JP2004336427A - Search circuit and packet transfer device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a search circuit capable of obtaining a right searched result when a search key is divided and an LPM search process is carried out. <P>SOLUTION: The search circuit includes a search control unit 2 for dividing an entered search key into a plurality of divided search keys and performs longest identification search by using corresponding tables 1a to 1c from a high order divided search key which includes head data in the search key, and tables 1a to 1c provided as many as the divided search keys for storing entries including prescribed result information corresponding to the divided search keys. The tables 1a to 1c have dummy entries for avoiding a case that the result information on the divided search key is no longer acquired by dividing the search key. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an address search circuit that determines a transfer destination of a packet or the like based on an address written in a header of the packet, and a packet transfer device having the address search circuit.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, a router that transfers a packet uses a CAM (Content Addressable Memory, associative memory) or the like in order to perform a high-speed search for obtaining information such as a transfer destination port from a packet destination IP (Internet Protocol) address. It is realized using devices. In addition, by combining with a tree which is a general search algorithm, a search exceeding a search key length supported by the device is supported, and search of information such as a transfer destination port from a long address is facilitated. For example, the destination address of a packet, which is a search key in a router, is divided into arbitrary numbers, and a table of a type corresponding to each of the divided search keys is provided. In this table, data corresponding to the divided search key or another data is stored. A table search device that stores a pointer or an index indicating the table of (1) and implements a longest match (Longest Match) (hereinafter referred to as LPM) search has been proposed (for example, see Patent Document 1).
[0003]
[Patent Document 1]
JP 2001-326679 A (Page 2, FIG. 2)
[0004]
[Problems to be solved by the invention]
However, when an LPM search is performed by dividing a search key into a plurality of pieces as in the above-described conventional technique, there is a problem that a mishit occurs. FIG. 24 is a diagram for explaining a problem of the conventional LPM search. Here, an example will be described in which a destination of a packet is determined using a destination IP address (hereinafter, referred to as an address) stored in an IPv6 packet. In addition, in order to simplify the description, a case where a search is performed by dividing a search key into two will be described as an example.
[0005]
FIG. 24A shows the contents of the table in the case where the search key is not divided. The "search target information" of the two entries A and B contains the address A, which is the "result" of each transfer destination. B is associated therewith. Further, an item “mask length” is provided for each entry. That is, in the case of a packet having the address “4000: 0102: 0304: 0506: 0708: 090a: 0b0c: 0d04” (in hexadecimal notation. The following address is also indicated in hexadecimal notation). , Since the packet matches the entry A, the packet is output (transferred) to the address A, and the address obtained as a result of masking with 48 bits is the address "4000: 0102: 0304: 0000: 0000: 0000: 0000: 0000: 0000". Indicates that the data is output (transferred) to the address B. Note that the entries A and B are the same up to the 48th bit, but in the LPM search, the result of the search target information that has a longer match with the search key has priority.
[0006]
FIG. 24B shows a table obtained by dividing the table shown in FIG. 24A into two parts. The first table has “search target information” including the upper 64 bits of the divided table and the corresponding table. The “mask length” and the “result data” are stored. The second table stores “search target information” including the lower 64 bits of the address of the divided search key, and the corresponding “mask length” and “result”. "Data" is stored. Although not shown, the entry A of the first table and the entry A of the second table are associated by an index or the like. The address stored in the entry B is stored in the first table but is stored in the second table since the destination can be determined by masking the upper 48 bits of the 128 bits. Is not stored.
[0007]
For example, in the case of the packet 1 having the address “4000: 0102: 0304: 0506: 0708: 090a: 0b0c: 0d04”, the search key address is the first divided search key “4000: 0102: 0304: 0506”. It is divided into a second divided search key “0708: 090a: 0b0c: 0d04”. First, the first divided search key matches the entry A in the first table, and information to be searched in the second table is obtained. Then, when the second divided search key is searched in the second table, it matches the entry A, and the address A is obtained as a result. In the case of the packet 2 having the address “4000: 0102: 0304: 0000: 0708: 090a: 0b0c: 0d04”, the search key address is the first divided search key “4000: 0102: 0304: 0000”. It is divided into the second divided search key “0708: 090a: 0b0c: 0d04”, and the first divided search key matches the entry B of the first table, so that the address B is obtained as a result.
[0008]
On the other hand, in the case of the packet 3 having the address “4000: 0102: 0304: 0506: 0708: 090a: 0b0c: 0000”, the address which is the search key is the first divided search key “4000: 0102: 0304: 0506”. And the second divided search key “0708: 090a: 0b0c: 0000”, but the search result cannot be obtained. That is, the first divided search key matches the entry A of the first table, and the information to be searched is obtained in the second table. However, when the second table is searched by the second divided search key, the second There is no entry in the table that matches the search condition, resulting in a mishit. However, since this packet 3 is a packet matching the entry B from FIG. 24A, it is necessary to actually obtain the address B as a result. As described above, in the case where the inclusion relationship exists between the entry A and the entry B, if the search key is divided and the LPM search is performed, there is a problem that such a miss occurs.
[0009]
The present invention has been made in view of the above, and provides a search circuit capable of obtaining a correct search result even when a search key is divided and performing an LPM search, and a packet transfer device using the search circuit. The purpose is to:
[0010]
[Means for Solving the Problems]
In order to achieve the above object, a search circuit according to the present invention divides an input search key into a plurality of divided search keys, and stores a corresponding table from a higher-order divided search key including the leading data in the search key. A search control means for performing a longest match search in order using, and obtaining a desired result information; and a table provided as many as the number of the divided search keys, and an entry including predetermined result information corresponding to the divided search key. And a table for storing, wherein the table has a dummy entry for preventing the result information for the divided search key from being lost due to the division of the search key.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, preferred embodiments of a search circuit and a packet transfer device according to the present invention will be described in detail with reference to the accompanying drawings.
[0012]
Embodiment 1 FIG.
FIG. 1 is a block diagram schematically showing a configuration of a search circuit according to the present invention. This search circuit divides a search key into a plurality of divided search keys by a plurality of tables (search engines) 1a to 1c including a CAM for performing an LPM search, and performs a search using each of the tables 1a to 1c. A search control unit 2 to be executed, a search key division / tag management unit 3 for setting entries (search keys and results) for which entry addition / deletion has been requested from a management device (not shown) for managing a search circuit in the tables 1a to 1c. , And includes an inclusion relation detection / dummy entry management unit 4 which manages necessary information such as entry inclusion relations and provides the information at the time of setting by the search key division / tag management unit 3. In FIG. 1, the table such as the CAM is divided into a plurality of tables, but may be logically divided into a plurality of tables. Further, the case where three tables 1a to 1c are used is shown, but the present invention is not limited to this. Further, in the following description, a case where a destination IP address (address) stored in a packet is used as a search key and an output destination for transferring the packet to this address is searched will be described as an example. The application is not limited to this.
[0013]
The tables 1a to 1c are used to perform an LPM search for the search key (address) input to the search circuit and obtain a result of the output destination, and are composed of tables obtained by dividing the search key. As shown in FIG. 24A, the address of an entry whose address and its output destination (result) are determined is divided into a plurality of addresses (hereinafter, referred to as divided search key corresponding data), and the first bit is included. .. Are stored as entries in the first table, the second table,... In this specification, when comparing two tables, a table closer to the first table is called a relatively higher table, and a table farther from the first table is relatively lower. It is called a table.
[0014]
FIG. 2 is a diagram illustrating an example of a table configuration. As shown in FIG. 2, the table 1 includes “search target information” in which data corresponding to a divided search key, which is an address having a length corresponding to the divided search key, is stored. An item including a "mask length" in which a mask length indicating the length of data that needs to be matched is stored, and a "result data" in which a result when the entry matches the entry stored in the "search target information" is stored. Consists of When the output destination can be derived from the entry stored in the “search target information” masked by the mask length in the “result data”, the output destination is stored, and the output destination cannot be derived. Stores a tag required for performing a search in the next table that exists below. In the "search target information", an address having a length corresponding to the divided search key is stored in the first table 1a, but in the tables below the second table 1b, a tag corresponding to the length corresponding to the divided search key is stored. The search target information in which the addresses are combined is stored.
[0015]
The search control unit 2 has a function of dividing a search key into two or more divided search keys, outputting a key for executing an appropriate search to each of the tables 1a to 1c, and obtaining a correct search result. The search key is divided into the search keys by the same number as the tables 1a to 1c, and the same bits as the addresses (split search key corresponding data) in the search target information stored in the tables 1a to 1c. It is done to divide by number. In this specification, a first divided search key, a second divided search key, a third divided search key, and so on are referred to in order from a divided search key including a leading bit of the divided divided search keys. Further, similarly to the table, when two divided search keys are compared, a divided search key closer to the first divided search key is called a relatively higher divided search key, and the first divided search key is used as the first divided search key. The distant divided search key is called a relatively lower divided search key. When the search control unit 2 obtains a tag as a result of outputting the first divided search key to the first table 1a, for example, the combination of the tag and the next second divided search key is used as the first divided search key. 2 to the table 1b. When the output destination of the packet, which is the final result, is obtained from a certain table in this way, the search processing ends.
[0016]
The search key division / tag management unit 3 is configured to execute the inclusion relationship detection / dummy entry management unit 4 when a request for adding or deleting a new entry (divided search key corresponding data, result data) to the tables 1a to 1c is made. It has a function to set the search control unit 2 and the tables 1a to 1c so that the search can be performed correctly based on the instruction.
[0017]
When an entry is added to the tables 1a to 1c, the inclusion relation detection / dummy entry management unit 4 investigates the inclusion relation between the added entry and the entry in the tables 1a to 1c, etc. It manages whether or not the search can be performed, and if necessary, instructs the search key division / tag management unit 3 to add or delete a dummy entry based on the necessary information such as the inclusion relation. It has a function to provide.
[0018]
Note that the dummy entry is used to prevent a search result such as the packet 3 described above in the problem of the conventional LPM search from becoming a mishit and to obtain an original transfer destination. This is an entry to be inserted into Table 1. Specifically, when there is an inclusion relationship between the entry to be added and the entry in Table 1, an arbitrary address is added to the tag stored in the result data of the included entry. Has a data structure that includes the “division target information”, the “mask length” using the tag length as the mask length, and the “result data” that stores the output destination that is the final result information of the enclosing entry. .
[0019]
Next, the operation of the inclusion relation detection / dummy entry management unit 4 will be described with reference to FIGS. The inclusion relationship detection / dummy entry management unit 4 performs two processes of adding and deleting entries in consideration of the inclusion relationship with the entries already stored in the tables 1a to 1c. The addition will be described, and then the deletion of the entry will be described.
[0020]
FIG. 3 is a flowchart showing the overall processing procedure when adding an entry. First, when there is a request to add an entry from the management device, the inclusion relation detection / dummy entry management unit 4 sets the search key of the entry to be added to the first to the first of the divided search key lengths specified in the respective tables 1a to 1c. The data is divided into third divided search key corresponding data (step S1). Next, it is determined how the entry to be added has an inclusion relationship with other entries, and it is determined whether or not a dummy entry is necessary, and if so, what kind of dummy entry is required ( Step S2). Next, an entry is added to each of the tables 1a to 1c (step S3), and it is determined whether the entry addition process is completed or a dummy entry is added based on the result of step S2 (step S4). If it is determined that it is not necessary to add a dummy entry (No in step S4), the process is terminated, and if it is determined that a dummy entry needs to be added (Yes in step S4), Then, a dummy entry is added from the result obtained in step S2 (step S5), and the process ends.
[0021]
Here, details of the inclusion relation detection processing in step S2 will be described. 4 to 8 are diagrams showing the contents of determination of inclusion relation detection. It should be noted that these diagrams showing the contents of the judgment of inclusion relation detection are the same as the judgment of inclusion relation detection performed when an entry is deleted, which will be described later. ing. FIG. 4 is for determining whether or not a dummy entry is necessary based on the inclusion relationship between the entry to be added and the entry already stored in the table. This inclusion relation is determined before the search key of the entry to be added is divided. If the entry to be added (hereinafter, referred to as an additional entry) does not include another entry and is not included in another entry, the dummy entry is not necessary.
[0022]
If the additional entry does not include another entry and is included in another entry, the presence / absence of a dummy entry is determined in accordance with the inclusion relationship 1 shown in FIG. FIG. 5 is a diagram illustrating a result of determining whether or not a dummy entry is necessary based on the relationship between the mask length of an additional entry and the mask length of an entry including the additional entry in the inclusion relationship 1. Here, the area 1 refers to a case where the mask length is equal to or less than the length of the divided search key corresponding data stored in the first table 1a (hereinafter, referred to as a first divided search key length). Indicates that the mask length is longer than the first divided search key length, and the first divided search key length and the length of the divided search key corresponding data stored in the second table 1b (hereinafter referred to as the second divided search key Area 3), and region 3 refers to a case where the mask length is longer than the sum of the first divided search key length and the second divided search key length. When the additional entry is included in a plurality of entries, the determination in FIG. 5 is applied to the entry having the longest mask length.
[0023]
For example, if both the mask length of the additional entry and the mask length of the entry including the additional entry are in the area 1, there is no need to add a dummy entry. Further, when the mask length of the additional entry is in the area 2 and the mask length of the entry including the additional entry is in the area 1, the processing of “addition 1-1” of the dummy entry shown in FIG. Will be determined. FIG. 8 is a diagram that specifies the result of the dummy entry addition determination and the processing content thereof. For example, in the case of “addition 1-1”, a process of adding a dummy entry of an entry including an additional entry to the second table 1b is selected.
[0024]
In FIG. 4, when the additional entry includes another entry and is not included in the other entry, the presence / absence of a dummy entry is determined corresponding to the inclusion relationship 2 shown in FIG. You. In FIG. 6, it is determined whether or not a dummy entry is necessary based on the relationship between the mask length of the additional entry and the mask length of the entry included in the additional entry. Then, as a result of this determination, it is determined that addition or change processing of the dummy entry defined in FIG. 8 is necessary. Here, when the additional entry includes a plurality of entries doubly, the determination in FIG. 6 is applied to the entry having the shortest mask length. Further, when the additional entry includes a plurality of entries but is not in a relationship of doubly including, the determination in FIG. 6 is applied to each entry (that is, the plurality of entries are included). When A and B and the additional entry C are not A⊂C, B⊂C and neither A⊆B nor B⊆A, the determination of FIG. 6 is performed for both the entry A and the entry B).
[0025]
Further, in FIG. 4, when the additional entry includes another entry and is included in another entry, the presence / absence of a dummy entry is determined corresponding to the inclusion relationship 3 shown in FIG. . FIG. 7 is a diagram for determining whether or not a dummy entry is necessary based on the relationship between the mask length of the additional entry, the mask length of the entry included in the additional entry, and the mask length of the entry including the additional entry. is there. Then, as a result of this determination, it is determined that addition or change processing of the dummy entry defined in FIG. 8 is necessary. Here, when the additional entry includes a plurality of entries doubly, the determination in FIG. 7 is applied to the entry having the shortest mask length. If the additional entry includes a plurality of entries but does not have a double inclusion relationship, the determination of FIG. 7 is applied to each entry (ie, a plurality of entries). If the entries C to be added, A and B, are not A⊂C, B⊂C and neither A⊆B nor B⊆A, the determination in FIG. 7 is performed for both the entries A and B). Further, when the additional entry is included in a plurality of entries, the determination in FIG. 7 is applied to the entry having the longest mask length.
[0026]
As described above, the inclusion relationship between the additional entry and the entry already stored in the tables 1a to 1c is determined in consideration of the mask length between the entries. Various algorithms for obtaining the above-described inclusion relation can be considered. For example, a method disclosed in Japanese Patent Application Laid-Open No. 11-103313 can be applied.
[0027]
Next, the details of the entry addition process in step S3 in FIG. 3 will be described. Here, the processing procedure is shown on the premise that the entry has already been checked by the management device so as not to add the already registered entry.
[0028]
FIG. 9 is a flowchart illustrating a procedure for adding an entry. First, the inclusion relation detection / dummy entry management unit 4 determines whether or not the same entry as the first divided search key corresponding data of the entry exists in the first table for adding the entry (step S101). ). At this time, it is different from a normal search in that not only the first divided search key corresponding data but also the mask length is taken into consideration. You have to). When it is determined that there is no entry that is the same as the first divided search key corresponding data (No in step S101), the mask length of the additional entry is determined (step S102). If the mask length of the additional entry is “mask length ≦ first divided search key length (area 1)” (Yes in step S102), the entry is added to the first table 1a (step S102). S103). This flow is a normal flow when the mask length of the additional entry is equal to or less than the first divided search key length.
[0029]
When it is determined in step S102 that the mask length is not the area 1 (in the case of No in step S102), the inclusion relation detection / dummy entry management unit 4 uses the unused second table 1b for the second table 1b. A tag (hereinafter, referred to as tag 1) is obtained (step S104), and the first divided search key corresponding data is set as “search target information”, and the obtained tag 1 is set as an entry to be set as “result data” in the first table. 1a (step S105). Next, the mask length of the additional entry is determined (step S106). At this time, it is determined that the mask length of the additional entry is “first divided search key length (area 1) <mask length ≦ first divided search key length + second divided search key length (area 2)”. If it is determined (Yes in step S106), in the second table 1b, the combination of the tag 1 and the data corresponding to the second divided search key is set to “search target information”, and the result of the additional entry is set to “result data”. , An entry is added (step S107). This flow is performed when the mask length of the additional entry is longer than the first split search key length and is equal to or less than the sum of the first split search key length and the second split search key length. This is a flow when the same entry has not been registered yet.
[0030]
If it is determined in step S106 that the mask length of the additional entry is not the area 2 (No in step S106), the inclusion relation detection / dummy entry management unit 4 uses the third table 1c that is not used. (Hereinafter, referred to as tag 2) (step S108), the combination of tag 1 and the data corresponding to the second divided search key is referred to as “search target information”, and the obtained tag 2 is referred to as “result data”. An entry to be added is added to the second table 1b (step S109). Then, the acquired tag 2 and the third divided search key corresponding data are added to the third table 1c as “search target information” and the result of the additional entry is set as “result data” (step S110). This flow is performed when the mask length of the additional entry is longer than the sum of the first divided search key length and the second divided search key length, and the first divided search key corresponding data + the second divided search key corresponding data This is a flow in the case where the same entry as has not been registered yet.
[0031]
On the other hand, if an entry already exists in the first table 1a in step S101 (Yes in step S101), it is determined whether the result of the entry is a tag (step S111). If it is determined that the result of the entry is not a tag (No in step S111), the entry is deleted from the first table 1a (step S112). This entry is for the case where the mask length is equal to the length of the area 1, and a dummy entry for this entry will be added later. After that, the processing after step S102 described above is performed. This flow is a special case flow in which there is already the same entry as the entry to be added to the first table 1a, and the mask length of the entry is equal to the first divided search key length.
[0032]
In such a situation, when the first divided search key corresponding data is equal and the mask length is the area 2 or 3, the tag information must be registered as “result data” in the original entry. The entry is deleted in step S112. However, although the entry is deleted here, a process of adding an entry in place of the deleted entry to the second table 1b is performed by a dummy entry addition process performed later.
[0033]
When it is determined in step S111 that the result of the entry is a tag (Yes in step S111), the inclusion detection / dummy entry management unit 4 sets the tag value (hereinafter, this tag value is also referred to as the tag value). Tag 1) is obtained from the first table 1a (step S113). Next, the mask length of the additional entry is determined (step S114). If it is determined that the mask length of the additional entry is not the area 1 (No in step S114), it is determined whether or not there is an entry of the acquired tag 1 + second divided search key corresponding data in the second table 1b. Is determined (step S115). If it is determined that there is no entry of the acquired tag 1 + second divided search key corresponding data in the second table 1b (No in step S115), the number of tags 1 used in the second table 1b is calculated. The value is incremented, and the processing after step S106 is performed. This flow is performed when the mask length is longer than the first divided search key length, when an entry having the same value as the first divided search key corresponding data is already registered, and when the second divided search key corresponding data is Is a flow when an entry of the same value has not been registered yet.
[0034]
If it is determined in the above-described step S115 that there is an entry of tag 1 + second divided search key corresponding data in the second table 1b (Yes in step S115), the result of the entry is a tag. It is determined whether or not it is (step S117). If it is determined that the entry is not a tag (No in step S117), this entry is deleted from the second table 1b (step S118). After that, the processing after step S106 described above is performed. This flow is a flow in the case where there is already an entry in the first and second tables 1a and 1b, and the mask length of the entry is equal to the sum of the first divided search key length + the second divided search key length. is there.
[0035]
If it is determined in step S117 that the entry result is a tag (Yes in step S117), the tag value (hereinafter, this tag value is also referred to as tag 2) is obtained (step S119). . Next, the mask length of the additional entry is determined (step S120). If the mask length of the additional entry is not the area 2, that is, if the mask length is the area 3 (No in step S120), the use number of the tag 2 in the third table 1c is incremented, and the processing in step S110 is performed. . This flow is for the case where the mask length is longer than the sum of the first divided search key length + the second divided search key length, and entries of the same value are registered up to the first and second divided search key corresponding data. If it is.
[0036]
On the other hand, when it is determined in step S120 that the mask length of the additional entry is the area 2 (in the case of Yes in step S120), the process ends without adding an entry. The processing for this entry is performed when adding a dummy entry.
[0037]
Also, when it is determined in step S114 that the mask length of the additional entry is the area 1 (in the case of Yes in step S114), the process ends without adding an entry. The processing for this entry is performed when a dummy entry is added.
[0038]
As described above, entries are added to the first to third tables 1a to 1c.
[0039]
Next, the dummy entry adding process in step S5 of FIG. 3 will be described. FIG. 8 defines the contents of processing for adding a dummy entry to the result of the inclusion relation determination processing of FIGS. 5 to 7. After the entry addition process shown in FIG. 9 is completed, a dummy entry addition process or a change process is performed on each of the tables 1a to 1c based on the determination result of the inclusion relation in step S2 regarding the additional entry and the processing content of FIG. .
[0040]
As described above, when a new entry is added to the tables 1a to 1c, a mishit at the time of search that occurs because the inclusion relationship between the additional entry and the entry already stored in the table is not considered is considered. Dummy entries to prevent this are added as needed.
[0041]
Next, the deletion of an entry by the inclusion relation detection / dummy entry management unit 4 will be described. FIG. 10 is a flowchart showing the entire processing procedure when deleting an entry. First, when there is an entry deletion request from the management device, the search key is divided into first to third divided search key corresponding data having the divided search key lengths specified in the tables 1a to 1c, respectively (step S11). Next, it is determined whether the dummy entry becomes unnecessary by deleting the entry based on the inclusion relation with other entries (step S12), and the entries of the respective tables 1a to 1c are deleted (step S13). .
[0042]
Thereafter, based on the result of step S12, it is determined whether to terminate the entry deletion processing or further delete the dummy entry (step S14). If there is no deletion of the dummy entry (No in step S14), the process ends. If a dummy entry has been deleted (Yes in step S14), unnecessary dummy entries are deleted (step S15), and the process ends. Note that the inclusion relation detection processing in step S12 is the same as the case described with reference to FIGS. 4 to 7 in the entry addition processing, and thus description thereof will be omitted. In this case, however, FIG. 12 that defines the content of the deletion process is used instead of FIG. 8 that defines the content of the addition process.
[0043]
Here, the details of the entry deletion processing in step S13 described above will be described. Note that the processing procedure is shown on the assumption that the management apparatus checks the entry so that the unregistered entry is not deleted.
[0044]
FIG. 10 is a flowchart illustrating the procedure of the entry deletion process. First, in order to delete an entry, the inclusion relation detection / dummy entry management unit 4 determines whether the result of the entry to be deleted in the first table 1a is a tag (step S201). If it is determined that the result of the entry in the first table 1a is not a tag (No in step S201), the entry of the first divided search key corresponding data is deleted from the first table 1a (step S213). ), End the entry deletion process. This flow is a normal flow when the mask length of the entry to be deleted is the area 1.
[0045]
Next, in step S201, if it is determined that the result of the entry in the first table 1a is a tag (Yes in step S201), the inclusion relation detection / dummy entry management unit 4 Is determined (step S202). Here, when it is determined that the mask length of the entry to be deleted is not the area 1 (No in step S202), it is determined whether or not the result of the entry in the second table 1b is a tag (step S202). S203). When it is determined that the result of the entry in the second table 1b is not a tag (No in step S203), the entry of the data corresponding to the second divided search key is deleted from the second table 1b (step S209). ). Thereafter, the number of used tags 1 in the second table 1b is decremented (step S210), and it is determined whether or not the number of used tags 1 has become 0 (step S211). If the number of uses of the tag 1 has become 0 (Yes in step S211), the corresponding tag 1 is released (step S212), and the processing in step S213 described above is performed. This flow is a flow when the mask length of the entry to be deleted is the area 2 and there is no other entry having the same value as the first divided search key corresponding data.
[0046]
On the other hand, if it is determined in step S211 that the number of uses of the tag 1 is not 0 (No in step S211), the process ends. This flow is a flow when the mask length of the entry to be deleted is the area 2 and there is another entry having the same value as the first divided search key corresponding data.
[0047]
If it is determined in step S203 that the result of the entry in the second table 1b is a tag (Yes in step S203), the mask length of the entry to be deleted is determined (step S204). If it is determined that the mask length is not the area 2 (No in step S204), the entry is deleted from the third table 1c (step S205). Thereafter, the number of used tags 2 in the third table 1c is decremented (step S206), and it is determined whether the number of used tags 2 has become 0 (step S207). If the number of uses of the tag 2 has become 0 (Yes in step S207), the corresponding tag 2 is released (step S208), and the processing from step S206 described above is performed. This flow is a flow in a case where the mask length of the entry to be deleted is the entry of the area 3 and there is no other entry having the same value up to the second divided search key corresponding data.
[0048]
On the other hand, if it is determined in step S207 that the number of uses of the tag 2 is not 0 (No in step S207), the process is terminated. This flow is a flow when the mask length of the entry to be deleted is the area 3 and there is another entry having the same value up to the second divided search key corresponding data.
[0049]
If it is determined in step S204 that the mask length is the area 2 (Yes in step S204), the process ends. This flow is when the mask length of the entry to be deleted is equal to the sum of the first divided search key length + the second divided search key length, and there is another entry having the same value up to the second divided search key corresponding data. It is a flow of a case. Since the entry itself to be deleted is registered as a dummy entry, it is deleted in the dummy entry deletion processing step.
[0050]
If it is determined in step S202 that the mask length is the area 1 (Yes in step S202), the process ends. This flow is a flow when the mask length of the entry to be deleted is equal to the sum of the first divided search key lengths and there is another entry having the same value as the first divided search key corresponding data. Also in this flow, this entry itself is deleted in the process of deleting the dummy entry.
[0051]
As described above, the entry is deleted from the first to third tables 1a to 1c.
[0052]
Next, the dummy entry deletion processing in step S15 in FIG. 10 will be described. FIG. 12 shows the contents of deletion of the dummy entry. FIG. 12 defines the contents of the process of deleting the dummy entry with respect to the result of the inclusion relationship determination process of FIGS. 4 to 7. After the entry deletion process shown in FIG. 11 is completed, a dummy entry deletion process or a change process is performed on each table based on the determination result of the inclusion relation in step S12 regarding the entry to be deleted and the processing content of FIG.
[0053]
As described above, when an entry is deleted from the tables 1a to 1c, the dummy entry is deleted including the inclusion relation between the entry to be deleted and the entry already stored in the table. .
[0054]
In the above description, the case where the search key is divided into three is described as an example. However, even when the search key is divided into two or four or more, entry addition or deletion processing can be realized by the same processing procedure. .
[0055]
Here, a dummy entry addition process, a deletion process, and a search process when a dummy entry is added will be described with specific examples. First, the process of adding a dummy entry will be described by taking as an example a case where a new entry is added to two tables in which one entry is already stored.
[0056]
FIG. 13 is a diagram showing how entries are added when there are two tables. FIG. 13A shows the state of the table before adding a new entry. Here, an entry corresponding to the entry A of the table of FIG. 24A is stored in the first and second tables 1a and 1b. In these tables 1a and 1b, the address (search key) of the table of FIG. 24A is "4000: 0102: 0304: 0000: 0000: 0000: 0000: 0000", the mask length is 48 bits, and the transfer is performed. When adding the entry B whose address is the address B, first, the inclusion relation detection / dummy entry management unit 4 sets the address (search key) to the first divided search key “4000: 0102: 0304: 0000”. It is divided into a second divided search key “0000: 0000: 0000: 0000”.
[0057]
When examining the inclusion relationship between the additional entry and the entries stored in the tables 1a and 1b, the additional entry includes the entry A but does not include other entries, and therefore, corresponds to the inclusion relationship 2 from FIG. I do. Further, since the mask length (48 bits) of the additional entry is equal to or less than the first divided search key length (64 bits), and the mask length of entry A is 128 bits and is in area 2, the "add dummy entry" in FIG. 2-1 ".
[0058]
In the first table 1a, the same entry as the first divided search key corresponding data does not exist (No in step S101), and the mask length (48 bits) is in the area 1 (Yes in step S102). ), The first divided search key corresponding data is added to the first table 1a. Here, as shown in FIG. 13B, “4000: 0102: 0304: 0000” is added to “search target information”, “48” is added to “mask length”, and “48” is added to “result data”. “Address B” is added.
[0059]
Then, the result of the inclusion relationship is “addition of dummy entry 2-1”, which performs the processing of “addition of dummy entry of added entry to second table” from FIG. Therefore, in the second table 1b, the “search target information” is “tag α + 0000: 0000: 0000: 0000”, the “mask length” is “n” which is the number of bits of tag α, and the “result A dummy entry whose “data” is “address B” is added. The result is shown in FIG. Note that the numerical value added after the tag α of the “search target information” can be set to an arbitrary numerical value because the “mask length” is the number of bits of the tag α.
[0060]
Next, a case where an entry is deleted from two tables will be described with reference to FIG. FIG. 14 shows how entries are deleted from the two tables, and FIG. 14A corresponds to the state shown in FIG. 13C. As described above, the entry B to be deleted corresponds to the inclusion relation 2 in FIG. 4 with respect to the other entries in the tables 1a and 1b, and corresponds to “dummy entry deletion 2-1” in FIG.
[0061]
Since the "result data" of the entry B of the first divided search key corresponding data in the first table 1a is not a tag (No in step S201), the first divided search key corresponding data The entry is deleted (step S213). This state is shown in FIG.
[0062]
Further, the processing of “dummy entry deletion 2-1” to “deletion of dummy entry of deleted entry in second table” in FIG. 12 is executed. Therefore, the dummy entry of the entry deleted in the first table 1a, that is, “search target information” is “tag α + 0000: 0000: 0000: 0000”, “mask length” is “n”, and “result data” Is deleted from the second table 1b. The result is shown in FIG.
[0063]
Next, a search procedure using a table having dummy entries will be described with reference to FIG. FIG. 15 shows an example in which the first embodiment is applied to the conventional example shown in FIG. 24, which is the same as FIG. 13C, and is inserted into the second table 1b by the search circuit. Stored dummy entry. In this dummy entry, a mask length is set so that portions other than the tag are not compared, and the “result data” stores the same result as that of the entry B. By doing so, the packet 3 with the search key “4000: 0102: 0304: 0506: 0708: 090a: 0b0c: 0000” that could not be searched correctly in the conventional example of FIG. The key “4000: 0102: 0304: 0506” hits the entry A of the first table 1a to acquire the tag α, and then the tag α + the second divided search key “tag α + 0708: 090a: 0b0c: 0000” "Hits the dummy entry of the second table 1b whose mask length is the length of the tag α, so that it is possible to finally obtain the address B which is a correct result.
[0064]
By adding a dummy entry in this manner, the search control unit 2 and the table 1 can correctly search even in an LPM search in which a long search key is divided by exactly the same processing as before.
[0065]
According to the first embodiment, the inclusion relation between the entry to be newly added / deleted and the entry stored in the table 1 is obtained. Since addition / deletion and change of the dummy entry are performed in 1a to 1c, it is possible to correctly perform the search even in the LPM search divided into a plurality.
[0066]
Embodiment 2 FIG.
In the first embodiment, by adding a dummy entry, a plurality of divided LPM searches can be correctly performed. In the second embodiment, a flag is added to the contents of the entry, and this search is performed at the time of the search. A description will be given of a search circuit capable of correctly performing a search even in a plurality of divided LPM searches by selecting a result according to the value of a flag.
[0067]
FIG. 16 is a diagram showing a configuration of a search circuit table according to the second embodiment of the present invention. This table includes items of a “search key” serving as a key when performing a search and a “search result” storing a result for an entry that matches the search key. The “search key” includes items of “search target information” including a combination of a tag and a divided search key, and “mask length” indicating a mask length. The “search result” includes a “processing continuation flag” indicating whether or not to search for the next table, a “tag” storing a tag that is a part of a key when searching for the next table, It is configured to include “result data” in which the final result is stored and “data valid flag” indicating whether the result data is valid. In the case of the item “search target information”, in the case of the highest table, only the divided search key corresponding data is stored.
[0068]
In the following description, in the flags stored in the “processing continuation flag”, “1” indicates a case where the next table is searched, and a numerical value other than “1” (for example, “0”) is , The case where the next table is not searched. Further, in the flags stored in the “data valid flag”, “1” indicates valid, and a numerical value other than “1” (for example, “0”) indicates invalid. This definition is made in the description of this specification, and each flag may be defined differently from the above.
[0069]
In the second embodiment, by setting the "data valid flag" to valid and registering the data of the content added by the dummy entry in the "result data", the same effect as when the dummy entry is added can be obtained. I have to. The other configuration of the search circuit according to the second embodiment is the same as that of FIG. 1 of the above-described first embodiment, and a description thereof will be omitted.
[0070]
The entry addition / deletion processing in the second embodiment is performed in substantially the same procedure as in the first embodiment, but the processing for adding a dummy entry is performed by setting the “data valid flag” to be valid. The difference between the process of registering the result data (result information) of the included entry in the “result data” of the included entry and the process of deleting the dummy entry is the “data valid flag”. The difference lies in the process of invalidating.
[0071]
Next, a search method by the search circuit will be described. FIG. 17 is a flowchart of a search procedure according to the second embodiment of the search circuit according to the present invention. Also in this description, a case where the search key is divided into three will be described as an example. First, the search control unit 2 searches the first table 1a using the first divided search key obtained by dividing the search key (step S301), and determines whether there is an entry (step S302). . If there is an entry (Yes in step S302), it is determined whether the processing continuation flag of the entry is 1 (step S303), and if the processing continuation flag is 1 (Yes in step S303). ), A search is performed using the combination of the tag obtained as a result and the second divided search key using the second table 1b (step S304), and it is determined whether or not there is an entry in the second table 1b (step S304). Step S305). If there is an entry in the second table 1b (Yes in step S305), it is determined whether or not the processing continuation flag of the entry is 1 (step S306), and if the processing continuation flag is 1 (step S306). In the case of Yes in step S306), the third table 1c is searched using the combination of the tag obtained as a result and the third divided search key (step S307), and whether or not there is an entry in the third table 1c Is determined (step S308). If there is an entry in the third table 1c (Yes in step S308), it is determined whether the data valid flag of the third search result is 1 (step S309), and the data valid flag is set to 1 In the case of, the third search result is output as a valid result (step S310), and the search processing ends.
[0072]
If the processing continuation flag of the entry in the second table 1b is not 1 in step S306 (No in step S306), or if there is no entry in the third table 1c in step S308 (No in step S308) ), It is determined whether or not the data valid flag of the second search result is 1 (step S311). If the data valid flag of the second search result is 1 (Yes in step S311), the second search result is output as a valid result (step S312), and the search process ends.
[0073]
Further, if the processing continuation flag of the entry in the first table 1a is not 1 in step S303 (No in step S303), if there is no entry in the second table 1b in step S305 (no in step S305) Alternatively, if the data valid flag of the second search result is not 1 in step S311 (No in step S311), it is determined whether the data valid flag of the first search result is 1 (step S313). . If the data valid flag of the first search result is 1 (Yes in step S313), the first search result is output as a valid result (step S314), and the search process ends.
[0074]
If there is no entry in the first table 1a in step S302 (No in step S302), if the data valid flag of the third search result is not 1 in step S309 (no in step S309), or If the data valid flag of the first search result is not 1 in S313 (No in Step S313), the result is output as a mishit (Step S315), and the search process ends.
[0075]
By performing the processing in such a procedure, for example, in the search when a packet is input, if a miss is found in the first search (No in step S302), it is determined to be a miss. In the case of a mishit in the first or third search (steps S304 and S307) (No in each of steps S305 and S308), it is determined whether the data valid flag of the previous search result is valid. (Steps S311 and S313) If the result is invalid, it is regarded as a mishit. If the result is valid, it is possible to output the retrieved result as a final result. Note that the search is performed when the processing continuation flag is 1 in the previous search except for the first time (in the case of Yes in each of steps S303 and S306) (steps S301, S304, and S307).
[0076]
In the above description, the case where the search key is divided into three is described as an example. However, even when the search key is divided into two or four or more, entry addition or deletion processing and search processing are realized by the same processing procedure. can do.
[0077]
FIG. 18 is a diagram showing a case where the example described in FIG. 13 is applied to the case of the second embodiment. As shown in FIG. 18, when there is an inclusion relationship between the divided search key corresponding data of the entries A and B, the “tag” of the included entry A obtained from the first table 1 a The tag used for the next search in the table 1b is stored, the address B which is the result data (result information) of the included entry B is stored in the "result data", and the "data valid flag" is validated ( 1). With such a data structure, in the case of the packet 3 having the above-mentioned address “4000: 0102: 0304: 0050: 090a: 0b0c: 0000”, the first divided search key “4000: 0102: 0304: 0506” Since it matches the entry A in the first table 1a, the tag α is obtained as a “tag”, and the “processing continuation flag” is 1, so that the search in the second table 1b is further continued. However, since the second divided search key “0708: 090a: 0b0c: 0000” does not match any entry in the second table 1b, returning to the first table 1a, the “data valid flag” is set to 1 ( Valid), correct search can be performed by using the address B stored in the “result data” as the result of the packet 3.
[0078]
According to the second embodiment, the inclusion relation of the entry to be newly added / deleted is determined, and the data valid flag and the resulting data are registered in the tables 1a to 1c based on the determination result. Even in the case of a mishit, if the data valid flag of the previous search result is valid, the search result can be determined. As a result, the search can be correctly performed even in the LPM search divided into a plurality.
[0079]
Embodiment 3 FIG.
In the third embodiment, a case will be described in which the search circuit described in the first and second embodiments, which includes a plurality of search control blocks 2 for implementing search pipeline processing, is applied to a packet transfer device. I do.
[0080]
FIG. 16 is a block diagram showing the overall configuration of the packet transfer device according to the present invention. This packet transfer device connects a plurality of interfaces 10a, 10b,... For searching for the termination of the physical layer and the transfer destination of the packet, and the respective interfaces 10a, 10b,. A switch 11 is provided. The interfaces 10a, 10b,... Include a physical layer processing unit 12 for terminating the physical layer, a receiving side packet processing unit 13 for searching for a destination of a received packet, and a packet transmitted from the switch 11. It is provided with a transmission side packet processing unit 14 for encapsulating the data.
[0081]
FIG. 17 is a block diagram illustrating the configuration of the receiving-side packet processing unit. As shown in FIG. 17, the receiving-side packet processing unit 13 receives a packet transmitted from the physical layer processing unit 12 and checks the header and the like. Header extraction unit 21 for extracting appropriate information, search circuit 22 for searching the output destination of the packet from the extracted header, header insertion unit 23 for adding (inserting) the search result to the packet, and packet in a format that switch 11 can exchange. It comprises a packet output unit 24 for conversion.
[0082]
The search circuit 22 corrects the search results by dividing the search keys into first to third tables (search engines) 1a to 1c composed of a CAM or the like and outputting appropriate keys to the tables 1a to 1c. The first to third search control units 2a to 2c for determining the The functions of the first to third search control units 2a to 2c are basically the same as those of the search control unit 2 described in the first and second embodiments. It has a function of notifying the search control unit of the next stage of the result and the key.
[0083]
FIG. 21 is a diagram illustrating an outline of processing timing in the receiving-side packet processing unit. As shown in FIG. 21, it is assumed that packets 1, 2, 3, 4,... Are sequentially input to the packet input unit 20 of the receiving-side packet processing unit 13. First, in the first search control unit 2a, a search process using the first table 1a of the packet 1 is performed. At the next timing, a search result by the first search control unit 2a is output to the search control unit 2, and a search process using the second table 1b of the packet 1 by a combination of the search result and the second divided search key Is performed. At the same time, the first search control unit 2a performs a search process using the first table 1a of the next packet 2. Thereafter, in the same manner, the respective search control units 2a to 2c simultaneously execute search processes for different packets simultaneously, and when the search results are obtained in the third table 1c, output to the header insertion unit 23 in order from packet 1. Is done. Here, although the case where there are three tables 1a to 1c and three search control units 2a to 2c has been described, these may be configured as two or four or more.
[0084]
According to the third embodiment, since a plurality of search control units 2a to 2c are provided and configured to be able to execute pipeline processing, a packet transfer device capable of realizing the same throughput as in the case of searching for a short search key Can be realized.
[0085]
Embodiment 4 FIG.
In the first to third embodiments, the search key is divided into a plurality of divided search keys, the first table is searched using the first divided search key, and the search result and the remaining second divided search key are used. A search key having a long data length has been searched by sequentially performing processing such as searching the next second table using a new search key with the same key. However, in the fourth embodiment, each table is searched in parallel with the divided search key, and the table is searched using a combination of the searched result data (tags) as a key. The case where the result is obtained will be described.
[0086]
FIG. 22 is a diagram schematically showing an outline of a search method according to the fourth embodiment of the present invention. The search key input to the search circuit is divided into a plurality of divided search keys, each of which is simultaneously input to the corresponding table. Here, a case of dividing into two divided search keys will be described as an example. FIG. 23 is a diagram showing the contents of entries stored in each table of FIG. The contents stored in these tables 1a and 1b are the same as those in the example of FIG. 13 described above. The first table 1a stores the upper 64 bits of the address of the search key, and the second table 1b stores the lower 64 bits of the address of the search key. The search key is split into two split search keys, and each of the split search keys is searched simultaneously by the first and second tables 1a and 1b. Then, the tag is output to the search control unit 2 as a search result.
[0087]
Next, the search control unit 2 outputs a combination of tags output from the tables 1a and 1b to the final table 1d as a final search key, and obtains a search result from the final table 1d. As shown in FIG. 23, the final table 1d stores, as entries, a tag combination, a mask length designating the tag length, and the result data for the tag combination. Note that the first and second tables 1a and 1b in FIGS. 22 and 23 correspond to the first table in the claims, and the final table 1d similarly corresponds to the second table.
[0088]
For example, when the packet 1 having the address “4000: 0102: 0304: 0506: 090a: 0b0c: 0d0e” described above with reference to FIG. 23 is input to the search circuit, the first divided search key “64 bits” 4000: 0102: 0304: 0506 "and a second divided search key" 0708: 090a: 0b0c: 0d0e "of lower 64 bits, and the first divided search key is stored in the first table 1a in the second table. The split search key is input to each of the second tables 1b. As a result, "tag a1" is output because it corresponds to entry 1 in the first table 1a, and "tag a2" is output because it corresponds to entry 2 in the second table 1b. Then, the final search key “tag a1 + tag a2” combining these tags is input to the final table 1d. In the final table 1d, since the entry A corresponds to the entry A, the address A is output as “result data”.
[0089]
In the case of the packet 2 having the address “4000: 0102: 0304: 0000: 0708: 0090a: 0b0c: 0d04”, the first divided search key “4000: 0102: 0304: 0000” is stored in the first table. Since it matches the entry B of 1a, the tag b1 is obtained. Since the second divided search key “0708: 090a: 0b0c: 0d04” matches the entry A of the second table 1b, the tag a2 is obtained. When the final table 1d is searched with “tag b1 + tag a2” combining these tags, address B is obtained as “result data” because the result corresponds to entry B (entry B has a mask length of one tag length and a tag length of one tag). Therefore, if the upper tag is tag b1, whatever the lower tag is, it corresponds to entry B).
[0090]
Further, in the case of the packet 3 having the above-mentioned address “4000: 0102: 0304: 0506: 0070: 090a: 0b0c: 0000”, the first divided search key “4000: 0102: 0304: 0506” is stored in the first table. Since it matches the entry A of 1a, the tag a1 is obtained, and the second divided search key “0708: 090a: 0b0c: 0000” has no corresponding entry in the second table 1b, so that a miss hit occurs. In the case of a mishit, an unused tag value (x in this example) is assigned to the entry. Lastly, the final table 1d is searched for “tag a1 + tag x” combining these tags, and as a result, it corresponds to a dummy entry, so that address B is obtained as “result data”.
[0091]
However, in this case as well, simply by searching the tables 1a, 1b, 1d in which the address of the entry is divided and registered by using the divided search key, the search can be performed correctly as in the case of the conventional problem shown in FIG. In some cases, the inclusion relationship between the entries is checked as in the first and second embodiments, and if the inclusion relationship exists, it is necessary to add a dummy entry or set a flag in a necessary table. .
[0092]
In the above description, the case where the search key is divided into two is described as an example. However, even when the search key is divided into three or more, entry addition or deletion processing can be realized by the same processing procedure.
[0093]
According to the fourth embodiment, a tag is set as “result data” for “search target information”, and a tag value obtained by simultaneously searching a divided search key obtained by dividing a search key into a plurality of tables in respective tables. The search is performed using the final table using a combination of the search keys as a search key. Therefore, even a plurality of divided LPM searches can be correctly searched, and the search time can be shortened.
[0094]
The packet transfer device can be configured by installing the search circuit shown in the first, second, and fourth embodiments in the search circuit 22 in the reception-side packet processing unit 13 shown in FIG. In particular, in the case of the fourth embodiment, since the divided search keys obtained by dividing the search key are simultaneously searched, even the long search key can achieve the same throughput as the short search key.
[0095]
【The invention's effect】
As described above, according to the present invention, the inclusion relationship between an entry to be newly added / deleted and the entry stored in the table is obtained. As a result, when the inclusion relationship exists between the two, Since the configuration is such that a dummy entry is added, deleted, or changed in the table, even in an LPM search in which a search key is divided into a plurality of parts, it is possible to avoid that result information for the divided search key is not obtained due to the division of the search key. This has the effect that the search can be performed correctly.
[Brief description of the drawings]
FIG. 1 is a block diagram schematically showing a configuration of a search circuit according to the present invention.
FIG. 2 is a diagram illustrating an example of a configuration of a table.
FIG. 3 is a flowchart illustrating an overall processing procedure when an entry is added;
FIG. 4 is a diagram showing conditions for determining an inclusion relationship.
FIG. 5 is a diagram showing conditions for determining whether a dummy entry is necessary in inclusion relation 1 in FIG. 4;
FIG. 6 is a diagram showing conditions for determining whether a dummy entry is necessary in inclusion relation 2 in FIG. 4;
FIG. 7 is a diagram showing conditions for determining whether or not a dummy entry is necessary in inclusion relation 3 in FIG. 4;
FIG. 8 is a diagram that defines the result of addition determination of a dummy entry and the processing content thereof.
FIG. 9 is a flowchart illustrating a procedure for adding an entry.
FIG. 10 is a flowchart illustrating an overall processing procedure when deleting an entry.
FIG. 11 is a flowchart illustrating a procedure of an entry deletion process.
FIG. 12 is a diagram showing the contents of deletion of a dummy entry.
FIG. 13 is a diagram showing a state of entry addition when there are two tables.
FIG. 14 is a diagram showing how entries are deleted when there are two tables.
FIG. 15 is a diagram showing a configuration of a table according to the present invention.
FIG. 16 is a diagram showing a configuration of a search circuit table according to a second embodiment of the present invention;
FIG. 17 is a flowchart illustrating a search processing procedure according to a second embodiment of the search circuit according to the present invention;
FIG. 18 is a diagram showing how an entry is added when there are two tables.
FIG. 19 is a block diagram showing an overall configuration of a packet transfer device according to the present invention.
FIG. 20 is a block diagram illustrating a configuration of a received packet processing unit.
FIG. 21 is a diagram showing an outline of processing timing in a receiving-side packet processing unit.
FIG. 22 is a diagram schematically showing an outline of a search method according to Embodiment 4 of the present invention.
FIG. 23 is a diagram showing an example of a table according to the fourth embodiment of the present invention.
FIG. 24 is a diagram for explaining a problem of a conventional LPM search.
[Explanation of symbols]
1 table (search engine), 2 search control section, 3 search key division / tag management section, 4 inclusion relation detection / dummy entry management section, 10 interfaces, 11 switches, 12 physical layer processing section, 13 reception side packet processing section, 14 transmission side packet processing unit, 20 packet input unit, 21 header extraction unit, 22 search circuit, 23 header insertion unit, 24 packet output unit.

Claims (9)

The input search key is divided into a plurality of divided search keys, and the longest match search is performed in order using the corresponding table from the upper divided search key including the leading data in the search key, and the desired result information is obtained. Search control means for obtaining
A table provided as many as the number of the divided search keys, and a table for storing entries including predetermined result information corresponding to the divided search keys;
A search circuit comprising:
The search circuit according to claim 1, wherein the table has a dummy entry for preventing the result information for the divided search key from being lost due to the division of the search key. The table, search target information comprising a combination of the divided search key and data obtained from another table or the divided search key corresponding to the search target of the divided search key, the divided search key is the search target information and Includes a mask length indicating a length that needs to be matched, and any result data of a predetermined result information corresponding to the search target information or a tag for searching in combination with a divided search key in another table Store the entry,
The dummy entry is, when the first entry in the table includes the second entry in the table, search target information having a tag of the result data corresponding to the second entry at the head. And a mask length that specifies the length of the tag, and the result information of the first entry, which is lower than a table in which an inclusion relationship occurs between the search target information of the first and second entries. 2. The search circuit according to claim 1, wherein the search circuit is stored in a table. The input search key is divided into a plurality of divided search keys, and the longest match search is performed in order using the corresponding table from the upper divided search key including the leading data in the search key, and the desired result information is obtained. Search control means for obtaining
A table provided as many as the number of the divided search keys, search target information comprising a combination of a tag obtained from the divided search key corresponding data or another table that is a search target of the divided search key and the divided search key, Includes a mask length indicating the length that the divided search key needs to match the search target information, and predetermined result information corresponding to the search target information or result data including a tag for searching in another table A table to store the entries,
When an entry is added to the table, an inclusion relation between the additional entry and an entry stored in the table is checked, and if there is an inclusion relation, the dummy entry is added to the necessary table together with the entry. Relationship detection / dummy entry management means;
A search circuit comprising: The inclusion relation detection / dummy entry management means, when there is an inclusion relation between the additional entry and the entry of the table, the search target information having a tag of the result data corresponding to the included entry at the beginning, The dummy length including the mask length designating the length of the tag and the result information of the included entry is stored in a table lower than the table in which an inclusion relationship occurs between the additional entry and the search target information of the entry of the table. The search circuit according to claim 3, wherein the search circuit is added. The input search key is divided into a plurality of divided search keys, and the longest match search is performed in order using the corresponding table from the upper divided search key including the leading data in the search key, and the desired result information is obtained. Search control means for obtaining
A table provided as many as the number of the divided search keys, search target information comprising a combination of a tag obtained from the divided search key corresponding data or another table that is a search target of the divided search key and the divided search key, A mask length indicating a length that the divided search key needs to match with the search target information, a tag corresponding to the search target information, which is a search result in another table, and a mask length corresponding to the search target information A table storing results information that is a final result and an entry including a data valid flag indicating whether the result information is valid;
With
When there is no entry in the table that matches the split search key, the search control unit finalizes the result information of the entry in which the data valid flag is valid in the table searched before this table. A search circuit characterized in that a search result is obtained as a basic result. In a case where an entry is added to the table, when an inclusion relationship occurs between the additional entry and the entry of the table, the result information of the entry included in the result information of the included entry is stored, and the data valid flag is stored. 6. An inclusive relation detection / dummy entry management means for executing a process for validating the information for a table in which an inclusive relation occurs between the additional entry and the search target information of the entry of the table. Search circuit described in. The search control means is independent from the upper table having the search target information including the leading data of the search target information of the entry before being divided, in order from the lower table having the search target information including the last data. In order to be able to perform a search, a plurality of tables are provided for each table,
7. The search control means connected to an upper table has a function of notifying a search result to a search control means connected to a lower table of the table. The search circuit according to any one of the above. Search control means for dividing the input search key into a plurality of divided search keys, performing a longest match search for each of the divided search keys, and obtaining desired result information;
A first table for storing an entry including a predetermined tag corresponding to the divided search key, the first table being a table provided by the number of the divided search keys;
A second table storing an entry including predetermined result information corresponding to a search key obtained by combining the tags output by the first tables;
A search circuit comprising:
The search circuit, wherein the second table has a dummy entry for preventing the result information for the divided search key from being lost due to the division of the search key. A plurality of communication interfaces for transmitting and receiving packets in a predetermined communication method, and a switch for connecting these communication interfaces and exchanging packets between the communication interfaces, comprising a packet received via the communication interface, In a packet transfer device for searching for and transferring a communication interface of an output destination from the destination IP address,
The communication interface includes:
Receiving-side packet processing means for searching for an output destination of the packet by the search circuit according to any one of claims 1 to 8 based on a destination address of the received packet and sending the packet to the switch,
Sender-side packet processing means for encapsulating and transmitting a packet received from the other communication interface via the switch,
A packet transfer device comprising:

Images