JP5817458B2 - Transfer processing device - Google Patents

Description

  The present invention relates to a transfer processing apparatus and can be applied to, for example, an application processor that performs complex frame processing and control and a fast path processor that bridges and transfers frames at high speed.

  When a frame transfer is performed from a port with a high link speed to a port with a low link speed in a communication apparatus having a transfer processing apparatus that performs transfer processing, etc., it is necessary to shape the data from the port with a high link speed and adjust the speed. There is. As is well known, a buffer is used for shaping (see Patent Document 1). Conventionally, a frame loss is suppressed by using a hardware-mounted buffer (for example, a frame buffer provided in an internal switch).

JP 2008-205932 A

  However, since the buffer memory for adjusting the input / output timing in the internal switch is used as a frame buffer for shaping, the capacity of the buffer is small when viewed from the shaping function. Therefore, the burst tolerance was low. In addition, since the buffer capacity is small, it is difficult to cope with dynamically changing the viewpoint of queuing.

  On the other hand, in order to achieve greater burst tolerance and to enable dynamic queuing, there has been a problem that the hardware configuration around the buffer has to be changed, such as increasing the buffer.

  Therefore, there is a demand for a transfer processing device that can dynamically change the queuing of frames without changing the hardware configuration and can ensure burst tolerance.

  The present invention relates to a transfer processing apparatus for transferring a frame input from an input / output port to a destination input / output port. (1) a first storage unit storing frame information to be transferred at high speed; and (2) high-speed transfer. A frame buffer for buffering frames to be transmitted; (3) a high-speed transfer processing unit that incorporates a plurality of software queues and performs high-speed transfer processing while storing the queued high-speed transfer frames in the frame buffer; ) Second storage unit storing frame information for non-high-speed transfer; and (5) Performing transfer processing for non-high-speed transfer frames and whether non-high-speed transfer frames can be changed to high-speed transfer frames. A bridge processing unit that issues an offload setting request when the change is possible, and (6) a high speed when the offload setting request is issued And having a offload setting processing unit for dynamically allocating a software queue to be applied to the frame to be changed to send the frame.

  According to the present invention, it is possible to realize a transfer processing apparatus that can dynamically change the queuing of frames without changing the hardware configuration and can ensure burst tolerance.

It is a block diagram which shows the structure of the transfer processing apparatus of 1st Embodiment. It is explanatory drawing of the software queue which the high-speed transfer process part in the transfer processing apparatus of 1st Embodiment controls queuing. It is explanatory drawing which shows the example of a setting of the priority with respect to the software queue in the transfer processing apparatus of 1st Embodiment. It is explanatory drawing of the flame | frame made into the transfer object by the high-speed transfer process part and bridge process part in the transfer processing apparatus of 1st Embodiment. FIG. 7 is a sequence diagram illustrating an operation of automatically determining a destination port of a frame and assigning a software queue at the time of frame transfer in the transfer processing device according to the first embodiment. FIG. 6 is a sequence diagram illustrating an operation of automatically determining a frame transmission source port and allocating a software queue when transferring a frame in a direction opposite to a direction in which burst tolerance is desired in the transfer processing device according to the first embodiment. FIG. 6 is a sequence diagram illustrating an operation of automatically setting a shaping band (shaping rate) of a shaper in the transfer processing device according to the first embodiment. FIG. 10 is a sequence diagram illustrating an operation of automatically determining frame priority at the time of frame transfer and assigning a software queue in the transfer processing device according to the second embodiment.

(A) First Embodiment Hereinafter, a first embodiment of a transfer processing apparatus according to the present invention will be described with reference to the drawings.

(A-1) Configuration of First Embodiment FIG. 1 is a block diagram illustrating a configuration of a transfer processing device 1 according to the first embodiment mounted in a communication device. In the following description of the configuration, portions related to embodiments described later are also described without distinction.

  Here, the communication device corresponds to, for example, a wired communication function, and includes a reception unit, a transmission unit, and the like in addition to the transfer processing device 1 of the first embodiment. The communication device is, for example, a VoIP (Voice over IP) gateway or a home gateway.

  The transfer processing apparatus 1 according to the first embodiment is configured, for example, as a whole by a one-chip semiconductor integrated circuit, and is roughly divided into a CPU (Central Processing Unit) 2 and an internal switch (internal SW) 3. Yes. The CPU 2 includes an application processor (hereinafter referred to as AP) 10 that performs complex frame processing and control, and a fast path processor (hereinafter referred to as FP) 20 for bridge-transferring frames at high speed. The internal switch 3 is a layer 2 switch for relaying, for example, the CPU 2 and a LAN (Local Area Network) port / WAN (Wide Area Network) port. The CPU 2 and the internal switch 3 are connected by RGMII (Reduced Gigabit Media Independent Interface) or the like, and LAN and WAN are connected separately.

  On the AP 10, a general-purpose OS (Operating System) having a network stack such as Linux (registered trademark) operates, and, for example, a DHCP (Dynamic Host Configuration Protocol) server / client, a PPPoE (PPP over Ethernet). ) Applications such as a client / server and a VoIP application operate.

  The FP 20 realizes high-speed transfer of a frame that has entered from a wired WAN interface or a wired LAN interface.

  The FP 20 includes a high-speed transfer processing unit 21, a first address storage unit 22, and a first frame buffer 23. Here, the high-speed transfer processing unit 21 has a configuration in the CPU 2, and performs software processing in accordance with a program stored in a program memory in the CPU 2.

  The high-speed transfer processing unit 21 determines whether or not high-speed transfer (high-speed bridge transfer) is set by inspecting a packet header such as a MAC address or an IP address, and when high-speed transfer is set In this method, the header of the frame is converted to transfer the frame at high speed. When a frame for which high-speed transfer is not set is received, the high-speed transfer processing unit 21 transfers the frame to the AP 10. The high-speed transfer processing unit 21 uses information held in the first address storage unit 22 for determining whether high-speed transfer is set or for converting the frame header.

  In addition, the high-speed transfer processing unit 21 determines a software queue to which the transfer frame is queued when transferring the frame, and a software queue number (described later) of the first address storage unit 22 is used. Use the information. It is also possible to shape to a set rate through such a queuing process.

  FIG. 2 is an explanatory diagram of a software queue in which the high-speed transfer processing unit 21 controls queuing. Although the frame data itself is stored in the first frame buffer 23, a software queue that queues only frame specifying information for controlling the first-in first-out of the frame is provided in the high-speed transfer processing unit 21.

  Frames (information) input from the WAN port or LAN port and given to the high-speed transfer processing unit 21 via the internal switch 3 are queued in the software queues SQ0 to SQN corresponding to the target WAN port or LAN port. The frames output from the software queues SQ0 to SQN are given to the shapers SH0 to SHN respectively corresponding to the software queues SQ0 to SQN. The frames output from the shapers SH0 to SHN are supplied to the target WAN port or LAN port via the internal switch 3.

As the software queues SQ0 and SQ1 share the shaper SH0, a plurality of software queues (for example, software queues having different priorities) can share one shaper. Also, each software queue SQ0 to SQN It is possible to set a priority, and it is possible to transfer a frame according to the set priority.

  FIG. 3 is an explanatory diagram showing a setting example of priority for the software queue. The example of FIG. 3 has four LAN ports # 1 to # 4 that can be set to 1 Gbps, 100 Mbps, and 10 Mbps as input / output ports, and one WAN port that is fixed at 1 Gbps and cannot be set. The priorities are set in four stages according to DSCP (Differentiated Services Code Point) values included in the IP header, and five shapers are provided corresponding to five input / output ports. The number of software queues is 20, which is obtained by multiplying the number of input / output ports by the priority level. The first line L1 in FIG. 3 describes the setting for the software queue “0”. The software queue “0” is for queuing the lowest priority frame having the DSCP value “46” among the frames addressed to the WAN port, and there is no corresponding shaper. Further, the eighth line L8 in FIG. 3 describes the setting relating to the software queue “7”. The software queue “7” is for queuing the highest priority frame whose DSCP value is a value other than “46”, “32”, and “8” among the frames addressed to the LAN port # 1. The corresponding shaper is the shaper “1”, and this shaper “1” can be set to 0, 100 Mbps, and 10 Mbps as the shaping band.

  As shown in the fifth line L5 to L8 in FIG. 3, the numbers of software queues for queuing frames addressed to the same LAN port # 1 are “4”, “5”, “6”, and “7”. It is a serial number, and the four priority levels of “lowest priority”, “low priority”, “high priority”, and “highest priority” are “0”, “1”, “2”, “3”. When digitized, software queue numbers “5”, “6”, and “7” that queue “low priority”, “high priority”, and “highest priority” frames queue “lowest priority” frames. The priority numbers “1”, “2”, and “3” are added to the software queue number “4” to be processed. In the second embodiment to be described later, such software queue number assignment is used.

  Settings as shown in FIG. 3 are made in advance for the high-speed transfer processing unit 21, and the high-speed transfer processing unit 21 controls the queuing process and the shaping process according to the setting contents.

  The high-speed transfer processing unit 21 determines whether or not high-speed transfer is set by inspecting a packet header such as a MAC address or an IP address. The frame header is converted to transfer the frame at high speed. When a frame for which high-speed transfer is not set is received, the high-speed transfer processing unit 21 transfers the frame to the AP 10.

  The first address storage unit 22 holds information on frames that the FP 20 should perform high-speed transfer processing. The entry as a storage unit of the first address storage unit 22 includes: [1] destination MAC address, [2] source MAC address, [3] destination IP address, [4] source IP address, [5] protocol, [6] Destination port, [7] Source port, [8] Ether type, [9] Input interface, [10] Output interface, [11] Software queue number, [12] Hash value.

  The protocol defines TCP, UDP and the like. The ether type is for determining whether it is IPv4 or IPv6, and this item may or may not be present. The input interface is used to determine from which interface the input has been received. The output interface is used to determine which interface to transfer to. The software queue number designates to which software queue number a queue that matches the items [1] to [10] described above is queued. The hash value is a value for improving the search speed with respect to the table adopted in the first address storage unit 22, and if the entry in the table can be uniquely determined, the above-described [1] to [11]. The values of all or some of the items may be used for calculating the hash value. The hash value may or may not be present.

  An example of entry setting will be described below. However, in the following, only the setting contents for item 3] destination IP address, [4] source IP address, [8] ether type, [9] input interface, and [10] output interface will be mentioned.

  For example, it is assumed that a wired LAN terminal A and a wired WAN terminal B are connected to a communication apparatus having the transfer processing apparatus 1 and transmit / receive an IPv4 frame. In this case, two bidirectional entries are described in the first address storage unit 22 as described below.

  [Entry of terminal B → terminal A] describes destination IP address = terminal A, transmission source IP address = terminal B, ether type = IPv4, input interface = WAN, and output interface = LAN.

  [Entry of terminal A → terminal B] describes destination IP address = terminal B, transmission source IP address = terminal A, ether type = IPv4, input interface = LAN, and output interface = WAN.

  Further, for example, it is assumed that the wired LAN terminal A and the wired WAN terminal B are connected to a communication apparatus having the transfer processing apparatus 1 and transmit / receive an IPv6 frame. In this case, a total of four entries, each having two entries with different ether types, are described in the first address storage unit 22 in both directions as described below.

  [First entry of terminal B → terminal A] describes destination IP address = terminal A, source IP address = terminal B, ether type = IPv4, input interface = WAN, and output interface = LAN.

  [First entry of terminal A → terminal B] describes destination IP address = terminal B, transmission source IP address = terminal A, ether type = IPv4, input interface = LAN, and output interface = WAN.

  [Second entry of terminal B → terminal A] describes destination IP address = terminal A, source IP address = terminal B, ether type = IPv6, input interface = WAN, and output interface = LAN.

  [Second entry of terminal A → terminal B] describes destination IP address = terminal B, transmission source IP address = terminal A, ether type = IPv6, input interface = LAN, and output interface = WAN.

  The first frame buffer 23 is a buffer that temporarily stores frame data when the high-speed transfer processing unit 21 transfers a frame. The frames queued in the software queue are held in the first frame buffer 23. As described above, a plurality of software queues can be provided for each frame priority and each frame transfer direction. As the first frame buffer 23, a DRAM (Dynamic Random Access Memory) can be applied, and a dynamic area can be secured, so that the size of each software queue can be configured.

  The AP 10 includes a bridge processing unit 11, a second address storage unit 12, and an offload setting processing unit 13. Here, the bridge processing unit 11 and the offload setting processing unit 13 are configured in the CPU 2 and are processed in software according to a program stored in a program memory in the CPU 2.

  The bridge processing unit 11 corresponds to, for example, a bridge processing unit in the network stack of the Linux kernel, and transfers frames. The bridge processing unit 11 searches the built-in MAC table list, determines the output destination interface by comparing with the destination MAC of the frame, and performs bridge transfer. Further, the bridge processing unit 11 has a function of monitoring each aging timer of an entry held in the second address storage unit 12 and deleting the entry when a predetermined time has elapsed.

  FIG. 4 is an explanatory diagram of frames to be transferred by the high-speed transfer processing unit 21 and the bridge processing unit 11.

  A frame is input from one of the input / output ports, transferred at the high-speed transfer processing unit 21 and output to another input / output port, and input from one of the input / output ports to perform high-speed transfer processing. There is an exception frame that is given to the bridge processing unit 11 via the unit 21, transferred by the bridge processing unit 11, and output to another input / output port via the high-speed transfer processing unit 21.

  The exception frame includes (1) a frame originated from the own device or a frame addressed to the own device, and (2) a packet of a type not compatible with high-speed transfer of the FP 20 (an ICMP (Internet Control) which is a broadcast packet, a multicast packet, or a control packet (Message Protocol) packet), (3) a frame before high-speed transfer setting (offload setting) described later is set, and the like. Since this exceptional frame is transferred on the AP 10 that assumes that various applications such as an HTTP (Hyper Text Transfer Protocol) server also operate, there is only a relatively low transfer capability, and the frame transfer process itself is the same. May cause processing delay of the application, or may cause deterioration in frame transfer performance / transfer delay.

  On the other hand, the high-speed transfer frame can realize high-speed transfer by the FP 20 with a minimum AP load. Therefore, as described later, high-speed transfer setting (offload setting) is enabled, and a part of the exception frame can be converted into a high-speed transfer frame.

  The second address storage unit 12 is a table that holds each stream information that the network stack of the AP 10 has. The entries in the table are: [1] destination MAC address, [2] source MAC address, [3] destination IP address, [4] source IP address, [5] protocol, [6] destination port, [7] transmission It has items such as original port, [8] Ether type, [9] input interface, [10] output interface, [11] software queue number, [12] hash value.

  The offload setting processing unit 13 receives a request from the bridge processing unit 11 and determines whether or not a frame that is about to be bridge-transferred can be set to offload in the FP 20. That is, the offload setting processing unit 13 determines whether or not the processing of the bridge processing unit 11 can be reduced by executing the transfer processing to be executed by the bridge processing unit 11 on the FP 20 side. For example, when the transfer frames are the following frames F1 and F2, the offload setting processing unit 13 determines that offload is not possible.

F1: Frame originating from own device or frame destined for own device F2: Type of packet not compatible with high-speed transfer in FP20 (broadcast packet, multicast packet, ICMP packet, etc.)
When the offload setting processing unit 13 determines that the offload setting is possible, the offload setting processing unit 13 performs the offload setting in the FP 20 (more precisely, the first address storage unit 22). The offload setting processing unit 13 does not execute any processing when it is determined that the offload setting is impossible.

  The offload setting processing unit 13 also holds whether or not each entry in the second address storage unit 12 has already been offloaded to the FP 20 in order to avoid duplicate setting of offload.

  In addition, the offload setting processing unit 13 performs the following determination processes (a) and (b), and performs saving to the second address storage unit 12 and offload setting according to the determination result.

  (A) It is determined to which port the transfer frame should be transferred. For example, the determination is made by comparing the information in the MAC address storage unit 32 of the internal switch 3 with the destination MAC address of the transfer frame.

  (B) The priority level of the transfer frame is determined. For example, it is determined by collating frame header information (CoS (Class of Service) value, DSCP value, IP address, MAC address, port number, etc.).

  Further, the offload setting processing unit 13 has a function of detecting the link speed by receiving an interrupt when each LAN port / WAN port is linked up and notifying the high-speed transfer processing unit 21 of the link speed. By doing so, the high-speed transfer processing unit 21 of the FP 20 can shape the frame addressed to each port according to the link speed. For example, the high-speed transfer processing unit 21 includes a conversion table in which a link speed range and a shaping band are associated with each other, and can set a shaping band corresponding to the link speed.

  The internal switch 3 has a second frame buffer 31 and a MAC address storage unit 32.

  The second frame buffer 31 is a buffer that temporarily stores frame data when the internal switch 3 transfers a frame. Compared with the first frame buffer 23 of the FP 20, the capacity is small, and the buffer size cannot be configured. Therefore, in the first embodiment, the first frame buffer 23 of the FP 20 is used for queuing transfer frames.

  The MAC address storage unit 32 is a table that holds the following neighbor information included in the bridge processing unit 11 in the AP 10. Neighbor information includes, for example, [1] MAC address (source MAC address of received frame), [2] LAN port / WAN port number (LAN port / WAN port number that received the frame; terminal having a certain MAC address 3) Aging timer (a timer value that is incremented every certain period; if this value exceeds a preset threshold after a certain period of time, this Entry is aged (deleted)), [4] static flag (flag indicating whether or not the entry is statically set by the user or the like; the statically set entry is flag ON and excluded from aging) , [5] Hash value (provided to improve table search speed, and uniquely determines an entry in the table If it is possible, may be using the hash value calculation some of the value of the item) and the like. Note that there is no need for static flags or hash values.

(A-2) Operation of the first embodiment The transfer processing device 1 of the first embodiment applies a software queue, and the data of the queued frame itself is the first frame of the FP 20. One feature is that the data is stored in the buffer 23.

  In addition, it is characterized by how to assign the transfer frame to which software queue, and the shaping rate setting is also characteristic. In the following, the assignment operation to the software queue and the setting operation of the shaping rate will be sequentially described.

  First, the operation of automatically determining the destination port of a frame at the time of frame transfer and automatically allocating it to each software queue will be described with reference to the sequence diagram of FIG. In FIG. 3 described above, the priority is also described so as to affect the operation of assigning. However, the assignment using the priority will be described in the operation section of the second embodiment to be described later.

  When the bridge processing unit 11 detects transfer of a frame (step 100), it sends an offload setting request including header information of the frame to the offload setting processing unit 13 (step 101).

  At this time, the offload setting processing unit 13 obtains information of the MAC address storage unit 32 after confirming that the frame to be bridge-transferred can be set offload (step 102), and the transfer frame The destination port number is obtained by checking the destination MAC address of the destination, and as a result, the number of the software queue to be queued is obtained (step 103).

Thereafter, the offload setting processing unit 13 stores the entry data in the second address storage unit 12 in accordance with the acquired contents described above (step 104) and sets the entry as a high-speed transfer target in the high-speed transfer processing unit 21. (Offload setting) is performed (step 105).
Accordingly, the high-speed transfer processing unit 21 adds an entry to the first address storage unit 22 with the frame as a high-speed transfer target (step 106).

  As described above, it is possible to automatically determine the destination port of the frame at the time of frame transfer and assign it to be automatically queued in each software queue, and to ensure burst tolerance.

  Next, an allocation operation that can cope with a case where a frame suddenly flows in a burst state without setting the software queue will be described. In the above-described case, since the high-speed transfer processing unit 21 of the FP 20 has not been set, there is a possibility that a burst frame is lost in the bridge processing unit 11 of the AP 10 that can perform only low-speed traffic transfer. . For this reason, a software queue can be set in advance when transferring a frame in the direction opposite to that of a frame that requires burst tolerance, so that it can cope with a case where a frame flows suddenly and in bursts. Here, a frame that requires burst tolerance is a frame that flows from a port with a high link speed to a port with a low link speed.

  In the following, the operation of automatically determining the frame source port and assigning it to be queued in each software queue at the time of frame transfer in the direction opposite to the direction in which burst tolerance is desired is performed as shown in the sequence of FIG. This will be described with reference to the drawings.

  When the bridge processing unit 11 detects the transfer of the frame in the direction opposite to the direction in which the burst tolerance is desired (step 150), the bridge processing unit 11 sends an offload setting request in the reverse direction including the header information of the frame to the offload setting processing unit 13. It is sent out (step 151).

  At this time, the offload setting processing unit 13 obtains information of the MAC address storage unit 32 after confirming that the frame to be bridge-transferred can be offloaded (step 152), and the transfer frame. The destination MAC number in the reverse direction is obtained by collating the source MAC address of, and as a result, the number of the software queue to be queued is obtained (step 153).

Thereafter, the offload setting processing unit 13 stores the entry data in the reverse direction in the second address storage unit 12 in accordance with the acquisition contents described above (step 154) and the high-speed transfer processing unit 21 in the reverse direction at high speed. Entry setting (offload setting) as a transfer target is performed (step 155). Accordingly, the high-speed transfer processing unit 21 adds an entry to the first address storage unit 22 with a frame flowing in the opposite direction to the frame as a target for high-speed transfer (step 156).

  Next, an operation of automatically setting a shaping band (shaping rate) of a related shaper when a link up of each LAN port / WAN port is detected will be described with reference to the sequence diagram of FIG.

  When the internal switch 3 detects a link up of a certain input / output port (step 200), it sends a link up interrupt notification to the offload setting processing unit 13 (step 201). At this time, the offload setting processing unit 13 recognizes the link up of the input / output port and reads the register value related to the communication speed of the internal switch 3 (step 202), thereby acquiring the link speed of the input / output port. (Step 203). Then, the offload setting processing unit 13 determines a shaping rate to be set for the shaper of the high-speed transfer processing unit 21 of the FP 20 from the acquired link speed (step 204), and causes the shaper of the high-speed transfer processing unit 21 to set it. Thereby, high-speed frame transfer according to the link speed can be performed.

(A-3) Effects of the First Embodiment As described above, according to the first embodiment, the buffer size that can only be dealt with by changing the hardware configuration in the past can be obtained using the software queue. By securing dynamically, it is possible to have greater burst tolerance without hardware change.

  Further, according to the first embodiment, even when a frame suddenly flows in a burst state when the software queue is not set, a frame in a direction opposite to a frame that requires burst tolerance is used. Since the software queue is set in advance at the time of transfer, burst-like frames can be transferred without frame loss.

  Furthermore, according to the first embodiment, even when the link speed is dynamically changed by auto-negotiation, the change in the link speed is dynamically reflected on the shaping rate of the shaper connected to each software queue. Thus, burst frame transfer without frame loss is possible.

(B) Second Embodiment Next, a second embodiment of the transfer processing apparatus according to the present invention will be described with reference to the drawings.

  The configuration of the transfer processing apparatus of the second embodiment can also be represented by FIG. 1 used in the description of the first embodiment.

  The transfer processing device of the second embodiment automatically determines the priority of a frame at the time of frame transfer, and assigns it to be automatically queued to each software queue, thereby ensuring burst tolerance and free. Thus, priority control can be performed.

  In the following, an operation for automatically determining the priority of a frame at the time of frame transfer and automatically allocating it to each software queue will be described with reference to the sequence diagram of FIG.

  When the bridge processing unit 11 detects transfer of a frame (step 250), it sends an offload setting request including header information of the frame to the offload setting processing unit 13 (step 251).

  At this time, the offload setting processing unit 13 obtains information of the MAC address storage unit 32 after confirming that the frame to be bridge-transferred can be offloaded (step 252), and the transfer frame The destination port number is obtained by checking the destination MAC address of the destination, and as a result, the software queue number is obtained (step 253).

  After that, the offload setting processing unit 13 calculates the priority from the DSCP value of the frame and adds it as an offset to the software queue number acquired in step 253 to obtain the software queue number to be finally used ( Step 254). For example, if the software queue number obtained in step 253 is “4” and the priority is “3”, “7” (= 4 + 3) is finally obtained as the software queue number to be used.

Then, the offload setting processing unit 13 stores the entry data in the second address storage unit 12 in accordance with the contents acquired as described above (step 255), and the high-speed transfer processing unit 21 performs high-speed transfer. Entry setting (offload setting) is performed (step 256). Accordingly, the high-speed transfer processing unit 21 adds an entry to the first address storage unit 22 with the frame as a high-speed transfer target (step 257).

  As described above, according to the second embodiment, the burst tolerance and priority control function that has been conventionally performed by hardware (internal switch) can be provided on software.

(C) Other Embodiments In the description of each of the above-described embodiments, various modified embodiments have been referred to. However, modified embodiments as exemplified below can be given.

  In each of the above embodiments, a DRAM is applied to the first frame buffer 23 of the FP 20, but any other storage medium such as SRAM (Static RAM) may be used as long as the storage medium can hold a frame. good.

  In each of the above embodiments, the “internal switch (internal SW)” is arranged on the apparatus and all LAN ports are connected via the interface of one CPU. Each LAN port / WAN port may be connected to the CPU via an independent interface as long as it can be determined whether the message is received at the WAN port.

  In each of the embodiments described above, the case where there is one “AP” is shown, but there may be two or more, and similarly, “FP” is not limited to one and may be two or more.

  In the second embodiment, the software queue to be applied is determined by the combination of the frame destination port and the priority. However, the software queue to be applied may be determined only by the frame priority.

  In the second embodiment, the number of software queues to be applied is limited to several according to the destination port of the frame, and finally one software queue to be applied is determined according to the priority of the frame. However, conversely, after limiting the number of software queues to be applied to several according to the priority of the frame, one software queue to be finally applied is determined according to the destination port of the frame. Anyway.

  DESCRIPTION OF SYMBOLS 1 ... Transfer processing apparatus, 2 ... CPU, 3 ... Internal switch (internal SW), 10 ... Application processor (AP), 11 ... Bridge processing part, 12 ... 2nd address storage part, 13 ... Offload setting processing part , 20 ... fast path processor (FP), 21 ... high-speed transfer processing unit, 22 ... first address storage unit, 23 ... first frame buffer, 31 ... second frame buffer, 32 ... MAC address storage unit.

Claims (5)

In a transfer processing device that transfers a frame input from an input / output port to a destination input / output port,
A first storage unit storing frame information for high-speed transfer;
A frame buffer for buffering high-speed transfer frames;
A high-speed transfer processing unit that incorporates a plurality of software queues and performs high-speed transfer processing while storing the queued high-speed transfer frames in the frame buffer;
A second storage unit storing frame information for non-high-speed transfer;
Bridge processing unit that performs transfer processing for non-high-speed transfer frames, determines whether non-high-speed transfer frames can be changed to high-speed transfer frames, and issues an offload setting request if change is possible When,
An offload setting processing unit that dynamically assigns a software queue to be applied to a frame to be changed to a high-speed transfer frame when an offload setting request is issued.   The transfer processing apparatus according to claim 1, wherein the offload setting processing unit assigns a software queue to be applied according to an input / output port of a destination of a frame to be changed to a high-speed transfer frame.   The transfer processing apparatus according to claim 1, wherein the offload setting processing unit assigns a software queue to be applied according to a priority of a frame to be changed to a high-speed transfer frame. The offload setting processing unit automatically determines a transmission source port of a frame when transferring a frame in a direction opposite to a frame transferred in a direction that requires burst tolerance, according to a request from the bridge processing unit , 4. The transfer processing device according to claim 1, wherein the transfer processing device is assigned so as to be automatically queued to each software queue . The high-speed transfer processing unit has a built-in shaper corresponding to each built-in software queue,
The offload setting processing unit detects a link speed of an input / output port, and sets a shaping rate of a shaper corresponding to a software queue related to the input / output port according to the detected link speed. The transfer processing device according to any one of 1 to 4.

Images