CN1322705C - A method of datum plane reset for forwarding equipment - Google Patents

Abstract

The invention discloses a data plane reset method of a forwarding device, comprising the following steps: (1) making the service processing part of the control plane unable to perceive the data plane reset; (2) making the control plane not perform operations other than the data plane reset operation Step; (3) the step of saving the data plane before resetting; (4) the step of resetting the data plane (preferably including replacing the microcode in the network processor); (5) the step of data recovery . The method of the invention has the advantages of short message forwarding interruption time, no loss of dynamic information learned by the system, and the like.

Description

A data plane reset method for forwarding equipment

technical field

The present invention relates to a system reset method of a forwarding device of a router and/or switch, in particular to a data plane reset method of the forwarding device.

technical background

In recent years, the forwarding equipment of some routers and switches has shifted from the original backplane bus switching method to the switching network structure. As shown in Figure 1, these devices are usually composed of a control plane and a data plane. The control plane is responsible for processing and forwarding control packets and a small amount of data packets, while the data plane is responsible for forwarding data packets to ensure normal traffic forwarding at high speed. .

The core device of the control plane is only the CPU (central processing unit), while the core devices of the data plane include network processors, switching networks, network coprocessors, and framers. The number of core devices and their complex processing make the probability of problems in the data plane much higher than that in the control plane.

In frame-type forwarding devices (including frame-type routers and frame-type switches), the control plane and data plane are located on different boards, and redundant design of the data plane and active/standby switchover can be used to deal with problems. data plane. In a box-type forwarding device, the control plane and the data plane are located on the same single board. Generally, the problem of the data plane can only be solved by resetting the entire system. In this way, the business of the entire system is interrupted. Generally speaking, the method of resetting the entire system to solve the problems in the data plane has the following disadvantages:

1. Message forwarding is interrupted for a long time, generally ranging from a few minutes to more than ten minutes, depending on the time required from system restart to configuration recovery;

2. The dynamic information such as routing and network topology learned by the original system is lost, and it will take a certain amount of time to recover;

3. When the forwarding device is used as the access server, the original online user information is lost. After restarting, the user must re-authenticate and go online to access the network normally;

4. When there is a bug in the microcode in the network processor, it cannot be replaced online without interrupting the business.

Contents of the invention

An object of the present invention is to provide a method for resetting a data plane of a forwarding device, by using the method, the data plane can be reset without service interruption.

Another object of the present invention is to provide a method for resetting the data plane of the forwarding device, which can replace the microcode in the network processor of the data plane online without service interruption.

Another object of the present invention is to provide a method for resetting the data plane of a forwarding device, which can quickly detect data plane anomalies.

In order to achieve the above object, the present invention provides a method for resetting the data plane of the forwarding device, comprising the following steps: (1) making the service processing part of the control plane unable to perceive the step of resetting the data plane; (2) making the control plane not reset the data plane Steps of operations other than the reset operation; (3) steps of saving the data saved on the data plane before reset; (4) steps of resetting the data plane; (5) steps of restoring data.

Preferably, the method of the present invention further includes a step of detecting data plane anomalies before the above-mentioned step of making the business processing part of the control plane unaware of data plane reset; the step of detecting data plane anomalies preferably uses the control plane and data plane The "handshake message mechanism" between planes is realized.

The "handshake message mechanism" means that the control plane sends a pre-agreed handshake message to the data plane every 0.5 seconds, and the microcode of the network processor returns a response message through the traffic manager to maintain the heartbeat. If no response is obtained for 3 consecutive handshake messages, it is considered that the data plane is abnormal.

Preferably, the step of resetting the data plane in the method of the present invention further includes the step of replacing the microcode of the network processor of the data plane. This step is preferably realized by changing the way of indicating the array pointer to load the microcode.

Preferably, the step of making the service processing part of the control plane unaware of the reset of the data plane and the step of making the control plane not perform operations other than the reset operation on the management plane are realized by stopping the CPU of the control plane from responding to all interrupts.

The pre-reset data stored in the data plane memory is not cleared during reset, and other pre-reset data stored in the control plane memory are stored.

In the step of resetting the data plane, the step of resetting the framer of the data plane includes the step of setting a stub function that enables the control plane to query port status.

When a problem occurs in the data plane, the technical solution of the present invention only resets the data plane without resetting the entire system, so it has the following beneficial effects:

1. The message forwarding interruption time is short, and the device can resume normal forwarding within 2 seconds.

2. The dynamic information learned when the system is running before reset will not be lost.

3. When the box device is used as the access server, the original online user information will not be lost. Users do not need to re-authenticate to go online, and will not perceive abnormalities.

Description of drawings

FIG. 1 is a schematic diagram of the composition of forwarding equipment based on a switching network;

Fig. 2 is a flow chart of data plane reset steps in an embodiment of the present invention.

Detailed ways

The principle and preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

When there is a problem with the data plane, the reasons may be:

1. Defects in key chips;

2. Defects in the microcode running in the key chip (network processor chip);

3. The chip is in an abnormal state due to unpredictable reasons.

The data plane reset technology restores the data plane to the normal working state by individually resetting the data plane, which can solve the above data plane problems without resetting the entire system. In the method for separately resetting the data plane of the present invention, the following points should be achieved:

1. The service processing part of the control plane must not be aware of the reset of the data plane. Otherwise, once the control plane detects that the physical layer connection of the port is disconnected, it will delete the interface route; if the box device is used as an access server, the control plane will also log off all users on the port for this reason;

2. The control plane must not perform other operations on the management plane during the data plane reset period. Otherwise, the business module of the control plane will think that there is an exception in the data plane, which will lead to an exception in the forwarding of the control message;

In the embodiment of the present invention, the above two points are guaranteed by turning off and on the interrupt of the CPU of the control plane before and after the reset of the data plane.

3. It must be able to save the data saved in the data plane before reset, so that it can be restored correctly after reset. In the embodiment of the present invention, the correct storage and restoration of data is ensured through hardware self storage and control plane assisted storage.

The present invention also provides a method for detecting the abnormality of the data plane within a short period of time (for example, less than 2 seconds), and then resets the data plane after detecting the abnormality of the data plane. Otherwise, if an abnormality in the data plane causes user detection and other control plane operations to fail, it can only be recovered by triggering the user to continue to perform online operations. In the embodiment of the present invention, a handshake message is used to quickly detect data plane anomalies.

In addition, in order to solve the data plane abnormality problem caused by the defect of the microcode in the network processor, it may be necessary to replace the microcode. That is, when the data plane is reset, the microcode patch (the microcode after the problem is modified) is loaded. In the embodiment of the present invention, the loading of different microcodes is controlled through the array pointer.

Embodiments of the present invention will be described in detail below in conjunction with FIG. 2 , and the accompanying drawings are only for illustration, not limitation of the present invention.

A data plane reset begins in the following two situations:

1. When the control plane detects that the data plane is abnormal, reset the data plane. In the present invention, the "handshake message mechanism" is used to detect the abnormality of the data plane. The "handshake message mechanism" means that the control plane sends a pre-agreed handshake message to the data plane every 0.5 seconds, and the microcode running in the network processor of the data plane returns a response message through the traffic manager to maintain the heartbeat. If no response is obtained for three consecutive handshake messages, it is considered that the data plane is abnormal and enters the data plane reset process.

2. The user triggers the reset of the data plane. When there is a problem with the operation of the device on the Internet and the developer locates a defect in the microcode, when the control plane has the characteristics of a patch mechanism, the manufacturer can issue a patch, and the user can modify the problematic running code online through the patch mechanism of the control plane. The mechanism does not need to interrupt the business), and the microcode after the modification of the defect is loaded into the system in the form of a control plane patch, activated, and saved. The data plane is reset by triggering the command line; when the control plane does not have a patch mechanism, other methods can be used, such as storing the microcode in the storage device of the control plane, and the user triggers the reset of the data plane to replace the microcode.

Memory chips in the data plane can be initialized either by microcode or by the control plane. When the system is powered on and initialized, the control plane is responsible for clearing the contents of the memory chip in the data plane. As the initial state of the memory chip, the microcode cannot clear the contents of the chip, otherwise, the contents of the memory chip will be cleared by the microcode during the reset operation. . And because these contents are very large (usually ranging from 1M to 10M), and the access speed is slow, when the data plane is reset, the content in the memory chip will not be lost (unless it is cleared), and the information is not saved in the control plane, so MUST NOT be cleared by microcode.

As shown in FIG. 2 , in the first embodiment of the present invention, when the control plane detects an abnormality in the data plane, it starts to reset the data plane. First apply for the memory space of the control plane to save the dynamic information of the data plane. These dynamic information include information stored in the network processor, stored in the network co-processor, stored in the traffic manager, configured in the switching network, and configured in the framer. All interrupts to the control plane are then turned off, protecting the site where the data plane resets. Then call the driving function (these functions complete the function of making the data plane work normally), and save the scene. The field information includes the information stored in the network processor, the network co-processor, the traffic manager, the switching network configuration, and the framer configuration information. The information stored in the memory chip is saved by itself, and the information stored in other devices is cleared by the chip itself during initialization.

In the next step, reset the network processor, network coprocessor, traffic manager, switching network, and framer of the data plane.

First, initialize the network processor. Load the network processor microcode.

Then initialize the network coprocessor and traffic manager; initialize the switching network and framer according to the saved configuration information.

Next, restore the scene. These contexts include information stored in the network processor, stored in the network coprocessor, stored in the traffic manager, configured in the switching network, and configured in the framer.

Because after the framer (such as the PHY device of Ethernet) is re-initialized, it generally takes about 1 second for the configuration to recover, and during this time, the port is in the disconnected state. The control plane cannot be made aware to avoid the control plane from performing related control operations, such as taking the online users who interrupt the port offline, etc. Therefore, the stub function for the control plane to query the port status must be enabled. When this function is not enabled, the control plane queries the is the real state of the framer. After it is enabled, the control plane queries the state of the framer saved before the reset, and keeps enabling it for about 3 seconds.

Finally, turn on the interrupts that were turned off by the control plane, and turn on the scene.

Still as shown in FIG. 2 , in the second embodiment of the present invention, the reset of the data plane is triggered by the user.

In the step of initializing the network processor, it is judged whether the microcode needs to be loaded, and the modified patched microcode is loaded according to the judgment result. The microcode to be loaded is in the data segment of the program space of the control plane in the form of a global variable (the program space stores the instructions executed by the CPU, the global data used, etc., and usually has data segments, code segments, etc.). When modifying the microcode of the patch, change the pointer indicating loading of the microcode, and load the microcode of the patch after modification.

Other steps are the same as those of the first embodiment, so they are not repeated.

It should be noted that, according to the integration of the chip, the data plane network processor, framer, switching network, network coprocessor, traffic manager, etc. can exist in the form of a separate chip, or the network processing chip can integrate traffic management, Switching network, network co-processing function, network processing chip, traffic management chip, switching network chip, network co-processing chip are packaged into one chip on the hardware. In addition, if a higher port density is required, the network processing chip can exist in the form of multiple chips on the hardware. For the above various forms, the data plane reset method of the present invention is equally applicable.

The present invention has been described in detail above, but those of ordinary skill in the art should realize that without departing from the scope and spirit of the present invention, various improvements, additions and substitutions are all possible, and all within the scope of the present invention within the scope defined by the claims.

Claims (9)

1. A data plane reset method for a forwarding device, comprising the following steps: (1) The steps of making the business processing part of the control plane unable to perceive the reset of the data plane; (2) The step of preventing the control plane from performing operations other than the reset operation on the data plane; (3) the step of saving the data saved by the data plane before resetting; (4) a step of resetting the data plane; and (5) Steps for data recovery. 2. The method according to claim 1, characterized in that, before the step of making the service processing part of the control plane unable to perceive the reset of the data plane, the step of detecting data plane anomalies is further included. 3. The method according to claim 2, wherein the step of detecting data plane anomalies is realized through a "handshake message mechanism" between the control plane and the data plane. 4. The method according to claim 3, wherein the "handshake message mechanism" means that the control plane sends a pre-agreed handshake message to the data plane every 0.5 seconds, and the microcode will pass through the traffic manager The response message is sent back to maintain the heartbeat. When no response is received for 3 consecutive handshake messages, the data plane is considered abnormal. 5. The method according to any one of claims 1 to 4, wherein the step of resetting the data plane further comprises the step of replacing the microcode of the network processor of the data plane. 6. The method according to claim 5, wherein the step of replacing the microcode of the network processor of the data plane is realized by changing an array pointer indicating loading of the microcode. 7. The method according to any one of claims 1 to 4, characterized in that, the step of making the service processing part of the control plane unable to perceive the reset of the data plane and the step of making the control plane not perform operations other than the reset operation on the management plane The step is realized by stopping the CPU of the control plane from responding to all interrupts. 8. The method according to any one of claims 1 to 4, wherein the data before reset stored in the memory of the data plane is not cleared when reset, and the data stored before the data plane is reset Other data is kept in the memory of the control plane. 9. The method according to any one of claims 1 to 4, wherein the step of resetting the framer of the data plane includes the step of enabling the control plane to query the stub function of the port state.

Images