TW201227327A - A analysis method of the data unit of the iSCSI - Google Patents

Abstract

A analysis method of the data unit of the iSCSI utilities to perform the request of the iSCSI. The analysis method comprises of the steps. Sending the data writing request to the target. The target prepare sends the data unit to the initial according to the data writing request. The initial generate a plurality of data output unit, and the target transfer tag is write into the SG list. The target looks up the correspond to the SG block according to the target transfer tag. The target takes the buffer form the SG blocks. The target performs a direct memory access order for writing the loading into the buffer.

Description

201227327 VI. Description of the invention: [Technical field to which the invention pertains] A method for analyzing a data unit, in particular, a method for analyzing a data unit of an Internet small-scale computer interface. [Prior Art] Internet Small Computer System Interface (_Γηβ Smau Computer Systems Interface ‘ iSCSI for short) is the Internet Engineering Task Force (Intemet)

Engineering Protocol Force (IETF) proposed communication protocol standards. Iscsi connects to multiple computer hosts of the network through the (Tmnsmission Control Protocol / Internet Protocol 'TCP/IP) protocol, and becomes a virtual network storage device by connecting a plurality of storage devices 130 dispersed in the Internet. The user accesses the network storage device as if accessing the local (4) device. She is in the traditional small (four) brain interface (SmaU Computer Systems Interface) (SCSI) and redundant disk array (system (4) she

Array of Independent Disks ’ RAID technology, iSCSI technology combines the high-speed transmission capability of SCSI, but is not limited to the storage device _ charging / backup capacity, unlimited expansion of the capacity of virtual storage devices. The virtual storage device storage area network (Storage Area Network 'SAN) is mainly composed of an initiator (dirty_) 11 〇 and a target 120 (target). Initiative! The 10 transfer access request establishes a session with the target terminal 12 (Session). Please refer to the "Graphic!" diagram, which is a schematic diagram of the transmission of the access request and data unit of the prior art. In the Internet small computer system interface protocol towel defines a good format of the package data unit, but most of its towels are secret agreement management and error recovery. Real 201227327 involves the transmission of data on the Internet small computer system interface only scSI command negative ^ (SCSIC ^ DU), data wheeling unit (DataInPDu), data output #T〇 (Data Out PDU) ^ it (Ready to Transfer PDU It takes a lot of time to resolve the data unit through each layer. Therefore, how to speed up the resolution processing of (4) units will effectively improve the overall efficiency. SUMMARY OF THE INVENTION The present invention is directed to providing a data request for processing a small computer interface of an Internet by providing an Internet-based small computer interface. The method for parsing the data unit of the small computer interface exposed by the rabbit has included the following steps: establishing communication between the initiator and the target, and sending the data write request to the target; the target reorganization engine (§聊邮^With this Engine) generates a ready to transmit data unit according to the data write request (κ2τρ〇 your target # end will transmit the ready to transmit data unit to the originating end; the originating end generates the unit group data output unit according to the ready to transmit data unit, At each target end of the data output unit, the label is transmitted. (4) (4) Position (4) Kiss writes the distributed/snow towel block (10) Age is fine. This sounds: the tribute output unit is transmitted to the target end; the target end finds the corresponding label according to the target end. The knife/concentration block, and obtain the host cache from the decentralized/concentrated block, and execute the direct access memory command to directly write the load content received by the target end. In the host cache; when the target end completes the write request, the target end sends a completion response unit (RSPPDU) to the initiator. 201227327 This book provides a data unit for the small computer interface to adjust the content of the data packet transmitted by the target side of the J-Qing, so that the overall system can be “higher”. Xiao Yuan, the ship is detailed. [Embodiment] When the eight-field end writes to the target, the target will issue a small computer; ι ® S ^ I t-^7L (SCSI Write CMD Protocol Data Unit (PDU) )) and the data 〇 PDU (Data 〇 PDU). The present invention proposes the operation of the town of the # ' ' 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 The architecture of the network packet of the Internet small computer interface includes: an originating end, a target end 22 and a storage device 230. The starting end 210 is used to issue an access request to the target end 220. In the initiator 210, a host bus adapter 24 is also included. (H〇st Bus Sheqing, referred to as the leg). The host sinks the listener ϋ 24G main (four) 对 TCMp 耽 and 5 packet parsing action Storage device 23Q is used for age target information. The access request from the initiator 210 is used by the target terminal 22 to access the target information in the storage device 23 according to the access material and transmit the target information to the initiator 210 to transmit the target terminal 220, the initiator 210, and the storage device. The 230 units can communicate with each other through the Internet, the internal network (imranet) or Wei. 201227327 The reference is shown in Figure 3, which is the operational flow of the present invention. schematic diagram. The parsing method of the present invention includes the following steps: Step S310: · Establish communication between the initiator and the target, and send a data write request to the target; Step S320. Target reorganization engine of the target (segment)

Engine) generating a ready to transmit data unit (R2TPDU) according to the data write request; step S330: the target end transmits the ready to transmit data unit to the originating end; step S340: the originating end generates a plurality of sets of data output orders according to the ready to transmit data unit, The target transfer tag of each data output unit is written into a scatter/gather block; step S350. The data output unit is transmitted to the target end; step S360: the target end transmits according to the target end The tag finds the corresponding scatter/ru concentration block and obtains the host buffer from the scatter/concentration block; Step S370: The target side performs direct access memory (Direct

The access, DMA) command is used to directly write the load content received by the target end into the host cache; and step S380: after the target end completes the write request, the target end sends a completion response unit (RSP PDU) to the initiator. . Initially, communication between the initiator and the target is established to send a data write request 201227327 to the target. Under the iSCSI architecture, the initiator first encapsulates the data write request (SCSIWriteCMDPDU). The data write request mainly includes the data length to be written by the command, and an initiator task tag (InitiatorTaskTag, ITT for short) is allocated to identify the command. In other words, each time the initiator wants to transmit a new data sequence, the initiator sends a target transmission label to the target, so that the target can store the new data sequence in other locations to avoid confusion of different data sequences. . When the target end receives a small computer interface write command from the initiator, the target end writes the corresponding information to the target end transfer label in the header of the preparation transfer unit. After receiving the data write request, the target end sends a Ready to Transfer PDU (R2T) to the initiator according to the local possible resource size and the write length negotiated during login (MaxBurstLength). In the ready to transmit data unit, this field through the ITT 1 is used to identify the correspondence between the prepared transport data unit and the data write request. The target end then sends the ready-to-delivery unit with the target-side transport label to the originator. The data units belonging to the same group of data sequences all have the same target delivery label. Because - a data write request may correspond to multiple ready to send data sheets. Therefore, the target end will have an initiator task thief to distinguish the different preparation data units within the scope of the writer command. In this way, the target end can check whether the received data unit has an error by checking whether the target end transmission label is the same. The target end performs the modification process of the 201227327 firmware of the tactile device fine path device of the data unit. The driver at the target end is responsible for passing the host buffer address to the firmware in the form of a message, and the format of the message is defined as follows: sg—addrO—the scatter/gather list of the host (sg list Basic address " Sg-dma-start - Index positioning of the first block of the scatter/gather list sg_dma_len - Number of consecutive locations of the scatter/gather list • Sg-Cnt - Total number of address lengths in the scatter/set list the amount

5 in Sg_add0 has been recorded as the starting address of the scatter/gather list, and the number of all scatter/gather lists is identified. The address of the scatter/gather list is passed to Boeing via sg-add〇. 1 read can be (10) access memory _ way to put the address information of these scatter / pin list _ should be the block towel 'to enable the county level · can take the corresponding SCSI generated information from the block, with direct access The memory goes to the data on the host. The blocks on the moving body do not need to be allocated by the !_ device. The driver only needs to know how many facets are called, the size of each block and the state of these blocks. The drive can know the scatter/gather list of the next command through the current state of use of the block. How many blocks from the first few are occupied, and the _ get these # messages will be processed. Please refer to the "Fig. 4" for the operation of the present invention for the search for the distribution/collection list. Buffers that store scatter/gather lists should be allocated by the drive, in the form of legs-task-sg-t. ‘First convert the scatter/gather list into the following structure. In this embodiment of the invention, this structure is 12 bytes. Knowing the number of the 201227327 scatter/gather list, it is possible to calculate how much space is needed and how many blocks need to be occupied on the blade. Typedef struct iscsi_task_sg_s { __uint32"addr_high; ___uint32 _t addrjow; ___uint32_t len; } iscsi-task_sg_t; As a general rule, the destination-side transport label consists of 32-bit fields. In the present invention, the target end transmits a tag to add information about the decentralized/concentrated block of the person to be written. A direct access memory command is issued from the target end to the connected storage device, and the payload content of the iSCSI data unit is directly written into the memory address of the corresponding host buffer address information. The segment reorganization engine parses after receiving the f-unit from the network card. The iSCSI layer only processes the data of the load unit in the fast path. This is because the load unit is carrying (4) materials, and the information about his negative fresh elements is only the basic protocol test. Therefore, the specific parsing and error handling scales are all done by the software unloading path coffee WPath). When the segment reorganization engine parses the data of the load unit, it can extract the bits of the iscsi generated block from the initiator task and then extract the address of the host buffer from the corresponding iscsI generation area. Saki-Jin sends the direct access memory to the surface of the memory and writes it to the storage device connected to the target. The present invention provides a solution for the data unit of the small-scale computer interface of the Internet. The method of the 201227327 method, the content of the (four) packet transmitted by the target end into the overall system can improve the access speed when writing.仃 整 , , , , , , , , , , , , , , , 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然 虽然The scope of patent protection of the invention shall be subject to the definition of the bribe of the towel. [Simple description of the map]

1 is a schematic diagram of transmission of an access request and a data unit of the prior art. FIG. 2 is a schematic diagram of the architecture of the present invention. Figure 3 is a schematic diagram of the operational flow of the present invention. Figure 4 is a schematic diagram of the operation of the present invention for finding a scatter/gather list. [Main component symbol description] 110 initiator 120 target terminal 130 storage device 210 initiator terminal 220 target terminal 230 storage device 240 host bus bar adapter

Claims (1)

201227327 Seven patent application scope: - an Internet small _ brain interface 赖解元 of an Internet small #0<<<<> method for processing: (4) μ seeking, analysis method package to establish an initiator and - target Sent to the target end; Na Yun (4) incoming request to send the target end - the segment reorganization engine transmits the data unit according to the preparation; the write request generates a = target end to transmit the ready transmission data unit to the originating end. , in each of the "multiple sets of data output single set block; ... a target end transfer tag is written to a scatter / target end; the data output order it is transferred to the = target end according to the target end transfer _ Find the middle block, f get a host cache in the scatter/concentrate block; and receive 亍: direct access memory command 'to write the target end, and wear the content directly to the host cache. 2 · The gateway of the gateway, the parsing side of the unit __ unit = the data of the Internet small computer interface described in the address of the address of the splitting block in the destination transport label Unit resolution Set the base address register on the target side. 12 201227327 side 4. Analyze the data unit of the Internet small computer interface as described in Item 3, where the steps of writing the scatter/concentration block The medium includes: writing an address offset in the target end transfer label. 5. The data unit of the Internet small computer interface as described in Item 4 is used to determine the method in which the target Wei Weng is biased The scatter/concentration block to which the shift amount and the miscellaneous address are to be accessed. 6. h. The Internet small computer method described in "Before", after the target end completes the write request, the second The parser sends out a completion response unit.