2004/17655, PCT/US/4/017655, PCT/US/4/17655, PCT/US2004/017655, PCT/US2004/17655, PCT/US2004017655, PCT/US200417655, PCT/US4/017655, PCT/US4/17655, PCT/US4017655, PCT/US417655
InventorsAk杰哈, R丹尼拉克, Pj居吉, Ta莫佛, S南达, A拉杰高帕兰, Pj赛登伯拉德
Applicant辉达公司
Export CitationBiBTeX, EndNote, RefMan
External Links: SIPO, Espacenet
Processing data for TCP connection using offload unit
CN 1698337 B
Abstract  translated from Chinese
本发明是一种处理在TCP连接上接收和传送数据的方法和装置。 The present invention is a method and apparatus for receiving and transmitting data over a TCP connection processing. 卸载单元处理接收到不存在特殊事例的数据,以便产生被直接上载到应用程序存储器的有效载荷数据。 Unloading unit to process the received data there is no special case, in order to generate the application memory directly uploaded payload data. 该卸载单元部分地处理接收到的存在特殊事例的数据并将部分处理过的接收的数据上载到存储于系统存储器内的缓冲区。 The unloading unit partially processing the received data and there is a special case of partially processed received data uploaded to the storage system to buffer memory. 该部分处理过的接收的数据然后被TCP堆栈进一步处理,以便产生被拷贝到应用程序存储器有效载荷数据。 The partially processed received data TCP stack is then further processed to produce memory is copied to the application data payload.
Claims(4)  translated from Chinese
1.一种处理用于委托TCP连接的帧的方法,包括: 接收用户缓冲区描述符,所述用户缓冲区描述符规定被分配给应用程序的存储器的第一部分中的多个用户缓冲区的位置,其中每个用户缓冲区被配置为存储用于所述委托TCP连接的处理的帧数据; 利用卸载单元处理所述帧的第一部分,以产生第一处理的帧数据,并且利用所述卸载单元处理所述帧的第二部分,以产生第二处理的帧数据; 将所述第一处理的帧数据上载至所述被分配给所述应用程序的存储器的第一部分中的用户缓冲区,并且将所述第二处理的帧数据上载至被分配给软件驱动程序的所述存储器的第二部分中的遗留缓冲区,所述软件驱动程序被配置为在所述卸载单元和TCP堆栈之间通信;以及利用在CPU上执行的所述TCP堆栈处理所述第二处理的帧数据,以产生第三处理的帧数据; 其中所述用户缓冲区描述符由所述TCP堆栈利用后接收缓冲区命令通过所述软件驱动程序发送给所述卸载单元,其中所述软件驱动程序包括所述后接收缓冲区命令内特定命令控制和/或状态信息字段内所述用户缓冲区的数量。 1. A method of processing commissioned frame TCP connection method, comprising: receiving a user buffer descriptor, described in the first part of the user buffer is allocated to meet the requirement for the application of a plurality of user memory buffers position, wherein each buffer is configured to store user frame data for processing the TCP connection delegate; handle unloading unit utilizing a first portion of the frame, to generate a first processed frame data, and the use of the unloading The second part of the frame unit processes to produce second processed frame data; on the first processing frame data carrier to the first portion of the application is assigned to the memory of the user buffer, between and will be uploaded to the memory is allocated to the software driver on the second part of the second frame of data processing in the left buffer, the software driver is configured to the unloading unit and TCP stack communication; and the use of the CPU to perform the TCP stack processing the second processed frame data to generate data for the third frame processing; and wherein the user buffer descriptor receive buffer after use by the TCP stack command sent by the software driver to the unloading unit, wherein the software drivers included within the receive buffer of the command-specific commands to control and / or the number of states in the information field of the user buffer.
2.一种处理用于委托和非委托TCP连接的帧的方法,包括: 接收用户缓冲区描述符,所述用户缓冲区描述符规定被分配给应用程序的存储器的第一部分中的多个用户缓冲区的位置,其中每个用户缓冲区被配置为存储用于所述委托TCP连接的处理的帧数据; 利用卸载单元处理委托TCP连接,所述卸载单元被配置为处理不存在特殊事例的帧,以产生处理的帧数据; 将所述处理的帧数据上载至被分配给所述应用程序的存储器的第一部分中的用户缓冲区; 利用在CPU上执行的TCP堆栈处理非委托TCP连接; 将用于所述非委托TCP连接的帧数据上载至被分配给软件驱动程序的所述存储器的第二部分中的遗留缓冲区,所述软件驱动程序被配置为在所述卸载单元和所述TCP堆栈之间通/[目;和利用在所述CPU上执行的所述TCP堆栈处理所有存在特殊事例的帧; 其中所述用户缓冲区描述符由所述TCP堆栈利用后接收缓冲区命令通过所述软件驱动程序发送给所述卸载单元,其中所述软件驱动程序包括所述后接收缓冲区命令内特定命令控制和/或状态信息字段内所述用户缓冲区的数量。 2. A process commissioned and non-commissioned frame TCP connection method, comprising: receiving a user buffer descriptor, described in the first part of the user buffer is allocated to meet the requirement for the application of the plurality of user memory location of the buffer, where each user buffer is configured to store the frame data processing of the TCP connection delegate; the use of the unloading unit commissioned TCP connection processing unit is configured to handle the particular case there is no frame of the unloading to generate a frame of data processing; the upper frame of the processed data is uploaded to the application assigned to the memory of the first part of the user buffer; the use of TCP stack on the CPU processing non-commissioned TCP connection; the for the non-commissioned uploaded to the memory is allocated to the software driver for the second part of the legacy frame data buffer on TCP connections, the software driver is configured to the unloading unit and the TCP pass between stacks / [mesh; and using the TCP stack executing on the CPU handles all there is a special case of the frame; wherein the user buffer descriptor from the receive buffer after the TCP stack through the use of command said software driver sends to the unloading unit, wherein the software driver is included within the receive buffer the command specific command control and / or the number of states in the information field of the user buffer.
3.一种设立委托连接的方法,该方法包括: 指定第一计算系统内的系统存储器的第一部分,用于在遗留缓冲区中存储帧有效载荷数据,其中所述系统存储器的所述第一部分被分配给软件驱动程序,所述软件驱动程序被配置为在专用硬件卸载单元和TCP堆栈之间通信; 指定所述系统存储器的第二部分,用于在用户缓冲区中存储帧有效载荷数据,其中所述系统存储器的所述第二部分被分配给应用程序; 在所述第一计算系统与第二计算系统之间建立TCP连接; 确定是否将所述TCP连接委托所述专用硬件卸载单元处理,以从所述第一计算系统的CPU卸载所述TCP连接处理; 在确定委托所述TCP连接时,在委托连接表中设置项; 接收用于所述委托连接的帧,并且确定用户缓冲区可用于存储帧有效载荷数据;和将所述帧有效载荷数据上载至所述系统存储器的所述第二部分中的在用户缓冲区信息中规定的位置。 3. A method delegate connection establishment, the method comprising: designating a first portion of the system memory within a first computing system for the legacy frame buffer storing the payload data, wherein said first portion of said system memory is assigned to the software driver, the software driver is configured to communicate between the unit and the TCP stack unloading in dedicated hardware; specifies the second part of the system memory for storing the frame payload data in the user buffer, wherein the second portion of the system memory is allocated to the application; establish a TCP connection between the first and second computing system computing system; determining whether the TCP connection entrust the dedicated hardware offload unit processes to unload the CPU from the first TCP connection processing computing systems; determining commissioned the TCP connection, the connection settings in the delegate list; receiving a frame connected to the commission, and to determine the user buffer can be used to store the payload data frame; and the position of the load to the system memory on the frame payload data in the second part of the provision of information in the user's buffer.
4.一种利用卸载单元处理入站和出站帧的方法,所述卸载单元具有委托连接表,所述方法包括: 初始化带有连接状态数据的所述委托连接表中的项,所述连接状态数据包括连接表索弓丨、序列号和确认号,所述连接表索引对应于由TCP堆栈选择用于由所述卸载单元处理的连接,所述TCP堆栈由主CPU执行; 接收具有用于从所述TCP堆栈传送的原型报头和数据的出站帧; 将所述出站帧中的数据分割成多个数据片段;以及对于每个数据片段,计算TCP检查和,确定传送序列号和传送确认号,更新存储于所述委托连接表中的序列号和确认号,并且输出包括所述TCP检查和、所述传送序列号、所述传送确认号、TCP时戳和用于在所述选择的连接上进行数据传送的所述数据片段的帧,其中,如果从所述TCP堆栈接收到的TCP时戳大于存储于所述委托连接中的TCP时戳,则所述TCP时戳等于所述从所述TCP堆栈接收到的TCP时戳。 An advantage of inbound and outbound frames methods unloading unit, the unloading unit has commissioned the connection table, the method comprising: initialization with the delegation connector connection status items in the table of data, the connection State data includes a connection table index 丨 bow, serial number and confirmation number, the connection table index corresponds to the TCP stack is selected for the connecting unit of the unloading process, the TCP stack is executed by the main CPU; receiving a used outbound frames transmitted from the TCP stack prototype header and data; the data station the frame is divided into a plurality of data segments; and for each piece of data, calculate TCP checksum to determine the transfer sequence number and transmission confirmation number, the update is stored in the connection table commissioned serial number and confirmation number, and output including the TCP checksum and the serial number of the transfer, the transfer confirmation number, TCP timestamp and for the selection data transfer frame connections on the data segment, which is greater than if the stamp is received from the TCP stack to the TCP stored at the time of the commission stamp TCP connections, when the TCP timestamp is equal to the received from the TCP stack to the TCP timestamp.
Description  translated from Chinese

利用卸载单元处理TCP连接数据的方法 Unloading unit utilizing TCP connection data processing method

[0001] 相关申请的交叉引用 Cross [0001] REFERENCE TO RELATED APPLICATIONS

[0002] 本申请主张2003年6月5日提交的名称为“硬件优化TCP/IP(H0T)处理”的美国临时专利申请号60/476,570的优先权,该申请与本申请为共同拥有的正在同时审查,并与本申请的发明人与受让人相同,其全部内容作为参考在此被纳入本申请。 [0002] This application claims the name of June 5, 2003 to submit the "Optimizing hardware TCP / IP (H0T) process" as a priority to U.S. Provisional Patent Application No. 60 / 476,570, which application is commonly owned with the present application It is simultaneously reviewed and the inventor and assignee of the present application is the same, the entire contents of which are hereby incorporated by reference herein.

技术领域 Technical Field

[0003] 本发明的一个或几个方面总体上涉及传输控制协议(TCP)处理,特别是涉及对基于TCP的通信的优化。 [0003] one or more aspect of the invention relates generally to the Transmission Control Protocol (TCP) process, particularly to the optimization of TCP-based communication.

背景技术 Background

[0004] 现有的TCP处理的例子是为了加速客户与服务器之间的数据传输而开发的各种系统和方法。 [0004] Examples of conventional TCP processing is to accelerate the transfer of data between the client and the server and the development of various systems and methods. 例如,在一台主处理机(如CPU)上执行的软件实现和用来从主处理机上卸载TCP处理的昂贵的专用硬件实现相比,成本低但速度慢。 For example, the software on a single host processor (such as CPU) on the implementation of the TCP implementation and to unload expensive, dedicated hardware processing from the main processor to achieve compared to the low cost but slower.

[0005]图1是现有技术的计算系统100的一个实施例的方块简图,计算系统100包括CPUllO和网络接口卡(NIC)150。 [0005] FIG. 1 is a block diagram of an embodiment 100 of a prior art computing system, computing system 100 includes a network interface card and CPUllO (NIC) 150. 计算系统100可以是台式电脑、服务器、手提电脑、掌上电脑、平板电脑、游戏控制台、移动电话、基于计算机的模拟器等等。 The computing system 100 may be a desktop computer, servers, notebooks, PDAs, tablet PCs, game consoles, mobile phones, computer-based simulator and so on. 将CPUllO连接到系统控制器120的总线112可以是前端总线(FSB)。 The CPUllO connected to the system controller 120 may be a front-side bus 112 bus (FSB). 相应地,计算系统100可以是基于网络集线器的体系结构,也被称为丨NTEL®网络集线器体系结构,其中系统控制器120是一个存储器控制器网络集线器,并且一个I/o桥140通过一集线器到集线器接口126与系统控制器120连接。 Accordingly, the computing system architecture 100 may be based on network hub, also referred 丨 NTEL® hub architecture, where the system controller 120 is a memory controller hub and an I / o bridge 140 through a hub 120 is connected to a hub interface 126 and the system controller. 系统控制器120通过存储器总线132与系统存储器130连接。 The system controller 120 is connected via a memory bus 132 and system memory 130. I/O桥140包括一个外围部件接口(PCI)总线182的控制器并且可以包括系统管理总线142和通用串行总线144等的控制器。 I / O bridge 140 includes a peripheral component interconnect (PCI) bus controller 182 and may include system management bus 142 and 144, etc. Universal Serial Bus controllers. I/O桥可以是一单一集成电路或单一半导体平台。 I / O bridge may be a single integrated circuit or single semiconductor platform. 已知的系统控制器120的例子包括丨NTEL® Northbridge。 Examples of known system controller 120 include 丨 NTEL® Northbridge. 已知的I/O桥140的例子包括INTEL® Southbridge和NVIDI 媒休和通信处理器(MCP)芯片。 Aware of the I / O bridge 140 Examples include INTEL® Southbridge and NVIDI media break and communications processors (MCP) chip.

[0006] NIC150可与一个或多个PCI设备180共享一个PCI总线182。 [0006] NIC150 with one or more PCI devices 180 to share a PCI bus 182. NIC150包括PCI接口145、专用处理机155、媒体存取控制器(MAC) 165、专用存储器160和与ETHERNET网络172接口连接的ETHERNET接口170。 NIC150 including PCI interface 145, dedicated processor 155, the media access controller (MAC) 165, ETHERNET dedicated memory interfaces 160 and 172 and ETHERNET network interface 170. NIC150的软件驱动程序119在NIC150和CPUllO上执行的应用程序117之间通信。 Communication between applications NIC150 117 software drivers 119 on NIC150 and CPUllO execution. 系统存储器130内配置有应用存储空间125、TCP堆栈存储空间145和驱动程序存储器空间135。 The system memory 130 is configured with application storage space 125, TCP stack storage memory space 145 and the driver 135.

[0007] NIC150内的专用处理机155被用于TCP处理,取代利用CPUllO执行TCP堆栈115来实施TCP处理。 Dedicated processor [0007] NIC150 within TCP processing 155 is used to replace the use CPUllO implementation of TCP stack 115 to implement TCP processing. 因此NIC150卸载CPU110,使CPUllO为其他应用程序释放出处理周期。 So NIC150 uninstall CPU110, so CPUllO release processing cycles for other applications. 同样,专用存储器160取代TCP堆栈存储空间145,释放出TCP堆栈存储空间145分配给其他应用程序。 Also, replace the TCP stack 160 dedicated memory storage space 145, the release of 145 TCP stack storage allocated to other applications. 但是,包括专用存储器160和专用处理机155在内的NIC150与CPUllO上执行的用于TCP处理的软件实现相比成本较高。 However, the software for TCP processing performed on include dedicated memory 160 and processor 155, including dedicated and CPUllO NIC150 achieve high cost compared.

[0008] 因此,有必要提供一种通过卸载主处理机的一些任务优化TCP处理并与专用硬件实现相比能够降低系统成本的部分硬件实现。 [0008] Thus, there is a need for a number of tasks from the main processor by offloading TCP optimization process with dedicated hardware can reduce the system cost compared to some of the hardware implementation. 发明内容 DISCLOSURE

[0009] 本发明方法的各种实施例包括利用卸载单元处理帧(frame)数据的第一部分以便产生第一处理帧数据,利用卸载单元处理帧的第二部分,以便产生第二处理帧数据,并利用在CPU上执行的应用程序处理第二处理帧数据以产生第三处理帧数据。 [0009] The various embodiments of the present invention comprises a method of utilizing a first portion unloading unit processing a frame (frame) data to generate a first processed frame data using the unloading unit processing the second frame portion, so as to generate the second processed frame data, and the use of the application processing executed by the CPU in the second processed frame data to produce third processed frame data.

[0010] 本发明方法的各种实施例包括利用卸载单元处理委托(delegated)TCP连接,利用CPU上执行的应用程序执行处理非委托TCP连接,并利用CPU上执行的应用程序处理所有存在特殊事例(special case)的帧。 [0010] Various methods of embodiments of the invention include the use of the unloading unit handling the entrusted (delegated) TCP connection, use an application executing on the CPU executes processing of non-commissioned TCP connection and use applications executing on the CPU handles all there is a special case (special case) frames. 卸载单元被配置为处理不存在特殊事例的帧。 Unloading unit is configured to handle special cases do not exist in the frame.

[0011] 本发明方法的各种实施例包括建立TCP连接并确定是否为硬件处理委托TCP连接。 [0011] Various methods of embodiments of the invention include the establishment of TCP connections and determine whether it is hardware-commissioned TCP connection.

[0012] 本发明方法的各种实施例包括接收委托连接表索引,接收来自TCP堆栈传送的原型报头和数据,利用委托连接表索引存取委托连接表项,基于传送数据的一部分计算TCP检查和(checksum),并输出包括TCP检查和和传送数据的该部分的帧。 [0012] The various embodiments of the inventive method includes receiving delegation connector table index, prototype header received from the TCP stack and data transmission using delegate access delegate connection connection table entry index is calculated based in part on TCP check and transmit data (checksum), and output includes a frame portion of the TCP inspection and and transmission of data.

[0013] 本发明各种实施例包括处理TCP连接数据的系统。 [0013] The various embodiments of the present invention includes a TCP connection data processing system implemented. 该系统包括TCP堆栈、软件驱动程序和卸载单元。 The system includes a TCP stack, software drivers and unloading unit. TCP堆栈被配置为处理接收到的存储于至少一个遗留缓冲区(legacybuffer)内的帧。 TCP stack is configured to process the received frame is stored in the at least one buffer (legacybuffer) within the left. 软件驱动程序被配置为接口连接TCP堆栈和卸载单元。 Software driver is configured TCP stack and unloading unit for the interface connection. 卸载单元被配置为处理委托连接上接收到的帧,以便产生有效载荷数据和部分处理的帧。 Unloading unit is configured to process the received delegation connector frame to produce a payload portion of the frame data and processing.

[0014] 本发明各种实施例包括用于TCP堆栈和卸载单元之间通信的系统。 [0014] various embodiments of the present invention comprises a system for communication between the TCP stack and the unloading unit. 该系统包括由TCP堆栈向卸载单元传送命令的装置和由卸载单元向TCP堆栈传送通知描述符的装置。 The system includes the TCP stack to send commands to the device and the unloading unit by the unloading unit to deliver the notification descriptor TCP stack devices.

[0015] 本发明涉及处理用于TCP连接的数据的新系统和新方法。 New systems and methods for data processing of TCP connections [0015] The present invention relates.

附图说明 Brief Description

[0016] 附图中显示了根据本发明的一个或更多的方面的实施例;但是,不应该认为附图是对本发明的限定,而仅是出于说明和理解本发明的目的。 [0016] shown in the drawings according to an embodiment of the present invention, one or more aspects; however, the figures should not be considered as limiting the present invention, but are for explanation and understanding only object of the present invention.

[0017] 图1是现有的包括主机计算机和网络接口卡的计算系统的一个示例性实施例的方块图。 [0017] FIG. 1 is a conventional one exemplary embodiment includes a host computer and a network interface card block diagram of a computing system.

[0018] 图2A和2B是根据本发明一个或几个方面的包括主机计算机的计算系统的示例性实施例的方块图。 [0018] FIGS. 2A and 2B according to the present invention, one or more aspects of the exemplary embodiment includes a host computer computing system block diagram.

[0019] 图3是根据本发明一个或几个方面的图2A和2B中所示的硬件优化TCP子单元的实施例的简图。 [0019] FIG. 3 is a diagram of one embodiment of the present invention or aspects of FIGS. 2A and 2B as shown in the hardware implementation of TCP optimization subunits.

[0020] 图4A是根据本发明一个或几个方面设立一个委托连接的方法的实施例的简图。 [0020] FIG. 4A is a diagram of an embodiment of an established connection delegate method according to the present invention, one or more aspects.

[0021] 图4B是根据本发明一个或几个方面接收一个帧的方法的实施例的简图。 [0021] FIG. 4B is a diagram of an embodiment of a received frame in accordance with one or several aspects of the method of the present invention.

[0022] 图4C是根据本发明一个或几个方面的慢启动序列的实施例的简图。 [0022] FIG. 4C is a diagram of an embodiment according to one or slow start sequence several aspects of the present invention.

[0023] 图5A是根据本发明一个或几个方面的在图2A和图2B中所示的应用程序存储空间中的用户缓冲区的实施例的简图。 [0023] FIG. 5A is a user or application shown in Figures 2A and 2B of storage space in several aspects of the present invention, a schematic example of implementation of the buffer.

[0024] 图5B是根据本发明一个或几个方面的用户缓冲区描述符的实施例的简图。 [0024] FIG. 5B is a diagram in accordance with one or several aspects of the present invention, a user buffer descriptor embodiment.

[0025] 图5C是根据本发明一个或几个方面的在图2A和图2B中所示软件驱动程序存储空间中的遗留缓冲区的实施例的简图。 [0025] FIG. 5C is a schematic embodiment or 2A and software drivers storage space left in the buffer shown in FIG. 2B according to several aspects of the present invention.

[0026] 图是根据本发明一个或几个方面的遗留缓冲区标签表的实施例的简图。 [0026] FIG. Is a diagram of an embodiment according to the present invention, one or more aspects of the legacy buffer tag table. [0027] 图5E是根据本发明一个或几个方面的遗留缓冲区描述符的实施例的简图。 [0027] FIG. 5E is a diagram of an embodiment according to the present invention, one or more aspects of the legacy buffer descriptor.

[0028] 图6A是根据本发明一个或几个方面的用于由应用程序向卸载单元传送命令的命令环(command ring)的简图。 [0028] FIG. 6A is a diagram according to the present invention, one or more aspects of the application for sending commands to the unloading unit command loop (command ring) of.

[0029] 图6B是根据本发明一个或几个方面的用于由卸载单元向应用程序传送连接信息的通知环(notification ring)的简图。 [0029] FIG. 6B is a diagram in accordance with one or more aspects of the present invention by the unloading unit for transmitting connection information to the application's notification ring (notification ring) of.

[0030] 图6C是根据本发明一个或几个方面的用于由应用程序向卸载单元传送接收缓冲区信息的接收描述符环(receive descriptor ring)的简图。 [0030] Figure 6C is based on one or more aspects of the present invention for receiving buffer information by the application to uninstall the unit transmits the received description diagram symbol ring (receive descriptor ring) of.

[0031] 图6D是根据本发明一个或几个方面的用于由应用程序向卸载单元传送缓冲区信息的传送描述符环(transmit descriptor ring)的简图。 [0031] FIG. 6D is a diagram describing character ring (transmit descriptor ring) in accordance with one or more aspects of the present invention is used by the application to the unloading unit transmit buffer information transmission.

[0032] 图7是根据本发明一个或几个方面的包括有图3所示硬件优化TCP子单元的一部分的方框图。 [0032] Figure 7 is one or more aspects of the present invention comprises the block diagram of TCP subunit 3 shows a part of a hardware optimization.

[0033] 图8A是根据本发明一个或几个方面的处理有效帧的方法的简图。 [0033] FIG. 8A is a diagram of the processing according to the present invention, one or a few aspects of the effective frame method.

[0034] 图SB是根据本发明一个或几个方面的处理失序帧的方法的简图。 [0034] FIG. SB is a diagram of the processing sequence frames in accordance with one or several aspects of the method of the present invention.

[0035] 图SC是根据本发明一个或几个方面的等待用户缓冲区的方法的简图。 [0035] FIG. SC is a diagram in accordance with one or wait for the user buffer method aspects of the present invention.

[0036] 图8D是根据本发明一个或几个方面的完成用户缓冲区处理的方法的简图。 [0036] FIG. 8D is a diagram of the method according to the invention one or more aspects of the completion of the processing of the user buffer.

[0037] 图9A是根据本发明一个或几个方面的确定通知的方法的简图。 [0037] FIG. 9A is a diagram in accordance with the present invention, one or several aspects of determining the notification method.

[0038] 图9B是根据本发明一个或几个方面的使用户缓冲区跟随遗留处理同步的方法步骤的简图。 [0038] FIG. 9B is a diagram of the synchronization buffer fill legacy processing method steps according to one or more aspects of the present invention allows a user.

[0039] 图10A和10B是根据本发明一个或几个方面的表示传送数据所使用的格式的简图。 [0039] FIG. 10A and 10B is a diagram showing the transmission of data in accordance with the format of one or more aspects of the present invention being used.

[0040] 图1lA和IlB是根据本发明一个或几个方面的编辑出站(outbound)帧的方法的实施例的简图。 [0040] FIG 1lA and IlB is a schematic of an embodiment of the present invention, according to one or more aspects of the editing process outbound (outbound) frame.

[0041] 图1lC是根据本发明一个或几个方面的在传送中生成的实施例的简图。 [0041] FIG 1lC is a schematic view of the present invention, one or more aspects of the transmission of the generated according to an embodiment.

具体实施方式 DETAILED DESCRIPTION

[0042] 为了更好地理解本发明,在以下的说明中对特定的细节进行了描述。 [0042] For a better understanding of the present invention, in the following description of the specific details described. 但是,对于本专业的技术人员来说,即使没有这些特定的细节仍然可以实现本发明。 However, those skilled in the art, even without these specific details are still present invention may be implemented. 此外,为了避免使本发明含混不清,对于一些公知的特点不再说明。 In addition, in order to avoid ambiguity of the present invention, some well-known features will not be explained.

[0043] 图2A和图2B是本发明计算系统200的典型实施例的方框图,该计算系统200包括CPUllO和硬件优化TCP(HOT)单元250。 [0043] Figures 2A and 2B are a block diagram of a typical computing system 200 of the present embodiment of the invention, the computing system 200 includes a hardware optimization CPUllO and TCP (HOT) unit 250. 在图2A中,卸载单元HOT单元250从CPU110卸载部分的TCP处理。 In Fig. 2A, unloading unit HOT Unit 250 offloads TCP processing section from the CPU110. CPUllO执行TCP堆栈215,该TCP堆栈215包括用来完成至少部分TCP处理的代码,特别是HOT单元250没有执行的TCP处理。 CPUllO implementation of TCP stack 215, which is used to complete the TCP stack 215 includes at least a portion of the TCP handling code, especially HOT TCP processing unit 250 does not perform. CPU110通过总线112与系统控制器120连接。 CPU110 120 connected by a bus 112 and the system controller. 系统控制器120通过系统总线132与系统存储器130连接。 The system controller 120 is connected via a system bus 132 and system memory 130. 系统存储器130包括TCP堆栈存储空间225、驱动程序存储空间235和连线表(CT) 245 (将在下面进一步说明)。 The system memory 130 includes storage space TCP stack 225, drivers, storage space 235 and netlist (CT) 245 (further described below). 系统控制器120通过集线器到集线器(hub-to-hub)接口126与1/0控制器240连接。 The system controller 120 through the hub-to-hub (hub-to-hub) interface 126 is connected with 1/0 controller 240.

[0044] 1/0控制器240包括PCI总线282的控制器,并可包括系统管理总线(SMBus) 142的控制器、通用串行总线(USB)144等等。 [0044] 1/0 controller 240 includes a PCI bus controller 282, and may include a System Management Bus (SMBus) controller 142, a universal serial bus (USB) 144, and so on. 在另一实施例中,1/0控制器包括PCI快速总线(Express bus)的控制器。 In another embodiment, the 1/0 controller including fast PCI bus (Express bus) controller. 1/0控制器还包括HOT单元250,有效地将HOT单元250与通过PCI总线282连接到I/O控制器240的设备分离。 1/0 controller further includes a HOT unit 250, effectively by HOT Unit 250 and the PCI bus 282 is connected to the I / O controller 240 of the separation device. 特别是,集线器到集线器接口126可以是通过系统控制器120将HOT单元250连接到系统存储器130的高速工业标准总线或专有总线。 In particular, the hub-to-hub interface 126 may be 120 to 250 connected to the system memory speed industry standard buses or proprietary bus 130 through the system controller HOT unit. 连接到I/O控制器240的设备共享PCI总线282的可用带宽,该带宽典型地低于集线器到集线器126的可用带宽。 Connected to the I / O controller PCI 240 devices to share the available bandwidth bus 282, the bandwidth is typically less than the hub to hub 126 available bandwidth. 与图1所示的NIC150和CPUllO之间的延迟(latency)相t匕,I/O控制器240内的HOT单元250的位置导致HOT单元250与CPUl 10和系统存储器130之间更少的延迟。 Delay (latency) NIC150 and CPUllO shown in Figure 1 between phase t dagger, location HOT unit 250 I / O controller 240 results in less delay between 250 and CPUl 10 and the system memory 130 HOT Unit . 在现有技术中,对于NIC和应用程序之间的通信,低延迟可能是关键的,如NIC150和通过驱动程序255的软件堆栈之间。 In the prior art, the communication between the NIC and applications, low latency may be critical, such as between NIC150 and through the driver software stack 255. 低延迟对于NIC150和CPUllO之间传递命令来说尤其重要,例如,为了通知存储于驱动程序存储空间135中的帧数据已经准备好被拷贝到应用程序存储空间125 了。 Low latency is particularly important to pass commands between NIC150 and CPUllO for example, in order to inform the driver stored in the storage space 135 in frame data is ready to be copied into the application memory space 125 a. 还有,因为集线器到集线器接口126和存储器总线132都分别支持比PCI总线282更高的带宽,HOT单元250与通过PCI总线282连接到I/O控制器240的设备相比具有对系统存储器130更高的带宽存取。 Also, because the hub-to-hub interface 126 and the memory bus 132 are respectively supported by more than 282 higher bandwidth PCI bus, HOT unit 250 and is connected to the I / O controller device 240 compared to 130 for the system memory via PCI bus 282 higher bandwidth access. 与通过较低带宽总线如PCI总线282连接到I/O控制器240的设备相比较而言,对系统存储器130更高的带宽存取使HOT单元250能够向应用程序存储空间227或驱动程序存储空间235以更快的速度传送接收到的帧,帧有时被称为“包”(packets)。 And through lower bandwidth bus such as PCI bus 282 to the I / O controller device 240. In contrast, the system memory 130 higher bandwidth access unit 250 can make the HOT application storage space to 227 or the driver store 235 space faster transmission received frame, the frame is sometimes called "package" (packets).

[0045] HOT单元250包括接口连接到输入/输出接口242的控制器。 [0045] HOT unit 250 includes an interface connected to the input / output interface 242 of the controller. 输入/输出接口242可将HOT 单元250 连接到物理层(PHY),如:802.3PHY、HPNA1.0PHY, HPNA2.0PHY 等等。 Input / output interface 242 may be HOT unit 250 is connected to the physical layer (PHY), such as: 802.3PHY, HPNA1.0PHY, HPNA2.0PHY like. 在另一实施例中,PHY包括在HOT单元250中,并且输入/输出接口242是一个以太网接口,如吉比特以太网。 In another embodiment, PHY includes 250 HOT units, and input / output interface 242 is an Ethernet interface, such as Gigabit Ethernet. I/O控制器240可以是单一集成电路或单一半导体平台。 I / O controller 240 may be a single integrated circuit or single semiconductor platform.

[0046] 图2B是包括有集成控制器220的计算系统200的另一实施例。 [0046] Figure 2B is another integrated controller includes a computing system 220 of embodiment 200. 集成控制器220至少执行系统控制器120和I/O控制器240的部分功能并包括HOT单元250。 The controller 220 performs the integrated system controller 120 and I / O controller 240, at least some of the features and includes HOT Unit 250. 集成控制器220还可包括附加接口控制器(未示),例如,SMBus、通用I/O(GPIO)、集成设备电路(IDE)等。 Integrated controller 220 may also include additional interface controller (not shown), for example, SMBus, general I / O (GPIO), integrated device electronics (IDE) and the like.

[0047] TCP堆栈215选择一个或更多TCP连接作为委托连接。 [0047] TCP stack 215 to select one or more TCP connections as delegate access. 委托连接就是HOT单元250处理的TCP堆栈215干涉最少的TCP连接。 HOT commission processing unit 250 is connected to the TCP stack 215 with minimal interference TCP connection. 未被委托的连接或需要特殊处理的委托连接被TCP堆栈215完全或部分地处理。 Connection or that require special handling commission is not committed 215 connections are fully processed TCP stack or in part. TCP堆栈215通过利用存储于系统存储器130中的驱动程序255在HOT单元250内初始化一个委托连接表中的项来设立一个委托连接,进一步详细说明如后。 215 by using the TCP stack stored in the system memory 130 drivers in the HOT 255 initialization unit 250 a delegate connection table entries to set up a delegate to connect, as described in further detail later. 驱动程序255是TCP堆栈215和HOT单元250之间有效的翻译程序,应TCP堆栈215的请求向HOT单元250发出命令。 Driver 255 is effective translation program between TCP stack 215 and HOT unit 250, TCP stack 215 should request to issue commands to HOT Unit 250. 驱动程序255还在接收到来自HOT单元250的通知后通知TCP堆栈215。 Notification driver 255 also receives notification from the TCP stack unit 250 HOT 215. 尽管TCP堆栈215和HOT单元250之间的通信是利用驱动程序255实现的,以后不再对驱动程序255详细说明。 Although the communication TCP stack between 215 and HOT Unit 250 is the use of the driver 255 implemented, will no longer be described in detail in the driver 255.

[0048] 与仅存储委托连接的连接状态数据的委托连接表不同,系统存储器130内的连接表245存储所有活动连接的连接状态数据。 [0048] Principal connection table connection to the storage status data connection only delegate different connection state data connection table 245 stores all activities within the system memory 130 connected. 因此,TCP堆栈215可将任何委托连接的处理假定为HOT单元250的请求。 Therefore, TCP stack 215 may be connected to any delegate process the request is assumed to be HOT unit 250. 由TCP堆栈215处理委托连接被称为“遗留处理”。 Commissioned by the TCP stack 215 process connection is called "legacy process."

[0049] 图3是根据本发明的一个或多个方面的图2A和图2B所示HOT单元250的方框图。 [0049] FIG. 3 is a block diagram 2A and 2B HOT unit 250 in accordance with one or more aspects of the present invention, FIG. DMA引擎310内的直接存储器存取(DMA)接口将I/O控制器240或集成控制器220内的一个或多个子单元连接。 A direct memory access DMA engine 310 (DMA) interface to the I / O controller 240 or one or more sub unit controller integrated 220-Fi. DMA接口被用来发送和接收系统存储器130和HOT单元250内子单元之间的数据,并用来发送和接收CPUllO和HOT单元250内子单元之间的数据。 DMA interface is used to send and receive data to system memory 130 and HOT Unit 250 between the inner sub-unit, and to send and receive data 250 CPUllO and HOT units within the sub-unit.

[0050] 传送引擎320包括被配置为解析和编辑出站帧的子单元,包括确认插入(acknowledgement insertion)和检查和循环冗余校验计算,产生出站巾贞。 [0050] 320 including delivery engine is configured to parse and edit a sub-unit station frame, including the identification of insertion (acknowledgement insertion) and check and cyclic redundancy check calculated to produce the outbound towel infidelity. 传送接口330包括一个或多个为传送存储出站帧的缓冲区和配置为通过输入/输出接口242将连接到HOT单元250的PHY接口连接的子单元。 Transmission interface 330 includes one or more of buffers and configuration storage outbound frames transmitted through the input / output interface 242 is connected to the subunit HOT unit 250 PHY interfacing. 在HOT单元250的另一实施例中,PHY被集成到HOT单元250内。 HOT unit 250. In another embodiment, PHY is integrated into the HOT unit 250. 传送引擎320与委托连接表(DCT)350连接,该委托连接表350中存储委托连接的连接状态数据。 Delivery engine 320 is connected with the commission table (DCT) 350 connection, the delegate connection table 350 is stored in the connected state commission connection data. 委托连接表350是一个存储资源,例如,随机存储器(RAM)、寄存器文件等。 Principal connection table 350 is a storage resource, e.g., random access memory (RAM), a register file and the like. 至少部分的委托连接的连接状态表还存储于连接表245中。 Connection Status Table entrust at least part of the connection is also stored in the connection table 245.

[0051] 委托连接表350中存储的状态信息可包括确定状态、连接地址、系统存储器缓冲区指针、连接跟踪标记、事件控制信息、传送窗口尺寸、接收窗口尺寸、时间戳数据等等。 [0051] Principal connection table 350 stored in the status information may include determining the status of the connection address, the system memory buffer pointer, the connection tracking marks, the event control information transmission window size, the receiving window size, time stamp data, and so on. 确定状态可包括待接收的下一预计序列号的序列号、控制及时生成确定的阈值等等。 Determine the state to be received may include the serial number of the next expected sequence number, control the timely formation to determine the threshold value and so on. 传送引擎320在帧处理处理中利用连接表索引、DCT索引读写委托连接表350的部分,以存取与委托连接相关的项。 Delivery engine 320 utilizing the frame processing process in connection table index, DCT index read commissioned connection table section 350, to access and commissioned connectivity-related entries. 存储于该项中的连接状态数据在委托连接处于活动状态时被TCP堆栈215、传送引擎320和接收引擎360更新,如在图7中所述。 Storing data in the connected state of the connection is active commissioned by the TCP stack 215, transfer engine 320 and receive engine 360 updates, as described in Fig. 7.

[0052] 接收接口370包括子单元,被配置为通过输入/输出接口242与连接到HOT单元250的PHY接口连接。 [0052] Interface 370 includes a receiving sub-unit is configured to input / output interface 242 is connected to the PHY interface unit 250 connected to the HOT. 接收接口370还包括一接收FIFO (先入先出)缓冲区,用于存储接收至IJ的被指定给接收引擎360的巾贞。 Reception interface 370 also includes a reception FIFO (first-in first-out) buffer for storing received to IJ is assigned to the received engine towel Zhen 360. 接收引擎360通过DMA引擎310向系统存储器130上载被部分处理的帧或仅上载TCP有效载荷数据,如后文详述。 Receiving engine 360 310 to the load on the system memory 130 is the frame partially treated or upload only TCP payload data, as hereinafter described in detail by the DMA engine.

[0053] 接收引擎360包括子单元,被配置为将传入帧解析并确定该帧是否有效,即,计算检查和、核对标记并识别出帧类型,例如,IP>DUP,TCP等。 [0053] The reception engine 360 includes a sub-unit is configured to parse incoming frames and determines whether the frame is valid, that is, calculate checksum verification mark and identify the frame type, for example, IP> DUP, TCP and so on. 当被解析的帧无效时被上载到驱动程序存储空间235内的遗留缓冲区中进行遗留处理。 When parsing invalid frames were left when processing uploaded driver storage space left within the buffer 235. 如果接收到的帧包含有带有TCP片段的IP包,TCP堆栈215被通知,并在执行请求的TCP处理后将被上传的帧从遗留缓冲区拷贝到应用程序存储空间227。 If the received frame contains an IP packet with a TCP segment, TCP stack 215 is notified, and the TCP after the process execution request is uploaded to the frame buffer is copied from the legacy application storage space 227.

[0054] 当被解析的帧被确定为有效时,接收引擎360提取源IP地址、TCP序列号(SN)、TCP确认(ACK)号、TCP源和目的端口号、TCP窗口尺寸、TCP报头长度等。 [0054] When the parsed frame is determined to be valid, the reception engine 360 extracts the source IP address, TCP sequence number (SN), TCP acknowledgment number (ACK), TCP source and destination port numbers, TCP window size, TCP header length and so on. 非委托连接上接收到的被解析的帧被上载到驱动程序存储空间235中的遗留缓冲区中处理。 Non-commissioned connected frames received parsed be uploaded to the driver storage space 235 left in the buffer processing. 委托连接上接收到的但又不是特殊事例的被解析的巾贞如失序号(out-of-order sequence number)和TCP标记压栈设置(TCP push flag set)等被处理,并且TCP有效载荷数据被上载到应用程序存储空间227中的用户缓冲区。 Received the delegation connector but not a special case to be resolved, such as loss of towel Zhen number (out-of-order sequence number) and TCP flags push settings (TCP push flag set), etc. are processed and TCP payload data It is uploaded to the user application storage space in the buffer 227. 将TCP有效载荷直接上载到应用程序存储空间227比通过驱动程序存储空间235上载有效载荷数据更为高效,因为不需要随后将TCP有效载荷由驱动程序存储空间235拷贝到应用程序存储空间227。 The TCP payload uploaded directly to application storage space than 227 storage space 235 through the driver upload payload data more efficient, because there is no subsequent TCP payload storage space 235 by the driver to the application a copy of the storage space 227.

[0055] 图4A是根据本发明的一个实施例的设立委托连接的方法步骤的流程图。 [0055] FIG. 4A is a flowchart of a method embodiment delegate connection establishment procedure according to one embodiment of the present invention. 步骤410中计算系统200利用本领域技术人员已知的处理方法通过3次握手建立TCP连接。 200 known to those skilled in the use of processing methods by three-way handshake to establish a TCP connection in step 410 the computing system. 步骤412中TCP堆栈215确定是否委托由HOT单元250处理的连接。 Step 412 TCP Stack 215 determines whether commissioned connected by HOT Unit 250 process. TCP堆栈215可基于委托连接的特性确定委托连接,如用户定义的连接类型的优先权、为连接规定的持续时间、为连接规定的帧速率、是否将连接用应用程序等,以便执行大量传送,连接所使用的特定TCP端口等。 TCP stack 215 may be determined based on the principal delegation connector connection features such as user-defined type of connection priority for the connection specified duration for the connection specified frame rate, whether the connection with the application, in order to perform a number of transmission, The use of a specific TCP connection ports.

[0056] 如果步骤412中TCP堆栈确定不应委托该连接用于HOT单元250进行处理,那么步骤414中TCP堆栈215在CT245中设立用于处理该连接的项并进入步骤422。 [0056] If step 412 determines TCP stack should not be entrusted to the connection processing unit 250 for HOT, then step 414 TCP Stack 215 to set up the connection entry and proceeds to step 422 for processing of the CT245. 如果步骤412TCP堆栈215确定应该委托连接用于HOT单元250进行处理,那么步骤416中TCP堆栈215向HOT单元250发出命令,在具有连接状态数据的DCT350中设立一个项。 If step 215 determines 412TCP stack should delegate connection processing unit 250 for HOT, then step 416 TCP Stack 215 HOT Unit 250 issues a command to set up an entry in DCT350 has a connection state data. 步骤418中TCP堆栈215确定是否向HOT单元250发出一个后接收缓冲区(PRB),如后文详述,向HOT单元250提供系统存储器130中一个或多个用户缓冲区的位置和尺寸。 Step 418 TCP Stack 215 determines whether to issue a post-receive buffer (PRB), as hereinafter described in detail to HOT Unit 250 to HOT Unit 250 provides a system memory 130, one or more user buffer's position and size. 如果步骤418中TCP堆栈215确定要发出PRB命令,那么步骤420中TCP堆栈215发出PRB命令。 If step 418 TCP Stack 215 determines to issue PRB command, then step 420 TCP Stack 215 PRB command issued. 如果在步骤418中TCP堆栈215确定不发出PRB命令,那么TCP堆栈进入步骤422。 If it is determined not to issue PRB command in step 418 TCP Stack 215, then proceeds to step 422 TCP stack. 步骤422中连接设立结束。 Step 422, the end of the connection establishment.

[0057] 图4B是图4A是根据本发明的一个实施例的接收帧的方法步骤的流程图。 [0057] FIG. 4B is a flow chart of FIG. 4A is a received frame method steps according to one embodiment of the present invention. 步骤424中,HOT单元250通过输入/输出接ロ242接收帧并可部分处理该帧,产生部分解析帧和报头数据。 In step 424, HOT Unit 250 via the input / output interface 242 receives the frame and ro handling portion of the frame, the frame generating section and a header parsing data. 步骤425中HOT单元250确定是否在委托连接上接收到了帧,并且,如果没有,步骤440中HOT単元250向ー个或多个遗留缓冲区上载部分处理帧,包括其整套数据链路层和网络层协议报头数据。 Step 425 HOT Unit 250 determines whether it has received the commission in connection frame, and, if not, step 440 HOT radiolabeling unit 250 ー or partially treated to upload multiple legacy frame buffer, including its suite of data link layer and the network layer protocol header data. 在步骤442中TCP堆栈215处理被上载到遗留缓冲区的被部分处理的帧。 TCP stack processing step 215 is uploaded to the legacy buffer is part of the processing of 442 frames.

[0058] 如果在步骤425中,HOT单元250确定委托连接上接收到了帧,那么在步骤426中HOT单元250完成对帧的解析,提取TCP有效载荷数据。 [0058] If, in step 425, HOT Unit 250 determines that the commission has received the frame is connected, then the complete resolution of the frame in step 426 HOT Unit 250, extracted TCP payload data. 在步骤427中HOT单元250确定用户缓冲区是否可用,如果可用,那么在步骤428中HOT单元250向一个或多个用户缓冲区中上载TCP有效载荷数据。 HOT unit 250 determines that the user buffer is available in step 427, if available, then upload in step 250 HOT unit to one or more user buffers 428 TCP payload data. 如果在步骤427中HOT单元250确定用户缓冲区不可用,那么在步骤430中HOT单元250向遗留缓冲区上载有效载荷数据的一部分并通知TCP堆栈215。 HOT Unit 250 determines if the user buffer is not available in step 427, then in the left HOT Unit 250 to upload payload data buffer part of step 430 and notifies TCP Stack 215. 在一个实施例中,该部分是由存储在与委托连接对应的DCT350内的项中的“启动限制”值规定的。 In one embodiment, the portion is stored in connection with the commissioning DCT350 corresponding entry in the "start limit" value specified. 该“启动限制”是ー个变量,其最大值可等于最大接收帧的尺寸,其最小值可由应用程序217或TCP堆栈215确定。 The "start limit" is ー variables, which may be equal to the maximum size of the largest of received frames, the minimum value by the application 217 or TCP stack 215 OK.

[0059] 步骤432中TCP堆栈215处理上载到遗留缓冲区的TCP有效载荷数据。 [0059] step 432 TCP Stack 215 processes the uploaded to the legacy buffer TCP payload data. 步骤434中HOT单元250确定TCP堆栈215为委托连接发出的ー个或多个PRB命令是否已经被处理。 Step 434 HOT Unit 250 determines whether the TCP stack ー 215 or more PRB commands issued by the delegate connection have been processed. 步骤436中HOT单元350将剩余TCP有效载荷数据上载到ー个或多个用户缓冲区。 Step 436 HOT Unit 350 remaining TCP payload data uploaded ー or more user buffers. 如果在步骤434中HOT单元250确定委托连接的ー个或多个PRB命令还没有被处理,那么在步骤438中HOT单元250将剩余TCP有效载荷数据上载到遗留缓冲区将通知TCP堆栈215。 If, in step 434 HOT Unit 250 determines commissioned connected ー or more PRB command has not been processed, then in step 438 HOT Unit 250 remaining TCP payload data uploaded to the legacy buffer will notify TCP stack 215. 在另ー实施例中,TCP堆栈215完成步骤434并在步骤438中TCP堆栈215指示HOT单元250将所有剰余TCP有效载荷数据上载到遗留缓冲区。ー In another embodiment, TCP stack 215, and indicates the completion of step 434 in step 438 TCP Stack 215 HOT Unit 250 to all 剰 than TCP payload data uploaded to the legacy buffer.

[0060] 在一个实施例中,消息信号中断(MSIs)为HOT单元250提供ー个利用多重中断向量发信号给其各种中断资源的机制。 [0060] In one embodiment, the message signaled interrupts (MSIs) to HOT Unit 250 provides ー a vector interrupt mechanism by multiplex signal interruption to their various resources. 利用MSI能够使主机的中断处理更为效率。 MSI enables the use of more efficient interrupt handling host. 在ー个实施例中,计算系统200利用多达8个中断向量。 In ー embodiment, the computing system 200 utilizes up to eight interrupt vectors.

[0061] 图4C是根据本发明ー个或几个方面的慢启动序列的实施例的简图。 [0061] FIG. 4C is a diagram of an embodiment ー or slow start sequence several aspects of the present invention. 慢启动和拥塞避免算法由TCP协议规定(如RFC793、RFCl 122及相关文件中提出的),因此是本领域技术人员所熟知的。 Slow start and congestion avoidance algorithms specified by the TCP protocol (such as RFC793, RFCl 122 and related documents presented), and therefore those skilled in the art. TCP使用慢启动和拥塞避免来确定某一给定连接的可用带宽。 TCP uses the slow start and congestion avoidance to determine the available bandwidth of a given connection. 两个变量ssthresh (慢启动阈值)和cwnd (拥塞窗ロ)确定委托连接的行为。 Two variable ssthresh (slow start threshold) and cwnd (congestion window ro) determine delegation connector behavior. HOT单元250上载ACK信息到TCP堆栈215,包括对接收到的ACK的计数。 HOT Unit 250 contained ACK information to the TCP stack 215, including a count of received ACK. HOT单元250可被TCP堆栈215配置为合并(coalesce) ACK,而不是将每个接收到的ACK都通知TCP堆栈215,以TCP堆栈215规定的频率将ACK上载到TCP堆栈215。 HOT unit 250 may be configured as a TCP stack 215 merge (coalesce) ACK, rather than have each received ACK notification TCP stack 215, 215 frequencies specified TCP stack TCP stack 215 will ACK upload. 合并ACK使HOT单元250减少HOT单元215通知TCP堆栈215委托连接状态的频率,有望在许多情况下提高性能。 The combined ACK reduction unit 250 so HOT HOT section 215 notice TCP stack 215 commissioned connected state frequency, is expected to improve performance in many cases. 特别是由于CPUllO不是为每个接收到的ACK都发生中断,对CPU的使用典型地减少了。 Especially since CPUllO not received for each ACK are interrupt occurs, the CPU usage is typically reduced. 对接收到的ACK的计数允许TCP堆栈215确定为每个通知接收到的ACK的数量以便实现慢启动和拥塞避免。 ACK received the count enable TCP stack 215 determines the number of ACK of each notification received in order to achieve slow start and congestion avoidance.

[0062] 步骤452中应用程序217为委托连接设置cwnd为I段(segment)并且TCP堆栈215输出传送缓冲区描述符到HOT单元250,如后面说明图6D时详细说明。 [0062] step 452 application 217 connection settings cwnd as I entrust segment (segment) and TCP Stack 215 outputs the transmission buffer descriptors to HOT Unit 250, later described as a detailed description Figure 6D. 步骤454中HOT单元250确定ACK是否被由目标接收到。 In step 454 HOT Unit 250 determines whether the ACK has been received by the target. HOT单元250在步骤454保持,直到ACK被接收到并且然后HOT单元250通知TCP堆栈215并进入步骤456。 HOT Unit 250 is maintained at step 454 until the ACK is received and then notify TCP Stack 250 HOT Unit 215 and proceeds to step 456. 在一个实施例中HOT单元250输出接收到ACK的计数给TCP堆栈215并且TCP堆栈215可计算在每个通知之间为委托连接接收到的ACK的数量。 In a number of cases ACK output HOT Unit 250 receives an ACK count to 215 and TCP stack TCP stack receives a delegation connector 215 can be calculated between each notification to implementation.

[0063] 步骤456中TCP堆栈215为委托连接确定cwnd是否大于或等于ssthresh,如果是,那么在步骤458中TCP堆栈215呈指数増加,即,打开,基于为委托连接接收到的ACK的数量的cwnd。 [0063] step 456 TCP Stack 215 delegate connection determine cwnd is greater than or equal to ssthresh, if it is, then exponentially to increase in step 458 TCP Stack 215, that is, open, based on the number of ACK is entrusted connection received the cwnd. 步骤458中TCP堆栈215还向HOT单元250传送缓冲区描述符并返回步骤454。 Step 458 TCP Stack 215 HOT Unit 250 is also transmitted to the buffer descriptor and returns to step 454.

[0064] 在一个实施例中,步骤458中TCP堆栈215将HOT单元250配置为通知TCP堆栈每个接收到的ACK。 [0064] In one embodiment, the step 215 TCP Stack 458 HOT Unit 250 is configured to notify TCP stack each received ACK. 在另ー实施例中TCP堆栈215将HOT单元250配置为通知TCP堆栈215一个接收到的ACK的计数,从而执行ー些ACK合并。 In another embodiment ー TCP stack 215 HOT unit 250 configured to notify TCP Stack 215 a received ACK counting implementation, to perform some ACK ー merger. 如果步骤456中TCP堆栈215为委托连接确定ssthresh小于cwnd,那么在步骤460中TCP堆栈215处理拥塞避免阶段。 If step 456 TCP Stack 215 is less than ssthresh delegation connector determine cwnd, then avoid the stage in step 460 TCP Stack 215 processes the congestion. 当拥塞避免被使用时cwnd线性打开,直到cwnd等于最大传送窗ロ尺寸或丢下包。 When congestion avoidance is used when cwnd linear open until cwnd size equal to the maximum transfer window ro or dropped packets.

[0065] 图5A是根据本发明ー个或几个方面的在图2A和图2B中所示的应用程序存储空间中的用户缓冲区的实施例的简图。 [0065] Figure 5A is according to the present invention ー or user aspects of the application shown in Figures 2A and 2B of storage space in the schematic example of implementation of the buffer. 每个用户缓冲区如用户缓冲区510、用户缓冲区512或用户缓冲区514都被分配在应用程序存储空间227中,用来接收HOT单元250上传的有效载荷数据。 Each user buffer such as user buffer 510, the user buffer 512 or the user buffer 514 are allocated in the application storage space 227 for receiving payload data HOT Unit 250 uploads. 物理存储器地址,如用户缓冲区地址515指出用户缓冲区510在应用程序存储空间227中的位置。 Physical memory addresses, such as the user buffer address 515 indicate the position of the user buffer 510 in 227 application storage space. 类似地,用户缓冲区地址520指示用户缓冲区512的位置,用户缓冲区地址525指示用户缓冲区514的位置。 Similarly, the user buffer address 520 indicating the user buffer 512, and the user buffer address 525 indicates the position of the user buffer 514. 用户缓冲区可被存储于应用程序存储空间227内的物理上互相毗邻的存储位置或存储于应用程序存储空间227内的物理上不相毗邻的存储位置。 User buffer can be stored in the application storage space within the next 227 physical storage locations to each other or within 227 stored in the physical storage space on applications not in adjacent memory locations.

[0066] 图5B是根据本发明ー个或几个方面的用户缓冲区描述符的实施例的简图。 [0066] FIG. 5B is a diagram of an embodiment of the present invention or user aspects ー buffer descriptor based. 每个用户缓冲区都具有由能够存储于该用户缓冲区中的字节数确定的对应长度。 Each user has the ability to store buffer to the user buffer length corresponding to the number of bytes determined. 例如,用户缓冲区510的长度就是用户缓冲区长度535。 For example, the user buffer length is 510 535 user buffer length. 用户缓冲区描述符是ー种数据结构,包括用户缓冲区地址,如用户缓冲区地址515和对应用户缓冲区长度,如用户缓冲区长度535。 User buffer descriptor is ー types of data structures, including user buffer address, such as user buffer address 515 and the corresponding user buffer length, such as user buffer length 535. 在另ー实施例中,用户缓冲区描述符可包括指示任何特殊处理等的描述符标记。ー In another embodiment, a user buffer descriptor may include an indication of any special handling descriptor tags. 用户缓冲区描述标记可在其他位中包括请求HOT单元350在有效载荷数据被上载到包括在PRB命令中的用户缓冲区发出通知命令的位 User buffer description tag in other bits HOT unit 350 includes a request is uploaded to the payload data included in the PRB command the user buffer notification command bits

[0067] 在又一实施例中,用户缓冲区描述符可包括用户缓冲区地址、用户缓冲区长度和用户缓冲区结束地址的任意組合。 [0067] In a further embodiment, a user buffer descriptor may include a user buffer address, the user buffer length and any combination of end user buffer address. 如前述,利用PRB命令由TCP堆栈215将用户缓冲区描述符提供给HOT单元250。 As mentioned above, the use of PRB command provided by the TCP stack 215 user buffer descriptor to HOT Unit 250. 向HOT单元250提供位于应用程序存储空间227的物理存储器地址使HOT単元250直接向应用程序存储空间227上载有效载荷数据。 Physical memory addresses to HOT Unit 250 provided in the application storage space 227 so HOT radiolabeling unit 250 directly to the application storage space 227 to upload payload data.

[0068] 应用程序217管理用户地址空间,其是由操作系统分配的虚拟地互相毗邻的地址空间。 [0068] application to manage user address space 217, which is assigned by the operating system virtual address space land adjoining each other. 当应用217将用户地址空间信息传送给TCP堆栈215时,TCP堆栈215请求操作系统锁定对应该用户缓冲区地址空间的存储器。 When the application 217 user address space information to the TCP stack 215, TCP stack 215 requests a lock on the operating system should the user buffer memory address space. 操作系统锁定存储器的数量井向TCP堆栈215返回对应系统存储器130的物理上互相毗邻的部分的ー个或多个物理地址(和长度)。 Operating system lock memory corresponding to the number of wells to return the system memory 130 to the TCP stack 215 physically adjacent to each other on one or more portions of the physical addresses ー (and length). HOT単元250存取的物理地址空间由TCP堆栈215管理而不必是物理上互相毗邻的。 HOT radiolabeling unit 250 accesses the physical address space managed by the TCP stack 215 is physically rather than adjacent to each other. TC堆栈215进行用户地址空间和物理地址空间之间的翻译转换。 TC 215 stack translate translate user address space and physical address spaces. 在另ー实施例中驱动程序255进行用户地址空间和物理地址空间之间的翻译转换。 In another embodiment of the driver-ー translate 255 converts the user's address space and physical address spaces. [0069] 图5C是根据本发明ー个或几个方面的在图2A和图2B中所示软件驱动程序存储空间中的遗留缓冲区的实施例的简图。 [0069] FIG. 5C is a schematic ー or embodiments 2A and software drivers storage space left in the buffer shown in FIG. 2B aspects of the present invention. 每个遗留缓冲区如遗留缓冲区550、遗留缓冲区552或遗留缓冲区554都被分配在驱动程序存储空间235中以便接收HOT単元250上载的部分处理的帧。 Each legacy buffer such as buffer 550 legacy, the legacy left buffer 552 or buffer 554 are allocated in the driver storage space 235 to receive the frame element 250 HOT radiolabeling contained partially processed. 物理存储地址,如遗留缓冲区地址555指示遗留缓冲区550在驱动程序存储空间235中的位置。 Physical memory addresses, such as the legacy 555 instruction buffer address buffer 550 legacy storage space 235 in the driver's position. 类似地,遗留缓冲区地址560指示遗留缓冲区552的位置并且遗留缓冲区地址565指示遗留缓冲区554的位置。 Similarly, the legacy buffer address buffer 560 indicates the position of the legacy and the legacy buffer address 552 565 554 indicating the position of the legacy buffer. 遗留缓冲区可被存储于驱动程序存储空间235内的互相毗邻的存储器位置或驱动程序存储空间235内不互相毗邻的存储器位置。 Buffer can be stored in the driver storage space adjacent to each other memory location or drivers storage space within 235 235 memory locations are not adjacent to each other left.

[0070] 图是根据本发明ー个或几个方面的遗留缓冲区标签表的实施例的简图。 [0070] is a schematic diagram according to the present invention or aspects ー legacy buffer tag table embodiment. 每个遗留缓冲区地址都与一唯一的标签相关联。 Each legacy buffer address is associated with a unique tag associated. 例如,遗留缓冲区地址555与标签575相关联,遗留缓冲区地址560与标签580相关,遗留缓冲区地址565与标签585相关。 For example, a legacy buffer address 555 is associated with a label 575, 560 legacy buffer address associated with the label 580, 565 associated with the legacy buffer address tag 585. 在一个实施例中,遗留缓冲区标签表590由驱动程序255维护并可被存储于驱动程序存储空间235中,或在另ー实施例中被存储在TCP堆栈存储空间225中。 In one embodiment, the legacy buffer label table 590 by the driver 255 to maintain and can be stored in the driver storage space 235, or in another embodiment ー TCP stack is stored in the storage space 225.

[0071]图5E是根据本发明ー个或几个方面的遗留缓冲区描述符的实施例的简图。 [0071] FIG. 5E is a diagram of an embodiment of the present invention or aspects ー legacy buffer descriptor based. 每个遗留缓冲区都具有能够存储在遗留缓冲区内字节的数目所決定的对应长度。 Each legacy buffer length has a corresponding number can be stored in the buffer byte legacy of the decision. 例如遗留缓冲区550的长度是遗留缓冲区570。 For example, the length of the left buffer 550 is a legacy buffer 570. 在另ー实施例中,所有遗留缓冲区的长度都相等。 In another embodiment ー, all legacy buffer length are equal. 遗留缓冲区描述符是ー种数据结构,包括遗留缓冲区地址,如遗留缓冲区地址555 ;对应遗留缓冲区长度,如遗留缓冲区长度570 ;和对应标签,如标签575。 Legacy buffer descriptor is ー types of data structures, including legacy buffer address, such as legacy buffer address 555; corresponding to the legacy buffer length, such as the legacy buffer length 570; and the corresponding tags, such as tags 575. 在另ー实施例中,遗留缓冲区描述符可包括任何遗留缓冲器地址的组合、遗留缓冲器长度、标签、和遗留缓冲器端地址。ー In another embodiment, the legacy buffer descriptor may include any combination of legacy buffer address legacy buffer length, labels, and implement legacy buffer end address. 遗留缓冲器描述符被驱动程序255利用接收(缓冲区)描述符环,提供给HOT単元250,以后还要详细说明。 Legacy buffer descriptor is the driver 255 using the reception (buffer) descriptor ring provided to HOT radiolabeling unit 250, described in greater detail later.

[0072] 驱动程序255与HOT単元250之间的通信通过存储在驱动程序存储空间235内的数据结构完成。 [0072] communication driver 255 and between 250 yuan HOT radiolabeling done through data structures are stored in the driver storage space 235. 环(ring)是指包括多个项的数据结构,其详述如后。 Ring (ring) is meant to include a plurality of items of data structures, such as after its detailing. 环被组织成项的环形队列,带有驱动程序255使用的指针和HOT単元250使用的另一指针。 Ring is organized into circular queue entry, another element pointers and HOT radiolabeling with drivers use 255 250 use. 每个环都存储在驱动程序存储空间235内毗邻的物理存储器中。 Each ring is stored in the driver storage space 235 adjacent physical memory.

[0073] 图6A是用于由应用程序255向HOT单元250传送命令并由HOT单元250向驱动程序255传送状态的命令环601的简图。 [0073] FIG. 6A is used to transfer commands from the application program 255 to the unit 250 by sketch HOT HOT unit 250 to the driver 255 delivery status of command ring 601. 命令环601用于初始化DCT350中的委托连接项并将用户缓冲区描述符提供给DCT350。 Command ring 601 is used to initialize DCT350 Principal connection entry and user buffer descriptor to the DCT350. 命令环601包括若干项,每个项在图6A中都以命令环项603表示。 Command ring 601 includes a number of items, each in FIG. 6A are in command ring entry 603. 姆个命令环项603都包括指示该项是归HOT单元250或驱动程序255所拥有的“own”位。 Tim command entries ring 603 includes instructions that are owned by HOT Unit 250 or Driver 255 has "own" bit. 启动吋,每个项中的“own”位都被初始化,被指示为驱动程序255所有,并且命令写指针607和命令读指针605在命令环601中是同一个项。 Start inches, each entry in the "own" bits are initialized, the driver was instructed all 255, and command the write pointer 607 and read pointer 605 commands in the command ring 601 is the same item. 当TCP堆栈215通过驱动程序255向一个项写入ー个命令吋,“ own”位被设置为指示该项为HOT单元250所有并且命令写指针607被改为指向命令环601中的下ー个指令环项603。 When the TCP stack 215 through the driver 255 is written to an entry ー command inch, "own" bit is set to indicate that the write pointer to HOT Unit 250 and ordered all 607 been changed to point to the next command ring 601 ー months instruction ring 603 entries. 当HOT单元250读出并完成处理时,设置一个被命令读指针605定址的项,指示该项为驱动程序255所有。 When HOT Unit 250 is read out and complete processing, set up a command read pointer 605 is addressed items, indicating that the driver 255 all. 不允许命令读指针605越过命令写指针607。 Not allowed to command the read pointer 605 over the command of the write pointer 607. 当命令读指针605或命令写指针607到达命令环601的尾项时,该指针卷入命令环601中的首项。 When the command read pointer 605 or 607 command write pointer reaches the command of end ring 601, the pointer is involved in command of the first ring 601 items. 本技术领域的专业人员能够理解还可以采用其他机制将命令送至HOT单元250,例如,命令连接表、FIFO、共享存储器方案等。 Those skilled in the art to understand other mechanisms may also be employed to the HOT command unit 250, for example, the command contacts list, FIFO, shared memory programs.

[0074] 除“own”位外,每个命令环项603还包括命令字段、DCT索引、特定命令控制和/或状态信息、特定命令数据等。 [0074] In addition to "own" bits, each command ring entry 603 also includes a command field, DCT index, specific commands to control and / or status information, specific command data. 如前述,DCT索引识别DCT350中与委托连接对应的项。 As mentioned above, the corresponding item identification DCT350 DCT index connected with the commission. 命令字段包括命令标识符,用于PRB命令、更新表项(UTE)命令、无效表项(ITE)命令、转储连接缓冲区表项(DCBTE)命令等命令。 Command field includes a command identifier for PRB command to update entries (UTE) command, invalid entries (ITE) command, dump the connection cache entry (DCBTE) command and other commands. 当一条命令被驱动程序255写入吋,特定命令控制/状态信息包括特定命令控制。 When a command is written to the driver 255 inches, the specific command and control / status information, including specific commands. 当命令被HOT単元350读出并更新吋,特定命令控制/状态信息被更新为包括特定命令状态。 When the command is out and updated inch HOT radiolabeling read 350 yuan, the specific command and control / status information is updated to include a specific command state. 特定命令数据被驱动程序255写入并被HOT単元350读出,详述如后。 Specific command data is written to the driver 255 yuan and 350 HOT radiolabeling readout, as described in detail later.

[0075] PRB命令被应用程序217使用,以便通过TCP堆栈215和驱动程序255将用户缓冲区描述符传递给HOT単元350。 [0075] PRB command is the application to use 217 to 215 by the TCP stack and driver 255 will be passed to the user buffer descriptor 350 HOT radiolabeling yuan. 每个用户缓冲区描述符为HOT単元350在应用程序存储空间227中指示ー个物理地址以便将有效载荷数据上载到其中。 Each user buffer descriptor to HOT radiolabeling $ 3.5 indicates ー physical address in the application storage space 227 so that the payload data uploaded it. TCP堆栈接收ー个或多个用户地址并确定对应的物理地址以便包括在用户缓冲区描述符中。 TCP stack receives ー one or more user addresses and determine the corresponding physical address to be included in a user buffer descriptor. TCP堆栈215可代表应用程序217为ー单一委托连接表项利用PRB命令通过驱动程序255发送ー个或多个用户缓冲区描述符。 TCP stack 215 may represent a single application 217 ー delegate connection entry utilization PRB command sent ー 255 or more users through driver buffer descriptor. 驱动程序255包括PRB命令内特定命令控制和/或状态信息字段内用户缓冲区的数量。 Driver 255 includes specific commands within the PRB command control and / or the number of states in the information fields for the user buffer. 因为驱动程序255没有必要的信息来确定HOT単元350已经上载了多少个先前被发送的用户缓冲区,所以HOT単元350在特定命令控制和/或状态信息字段中写入ー个指示PRB命令接受的用户缓冲区的数量。 Because the driver 255 without the necessary information to determine the HOT radiolabeling element 350 have uploaded a number of previously transmitted user buffer, so HOT radiolabeling element 350 in a specific order to control and / or status information field written ー an indication PRB accepted command the number of users of the buffer.

[0076] PRB命令中的特定命令控制和/或状态信息字段还包括“sync”位。 [0076] PRB command specific command and control and / or status information field also includes "sync" bit. TCP堆栈215在包括有被断言遗留标记的通知命令已经通过通知环611由HOT单元接收时(如后在图9B中详述),请求驱动程序255写入“sync”位。 TCP stack 215 includes left asserted labeled notification command has been notified by the ring 611 is received by HOT Unit (as described in detail later in FIG. 9B), requesting driver 255 writes "sync" bit.

[0077] 由驱动程序255构建的在PRB命令中的特定命令数据字段包括对应PRB命令中发送的第一缓冲区的首字节的起始TCP序列号和包括在PRB命令中的所有用户缓冲区的用户缓冲区描述符等。 [0077] Specific command data field by the driver 255 constructed in PRB command includes a first buffer of the first byte of the starting TCP sequence numbers corresponding to the PRB command sent and all user buffer included in the PRB command The user buffer descriptor like. 用户缓冲区描述符包括在应用程序存储空间227中规定位置的物理地址、用户缓冲区长度、指示任何特殊处理的描述符标记等等。 User buffer descriptor included in the application storage space 227 in a predetermined position of the physical address, the user buffer length descriptor tag indicating any special treatment and so on.

[0078] UTE命令被驱动程序255使用,用来更新DCT350中的项,还被用于设立委托连接和委托连接活动时更新连接数据。 [0078] UTE command by the driver 255 used to update DCT350 the item, also used to set up commission to update the connection data connections and connection activity delegate. ITE命令被用来将委托连接无效。 ITE command is used to delegate connection invalid. 当HOT单元250接收ITE指令吋,如果有必要,在清除对应ITE命令中规定的DCT索引的委托连接之前,它将等待传送引擎320和接收引擎360处理的完成(同时阻塞任何新的TCP处理的启动)。 Upon completion of HOT Unit 250 receives instruction ITE-inch, if necessary, before clearing the corresponding ITE specified in the order DCT index commissioned connection, it waits for delivery engine 320 and the engine 360 receiving treatment (while blocking any new TCP processing start up). DCBTE命令使HOT单元350向遗留缓冲区上载DCBTE命令中所包括的由DCT索引规定的项。 DCBTE command makes HOT Unit 350 to the legacy buffer ins provided by the DCT index DCBTE command included.

[0079] 驱动程序255可存取命令环601而不干扰PRB命令的传送或接收处理处理。 [0079] Driver 255 may access command ring 601 without interfering with the PRB command is transmitted or received treatment process. 这样就允许驱动程序255向HOT単元350及时提供新的用户缓冲区,从而提高接收帧能够被HOT単元350接受而不是被阻塞的可能性。 This allows the driver to provide a new user buffers 255 to 350 yuan timely HOT radiolabeling, thereby improving the reception frame can be accepted HOT radiolabeling $ 3.5 instead likely to be blocked.

[0080] 图6B是用于由HOT单元250向驱动程序255传送事件通知描述符的通知环(notification ring)的简图。 [0080] FIG. 6B is a diagram used to notify the descriptor by HOT Unit 250 to the driver 255 transmit the event notification ring (notification ring) of. 通知环611载有通过驱动程序255由HOT单元250向TCP堆栈215的连接信息。 Notify ring 611 carrying 255 through the driver information from the HOT unit 250 is connected to the TCP stack 215. 本领域技术人员应该理解,还可以使用其他机制由HOT单元250向TCP堆栈215交流信息,例如,通知描述符的链接表、FIFO、共享存储器方案等。 Those skilled in the art would understand, other mechanisms may also be used information from the TCP stack HOT Unit 250 to 215 AC, for example, notification descriptor linked list, FIFO, shared memory programs.

[0081] 通知环611包括多个项,图6B中所有项都用通知环项613表示。 [0081] notification ring 611 includes a plurality of entries, 6B, all items are 613 represented by notice in the ring. 每个通知环项613包括指示该项是被HOT单元250还是被驱动程序255所有的“own”位。 Each notification includes information indicating that ring entry 613 is HOT Unit 250 or 255 by the driver all "own" bit. 启动时,姆个项中的“own”位都被初始化,被指示为HOT单元250所有,并且通知写指针617和通知读指针615在通知环611中是同一个项。 When you start, Tim entries in the "own" bits are initialized, is indicated as a HOT unit 250 at all, and notifies the write pointer 617 and read pointer 615 notifications in the notification ring 611 is the same item. 当HOT单元250通过驱动程序255向一个项写入ー个命令吋,“own”位被设置为指示该项为驱动程序255所有并且通知写指针615被改为指向通知环611中的下ー个通知环项613。 When HOT Unit 250 through the driver 255 is written to an entry ー command inch, "own" bit is set to indicate that the driver 255 and notification of all the write pointer 615 is changed to point to the next notification ring 611 ー months Notify ring 613 entries. 当驱动程序255读出并完成处理被通知写指针617定址的项,“own”位被设置为指示该项为HOT单元250所有。 When the driver 255 is read out and complete the process by the write pointer 617 notification addressed items, "own" bit is set to indicate that the HOT unit 250 at all. 不允许通知读指针617越过通知写指针615。 Not allowed to inform the read pointer 617 over the write pointer 615 notice. 当通知读指针617或通知写指针615到达通知环611的尾项时,该指针卷入通知环611中的首项。 When the notice read pointer 617 or notification of the write pointer 615 reaches the end of the notification ring 611 item, the pointer is involved in the notification ring 611 in the first term.

[0082] 除“own”位外,每个通知环项613还包括通知标记字段、DCT索引、可选标记(该标记如果存在,向特定遗留缓冲区提供參考)、下ー预计序列号、最高接收到的ACK号、最近接收到的传送窗ロ尺寸、当前TCP时戳等等,用于DCT索引规定的委托连接。 [0082] In addition to "own" bits, each notification ring entry 613 also includes a notification tag field, DCT index, selectable marker (the marker, if present, to provide a reference to a specific legacy buffer), under ー expected sequence number, the maximum ACK number received recently received ro transfer window size, the current TCP timestamp, etc., commissioned by DCT index specified for the connection. 通知标记字段包括“遺留”标记、“压栈通知”标记、“双重ACK”标记、“序列号阈值”标记、“ACK阈值”标记、“请求缓冲区”标记等等。 Notification tag fields include "legacy" tags, "push notification" flag, "double ACK" tag "sequence number threshold" flag, "ACK threshold" flag, flag "request buffer" and so on. “遺留”标记在有效载荷数据或部分解析帧数据已经被HOT単元250上载到遗留缓冲区。 "Legacy" mark in the payload data or partially parsed frame data has been uploaded to the legacy HOT radiolabeling unit 250 buffers. “压栈通知”标记、“双重ACK”标记、“序列号阈值”标记、“ACK阈值”标记和“请求缓冲区”标记将在图9A中说明。 "Push notifications" tags, "double ACK" tag "sequence number threshold value" tag, "ACK threshold" flag and the "request buffer" flag will be described in FIG. 9A.

[0083] 当HOT単元250向遗留缓冲区上载有效载荷数据或被部分解析帧数据时可包括可选标记,以后在图8C中还有详细说明。 [0083] When the HOT radiolabeling unit 250 to the legacy buffer payload data uploaded or partially parsed frame data may include selectable marker, later in FIG. 8C also described in detail. 该标记是通过接收描述符环由驱动程序255接收的(后面将详细说明),并被用于将给定的通知与上载了有效载荷数据或被部分解析帧数据联系起来。 This flag is used by the receiving descriptor ring 255 is received by the driver (described in detail below), and for a given notification and upload the payload data or partially parsed frame data link. 驱动程序255可使用随通知接收的标记通过读出与该标记相关联的遗留缓冲区标签表590中的项找到遗留缓冲区在驱动程序存储空间235中的位置。 Drivers can use 255 mark with the notification received by reading labels and legacy buffer table 590 associated with the mark of the items found in the driver left the buffer memory 235 in place.

[0084] HOT単元250可利用通知环611及时通知驱动程序255要求进ー步由驱动程序255以对HOT单元250传送或接收处理影响最小的方式处理的连接条件。 [0084] HOT radiolabeling unit 250 may use the notification ring 611 255 claim promptly notify the driver by the driver steps into ー 255 to 250 to transmit or receive minimal impact on HOT unit processing manner of connection conditions. 通知环611的操作允许驱动程序255及时提供HOT単元350新用户缓冲区,从而提高接收帧能够被HOT单元250接受而非阻塞的可能性。 Notify ring 611 allows the driver to operate the timely provision of 255 350 yuan HOT radiolabeling new user buffer, thereby increasing the possibility of receiving a frame that can be accepted HOT Unit 250 and non-blocking.

[0085] 图6C是根据本发明ー个或几个方面的用于由通过驱动程序255由TCP堆栈215向HOT单元250传送接收缓冲区信息的接收描述符环621 (receive descriptor ring)的简图。 [0085] FIG. 6C is ー or aspects for the 255 by the TCP stack buffer 215 receives information transmitted to HOT Unit 250 receives the descriptor ring 621 by the driver (receive descriptor ring) of the diagram according to the present invention . 接收描述符环621被用于向HOT单元250提供遗留缓冲区描述符。 Receive descriptor ring 621 to HOT Unit 250 is used to provide a legacy buffer descriptor. 可由HOT单元250向遗留缓冲区上载若干种类型的数据,包括非TCP帧、非委托连接上接收到的帧、在委托连接上接收的包含有的异常(意外的标记、失序、无效检查和等)的帧,和由DCT350上载的连接数据。 By HOT Unit 250 to the legacy buffer contained several types of data, including non-TCP frames received on a non-commissioned connected to the frame, connected to receive the commission contains some exceptions (unexpected token, disorder, ineffective inspection and so on ) frame, and the connection data generated by the DCT350 upload. 本领域技术人员应该理解,还可以使用其他机制向HOT単元250提供缓冲区描述符,例如,通知描述符的链接表、FIFO、共享存储器方案等。 Those skilled in the art would understand, other mechanisms may also be used to provide 250 yuan HOT radiolabeling buffer descriptor, for example, notification descriptor linked list, FIFO, shared memory programs.

[0086] 接收描述符环621包括若干项,所有的项在图6C中都表示为接收描述符环项623。 [0086] receive descriptor ring 621 includes a number of items, all of the items in Fig. 6C are represented as receive descriptor ring 623 entries. 每个接收描述符项623都包括指示该项是由HOT单元250所有还是由驱动程序255所有的“own”位。 Each receive descriptor entry 623 includes instructions that all by HOT Unit 250 or 255 by the driver all "own" bit. “own”位的功能在对图6A的说明中已有阐述。 "Own" bit function in the description of FIG. 6A has been described. 接收描述符写指针627与命令写指针607的功能将同,接收描述符读指针625与命令读指针605的功能将同 Receive descriptor write pointer 627 and the write pointer 607 command functions will work with, receive descriptor read pointer 625 and read pointer 605 command functions will work with

[0087] 除“own”位外,每个接收描述符环项623还包括遗留缓冲区描述符、接收控制和/或状态字段等。 [0087] In addition to "own" bits, each receive descriptor ring 623 items also include legacy buffer descriptor, receiving control and / or status fields and the like. 如对图5E的说明中所述,遗留缓冲区描述符包括规定在驱动程序存储空间235中的物理位置、遗留缓冲区长度和可选标记。 As described in the description of FIG. 5E, a legacy buffer descriptor including provisions in the driver storage space 235 physical locations, the legacy buffer length and a selectable marker.

[0088] 当接收描述符环项623被驱动程序255写入吋,接收描述符环项623可与其他位包括:请求HOT単元350在数据被上载到接收描述符环项623中规定的遗留缓冲区上时发送中断的位。 [0088] When the receive descriptor ring 623 entries written by the driver 255 inches, receive descriptor ring 623 items with other bits include: requesting $ 3.5 HOT radiolabeling the data is uploaded to the receive descriptor ring 623 items prescribed legacy buffer transmit interrupt bit on the region. 当接收描述符环项623被HOT单元350读出并更新时,接收控制和/或状态信息被更新,以便在有效载荷数据或被解析的帧数据被上载到遗留缓冲区时包括连接状态。 When receiving descriptor ring 623 entries are read out and updated in 350 HOT unit, receiving control and / or status information is updated so that the frame data or parsed payload data including connection status when it is uploaded to the legacy buffer. 上载到遗留缓冲区被HOT単元350为非委托连接写入的接收控制和/或信息可包括接收帧指示器的尾端、超出最大帧尺寸指示器等。 Uploaded to the legacy buffer is HOT 350 non-commissioned radiolabeling element connected to receive the write control and / or information may include receiving the tail end of frame indicator, exceeds the maximum frame size indicators. 上载到遗留缓冲区被HOT単元350为委托连接写入的接收控制和/或信息可包括启动缓冲区指示器、用户缓冲区不可用指示器、接收帧尾端、超区域ACK接收到指示器等 Uploaded to the legacy buffer is HOT radiolabeling element 350 is connected to the control commission received written and / or information may include starting buffer indicator, the user buffer unavailable indicator received frame ends, supra-regional ACK received indicators, etc.

[0089] 图6D是根据本发明ー个或几个方面的用于通过驱动程序255由TCP堆栈215向HOT单元250传送传送缓冲区信息的传送描述符环(transmit descriptor ring)的简图。 [0089] Figure 6D is a TCP stack 255 by the buffer 215 transmits the information to the transmission unit 250 transmits HOT descriptor ring (transmit descriptor ring) by the driver of the present invention or aspects ー accordance schematic. 该传送描述符环631被用于向HOT単元250提供传送缓冲区描述符。 The transfer descriptor ring 631 is used to HOT radiolabeling element 250 provides the transmit buffer descriptor. 本领域技术人员应该理解,还可以使用其他机制向HOT单元250提供缓冲区描述符,例如,通知描述符的链接表、FIFO、共享存储器方案等。 Those skilled in the art would understand, other mechanisms may also be used to provide HOT Unit 250 buffer descriptor, for example, notification descriptor linked list, FIFO, shared memory programs.

[0090] 传送描述符环631包括若干项,姆个项都在图6D中用传送描述符环项633表不。 [0090] transfer descriptor ring 631 includes a number of items, Salim entries have been transferred descriptor ring 633 entries in the table is not used in Fig. 6D. 姆个传送描述符项633包括“own”位,指示该项是由HOT单元250所有或是由驱动程序255所有。 Farm a transfer descriptor entry 633 includes "own" bit to indicate that all by HOT Unit 250 or 255 by the driver at all. “ own”位的功能在对图6A中的描述中已经说明。 "Own" bit function in the description of FIG. 6A has been described. エ传送描述符写指针637的功能与命令写指针607的功能相同,传送描述读指针635的功能与命令读指针605的功能相同。 EVAL transfer descriptor write pointer 637 functions and command functions the same as the write pointer 607, the transfer function described in the read pointer 635 and 605 of the same command read pointer function.

[0091] 除“own”位外,每个传送描述符环项633都包括传送缓冲区描述符、DCT索引、特定命令传送控制、传送控制/状态字段、传送缓冲区字节计数等。 [0091] In addition to "own" bits, each transfer descriptor ring 633 items includes transfer buffer descriptor, DCT index, specific command transmission control, transmission control / status field byte transmit buffer count and the like. 传送缓冲区描述符包括规定了在的应用程序空间227或存储了待传送帧数据的TCP堆栈存储空间225中的物理地址。 Transmission buffer descriptor including provisions in the application space 227 to be transmitted or stored in the frame data storage space TCP stack 225 physical address. HOT单元250利用该物理地址从驱动程序存储空间235中读出帧数据。 HOT unit 250 using the physical address 235 is read out from the driver frame data storage space. 特定传送控制可包括为传送引擎320的请求以便保留DCT350中的帧的首字节的序列号。 Specific transport control may include a request for the transfer engine 320 in order to preserve the first byte of the serial number DCT350 the frame. 当为了帧而接收到ACK时,HOT单元250可生成通知命令。 When the order is received in the frame ACK, HOT unit 250 may generate a notification command.

[0092] 传送控制和/或驱动程序255写入的状态字段可包括传送帧指示器的末端、启动TCP分段的一位、启动HOT单元250的TCP检查和计算的ー个或多个位、TCP分段中使用的最大段尺寸等等。 [0092] transmission control and / or drivers written state field 255 may include an indicator of the end of the transmission frame, start a TCP segment, TCP checks start HOT Unit 250 and calculated ー one or more bits, TCP segment using maximum segment size, and so on. 当传送描述符环项633被HOT単元250读出并更新时,特定传送控制和/或状态信息被更新以便包括特点传送状态。 When sending descriptor ring 633 entries are read and updated HOT radiolabeling unit 250, the specific transport control and / or status information is updated to include the characteristics of the transmission status. 特定传送状态可包括载体指示器的丢失、传送重试计数、重试错误等等。 Specific delivery status indicator may comprise a carrier loss, transmission retry count, retry error and so on.

[0093] 图7是根据本发明ー个或几个方面的包括有图3所示HOT単元250 —部分的方框图。 [0093] According to the present invention, FIG. 7 is a ー or several areas including radiolabeling element shown in Figure 3 HOT 250-- block diagram of portions. DCT350包括用于处理通过命令环601接收来自驱动程序255的CMD单元710。 DCT350 comprising ring 601 for processing by the command CMD receiving unit 255 from the driver 710. 委托连接信息存储在连接缓冲区表(CBT)715、连接数据表(⑶T)720和连接匹配表(CMT)725中的DCT350内。 Delegation connector 720 and connection information is stored matching table (CMT) in connection buffer table (CBT) 715, the connection data table (⑶T) within the 725 DCT350. CBT715、CDI720和CMT725内的委托连接项可被CMD单元710写入。 CBT715, CDI720 commissioned CMT725 and connections within the 710 items that can be written to CMD unit. CMT725存储委托连接标识信息并且CMT725在委托连接被设立时被CMD单元710写入。 CMT725 connection identification information storage delegation and CMT725 is written in 710 when the connection is established commission CMD unit. 在CMT725中,对应委托连接的项保持对应,只要连接保持委托或直到连接终止。 In CMT725 the item corresponding to entrust Remain connected as long as the connection remains delegate or until the connection is terminated. CMT725中的项包括目标IP地址、源IP地址、源TCP端ロ、目标TCP端ロ等。 CMT725 The entry includes the destination IP address, source IP address, source TCP end ro, ro and other target TCP end.

[0094] 当委托连接被建立时,CMD单元710初始化⑶T720中的项。 [0094] When the delegate connection is established, the term CMD unit 710 in the initialization ⑶T720. ⑶T720中的项包括委托连接的委托连接状态信息,如预期的序列号、ACK号、时戳数据、未承认帧(unacknowledged frames)的计数等。 ⑶T720 The items include the entrustment of the connection state connections, such as the expected sequence number, ACK number, time stamp data, frame (unacknowledged frames) count did not recognize the like. 当巾贞被构造为在委托连接传送时,0)1720中的项内的字段被传送引擎320读出并选择性地修改。 When the towel Zhen is configured to transmit commissioned connection, 0) in the 1720 entry in the field is transmitted engine 320 reads and selectively modified. 同样地,当委托连接上的进入帧被处理吋,CDT720中的项内的字段被接收引擎360读出并有选择地修改。 Similarly, when commissioned into the frame on the connection being processed inch field CDT720 the item is received within the engine 360 is read out and selectively modified. 当PRB命令为委托连接被接收到吋,CBI715内的项被CMD单元710用一个或多个用户缓冲区描述符写入。 When connected to a delegate PRB command is received inches term CBI715 CMD unit 710 is within one or more of the user buffer descriptor is written. 用户缓冲区信息被接收引擎360内的缓冲区上载单元745读出并有选择地修改。 User buffer information is received engine load unit 745 within the buffer 360 is read out and selectively modified.

[0095] 委托连接信息已经被分配在CBI715、⑶1720和CMT725之间,所采取的方式是断开由影响CBI715的用户缓冲区张贴(postings)到存储于CDI720的状态信息。 [0095] Principal connection information has been distributed among CBI715, ⑶1720 and CMT725, the way was taken off by the impact CBI715 user buffer posted (postings) to the state of the information stored in CDI720. 还有,由于状态信息是被接收引擎360基于最新接收到的帧更新的,传送引擎320和TCP堆栈215能够在帧构建期间存取当前状态信息。 Also, since the state information is received engine 360 based on the latest update of the received frame, and transmits the engine 320 during TCP stack 215 can access the status information of the current frame is built. 类似地,由于状态信息也是传送引擎320基于最新传送的帧更新的,接收引擎360和TCP堆栈215能够在帧构建期间存取当前状态信息。 Similarly, since the engine 320 transmits the state information is transmitted based on the latest update of the frame, the engine 360 and the receiving TCP stack 215 can access the status information of the current frame during build.

[0096] 在接收接ロ370内,缓冲区,接收FIF0730缓冲进入帧。 [0096] In the receiving access ro 370, buffer, receiving FIF0730 into the frame buffer. 接收接ロ370输出帧和有效帧指示器到接收引擎360内的解析单元735。 Ro then received 370 valid frames and output frame pointer to the parsing unit 360 receives the engine within 735. 预解析単元735解析有效帧,产生被部分解析的帧,并读出CMT725以确定是否该帧被接收到委托连接。 735 yuan pre-analysis radiolabeling resolve valid frame to generate frame is partially resolved, and reads CMT725 to determine whether the frame is connected to the commission received. 预解析単元735输出被部分解析的帧到解析単元740。 Pre-analysis radiolabeling yuan frame portion 735 outputs are resolved to resolve radiolabeling 740 yuan. 解析単元740确定每个被部分解析帧的协议类型,即,TCP、UDP、IP等,并选择性地解析该被部分解析的巾贞,产生解析巾贞和被部分解析的巾贞。 Analytical radiolabeling element 740 is determined and each is resolved towel Zhen Zhen towel partially resolved partially resolving frame protocol type, that is, TCP, UDP, IP, etc., and optionally resolves the partially resolved towel Chen, produce. 解析单元740读出CDI720,确定是否存在ー个或多个特殊事例,并输出被部分解析的帧、解析帧或帧,到缓冲区上载单元745。 Parsing unit 740 reads CDI720, to determine whether one or more specific instances ー exist, and outputs the parsed frame portion, analytical frame or frames onto the buffer loading unit 745. 解析単元740还选择性地设置存储在如通知単元750内寄存器的存储器件内的通知标记,如后详述。 Analytical radiolabeling element 740 also selectively set such notification tag stored in the internal register 750 yuan notice radiolabeling of memory device, such as described in detail later.

[0097] 缓冲区上载单元745读出CBI715并选择性地写入CBI715和CDI720。 [0097] the buffer load unit 745 reads CBI715 and selectively writes CBI715 and CDI720. 缓冲区上载単元745通过DMA引擎310上载帧、部分解析帧和解析帧到系统存储器130。 Upload radiolabeling yuan buffer DMA engine 745 through 310 load frame, part of the analytical frame to frame and resolve system memory 130. 缓冲区上载单元745基于存储于CBI715中的用户缓冲区描述符中或通过接收描述符环261接收到的来自驱动程序255的遗留缓冲区符中的数据规定写入系统存储器130的位置。 Based on the buffer load unit 745 is stored in CBI715 user buffer descriptor or write to system memory by receiving description data specifier ring 261 operator received a legacy from the driver buffer 255 position 130. 类似地,传送引擎320基于通过传送描述符环631接收的来自驱动程序255的传送缓冲区描述符规定系统存储器130中的读出位置。 Similarly, the transfer engine 320 based on the received symbol ring 631 described by transmitting the transmit buffer 255 from the driver's website provides system memory 130 reading position description. 通知单元750通过到驱动程序255的DMA引擎310输出通知到通知环611。 Notification unit 750 to the driver through DMA engine 310 outputs a notification to the notification ring 255 611.

[0098] 限制数量连接的委托连接信息被存储于CMT725,并在达到该限制数量时超出连接的连接信息仅被存储于系统存储器130内的CT245 (图2A或2B)中。 [0098] Principal connection information limits the number of connections in the connection information is stored CMT725, and this limit is reached when the number of connections exceeds CT245 only be stored in system memory 130 (FIG. 2A or 2B) of. 在一个实施例中,HOT単元250上载部分处理帧到驱动程序存储空间235中的遗留缓冲区,而非为了处理用于超出连接的进入和外出帧而存取CT245的连接信息。 In one embodiment, the carrier element 250 on HOT radiolabeling partially processed frames to the driver storage space left in the buffer 235, rather than for processing for inbound and outbound connections beyond the frame while the access connection information CT245. 在另ー实施例中,HOT单元250将DCT350作为高速缓存并根据需要存取CT245以便定位相对连接数据。 In another embodiment ー, HOT DCT350 unit 250 as a cache and the access CT245 needed to locate data relative to the connection. TCP堆栈215独立于HOT单元250执行的额外进入或外出帧的处理完成对部分处理的帧的处理。 TCP stack additional 215 independent execution unit 250 enters in HOT or out frame processing is completed partially processed frame processing. 因此,超出连接的遗留处理典型地以等于或高于在没有HOT単元250的情况下的计算系统200中的帧处理速率进行处理。 Accordingly, beyond the legacy processing is typically connected to the frame is equal to or higher than the processing rate calculation unit 250 the system does not HOT radiolabeling case 200 for processing. 该速率可能是更好的,因为帧是由HTO単元250部分处理的并且在接收接ロ370执行的帧有效性也被上载到驱动程序存储窨235。 The rate may be better, since the frame is part of the processing unit 250 HTO radiolabeling and 370 frames at the receiving access ro effectiveness of the implementation can be uploaded to the driver store scenting 235. 还有,驱动程序255和TCP堆栈215的遗留处理可以与HOT単元250的传送处理或接收单元250的接收处理同时进行。 Also, the driver 255 and the TCP stack legacy handle 215 may be performed simultaneously with the reception processing unit 250 HOT radiolabeling transmission or reception processing unit 250.

[0099] 图8A是根据本发明ー个或几个方面的处理有效帧的方法的简图。 [0099] FIG. 8A is a diagram or deal with several aspects of effective methods of ー frame according to the present invention. 在步骤801中有效帧被接收引擎360内的预解析单元735接收。 In step 801 valid frames are received within 360 engine pre-analysis unit 735 receives. 步骤803中预解析单元735确定该有效帧是否为TCP帧,如果不是,在步骤805中完成本领域技术人员所熟知的非TCP处理。 Step 803 pre-parsing unit 735 determines whether the frame is valid TCP frame, if it is not completed in step 805 are well known to those skilled in the processing of non-TCP. 在一个实施例中,接收引擎360完成接收的UDP帧的处理。 In one embodiment, the receiver engine 360 to complete the processing of the received UDP frames. 在另ー实施例中,接收引擎360完成另ー协议的处理。ー In another embodiment, the engine 360 finishes processing the received protocol ー other. 在又一实施例中,接收引擎360上载其他协议的帧到驱动程序存储空间235并通知驱动程序255。 In yet another embodiment, the receiving load on the engine frame 360 other agreement to the driver storage space 235 and signals the driver 255.

[0100] 如果在步骤803中,预解析単元735确定有效帧是TCP帧,那么在步骤807中,预解析单元735由CMT725中读ー个或多个项。 [0100] If, in step 803, the pre-analysis to determine the effective frame radiolabeling element 735 is TCP frame, then in step 807, the pre-analysis unit 735 is read by the CMT725 ー one or more items. 步骤809中预解析单元735确定TCP帧(以下称为“该帧”)是否在委托连接被接收,即,是否该帧与CMT725中的项匹配。 Step 809 pre-parsing unit 735 determines TCP frame (hereinafter referred to as "the frame") is being received in the delegate connection, i.e., whether the matching entry in the frame and CMT725. 预解析単元735从该帧中提取目标IP地址、源IP地址、源PCT端口和目标TCP端ロ并利用这些值在CMT725中查找匹配项。 735 yuan pre-analysis radiolabeling extract the target IP address from the frame, the source IP address, source and destination TCP port PCT ro terminal and use these values to find a match in CMT725 in. 匹配是指连接已经被委托。 Match is the connection has been commissioned. 如果步骤809中,预解析单元735确定该巾贞未在委托连接上被接收到,那么在步骤813中该帧的遗留处理完成。 If step 809, the pre-analysis unit 735 determines that the towel is not the delegate Jeong connection is received, then in step 813 to complete the frame of legacy processing. 预解析単元735借助解析单元740通过输出该帧到缓冲区上载单元745和指示该帧未在委托连接上接收来开始进行遗留处理。 Radiolabeling element 735 by means of a pre-analytical analysis unit 740 through the output of the frame buffer to the loading unit 745 and indicating that the frame is not connected to the commission to start receiving the legacy process. 缓冲区上载单元74通过DMA引擎310上载至少被部分解析的帧到驱动程序存储空间235并通知带有遗留处理的驱动程序255,如后详述。 Upload unit 74 through the buffer contained at least partially parsed frame memory 235 to the driver and signals the driver with the left handle 255, as described in detail later on DMA engine 310.

[0101] 在步骤809中,预解析単元735确定该帧在委托连接上被接收,然后在步骤811预解析単元735输出部分处理帧到解析単元740。 [0101] In step 809, the pre-analytical radiolabeling element 735 determines that the frame is connected to receive the commission, and then in step 811 the pre-analysis radiolabeling yuan output section 735 process to resolve radiolabeling frame element 740. 步骤811中解析単元740解析部分处理的帧,产生解析帧并确定是否存在特殊事例,例如,IP或TCP选择、无效标记等,如果是这样的,步骤812中解析单元输出解析帧到缓冲区上载单元745,指明存在特殊事例。 Step 811 resolves radiolabeling element 740 analysis section frame processing, generates analytical frame and determining whether specific instances exist, e.g., IP or TCP selecting invalid tag, etc., if so, in step 812 the parsing unit outputs parsed frame to Buffer Upload unit 745, indicating there is a special case. 步骤812中,缓冲区上载单元745在CBI720的项中设置“sync request”标记对应委托连接并刷新对应委托连接的CBI715中的项中的任何用户缓冲区描述符。 In step 812, the buffer loading unit 745 is set "sync request" mark corresponds to entrust the corresponding connection and refresh CBI715 commissioned items connected to any user buffer descriptor in CBI720 the item. 步骤813中,缓冲区上载单元745通过DMA引擎310上载解析帧到驱动程序存储空间235并通知驱动程序255遗留处理请求。 Step 813, the carrier unit 745 through the buffer DMA engine 310 to the driver load analytical frame memory 235 and notifies the driver 255 legacy process the request. 步骤812中为委托连接设置“sync request”标记表明委托连接是使用遗留处理被处理的。 Step 812 delegate connection settings "sync request" flag indicates that the connection is using the legacy commission process to be treated for. 接收引擎360不接受委托连接的用户缓冲区描述符命令,直到sync request标记被未来的缓冲区张贴事件清除,后面将结合图9A进ー步说明。 Engine 360 receives the user does not accept the delegation connector buffer descriptor command until the next sync request flag is cleared buffer posted events later in connection with FIG. 9A into ー-step instructions.

[0102] 如果在步骤811中解析单元740确定不存在特殊事例,那么在步骤815中解析单元740读出⑶1720中对应委托连接的项。 [0102] If the resolution unit 740 in step 811 determines the particular case does not exist, then the analytic reading unit 740 corresponding item ⑶1720 commissioned connected in step 815. 步骤817中解析单元740和缓冲区上载单元745确定哪个存储于通知単元750中的通知标记被设置(如有),后面将结合图9A具体说明。 Step 817 analysis unit 740 and the buffer 745 loading unit to determine which storage element 750 in the notification radiolabeling notification flag is set (if any), followed by specific instructions in conjunction with Figure 9A. 步骤819中解析単元740确定由TCP帧中提取的序列号(SN)是否不等于存储于⑶1720中对应委托连接的项中的序列号(DCT SN),如果是这样,步骤821中解析单元740请求失序恢复。 Step 819 Analytical radiolabeling element 740 is determined to extract from the TCP frame sequence number (SN) is not equal to the storage in ⑶1720 of items connected to the serial number of the corresponding commission (DCT SN), and if so, step 821 resolution unit 740 requests disorder recovery. 后面将结合图SB进ー步说明失序恢复。 Later in conjunction with FIG SB-step instructions into ー disorder recovery.

[0103] 如果在步骤819中解析单元740确定SN等于DCT SN,那么在步骤823中解析单元740输出解析帧到缓冲区上载单元745。 [0103] If, in step 819 determines the parsing unit 740 SN equal DCT SN, then in step 823, the output unit 740 resolves to resolve frame buffer loading unit 745. 步骤823中缓冲区上载单元745读出CBI715中对应委托连接的项。 Buffer Upload Unit 745 in step 823 reads out the corresponding item CBI715 delegate connection. 步骤825中,缓冲区上载单元745确定用户缓冲区是否可用。 In step 825, the buffer loading unit 745 determines that the user buffer is available. 术语“用户缓冲区”与术语“HOT缓冲区”是可互換的。 The term "user buffer" with the term "HOT buffers" are interchangeable. 如果HOT缓冲区不可用,那么步骤827中缓冲区上载单元745或者等待HOT缓冲区变为可用或通过DMA引擎310上载解析TCP帧到遗留缓冲区,后面将结合图8C对此进ー步说明。 HOT buffer if not available, then step 827 is contained in the buffer unit 745 HOT or wait for the buffer to become available or by the DMA engine 310 contained resolve to legacy TCP frame buffer, this later in connection with FIG. 8C into ー-step instructions.

[0104] 如果在步骤825中缓冲区上载单元745确定HOT缓冲区可用,那么在步骤829中缓冲区上载单元745完成解析帧的处理并上载至少有效载荷数据的一部分到HOT缓冲区,后面将结合图8D对此进ー步说明。 [0104] 745 to determine if HOT buffers available in the buffer step on the loading unit 825, then the buffer unit 745 to complete the upload process analytical frame at step 829 and upload at least a portion of the payload data to the HOT buffers, later combined Figure 8D of this into ー-step instructions. 在步骤831中,在上载有效载荷数据的部分到HOT缓冲区后,缓冲区上载单元745确定是否在解析帧中存在额外的有效载荷数据,如果是这样,重复步骤825和829。 In step 831, the portion of payload data uploaded to the HOT buffer, the load on the buffer unit 745 determines whether there is extra payload data in the analysis frame, and if so, repeat steps 825 and 829. 如果在步骤831中缓冲区上载单元745确定所有的有效载荷数据都已经被上载到ー个或多个HOT缓冲区了,那么在步骤833中缓冲区上载单元确定解析帧中TCP “压栈”标记是否已经被断言。 If the carrier unit 745 determines that all the payload data has been uploaded ー HOT or more buffers in the buffer in step 831, then the frame is determined to resolve TCP "push" flag in step 833 Buffer Upload Unit whether it has been asserted. 如果在步骤833中缓冲区上载单元745确定TCP “压栈”标记被断言,缓冲区上载单元745为CBI715中对应委托连接的项设置“sync request”标记并刷新所有CBI715中对应委托连接的项中的缓冲区描述符。 If the buffer is uploaded unit 745 determines TCP "push" flag is asserted at step 833, the buffer loading unit 745 for an item all CBI715 corresponding delegate entry to entrust connection connection "sync request" flag and refresh in the corresponding CBI715 The buffer descriptor. 如果在步骤833中缓冲区上载单元745确定TCP “压栈”标记没有被断言,那么接收引擎360进行步骤801。 If the carrier unit 745 determines TCP "push" on the buffer in step 833 flag is not asserted, the receiver engine 360 proceeds to step 801.

[0105] 图SB是根据本发明ー个或几个方面的处理失序帧的方法的简图。 [0105] FIG. SB is a diagram of a method or process sequence frames ー several aspects of the present invention. 本领域技术人员所容易理解,图8B中的方法步骤构成了执行图8A的步骤821的ー种方式。 The skilled artisan readily appreciate that the method steps in Figure 8B constitute steps in FIG. 8A ー ways 821. 失序恢复处理包括SN大于存储在DCT SN中的值所预期的(例如,由ー个或多个丢失帧造成的)情况或SN小于DCT SN (例如,由于传输超时或丢失ACK造成的帧的重复传送)的情況。 SN disorder restoring processing includes values stored in the DCT SN is greater than the expected (e.g., by a loss of one or more frames ー result) or SN smaller than the case of DCT SN (e.g., due to transmission timeout or lost frames due to the repeated ACK transfer) situation. 当SN大于DCT SN时,接收引擎360执行“快速恢复”(fast recovery)算法,包括传输最多3个连续相同构造的ACK (每个都对应按顺序接收到的最后帧),使为委托连接的用户缓冲区失效(刷新),并上载整个帧到ー个或多个遗留缓冲区。 When SN is greater than the DCT SN, received engine 360 to perform "quick recovery" (fast recovery) algorithms, including transport up to three consecutive identical structure of ACK (each pair should be sequentially received last frame), so that the delegate connection user buffer failure (refresh), and upload the entire frame to ー or more legacy buffers. 当SN小于DCT SN时,接收引擎360为帧传送ACK,为委托连接无效用户缓冲区,并上载整个帧到ー个或多个遗留缓冲区。 When SN is less than DCT SN, received 360 engine for the frame transmission ACK, the delegate connection invalid user buffer, and upload the entire frame to ー or more legacy buffers.

[0106] 在步骤830中解析单元740确定从帧中提取的SN是否小于由图8A的步骤815中的⑶1720读取的DCT SN,如果是这样,在步骤832中解析单元740向传送引擎320发出信号,生成帧的ACK。 [0106] In step 830, the parsing unit 740 determines the SN extracted from the frame is smaller than the step 815 of FIG. 8A ⑶1720 read DCT SN, and if so, in step 832 the parsing unit 740 issued to delivery engine 320 signal, generates a frame of ACK. 步骤832中,解析单元740还输出解析帧到缓冲区上载单元745并执行步骤838。 Step 832, the parsing unit 740 also outputs the parsed frame buffer 745 to the carrier unit and the step 838. 如果在步骤830中解析740确定从帧中提取的SN不小于DCT SN,那么步骤834中,解析单元740确定在步骤815中由⑶1720读出的“Fast ACK”值小于3,如果是这样,解析单元740给传送引擎320发信号,为步骤836中的解析帧生成ACK。 If the resolution is determined in step 830 740 SN extracted from the frame is not less than DCT SN, then in step 834, the parsing unit 740 determines in step 815 is read out from ⑶1720 "Fast ACK" value of less than 3, and if so, parsing unit 740 to transmit engine 320 signals, generates an ACK to step 836 resolution frames. 还有在步骤836中,解析单元740输出解析帧到缓冲上载单元745并指示缓冲区上载单元745将存储于CDI720中的“Fast ACK”递增。 Also in step 836, the output parsing unit 740 to the frame buffer to resolve the loading unit 745 and instructs the load unit stored in the buffer 745 CDI720 the "Fast ACK" increments.

[0107] 步骤838中缓冲区上载单元745刷新任何对应存储于CBI715中的委托连接的HOT缓冲区。 Buffer Upload Unit [0107] 838 745 refresh any correspondence stored in CBI715 Principal connection HOT buffer. 步骤840中缓冲区上载单元745设置CBI715中对应委托连接的“sync request”标记并有选择地为委托连接更新连接存储于⑶1720中的状态数据,例如,Fast ACK、DCTSN、ACK号等。 In step 840 Buffer Upload Unit 745 set CBI715 connection corresponding commission "sync request" tag and selectively connected to entrust Update Connection stored in ⑶1720 the state data, for example, Fast ACK, DCTSN, ACK number. 步骤813中,缓冲区上载单元745通过DMA引擎310上载解析TCP帧到驱动程序存储空间235并通知驱动程序255遗留处理请求。 Step 813, the carrier unit 745 through the buffer DMA engine 310 contained resolve TCP frame memory 235 to the driver and inform the driver 255 legacy process the request.

[0108] 图SC是根据本发明ー个或几个方面的等待用户缓冲区的方法的简图。 [0108] FIG ー SC or diagram is waiting for the user buffer method aspects of the present invention. 本领域技术人员所容易理解,图8C中的方法步骤构成了执行图8A的步骤827的ー种方式。 The skilled artisan readily appreciate that the method steps in FIG. 8C constitutes the steps of FIG. 8A ー ways 827. 接收引擎360等待用户缓冲区而不是上载解析帧或部分解析帧到遗留缓冲区,因为数据上载用户缓冲区不需要通过TCP堆栈由驱动程序存储空间235拷贝到应用程序存储空间227。 Engine 360 receives the buffer instead of waiting for the user to upload or partially resolving analytical frame buffer frame to the left, because the data does not need to upload the user buffer by the driver via TCP stack memory 235 are copied to the application storage space 227. 还有,ー但使用遗留缓冲区处理委托连接,TCP堆栈215必需响应“sync request”标记,导致在委托连接的解析巾贞或部分解析巾贞会被接收引擎360上载到用户缓冲区之前,“sync request”标记被清空。 Also, ー but using legacy buffer handling commissioned connection, TCP stack 215 necessary response "sync request" mark, resulting in a towel before parsing delegate Jeong-linked or partially resolving the towel Zhen will be uploaded to the user buffer 360 receives the engine. " sync request "flag is cleared.

[0109] 步骤850中,缓冲区上载单元745确定图8A中的步骤823中读出的对应委托连接的项中是否设置了“ request buffer”。 [0109] step 850, the carrier unit 745 determines the buffer corresponds to step 823 of FIG. 8A delegate read out the connection entry is set up "request buffer". 如果“request buffer”标记没有被设置,那么步骤852中缓冲区上载单元745设置CBI715中对应委托连接的项中的请求缓冲区标记并初始化缓冲区请求计时器到一个存储在寄存器中的值。 If the "request buffer" flag is not set, then step 852 Buffer Upload Unit 745 set CBI715 corresponding delegate entry in the connection request buffer mark and initializes the buffer request timer to the value stored in the register. 在另ー实施例中缓冲区请求计时器被初始化到一个存储在CBI715中对应委托连接的项中值。 In another embodiment ー buffer request timer is initialized to a value stored in the corresponding entry in the CBI715 the delegation connector. 如果在步骤850中,缓冲区上载单元745确定“request buffer”标记被设置,那么缓冲区上载单元745进入步骤862。 If, at step 850, the buffer loading unit 745 determines "request buffer" flag is set, then the buffer loading unit 745 proceeds to step 862.

[0110] 步骤854中缓冲区上载单元745上载一定数量的由“startup limit”(启动限制)值通过DMA引擎310确定到遗留缓冲区的字节。 [0110] step 854 Buffer Upload Unit on 745 contained a number of the "startup limit" (Start limit) value is determined by the DMA engine 310 to a legacy buffer bytes. 该启动限制值由TCP堆栈215初始化,被存储于⑶1720中对应委托连接的项中。 The start limit values initialized by the TCP stack 215, is stored in the corresponding entry ⑶1720 delegate connection. 步骤856中,缓冲区上载单元745设置“requestbuffer”标记存储于通知単元750中并且通知単元750通过通知环向驱动程序255发送通知。 Step 856, the carrier unit 745 is provided on the buffer "requestbuffer" mark is stored in the notification radiolabeling 750 yuan and 750 yuan radiolabeling notification by notifying the ring 255 to the driver sends a notification. 该通知包括与相关遗留缓冲区描述符的标签字段中使用的标签值相同的标签值。 The notification includes the value of the same label label label field value associated with legacy buffer descriptor in use. 通知単元750在发送通知后清空通知标记。 Notify radiolabeling element 750 after sending the notification Empty notification tag. 采用本领域技术人员公知的技术以确保驱动程序255收到对应通知前解析帧被上载到驱动程序存储空间235。 Using well-known to those skilled in the art to ensure that pre-parsed frame driver 255 receives the corresponding notification is uploaded to the driver memory 235.

[0111] 步骤858中缓冲区上载单元745确定指示接收FIF0730的“ fullness”值是否大于限定值,例如,“高水位”标志,并且,如果是这样缓冲区上载单元745进入步骤862。 Buffer Upload Unit [0111] Step 858 determines 745 indicating the reception FIF0730 of "fullness" value is greater than the limit value, for example, the "high water" mark, and, if it is on such a buffer loading unit 745 proceeds to step 862. 在一个实施例中,高水位值是固定的。 In one embodiment, the high water level is fixed. 在另ー实施例中,高水位值存储于驱动程序255编程的寄存器中。ー In another embodiment, the high level value is stored in the driver program 255 registers. 如果在步骤858中,缓冲区上载单元745确定指示接收FIF0730的“fullness”的值不大干“高水位”值,那么在步骤860中,缓冲区上载单元745确定缓冲区请求计时器已经期满,那么在步骤862中缓冲区上载单元745设置存储于CBI715的“sync request”标记和存储于通知単元750中的遗留标记。 If, at step 858, the carrier unit determines the buffer 745 instructs the "fullness" of value FIF0730 little dry "high water mark" value, then in step 860, the buffer loading unit 745 determines the buffer request timer has expired Legacy mark, then in step 745 Buffer Upload Unit 862 settings are stored in that CBI715 the "sync request" tag and stored in the notification radiolabeling of 750 yuan. 步骤813中,缓冲区単元745通过DMA引擎410上载解析帧到到驱动程序255,通知単元750清空通知标记,并且接收引擎360返回图8A中的步骤801。 In step 813, the buffer 745 by radiolabeling membered upload DMA engine 410 to parse the frame 255 to the driver, the notification radiolabeling empty notification tag element 750, and receives the engine 360 returns step 801 in FIG. 8A. 通知包括由相关遗留缓冲区描述符的标签字段中使用的标签值相同的标签值。 Notice including labeling the value used by the relevant label field legacy buffer descriptor in the same tag value. 采用本领域技术人员公知的技术以确保驱动程序255收到对应通知前解析帧被上载到驱动程序存储空间235。 Using well-known to those skilled in the art to ensure that pre-parsed frame driver 255 receives the corresponding notification is uploaded to the driver memory 235.

[0112] 如果步骤860中缓冲区上载单元745确定缓冲区请求计时器没有期满,步骤864中缓冲区上载单元745确定用户缓冲区是否可用,S卩,如果应用程序已经通过命令环张贴了用户缓冲区。 [0112] Buffer Upload Unit 745 determines if step 860 buffer request timer has not expired, in step 864 Buffer Upload Unit 745 determines the user buffer is available, S Jie, if the application has been posted by a user command ring buffer. 如果用户缓冲区不可用,缓冲区上载单元745完成解析帧的处理并在步骤829中上载有效载荷到用户缓冲,如后面对图8D中进ー步详述。 If the user buffer is not available, the buffer loading unit 745 completes processing analytic frame and upload to the user buffer payload, such as into the rear face of FIG. 8D ー steps detailed in step 829. 步骤829后,接收引擎360回到图8A的步骤831中。 Step 829, the receiving engine 360 back to step 831 of FIG. 8A.

[0113] 图8D是根据本发明ー个或几个方面的完成用户缓冲区处理的方法的简图。 [0113] FIG. 8D is a diagram of the method according to the invention or aspects ー complete user buffer processed. 本领域技术人员所容易理解,图8D中的方法步骤构成了执行图8A的步骤829的ー种方式。 The skilled artisan readily appreciate that the method steps in FIG. 8D constitutes the steps of FIG. 8A ー ways 829. 当HOT单元250接收进入数据时,应该将ACK发送到发送者(目标连接),以确保指示应该有多少数据送到HOT单元250接收窗ロ保持足够的打开宽度以允许接收路径的饱和。 When HOT Unit 250 receives the incoming data, it should send an ACK to the sender (connection target), to ensure that should indicate how much data is sent to HOT Unit 250 receives the window open ro maintain sufficient width to allow saturation of the receive path. 当接收到的帧没有及时被承认时,接收窗ロ可能需要被关闭,否则发送者内的重发计时器可能会期满,导致发送者重发ー个或多个帧。 When the received frame is not to be recognized in time, the reception window ro may need to be closed, otherwise the sender within the retransmission timer expires may lead sender retransmit ー one or more frames.

[0114] 除对发送者传送ACK之外,驱动程序255在帧被基于序列号和计时器等被HOT单元250接收到时被通知。 [0114] In addition to the transfer ACK to the sender, the driver 255 is when the serial number and timers are received HOT Unit 250 to be notified based on the frame. ⑶1720通过由接收的TCP有效载荷尺寸增加DCT SN被更新,对未承认帧的计数递增,并且向接收到的帧提取的最近接收到的TCP时戳被存储到为了委托连接的⑶1720,如果TCP时戳选择被适合地包括在接收到的帧中。 ⑶1720 received by the TCP payload size increase DCT SN is updated to not recognize the frame count is incremented, and the received frames extracted recently received TCP timestamp is stored in the order entrusted ⑶1720 connection, if TCP time stamp selection is suitably included in the received frame.

[0115] 步骤876中解析単元740确定未承认帧的计数是否大于未承认帧的限制值,如果是这样,进入步骤880。 [0115] Step 876 yuan resolve radiolabeling frame 740 is determined not to admit the count is greater than the limit value does not recognize frames, and if so, proceeds to step 880. 未承认帧限制值为连接存储于⑶1720中并且由TCP堆栈215确定。 It did not recognize the connection limit is stored in the frame ⑶1720 in and determined by the TCP stack 215. 在另ー实施例中,解析単元740确定委托连接上接收到的未承认帧的计数是否大于或等于步骤876中的未承认帧限定值。 In another embodiment ー, 740 yuan resolve radiolabeling determine entrust connection received unacknowledged frame count is greater than or equal to 876 frames defined step value is not recognized. 在另ー实施例中,缓冲区上载单元745确定未承认帧的计数是否大于未承认帧的限定值。ー In another embodiment, the buffer loading unit 745 determines whether or not to admit the frame count is greater than the limit value unacknowledged frame.

[0116] 如果在步骤876中解析单元740确定未承认帧的计数小于或等于未承认帧限定值,那么解析単元740确定在步骤878中传送计时器是否已经期满。 [0116] If in step 876 the parsing unit 740 determines not to admit the frame count is less than or equal to the frame defined value is not acknowledged, then resolve radiolabeling element 740 to determine whether the transmission timer has expired in step 878. 传送ACK计时器被配置为在发送者的重传送计时器期满前期满,目的是将由于发送者未能及时接收ACK最小化不必要的重新传送。 ACK transmit timer is configured to expire before the sender's heavy transmission timer expires, the purpose is to due to the failure to receive the ACK sender minimizing unnecessary re-transmission. 在一个实施例中,传送ACK计时器的期限对于所有委托连接是ー个常数。 In one embodiment, the transmission ACK timer deadline for all delegate connection ー constant. 在另ー个实施例中,传送ACK计时器的期限可由TCP堆栈为每个委托连接的期限编程。 In another embodiment, a time limit ー transmit ACK timer can be programmed as a term TCP stack connected to each delegate.

[0117] 如果在步骤878中解析単元740确定传送ACK计时器已经期满,那么解析単元740向传送引擎320发信号,为步骤880中的解析帧生成ACK并且传送引擎320输出解析帧到缓冲区上载单元745。 [0117] If the resolution radiolabeling element 740 in step 878 determines the transmission ACK timer has expired, then resolve radiolabeling 740 yuan to 320 signal transmission engine, frame generating ACK to step 880 and transmit engine 320 resolution output to resolve frame buffer Upload unit 745. 步骤882中,缓冲区上载单元745通过将其设置零将更新为连接存储于⑶1720的项中的未承认帧计数并更新“ last ACK sent”值到由帧提取的SN值。 Step 882, the carrier unit on the buffer 745 by setting it to zero will be updated to connect items stored in the ⑶1720 not recognize the frame count and update the "last ACK sent" value to the SN value extracted by the frame. 缓冲区上载单元745还更新连接状态数据如递增ACK号和递增序列号等,并在进入步骤886前重置传送ACK计时器。 Upload a buffer unit 745 also updates connection state data such as increasing number ACK and increasing sequence number, etc., and proceeds to step 886 to reset the transfer before ACK timer.

[0118] 如果在步骤878中缓冲区上载单元745确定传送ACK计时器没有期满,那么缓冲区上载单元745更新步骤884中在CDI720中对应委托连接的项,例如,通过更新未承认帧的计数等。 [01] If in step 878 Buffer Upload Unit 745 determines transmit ACK timer has not expired, then the buffer loading unit 884 in step 745 to update the corresponding entry in CDI720 delegation connector, for example, by updating count of frames not acknowledged in and so on.

[0119] 步骤886中有效载荷数据被缓冲区上载单元745通过DMA引擎310上载到应用程序存储空间227中的HOT缓冲区TCP。 [0119] Step 886 payload data is uploaded on the buffer unit 745 by DMA engine 310 uploaded application storage space 227 HOT buffer TCP. 步骤888中通知单元750确定是否设置了任何通知标记,并且,如果是这样,在步骤890中,通知単元750通过通知环发送通知到驱动程序255。 Step 888 determines whether the notification unit 750 mark set any notice, and, if so, in step 890, the notification radiolabeling 750 yuan to send notifications via the notification ring to the driver 255. 通知单元750构造事件通知描述符,包括通知标记、传送窗ロ尺寸、SN、最后的ACK号、TCP时戳值、来自遗留描述符的标签值等等。 Notification unit 750 configured event notification descriptor, including notification marks ro transfer window size, SN, the last ACK number, TCP timestamp value, the tag value from legacy descriptors like. 通知単元750在发送通知后清空通知标记。 Notify radiolabeling element 750 after sending the notification Empty notification tag.

[0120] 通知単元750输出事件通知描述符到DMA引擎310,该DMA引擎310将事件通知描述符传送给存储中驱动程序存储空间235内的卸载事件通知环。 [0120] 750 yuan notice radiolabeling output event notification descriptor to the DMA engine 310, the DMA engine 310 event notification descriptor conveys to store driver unload event storage space 235 within the notification ring. 卸载事件通知环被组织为毗邻内存块形式的环形队列。 Unload event notification ring adjacent to the memory block is organized as a form of circular queue. HOT単元250写入卸载事件通知环并且驱动程序255读出卸载事件通知环。 HOT radiolabeling unit 250 writes unload event notification ring and the driver 255 reads unload event notification ring. TCP堆栈215可使用由卸载事件通知环读出的数据更新CT245,从而保持CT245和DCT350之间的一致。 TCP stack 215 may be used by the unload event notification ring read data update CT245, CT245 and to maintain consistency between DCT350. TCP堆栈215还可通过上载⑶1715中的项到ー个或多个遗留缓冲区保持CT245和DCT350之间的一致 TCP Stack 215 can also upload the item to ー ⑶1715 or more legacy buffers consistency between CT245 and DCT350

[0121] 步骤890之后,接收引擎360回到步骤801处理其它有效帧。 After the [0121] Step 890, the engine 360 returns to step 801 to receive treatment other valid frame. 如果在步骤888中,通知単元750确定ー个或多个通知标记不被设置,那么接收引擎360回到步骤801处理另 If, at step 888, the notification determining ー radiolabeling 750 yuan or more notification flag is not set, then the receiver returns to step 801 processing engine 360 other

一个有效中贞。 An effective in infidelity.

[0122] 图9A是根据本发明ー个或几个方面的确定通知的方法的简图。 [0122] FIG. 9A is a schematic diagram or method of determining aspects ー notified in accordance with the present invention. 本领域技术人员所容易理解,图9A中的方法步骤构成了执行图8A的步骤817的ー种方式。 The skilled artisan readily appreciate that the method steps in FIG. 9A constitute steps in FIG. 8A ー ways 817. 可使用阈值控制HOT单元250接收的到驱动程序255或TCP堆栈215的通信序列号频率,而非中断CPUllO以通过驱动程序255通知TCP堆栈215每个帧都已经被目标接收了。 You can use HOT threshold control unit 250 receives the driver to communicate serial number 255 or 215 TCP stack frequency, rather than interrupt CPUllO notice by drivers 255 215 TCP stack have been the target of each frame received. 类似地,ACK阈值或特定ACK号可被用于控制HOT单元250接收的到驱动程序255的通信,而非中断CPUllO通知驱动程序255每个传送帧的ACK都已经被接收了。 Similarly, ACK ACK threshold or a particular number can be used to control the HOT unit 250 receives the communication driver 255, rather than interrupt CPUllO notification driver 255 transmits each ACK frame have been received.

[0123] 帧处理期间减少中断CPUllO的频率使CPUllO能够解放出来用于执行其他应用程序,典型的是通过CPUllO执行的应用程序指令的数量从而提高这些应用程序的性能。 Reducing the frequency of interruption CPUllO during [0123] frame processing can make CPUllO freed for execution of other applications, typically through the application CPUllO number of instructions executed to improve the performance of these applications. 阈值允许在确定中断间的平衡时一定的灵活性,以通知TCP堆栈215接收连接状态并为委托连接传送连接状态。 Threshold allows determining when the balance between interrupts some flexibility to notice TCP stack 215 receives the connection state and to entrust connectionless transport connection status.

[0124] 步骤901中解析单元740确定传送窗ロ是否自右收缩。 [0124] Step 901 resolution unit 740 determines whether the transfer window ro shrink from the right. 当由帧提取的ACK号与由帧中提取的接收窗ロ尺寸的总和小于为委托连接存储在CDI720中的最大传输窗ロ尺寸时,解析単元740确定传送窗ロ自右收缩。 When the total number of ACK frames extracted from ro receive window size by frame extraction is less than the commission attached storage CDI720 maximum transmission window size when ro parse radiolabeling 740 yuan from the right to determine the transfer window ro contraction. 缓冲区上载单元745使用从帧中提取的传送窗ロ尺寸更新为委托连接存储于⑶1720的最大传送窗ロ尺寸。 Upload a buffer unit 745 extracted from the frame transfer window size is updated to entrust ro-attached storage to the maximum transfer window ⑶1720 of ro dimensions. 如果在步骤901中,解析单元740确定传送窗ロ自右收缩,那么在步骤903中,解析单元740在通知单元750中设定传送窗ロ通知标记。 If in step 901, the parsing unit 740 determines the transfer windows ro contraction from the right, then in step 903, the parsing unit 740 is set in the transmit window ro notification tag notification unit 750.

[0125] 步骤905中,解析单元740确定重复ACK (在一个或多个接收帧中相同ACK号)已经被接收,指示出目标被请求重新传送ー巾贞或多巾贞。 [0125] Step 905, the analysis unit 740 determines duplicate ACK (the same as in one or more received frame number ACK) has been received, indicating that the target is requesting retransmission ー towel or towel Zhen Zhen. 如果在步骤905中,解析单元740确定重复ACK已经被接收,那么在步骤903中解析单元740在通知单元750中设置“重复ACK通知”标记。 If in step 905, the parsing unit 740 determines duplicate ACK has been received, then in step 903 the parsing unit 740 is set in the notification unit 750 "Duplicate ACK notice" tag. [0126] 步骤907中,解析単元740确定SN是否大于阈值,例如,限定值、指示递增序列号的阈值。 [0126] Step 907, the parsing radiolabeling element 740 SN is greater than a threshold value to determine, for example, the limit value, indicating the threshold increasing sequence number. 递增序列号在委托连接被设置时由TCP堆栈215初始化并被缓冲区上载单元745在每当向驱动程序255发送通知时被更新。 The serial number is incremented when the commission was set up TCP stack 215 connection initialization and loading unit 745 whenever the notification is sent to the driver 255 is updated on the buffer is. 在一个实施例中,递增序列号通过将递增序列号増加ー个序列増加值而完成更新。 In one embodiment, the serial number is incremented by incrementing serial number to increase in sequence ー zo bonus and complete the update. 序列增加值可以是固定的或被TCP堆栈215编程的。 Sequences added value may be fixed or programmable TCP stack 215. 如果步骤907中,解析单元740确定SN大于阈值,那么序列号阈值标记在步骤903中设定。 If in step 907, the parsing unit 740 determines SN greater than the threshold value, then the sequence number flag setting a threshold value in step 903.

[0127] 步骤909中,解析单元740确定存储于⑶1720中的最后的ACK号(接收到的委托连接的最高级ACK号)是否大于ー个限定值,该限定值指示递增ACK号。 [0127] Step 909, the parsing unit 740 is stored in ⑶1720 determine the final number of ACK (ACK received the highest number of delegate connected) is greater than a limit value ー, the limit value indicating ACK number is incremented. 当委托连接被设定时ACK号被TCP堆栈215初始化并被缓冲区上载单元745在接收到ACK时即时更新。 When the connection is set to No. ACK commissioned by the TCP stack mounted unit 215 is initialized by the buffer 745 when receiving the ACK instant updates. 递增ACK号在委托连接被设定时被TCP堆栈215初始化并被缓冲区上载单元745在通知被送到TCP堆栈215时即时的更新。 ACK number is incremented when the connection is set to be commissioned 215 initialize and load the TCP stack buffer unit 745 in the notification is sent to the TCP stack 215 instant updates. 在一个实施例中,递增ACK号通过将递增ACK号增加ー个ACK増加值而完成更新。 In one embodiment, the incremental ACK number and complete the update by the increasing number one ACK ACK increase ー zo bonus. ACK增加值可以是固定的或被TCP堆栈215编程的。 ACK increase value may be fixed or programmable TCP stack 215.

[0128] 步骤909中,解析单元740还可确定存储于⑶1720中的最后ACK号是否大于另ー限定值,该另ー限定值指示TCP堆栈215编程的特定ACK号。 [0128] Step 909, the parsing unit 740 may also determine stored in ⑶1720 last ACK number is greater than the other ー limit, indicating that another ー TCP stack 215 defines a specific ACK number programmed. 如果在步骤909中,解析単元740确定最后ACK号大于限定值(指示递增ACK号)或另ー限定值(指示特定ACK号),然后在步骤903中设置ー个ACK阈值。 If, at step 909, the parsing radiolabeling 740 yuan to determine the final ACK number is greater than the limit value (indicating ACK incremented number) or another ー limit (indicating a specific ACK number), then set in step 903 ー a ACK threshold.

[0129] 步骤911中,解析単元740确定ー个或多个计时器是否期满。 [0129] Step 911, the parsing radiolabeling determine ー 740 yuan or more timer has expired. 接收ACK计时器被配置为在TCP堆栈215重传送计时器期满前期满,这是为了尽量減少不必要的重传送。 Receiving ACK timer is configured to expire before the TCP stack 215 heavy transmission timer expires, which is to minimize unnecessary re-transmission. 存储于寄存器中为了委托连接的接收ACK计时器的期限可被TCP堆栈215编程并可为委托连接基于往返时间。 Stored in the register in order to entrust to receive ACK timer duration can be connected to the TCP stack 215 can be programmed based on round-trip time for the delegation connector. 接收存储于寄存器中的委托连接的SN计时器的期限可被TCP堆栈215编程。 Receiving timer register is stored in the SN delegate period connected TCP stack 215 can be programmed. 在另ー实施例中,接收ACK计时器和接收SN计时器的期限被存储在CMT725中对应委托连接的项中。 Term In another embodiment ー receiving ACK timer and receive SN timer is stored in the corresponding entry CMT725 entrust connection.

[0130] 如果在步骤911中,解析単元740确定计时器已经期满,那么在步骤903中更新一个对应的通知标记并且解析单元740进入步骤913。 [0130] If, in step 911, the parsing radiolabeling element 740 determines that the timer has expired, then a corresponding update notification tag in step 903 and the analysis unit 740 proceeds to step 913. 例如,如接收SN计时器期满时,“sequence number threshold”(序列号阈值)标记被设置并且在接收ACK计时器期满时,“ACK threshold”标记被设置。 For example, when receiving SN timer expires, "sequence number threshold" (serial number threshold) flag is set and when receiving ACK timer expires, "ACK threshold" flag is set. 如果在步骤911中,接收引擎360确定没有计时器期满,那么解析単元740在步骤913中输出解析帧到缓冲区上载单元745并且缓冲区上载单元745确定压栈标记是否被从被断方的帧中提取。 If, at step 911, the reception engine 360 determines that no timer expires, then the output resolves radiolabeling 740 yuan to resolve the frame buffer in step 913 the loading unit 745 and the buffer load unit 745 determines whether the push marks from the off side frame extraction. 如果压栈标记被断言,那么压栈通知标记在步骤903中被设置并且缓冲区上载单元745进入图8A的步骤819。 If the push flag is asserted, then push notification flag is set in step 903 and the buffer 745 steps into the loading unit 819 of FIG. 8A.

[0131] 图9B是根据本发明ー个或几个方面的使用户缓冲区跟随遗留处理同步的方法步骤的简图。 [0131] FIG. 9B is ー or allowing users to follow the legacy buffer handle synchronization method steps diagram of several aspects of the present invention. 如前述,当解析単元740检测到帧具有大于预期序列号DCT SN的序列号时会利用遗留缓冲区处理委托连接。 As mentioned above, when resolves radiolabeling element 740 detects a frame having more than the expected sequence number DCT SN will use the serial number to connect legacy buffer handling delegate. 例如,当ー个或多个帧由于传送错误而丢失吋。 For example, when one or more frames ー transmission error due lost inches. 通知単元750为连接请求遗留处理并且缓冲区上载单元745将为了连接的HOT缓冲区无效和。 Notify radiolabeling element 750 as connection requests legacy processing and loading unit 745 to the buffer in order to connect the HOT buffers and invalid. 因为存储于⑶1720的SN没有改变,所有后续接收到的帧在发生重传送之前都会被认为离序。 Because the SN stored in ⑶1720 not changed, all subsequent received frames prior to the occurrence of re-transmission will be considered from the sequence.

[0132] 在接收到重新传送的帧之前,缓冲区上传单元745上载为连接接收到的帧到遗留缓冲器。 [0132] Before re-transmitted frame is received, the buffer unit 745 to upload payload of the received frame is connected to a legacy buffer. TCP堆栈215从遗留缓冲区将有效载荷数据由遗留缓冲区拷贝到用户缓冲区。 TCP stack legacy buffers 215 from the payload data is copied from the legacy buffer to the user buffer. 当重传送的帧被上载到遗留缓冲区吋,TCP堆栈215为所有按顺序接收到的帧发送ACK。 When re-transmitted frame buffer is uploaded legacy inch, TCP stack 215 for all received frames in order to send ACK. 传送引擎320为连接更新存储在⑶1720中的DCT SN。 Transfer engine 320 is updated and stored in connection ⑶1720 the DCT SN. 当序列内所有重新传送帧已经被上载到遗留缓冲区时,TCP堆栈215在发送ACK前张贴HOT缓冲区。 When all the re-transmission frame within the sequence has been uploaded to the legacy buffer, TCP stack 215 before sending ACK posted HOT buffer. 张贴HOT缓冲区使缓冲区上载单元745利用HOT缓冲区在不要求用户缓冲区的情况下为连接恢复处理进入帧。 Post HOT buffer contained unit makes use of the buffer 745 HOT buffer without requiring the user buffer connection recovery process into the case frame. [0133] 在步骤930中,CMD单元710通过DMA引擎310接收来自命令环的PRB命令。 [0133] In step 930, CMD unit 710 by DMA engine 310 receives commands from the command ring PRB. 该PRB命令包括在其它字段中对应连接的项的DCT索引和同步位。 The PRB command includes DCT indexes, and synchronization bit entries in other fields in the corresponding connection. 步骤932中,CMD单元710利用索引读出CBI715。 Step 932, CMD unit 710 using the index read out CBI715. 在步骤934中,CMD单元710确定CBI715中读出的同步请求标记是否被设置,如果是这样,步骤936中CMD单元710确定从PRB命令中的同步位是否被设置。 In step 934, CMD Unit 710 determines the synchronization request CBI715 read out flag is set, and if so, step 936 determines whether the CMD element 710 is provided from the synchronization bits PRB command. 如果在步骤934中CMD单元710确定ww CBT715的项中读出的“ sync request”标记没有被设置,那么CMD单元710在步骤938中清空CBI715中的“ sync request”标记,并进入步骤940。 Ww CBT715 710 determines if the entry read "sync request" in step CMD unit 934 flags were not set, then the CMD unit 710 empty CBI715 the "sync request" mark in step 938, and proceeds to step 940. 当“sync request”标记被清除后,连接可利用HOT缓冲区被处理。 When the "sync request" flag is cleared, the connection can use HOT buffer is processed. 如果在步骤936中,CMD单元710确定PRB命令中的同步位没有被设置,那么“sync request”不被清空,并且利用遗留处理继续处理连接。 If, in step 936, CMD Unit 710 determines the PRB command sync bit is not set, then the "sync request" is not empty, processing continues processing and the use of the legacy connection.

[0134] 传送引擎320包括用于卸载来自TCP堆栈215的出站帧处理子单元。 [0134] transfer engine 320 includes an uninstall outbound frame processing sub-unit 215 from the TCP stack. 例如,传送引擎320可被配置为执行TCP分段,计算TCP和Ipv4检查和,并编辑出站帧以背负ACK并包括委托连接的最近状态数据(由DCT350读出)。 For example, the transfer engine 320 may be configured to perform TCP segmentation, calculation and Ipv4 TCP checksum and edit outbound frames to carry the ACK and including the commissioning of recent state data connection (by the DCT350 readout). 驱动程序255或接收引擎360对DCT350的更新可被包括在传送中,如下详述。 Driver 255 or 360 pairs DCT350 receiving engine updates may be included in the transfer, as detailed below.

[0135] 图1OA是根据本发明ー个或几个方面的表示数据通过DMA引擎310由系统存储器130向HOT单元250传送所使用的格式的简图,例如,在“大发送”(large send)传送期间。 [0135] FIG 1OA is a ー or data representing aspects schematic format by 310 by the system memory 130 to HOT Unit 250 transmits a DMA engine is used, for example, in the "high transmission" (large send) according to the present invention during transmission. 字段1007是ー个特定群体MAC报头,如以太网报头。 Field 1007 is a particular group ー MAC header, such as Ethernet header. 字段1005是原型报头,包括MAC报头、IP报头和TCP报头,其又包括SN、传送ACK号、TCP时戳等。 1005 is the prototype header fields, including MAC header, IP header and TCP header, which also includes SN, transmit ACK number, TCP timestamp and so on. 字段1010是传送数据并且仅次于TCP堆栈存储空间225中。 Field 1010 is transmitting data and storage space behind TCP stack 225. 字段1007、字段1005和字段101的组合存放于系统存储器130 中。 Field 1007, a combination of field and field 101 1005 stored in the system memory 130.

[0136] 图1OB是根据本发明ー个或几个方面的表示数据传送引擎320向传送接ロ330传送所使用的格式的简图。 [0136] FIG 1OB is a diagram of the format used for transmission 330 according to the present invention, or aspects ー indicates data transfer engine 320 connected to the transmission ro. DMA引擎310由系统存储空间130读出图1OA中所示的格式并且传送引擎320在启动分段后产生图1OB中所示的格式并且字段1010的部分被包括在分段中,其中每个段都是ー个帧。 DMA engine 310 by the format shown in system memory 130 reads out 1OA and transmit engine 320 after starting the segment generated format shown in FIG. 1OB and part field 1010 is included in the segment, where each segment all ー frames. 在另ー实施例中TCP堆栈215生成协议报头并将协议报头存储于TCP堆栈存储空间225中,并且HOT单元250由应用程序存储空间227中读出传送数据。 In another ー implemented in TCP stack storage 225, and 227 HOT unit 250 reads out data transmission example TCP stack 215 generates protocol headers and protocol header storing application storage space.

[0137] 字段1015是ー个IP报头,字段1020是ー个TCP报头,字段1025是分段数据。 [0137] 1015 is ー field of IP header field 1020 is ー a TCP header field 1025 is segmented data. 本领域技术人员应该理解,图1OB中所示的格式是ー个TCP兼容格式。 Those skilled in the art would understand the format shown in Figure 1OB is a TCP ー compatible format. TCP报头包括传送ACK号,传送SN等。 TCP header includes transmission ACK number, and other transfer SN. 分段数据是字段1010中传送数据的一部分。 Segment data field 1010 is part of the transmitted data. 字段1030是另ー个TCP报头,包括可选的更新过的传送ACK号和更新过的传送SN等。 Field 1030 is another ー a TCP header, including an optional updated transmission ACK number and updated transmission SN like. 更新过的传送SN被包括在前一段中的数据字节数増加,该前一段与最大帧尺寸和数据链、网络、和传输层报头尺寸的区别相同。 SN updated transfer is included in the previous paragraph to increase in the number of data bytes, which differs from the previous period maximum frame size and data link, network, and transport layer header size is the same. 字段1035是分段数据,该分段数据是字段1010中传输数据的另一部分。 Field 1035 is segmented data, the segment data is another part of the transmission data field 1010.

[0138] 图1lA是根据本发明的一个实施例,用于分割图1OA中所示的ー个出站帧为包括编辑出站帧的片断的流程图。 [0138] FIG 1lA is according to one embodiment of the present invention, the ー outbound frames including a snippet editor outbound frames of the flowchart shown in Figure 1OA for segmentation. 在步骤1101中,DMA引擎310接收ー个来自TCP堆栈的经由驱动程序255的传送描述符。 In step 1101, DMA engine 310 receives from the TCP stack ー a transfer descriptor via driver 255. 该传送描述符包括一定位存储于系统存储器130的传送缓存物理地址,该传送描述符包括一原型报头和数据用于传送。 The transfer descriptor includes a positioning memory transfer buffer between the physical address in the system memory 130, which includes a prototype transfer descriptor header and data for transmission. 该传送描述符也波阿卡详细说明处理选项的控制位,一可选的DCT索引,控制位详细说明传送选项等。 The transfer descriptor also 波阿卡 detail control bit processing options, an optional DCT index, control bits delivery options described in detail like.

[0139] 在步骤1103中,DMA引擎310读取传送缓存并输出该传送描述符和传送缓存到传送引擎320。 [0139] In step 1103, DMA engine 310 reads the output of the transmission buffer and the transmission and transport buffer descriptor to transfer engine 320. 在步骤1109中,传送引擎320计算一基于从该原型报头中提取的IP报头数据的IP校验和。 In step 1109, the transfer engine 320 calculates a prototype based on extracts from the IP header in the IP header checksum data. 在步骤1111中,传送引擎320确定用于传送的数据的一部分,包括在分割后的最大的片断(通过在连接建立中的目标来设定)中的传送缓存中。 In step 1111, the transmission portion of the engine 320 is determined for transmission of data, including the largest in the divided fragments (through the target in connection establishment to set) the transfer cache. 在步骤1113中,传送引擎320构建ー用于传送的片断,其进ー步的描述请參阅图11B。 In step 1113, the transfer engine 320 constructed piece ー for transmission, which is described further into ー see Figure 11B.

[0140] 在步骤1131中,传送引擎320计算一基于从该原型报头中提取的TCP报头数据的TCP校验和、从DCT350中读取连接状态数据、以及在当前帧中的用于传送的数据的一部分。 [0140] In step 1131, the transfer engine 320 calculated based on data extracted from one of the prototype header TCP header TCP checksum data read from DCT350 data connection status, as well as in the current frame for transmission part of it. 经过计算的校验和被存储在该帧的TCP报头中。 After calculation of the checksum and TCP packets are stored in the frame header. 在步骤1133中,通过包括在该帧中数据大小和报头大小的差增加该传送SN,传送引擎320升级一存储在DCT350中的用于委托连接的传送SN。 In step 1133, by including in the frame difference data size and header size increases the transmission SN, transfer engine 320 transferred SN commissioned a memory upgrade for the connection of the DCT350. 在步骤1135中,传送引擎320输出ー构建帧到传送接ロ330,包括一计算过的TCP校验和。 In step 1135, the transmission output engine 320 to transfer the framing ー pick ro 330 comprises a computer over TCP checksum. 该传送接ロ330输出该构建帧。 The transport connection ro 330 outputs the constructed frame.

[0141] 图1lB是根据本发明的一个实施例,用于构建出站帧的方法步骤的流程图。 [0141] FIG 1lB is according to one embodiment of the present invention, a method for constructing a flow chart of the steps of the station frame. 本领域的熟练人员能够理解,图1lB的方法步骤是图1lA中步骤1113的延续。 Those skilled in the art will appreciate that the method steps are a continuation of FIG 1lB 1lA step 1113 in FIG. 使用存储在CDT720中的连接状态数据构建该出站帧;接下来,该出站帧包括最新的状态数据用于ー个连接,例如一相应于最近接收到的帧的ACK数值。 Connection state data stored in the construction of this CDT720 outbound frames; Next, the outbound frames including the latest state data for ー connections, such as a value corresponding to the most recently received ACK frame. 在传送中,包括在分割中,如有可能,ACKs是“背负式”,也就是,包括在用于传送的帧输出而不是直到ー帧被完全的分割并传送之后,等待输出离散的ACKs。 In the transfer, including the segmentation, if possible, ACKs is "piggyback", that is, it included in the frame for transmitting the output rather than until ー frame is completely divided and transferred after waiting for the output of discrete ACKs. 及时的ACKs保证对于发送器来说该接收窗ロ处于开放状态,以便该发送器能连续地传送数据。 ACKs ensure timely for the receiver to the transmitter window ro in the open state, so that the sender can transmit data continuously. 另外,当发送器的时钟失效时,推迟ACK传送直到一大发送(large send)完成可能会导致浪费的再传送,或者会导致在接收数据流中发生非必要的爆发,其由于ACKs被推迟而只能发生“大发送”。 In addition, when the transmitter clock failure, delayed ACK transmission until a high transmission (large send) complete waste may lead to the re-transmission, or will lead to unnecessary outbreaks occurred in the received data stream, which has been delayed since the ACKs and only "high transmission" occurs.

[0142] 在步骤1115中,传送引擎320使用具有传送请求的在步骤1101中接收到的DCT索引来确定该传送请求是否相应于ー委托连接,并且,如果为否,进行如图1lA的步骤1131。 Step [0142] In step 1115, transfer engine 320 having received DCT index at step 1101 to determine if the transfer request corresponding to the transfer request is commissioned ー connection, and, if NO, performed as shown 1lA 1131 . 否则,在步骤1117中,传送引擎320使用在步骤1101中接收到的索引访问CDI720,以获得用于委托连接的连接状态数据。 Otherwise, in step 1117, the transfer engine 320 received in step 1101 index access CDI720, in order to get a connection state data for delegate connection. 在步骤1119中,传送引擎320确定以传送SN用于该构建帧。 In step 1119, the transfer engine 320 determines to transfer SN for this construct frames. 当从TCP堆栈215中接收到的SN在数据流中较存储在⑶1720中的用于连接的传送SN晚时,传送引擎320设置该传送SN为TCP215接收的SN。 When the 215 received from the TCP stack to the SN in the data stream is stored in ⑶1720 representing the connection for transferring SN night, transfer engine 320 to set the transmission SN of TCP215 received SN.

[0143] 在步骤1121中,传送引擎320检查控制位详细指明处理选项和确定TCP215是否请求传送引擎320以在⑶1720中保存该帧第一位的SN。 [0143] In step 1121, the transmission control engine 320 check bits specify in detail the treatment options and determine whether to request a delivery engine 320 TCP215 to save the frame in the first bit ⑶1720 SN. 当一相应于该存储的SN的ACK收到时,该存储的SN被用于图9A中步骤907以控制TCP堆栈215的通知。 When a storage corresponding to the SN of the ACK received, the stored SN is used in step 907 in FIG. 9A notification 215 to control the TCP stack. 如果,在步骤1121中,传送引擎320确定TCP堆栈215请求用于ー特定ACK数值的通知,那么在步骤1123中,传送引擎320在⑶1720中保存该SN作为ー特定ACK数值用于该连接。 If, in step 1121, the transfer engine 320 determines TCP stack 215 ー request notification for specific ACK value, then the delivery engine 320 stored in step 1123 in ⑶1720 in the SN as ACK ー specific value for the connection.

[0144] 在步骤1125中,传送引擎320确定ーACK数值由于构建帧。 [0144] In step 1125, the transfer engine 320 determines ー ACK value due to build frames. 当该从TCP堆栈215接收到的该ACK数值在数据流中比在DCT存储的用于连接的SN晚时,传送引擎320设置该DCT SN为从TCP堆栈215中收到的ACK数值。 When this is received from the TCP stack 215 to the ACK value in the data stream than in the late DCT SN stored for the connection, the transfer engine 320 sets the DCT SN ACK value from the 215 received in the TCP stack. 传送引擎320也使者该存储在⑶1720中的用于连接的最后ACK数值为该DCTSN或者从TCP堆栈中接收的ACK数值较大的。 Delivery engine 320 Herald last ACK value in the memory is used to connect the ⑶1720 for DCTSN or received from the TCP stack ACK value larger.

[0145] 在步骤1127中,通过检查存储于⑶1720中的用于委托连接的连接数据状态,传送引擎320确定该TCP时戳选项是否为能。 [0145] In step 1127 by checking the connection data storage for delegate status of the connection, the transfer engine 320 determines that the TCP timestamp option is to be able to ⑶1720 in. 当TCP时戳选项为不能时,传送引擎处理如图1lA所示的步骤1131。 When TCP time stamp option is not, the delivery engine processing step shown in FIG. 1lA 1131. 否则,在步骤1129中,传送引擎320包括有在该构建的帧的TCP报头中的自由运行的时钟的电流值。 Otherwise, in step 1129, the transfer engine 320 includes a current value of the clock in the TCP header of the frame is constructed in the free-running. 传送引擎320也包括从TCP堆栈收到的更大的时戳以及存储在⑶1720中的用于连接的TCP时戳(最近的TCP时戳)。 Delivery engine 320 also includes a time stamp (when the most recent TCP timestamp) from the TCP stack received greater time stamp and stored in the connection of TCP ⑶1720 for. 当从TCP215接收到的TCP时戳比存储的用于连接的TCP时戳大吋,该存储的用于连接的TCP时戳被设置为从TCP堆栈215接收的TCP时戳。 Poke big inch timestamp stored for more than a TCP connection is received from TCP215 to TCP, this store is used to connect the TCP timestamp is set to receive from the TCP stack 215 TCP timestamp. 传送引擎320进行步骤1131以计算用于该构建帧的该TCP校验和。 Transfer engine 320 to step 1131 to calculate the build frames for the TCP checksum. [0146] 图1lC是根据本发明ー个或几个方面的,产生传送过程中的夹杂(也就是,背负式)的ACKs的方法步骤的流程图。 [0146] FIG 1lC is (that is, piggyback) ACKs flow chart of the method steps according to the present invention, one or more aspects ー, resulting in the transmission process inclusions. 图1lC所示的方法是通过接收引擎360作为被接收的帧被完成。 The method is illustrated in FIG. 1lC engine 360 by receiving a received frame is completed. 在步骤1145中,接收引擎360接收连续的TCP帧作为一委托连接。 In step 1145, the receiving TCP engine 360 receives consecutive frames as a delegate connection. 在步骤1147中,接受引擎360在相应于该委托连接的DCT350中,升级连接数据状态,例如,TCP时戳,SN,传送窗大小,以及类似数据。 In step 1147, accept the engine 360 corresponding to the commission in DCT350 connection, upgrade the status of the connection data, for example, TCP timestamp, SN, the transmission window size, and similar data. DCT SN被升级为接下来希望的输入SN。 DCT SN is upgraded to the next desired input SN.

[0147] 在步骤1149中,接受引擎360计算ーACK差,其是介于该SN与最后ACK数值(从DCT350中读出的)之间的差。 [0147] In step 1149, it accepts ー ACK difference calculation engine 360, which is situated between the SN and the difference between the final ACK value (read from DCT350 in). 在步骤1151中,接受引擎360确定该ACK差是否大于ー极限值,该极限值被TCP堆栈215编程以触发ー个接收帧的ー个ACK。 In step 1151, the acceptance of engine 360 determines whether the difference is greater than the ACK ー limit, this limit is programmed to trigger TCP stack 215 frames ー ー received an ACK. 如果该ACK差大于该极限值,接收引擎360执行步骤1157。 If the ACK difference is greater than this limit, the reception engine 360 to Step 1157. 否则,在步骤1153中,接收引擎360确定该DCT SN是否大于ー个门槛值,该门槛值是一递增序号或者一特定序号。 Otherwise, in step 1153, the reception engine 360 determines whether the DCT SN ー greater than a threshold, the threshold value is incremented by a number or a specific number. 如果该DCT SN大于该门槛值,接收引擎360执行步骤1157。 If the DCT SN is greater than the threshold value, the reception engine 360 to Step 1157. 否则,在步骤1155中,接收引擎360确定前述的传送ACK时钟是否失效,以及,如果没有失效,则升级传送ACK时钟以及未被接受的计数。 Otherwise, in step 1155, the reception engine 360 determines whether the aforementioned transfer ACK clock failure, as well as, if not failed, then upgrade the transmission and ACK clock count is not accepted. 如果,在步骤1155中,接收引擎360确定该传送ACK时钟已经失效,则在步骤1157中接收引擎360升级存储在DCT350中的用于委托连接的连接状态数据,例如,清除未被接受的计数,升级该最后已接受的数值,升级该递增的序号,等类似数据。 If, in step 1155, the receiver transmits an ACK engine 360 determines that the clock has expired, the reception engine upgrade is stored in step 360 in DCT350 in 1157 commissioned by the connection status for data connections, for example, to clear the count is not accepted, Upgrading the last accepted value, upgrade the increasing number, and other similar data. 接收引擎360也重置传送ACK时钟。 Engine 360 also receives the transmitted ACK clock reset. 在步骤1159中,接收引擎360用信号通知传送引擎320,以在用于传送的ー帧中包括一ACK,例如,通过背负一ACK。 In step 1159, the engine 360 receives the signaling transfer engine 320 to include an ACK frame in ー for transmitting, for example, by carrying an ACK.

[0148] 在允许驱动程序255或者TCP堆栈215具有灵活性的同吋,HOT单元350卸载接收的有效的用于委托连接的TCP帧的TCP处理,来确定ー个基于接收ACKs和时钟的中断的门槛值。 [0148] allowing drivers 255 or TCP stack with 215-inch flexible and effective process for TCP connections TCP frames delegate HOT unloading unit 350 received to determine ー a clock based on the reception of ACKs and interrupt threshold. 该门槛值可以被用于减少中断,空闲CPUllO以处理其他的应用程序。 The threshold value can be used to reduce disruption, idle CPUllO to handle other applications. 进一歩,HOT单元350产生用于传送和编辑背负ACKs的出站帧的ACKs,以计算TCP和IPv4校验和,并执行TCP片断。 Into a ho, HOT unit 350 generates ACKs for transferring and editing bear outbound frame ACKs, to calculate the TCP and IPv4 checksum and perform TCP segment. 一个发送器(sender)的ACKs的及时的产生和传送能够保持接收窗ロ的开放,在单向和双向通信中提高带宽的利用和減少不必要的再传送。 Timely generation and transmission of a transmitter (sender) of ACKs ro receive window can be kept open to improve bandwidth utilization and reduce unnecessary re-transmission of one-way and two-way communications. 最后,有效载荷数据上传至在应用程序存储空间227中的用户缓冲区,减少从驱动存储空间235拷贝数据到应用存储空间227的需要。 Finally, the payload data is uploaded to the application 227 user buffer memory space, reducing the need for storage space 235 from the drive to copy data to the application storage space 227. 当用于委托连接的使用者不可得到并且接收FIF0730满载时,遗留缓冲区(legacy buffer)可以被用于上载接收到的帧,并不是不接受输入的数据。 When the user is not available for the delegation connector and receives FIF0730 when fully loaded, and the remaining buffer (legacy buffer) can be used to upload received frames, not does not accept incoming data. 当为ー些来自于CPUllO的TCP处理提供卸载时,HOT单元250不依赖于大量的专用存储器或者专用的处理器。 When some of ー of TCP processing from CPUllO provide an uninstall, HOT unit 250 does not rely on a large number of dedicated memory or a dedicated processor. HOT单元250也从ー主处理器卸载ー些TCP处理并在接收输入数据时处理ー些多余的连接。 HOT Unit 250 also offloaded from the main processor ー ー some TCP processing and upon receiving input data processing ー these extra connections.

[0149] 在另一方面,本发明提供了一个用于TCP连接的帧处理方法。 [0149] In another aspect, the present invention provides a frame processing method for TCP connections. 该方法包括利用卸载单元处理该帧的第一部分,以产生第一处理的帧数据;利用该卸载单元处理该帧的第二部分,以产生第二处理的帧数据;利用在CPU上执行的TCP堆栈处理该第二处理的帧数据,以产生第三处理的帧数据。 The method includes processing a first portion of the unloading unit of the frame, to generate a first processed frame data; using the unloading unit processes the second portion of the frame, to generate the second processed frame data; using TCP executed on the CPU processing the second stack frame data processing, to produce a third processed frame data. 另ー方面,该方法进ー步包括在处理每一帧过程中确定是否有ー个特殊事例存在。ー another aspect, the method further comprises determining step ー process each frame in the process if there ー special case exists. 在其他方面,该方法进ー步包括上传该第一处理的帧数据到一用户缓冲区。 In other aspects, the method further comprises the step ー uploading the first processed frame data to a user buffer. 在另一方面,该方法还包括上传该第二处理的帧数据到ー遗留缓冲区。 In another aspect, the method further comprises uploading the second processed frame data to a legacy buffer ー. 在另一方面,该方法还进一歩包括传送ー在系统存储器中的遗留缓冲区的本地和ー软件驱动程序间通信的标签。 In another aspect, the method further comprises a label transfer into a ho ー legacy in the system memory buffer between local and ー software driver communication.

[0150] 在另一方面,本发明提供了ー种处理用于TCP连接的数据的系统。 [0150] In another aspect, the present invention provides a process for data types ー TCP connection systems. 该系统包括一TCP堆栈,其被配置为处理存储于至少ー个遗留缓冲区的接收到的帧;ー软件驱动程序,被配置为连接该TCP堆栈和卸载单元之间的接ロ;以及该卸载单元被配置为处理在委托连接上接收到的帧,以产生有效载荷数据和被部分处理的帧。 The system includes a TCP stack, which is configured to process at least ー stored in a legacy frame buffer received; ー software driver is configured to connect between the TCP stack and unloading unit access ro; and uninstall unit is configured to process the received connection frame on a commission to produce payload data and the processed frame portion. 在另一方面,该卸载単元被配置以处理没有特殊事例存在的帧。 On the other hand, the offload radiolabeling cell is configured to handle special cases there is no frame exists. 还在另一方面,该卸载単元被配置当一特殊事例被确定存在时通知TCP堆栈。 In yet another aspect, the offload radiolabeling element is configured as a special case to be determined notify the presence of TCP stack. 还在另一方面,该卸载単元被配置以上载部分地处理的帧到至少ー个遗留缓冲区。 In yet another aspect, the offload radiolabeling yuan frame is configured to upload to at least partially processing ー a legacy buffer. 还在另一方面,该卸载単元被配置上载有效载荷数据到至少ー用户缓冲区。 In yet another aspect, the offload radiolabeling cell is configured to upload the payload data to at least ー user buffer. 还在另一方面,该卸载単元被配置,在上载部分地处理的帧到遗留缓冲区或者上载有效载荷数据到用户缓冲区,以接收附加的帧。 In yet another aspect, the offload radiolabeling element is arranged in the frame part of the upload to handle legacy payload buffer or upload data to the user buffer to receive additional frames. 还在另一方面,该TCP堆栈提供具有相应于至少ー用户缓冲区的本地信息的卸载单元。 In yet another aspect, the TCP stack providing a user buffer corresponding to at least ー unloading unit of local information. 还在另一方面,该卸载単元包括一被配置的委托连接表以存储使用者信息和委托连接状态信息。 In yet another aspect, the offload radiolabeling unit includes a delegation connector being configured to store user information table and delegate connection status information.

[0151] 在另一方面,本发明提供了一种处理用于委托和非委托TCP连接的帧的方法。 [0151] In another aspect, the present invention provides a process for commissioned and non-commissioned frame TCP connection method. 该方法包括利用卸载单元处理委托TCP连接,该卸载単元被配置为处理不存在特殊事例的帧;利用在CPU上执行的TCP堆栈处理非委托TCP连接;和利用该CPU上执行的TCP堆栈处理所有特殊事例存在的帧。 The method includes the unloading unit to handle the commissioned TCP connection, the offload radiolabeling element is a frame configured to handle special cases that do not exist; the use of TCP stack processing on the CPU to perform a non-commissioned TCP connection; and the use of TCP stack executing on the CPU handles all special case exists frames. 还在另一方面,委托TCP连接的处理产生有效载荷数据。 In yet another aspect, commission processing TCP connections generate payload data. 还在另ー方面,该有效载荷数据通过卸载单元上载到ー委托连接。 Still another ー aspect, the payload data uploaded ー commissioned by offloading unit connection. 还在另一方面,该方法进ー步包括,当作为以委托连接的数据传送时,升级存储在卸载单元的连接状态信息。 In yet another aspect, the method further comprises the step ー, when a data transfer connection with commissioned, the connection status update information stored in the unloading unit.

[0152] 在另一方面,本发明提供了一个设立委托连接的方法。 [0152] In another aspect, the present invention provides a method to set up delegate access. 该方法包括:建立TCP连接;和确定是否通过硬件处理委托该TCP连接。 The method comprising: establishing a TCP connection; and to determine if hardware processing commissioned by the TCP connection. 还在另一方面,该方法进ー步包括在关于确定委托该TCP连接的委托连接表中建立一内容。 In yet another aspect, the method further comprises the step ー content on the establishment of a commission to determine the commission of the TCP connection table connection. 还在另一方面,该方法进ー步包括转移用于该委托连接的用户缓冲区信息到硬件。 In yet another aspect, the method further comprises the step ー transferred to the user buffer information of the commission connected to the hardware.

[0153] 在另一方面,本发明提供了一个用于TCP堆栈和卸载单元之间通信的系统。 [0153] In another aspect, the present invention provides a system for communication between TCP stack and unloading unit used. 该系统包括由TCP堆栈向该卸载单元传送命令的装置;和由该卸载単元向该TCP堆栈传送通知描述符的装置。 The system includes the TCP stack to the unloading unit transmission command means; radiolabeling yuan to the TCP stack transfer device and a notification descriptors from this uninstalled. 还在另一方面,该系统进ー步包括一用于从卸载单元向该TCP堆栈传送命令细节情况的装置。 In yet another aspect, the system further comprises a feed ー from unloading unit for sending commands to the TCP stack details of the device. 还在另一方面,该系统进ー步包括用于从TCP堆栈向该卸载单元传送接收缓存描述符的装置。 In yet another aspect, the system further comprises a feed ー from TCP stack to the unloading unit transmits the receive buffer descriptors devices. 还在另一方面,该系统进ー步包括从TCP堆栈向该卸载单元传送传送缓存描述符的装置。 In yet another aspect, the system further comprises intake ー from unloading unit transferred to the TCP stack buffer descriptor transfer device.

[0154] 在另一方面,本发明提供了利用卸载单元编辑出站(outbound)帧的方法。 [0154] In another aspect, the present invention provides the use of the unloading unit edit outbound (outbound) frame approach. 该方法包括接收委托连接表索引;接收用于从TCP堆栈传送的原型报头和数据;利用该委托连接表索引存取委托连接表项目;基于传送数据的一部分计算TCP检查和;和输出包括该TCP检查和的帧和该传送数据的一部分。 The method includes receiving a delegation connector table index; receiving a prototype header and data transmitted from the TCP stack; use the delegate access delegate connection connection table index sheet items; calculation based in part on TCP inspection and transfer of data; and an output and including the TCP part of the inspection and the frame and the transmission of data. 还在另一方面,该方法进ー步包括升级委托连接表内容。 In yet another aspect, the method further includes upgrading commissioned into ー connection table content. 还在另一方面,该方法也包括访问该连接表内容;计算一基于用于传送的数据的其他一部分的TCP校验和;以及输出一包括该TCP校验和的附加帧和另外用于传送的数据的另外一部分。 In yet another aspect, the method also includes access to the connection table content; calculating a checksum for TCP-based data transmission and other part; and an output and including the TCP checksum and the additional frames and another for transmitting Another part of the data. 还在另一方面,当一个目标具有一特定序号的得到认可的收条时,应用程序需要得到通知。 In yet another aspect, when the receipt of a recognized goal has a specific serial number, the application needs to be notified. 还在另一方面,该方法进ー步包括背在该帧中背负一回执。 In yet another aspect, the method further comprises the step ー back in the frame carrying a receipt.

[0155] 以上所述,仅是本发明的较佳实施例而已,并非对本发明作任何形式上的限制,虽然本发明已以较佳实施例揭露如上,然而并非用以限定本发明,任何熟悉本专业的技术人员,在不脱离本发明技术方案范围内,当可利用上述掲示的方法及技术内容作出些许的更动或修饰为等同变化的等效实施例,但是凡是未脱离本发明技术方案的内容,依据本发明的技术实质对以上实施例所作的任何简单修改、等同变化与修饰,均仍属于本发明技术方案的范围内。 [0155] described above, only the preferred embodiment of the present invention, it is not the invention of any formal restrictions, although the present invention has been described with preferred embodiments anyone familiar with the drawings described above, however, not intended to limit the present invention, Those skilled in the art, without departing from the scope of the present invention within the program, can be used as shown above kei methods and technical content to various omissions, substitutions and changes in the equivalent embodiments, but those who are not from the aspect of the present invention The content, based on technical essence of the invention is simple to modify any of the above embodiments when taken in equivalent changes and modifications as would fall within the scope of the present invention the technical solution.

Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US593716929 Oct 199710 Aug 19993Com CorporationOffload of TCP segmentation to a smart adapter
US642717128 Feb 200030 Jul 2002Alacritech, Inc.Protocol processing stack for use with intelligent network interface device
Classifications
International ClassificationH04L12/56, H04L12/28, G06F3/00, H04L29/06, G06F13/38
Cooperative ClassificationH04L47/225, H04L69/22, H04L47/12, H04L69/16, H04L69/161, H04L29/06, H04L69/166, H04L69/12, H04L69/10, H04L69/163
Legal Events
DateCodeEventDescription
16 Nov 2005C06Publication
16 Aug 2006C10Entry into substantive examination
30 Oct 2013C14Grant of patent or utility model
28 Jul 2017CF01