CN100438396C - System and method for realizing error return in iSCSI - Google Patents

Abstract

A system and method for implementing error return in the Internet Small Computer System Interface (hereinafter referred to as iSCSI) is provided. A virtual SCSI disk driver is established in the iSCSI module in the target machine. When the initiator machine sends a command to the target machine, the correctness of the command can be quickly identified without transmitting it to the SCSI layer through the file system, and the error command information can be immediately returned to the initiator machine.

Description

System and method for implementing error return in iSCSI

technical field

The present invention relates to a system and method for returning error information, in particular to a system and method for directly returning error command information by establishing a virtual SCSI disk device in a target device when applied to a storage network.

Background technique

A storage network refers specifically to a network that connects storage devices and servers. Since a storage network uses a network to connect storage devices in series, it is not only capable of high-speed transmission, but also free of The expansion and backup of the storage device are limited by the server itself, and the scalability and reliability are good. Fiber Channel is currently the most commonly used storage network connection technology representative.

Another emerging storage network technology is the iSCSI (Internet Small Computer System Interface) transmission protocol, which has just completed standard formulation. The main function of this transmission protocol is to transmit SCSI commands and data through iSCSI Protocol Data Units (PDU, Protocol Data Units) on the currently developed IP network. With the iSCSI protocol, the IP network will also be able to become a storage network. The biggest advantage of using an IP network as a storage network is that for users, they only need to replace their servers and storage devices with iSCSI interfaces, and use the original IP network switching devices to build a low-cost and high-quality system. storage network and are used as if they were "direct access devices". iSCSI can realize the operation of SCSI protocol on IP network, so that it can perform routing on high-speed Gigabit Ethernet.

The iSCSI transport protocol is a standard developed by the Internet Engineering Task Force (IETF, Internet Engineering Task Force), which defines the initiator machine (Initiator Machine), that is, the party that issues a data access request, usually the server or user computer, and the target Device (TargetMachine), that is, the party that receives and responds to the request, usually the storage device side, the transmission method between the two, including the format of the iSCSI PDU transmitted between the two, the establishment and termination of the Session, and The start and end of SCSI Task.

For traditional SCSI, the iSCSI protocol is based on the TCP/IP protocol, and then it is connected to the SCSI interface. Therefore, there is no need to be limited by the longest distance of SCSI, which can only reach 25 meters. The place where the network can be constructed, because the SCSI interface is used to transmit commands, and executing a SCSI command can only have two results: correct or wrong. After the initiating device transmits the command to the target device, it will wait for the execution result of the command. If there is an execution error, the target device must return an error message to the boot device to ensure that the boot device can operate normally.

Please refer to FIG. 1 , which is a schematic diagram of the process of transferring SCSI commands by iSCSI. When the starting device end 270 sends a command 101 whose format is a command description block (CDB, Command DescriptorBlock) from the SCSI layer 100 to the iSCSI layer 110, the iSCSI layer 110 then wraps up a layer of protocol data unit 111, and finally adds TCP/ The IP header (header) can be transmitted to the Internet to reach the target device end 280 after becoming an IP packet.

And when the target device end 280 receives this packet from the TCP/IP network 120, the TCP/IP header is first removed by the network layer, and then passed to the iSCSI layer 110 to remove the header of the PDU111, and finally the command is described The block format data 101 is passed to the SCSI layer 100 . Since the error information returned from the target equipment end 280 must be included in the IP packet, but if the target equipment end 280 adds the error information in the header of the protocol data unit, then the layered disassembly process at the initiation equipment end 270 will result in The error information is lost to the SCSI layer 100. As a result, the SCSI disk device of the boot device 270 will not get the result of command execution, so that the operation of the boot device 270 and the target device 280 will be abnormal, and even cause the system to stop. operational hazards.

Contents of the invention

The main purpose of the present invention is to provide a system for realizing error return in iSCSI, which can process the error information in the iSCSI module in the target device and return it to the SCSI layer of the boot device to solve the problems in the prior art. For the above problems, correctly handle the error information, and ensure that the error can be returned to the SCSI layer on the boot device side.

In order to achieve the above object, the present invention is a system for realizing error return in iSCSI. For an IP packet sent by an initiating device in the system, it needs to disassemble its protocol data unit via an iSCSI module in a target device. Afterwards, the remaining SCSI command description block format data and commands are interpreted by a SCSI disk device, and executed in the SCSI disk device. It is characterized in that the system also includes: a virtual SCSI disk device, located in the iSCSI module of the target device In, when the target device is connected to the boot device via the network system and contains a logical sector management/redundant disk array disk device connected between the iSCSI layer and the SCSI layer, more than one error is identified in the iSCSI module When the command is issued, it is immediately fed back to the boot device, so that each SCSI command can be confirmed in real time, avoiding information loss, and resending the IP packet.

The virtual SCSI disk device is used for interpreting a plurality of commands sent by a plurality of SCSI command description blocks.

The virtual SCSI disk device can add error information to each of the SCSI command description blocks.

The SCSI layer at the boot device side is used for interpreting each SCSI command description block added with error information by the virtual SCSI disk device.

After the virtual SCSI disk device checks that each command information in the SCSI command description block is correct, it is still used to deliver each command information to the SCSI disk device for execution.

The present invention also provides a method for realizing error return in iSCSI. The method comprises the following steps: receiving an IP packet sent by a starting device; removing a protocol data unit and a header of a SCSI command description block; Take out a plurality of device information of SCSI command in the description block; When this device information is normal, pass the device information of this SCSI command description block to a file system, after processing, send confirmation information to this starting device end; When finding the When the device information is wrong, add more than one error handling information in the SCSI command description block; and directly send back to the starting device side.

The device information is a command executed by a SCSI disk device.

When the boot device receives the reply of the error message, it immediately resends the SCSI command.

The virtual SCSI disk device is used to interpret the commands of the SCSI disk device: READCAPACITY (10), MODE SENSE (6), REPORT LUNS, INQUIRY, TEST UNITREADY, VERIFY (10), READ (10), WRITE (10).

That is to say, the present invention provides a system for realizing error return in iSCSI, which is characterized in that: in addition to the original iSCSI module, the iSCSI layer of the target device also includes a virtual SCSI disk driver (Virtual SCSI disk driver), The iSCSI protocol at the iSCSI layer can accurately feed back the error execution result of the SCSI command to the boot device without transmitting it to the SCSI layer, so that each SCSI command can be confirmed in real time, avoiding information loss, and achieving the goal The information exchange between the device side and the starting device side finally ensures the stable operation of the target device side and the starting device side.

As for the method of realizing error return in iSCSI, the following steps are included: first, at the target device end, receive the IP packet sent by the boot device; then, remove the header of the protocol data unit and the command description block; then, from the command description block Take out the SCSI command; judge whether the contained device information is normal? If not, then add error handling information in the command description block, otherwise just directly pass the command description block information to the file system (file system), and finally, send confirmation information to the boot device.

Regarding the characteristics and implementation of the present invention, the preferred embodiments are described in detail below in conjunction with the accompanying drawings.

Description of drawings

Fig. 1 is a known system structure diagram;

Fig. 2 is a system structure diagram of the present invention; and

Fig. 3 is a flow chart of the method of the present invention.

Wherein, the reference signs are explained as follows:

100-SCSI layer; 101-SCSI command (CDB)

110-iSCSI layer; 111-iSCSI command (PDU);

120-TCP/IP network; 200-SCSI disk device;

210-LVM/RAID device; 220-file system;

230-virtual SCSI disk device; 240-iSCSI module;

250-network system; 260-IP packet;

270-Start device side; 280-Target device side;

Step 301 - receiving the IP packet sent by the boot device;

Step 310 - remove the header of PDU and CDB;

Step 320—retrieve the SCSI command from the CDB;

Step 330-Is the device information normal? ;

Step 340—transfer CDB information to the file system;

Step 350—sending confirmation information to the boot device;

Step 360—add error handling information to CDB;

Detailed ways

The present invention proposes a system and method for implementing error return in iSCSI. The iSCSI module 240 in the target device 280 can directly process error information to prevent information loss.

Please refer to FIG. 1 , which is a schematic diagram of a known process of transferring SCSI commands by iSCSI. As shown in the figure, the operation of both the initiator device 270 and the target device 280 on the SCSI disk device 200 is actually an operation on the SCSI disk driver, that is, the SCSI layer 100, so the iSCSI layer 110 can describe the command Blocks are passed directly to the SCSI layer 100 disk driver, much as they have been done with native SCSI disks. For the original local SCSI disk, the operating system only needs to pass the SCSI command description block to the driver program of the SCSI layer 100. After receiving the SCSI command description block, the driver program of the SCSI layer 100 can independently complete the processing of the command. After that, Send the execution result back to the operating system. Moreover, if the execution of the SCSI command is wrong, the driver program of the SCSI layer 100 will automatically record the error information into the SCSI command description block, and finally return the SCSI command description block to the operating system.

Please refer to Fig. 2, which is a system structure diagram of the present invention. The difference from Fig. 1 is that the storage device at the back end of the system is a LVM (Logical Volume Manager)/RAID (Redundant Array of Inexpensive Disks) device 210, and the system structure inside the boot device 270 is the same as that in Fig. 1, so The boot device end 270 in this figure is only omitted and replaced by a block. The original iSCSI system is a simple SCSI disk device 200, so when the disk driver program based on the SCSI layer 100 processes SCSI commands, it will automatically add error information to the SCSI command description block, and the iSCSI module 240 does not need to judge the SCSI command description The content of the block can directly transfer the SCSI command description block to the disk driver of the SCSI layer 100 for processing.

After the disk driver of the SCSI layer 100 has processed the command description block, if an error occurs, the command description block has contained error information, and the iSCSI layer 110 only needs to pack the SCSI command description block into a protocol data unit and send it to Just start the device terminal 270. And the starting device end 270 analyzes the result sent from the target device end 280 after obtaining the SCSI command description block, and this part is the same as processing the local SCSI disk.

Therefore, for this machine to directly store the TCP/IP mode 120 through the network and connect an iSCSI disk device, the iSCSI layer 110 only needs to directly send the SCSI command description block to the boot device end 270, and all error handling and judgment are in the The target device side 280 and the boot device side 270 are internally implemented by the respective disk drivers of the SCSI layer 100 .

But if the iSCSI module 240 is to control the LVM/RAID device 210, it must be accessed through the device file, which is different from the original iSCSI system that directly manipulates the disk driver of the SCSI layer 100. It must first pass through the file system 220. In order to control the corresponding equipment. As shown in Figure 2, when the boot device terminal 270 sends an I/O (Input/Output) request, the iSCSI module 240 cannot directly send this request to the SCSI disk device 200, that is to say, the SCSI command description block cannot directly At the same time, the SCSI disk device 200 cannot upload the error information to the iSCSI module 240 . Therefore, we must let the iSCSI module 240 have a function that can complete error handling by itself.

We adopt in the iSCSI module 240 to create a virtual SCSI disk device 230 to replace the SCSI disk device 200 to handle the work of the error. Interpret READ CAPACITY(10), MODE SENSE(6), REPORT LUNS, INQUIRY, TEST UNIT READY, VERIFY(10), READ(10), WRITE(10) and other commands one by one. If abnormalities are found, directly modify the SCSI The command description block organizes the corresponding error information according to the specific problem.

Please refer to FIG. 3 , which is a flow chart of the method of the present invention. First, after the target device 280 receives the IP packet 260 sent by the initiator 270 (step 300), the network system 250 will first remove the IP header, and then pass the remaining data down to the iSCSI module 240. The header of the protocol data unit is removed, and what remains is the command description block format to be interpreted by the SCSI disk device 200. The original method is to wait for the SCSI disk device 200 to be transmitted to the bottom layer before judging whether the command is wrong.

And our way is to utilize in iSCSI module 240, set up a virtual SCSI disk device 230, it is identical with SCSI disk device 200 for the ability of judging command description block format, so can directly remove the header of command description block (step 310) , and read the SCSI commands from the command description block (step 320), and check these commands to judge whether there is an error (step 330).

If the result of inspection finds that there is abnormality, then directly modify the command description block, add error handling information (step 360) therein, and pass down the command description block adding error code to iSCSI module 240; if the inspection result does not find Abnormal situation, then the information of command description block is directly passed to file system 220 (step 340), after file system 220 communicates with LVM/RAID device 210 again, command description block is passed to SCSI disk device 200 to carry out at last; When command description If the block information is executed normally and without error, it will send back a reply message of normal execution to the boot device end 270; otherwise, it will send back the command description block previously added with the error handling information.

When the boot device end 270 receives the information replied by the target device end 280, it also disassembles layer by layer in the reverse order. Since the virtual SCSI disk device 230 organizes error information according to the SCSI protocol standard, it can be sent to the boot device end 270. It is judged normally, when the boot device end 270 reads the content of the command description block and finds that there is an error in the execution of the command, it will immediately cancel the execution of the command, otherwise, this command will make the boot device end 270 wait for the result. state, which may eventually cause the system to fall into a crashed state.

Although the present invention is disclosed above with the foregoing preferred embodiments, it is not intended to limit the present invention. Any person familiar with the similar technology can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, The scope of patent protection of the present invention shall be determined by the appended claims of this specification.

Claims (8)

1. A system for realizing error return in the Internet small computer system interface iSCSI, for an IP packet transmitted by an initiating device in the system, it needs to disassemble its protocol data unit through an iSCSI module in a target device Afterwards, the remaining SCSI command description block format data and commands are interpreted by a SCSI disk device, and executed in the SCSI disk device. It is characterized in that the system also includes: A virtual SCSI disk device, located in the iSCSI module of the target device, connected to the boot device through the network system, the target device includes a logical sector management/disk redundant array disk connected between the iSCSI module and the SCSI disk device device, when more than one error command is identified in the virtual SCSI disk device, it will immediately feed back to the boot device side, so that each SCSI command can be confirmed in real time, avoiding information loss, and resending the IP packet . 2. The system for implementing error return in iSCSI according to claim 1, wherein the virtual SCSI disk device is used for interpreting a plurality of commands contained in a plurality of SCSI command description blocks. 3. The system for implementing error return in iSCSI according to claim 1, wherein the virtual SCSI disk device can add error information into each SCSI command description block. 4. The system for implementing error return in iSCSI as claimed in claim 1, wherein the SCSI disk device at the boot device is used to interpret each SCSI command description block added with error information by the virtual SCSI disk device. 5. The system for realizing error return in iSCSI as claimed in claim 2, characterized in that, after the virtual SCSI disk device checks that each of the commands in the SCSI command description block is correct, it is still used to send each of the commands to Executed by SCSI disk devices. 6. A method for implementing error return in iSCSI, the method comprising the following steps: Receive an IP packet sent by the boot device; removing a protocol data unit and a header of a SCSI command description block; Taking out multiple device information of SCSI commands from the SCSI command description block; When the device information is normal, transfer the device information of the SCSI command description block to a file system; After the processing is completed, a confirmation message is sent to the boot device; When it is found that the device information is wrong, add more than one error handling information in the SCSI command description block; and directly send back to the starting device side. 7 . The method for implementing error return in iSCSI according to claim 6 , wherein the boot device immediately resends the SCSI command when receiving the error handling information reply. 8 . 8. The method for realizing error return in iSCSI as claimed in claim 6, characterized in that, a virtual SCSI disk device is set up to interpret operation commands: READ CAPACITY (10), MODESENSE (6), REPORT LUNS, INQUIRY, TEST UNIT READY, VERIFY(10), READ(10), WRITE(10).

Images