JP4772648B2 - Data receiving method, data receiving apparatus and computer system - Google Patents

Description

  The present invention relates to a data receiving method executed in a system constituted by a data receiving device and a computer system, and a data receiving device and a computer system used for realizing the data receiving method, and particularly to expensive hardware. A data receiving method for realizing that a computer system can receive data at high speed from outside without preparing or rewriting a program, and a data receiving apparatus and a computer system used for realizing the data receiving method, About.

  In today's computer systems, high-speed data exchange with the outside is an important issue that leads to high speed of the entire system. Especially in a cluster system where small computer systems such as personal computers are connected via a network and used as one large computer system as a whole, data transmission via the network is an important factor that determines the calculation speed of the entire system. ing.

  For this reason, the data communication performance of the computer system has been improving year by year, and the communication speed has increased at a double rate in about three years here. This is faster than the speed of computer computing devices.

  As the data communication speed increases, the handling speed of received data, which has hardly been a problem in the past, becomes a very important issue.

  Conventionally, the performance of a computing device is very high compared to the speed at which data arrives from the network, so even if complex processing is performed on received data, there is almost no overhead.

  At present, however, the relationship is broken, and there are frequent situations where the computer system is full of processing received data.

  This not only reduces the speed of data reception, but also subtracts the secondary performance degradation that would otherwise prevent processing that would have been possible on this computer system. It is an important problem that will greatly reduce it.

  In order to deal with such problems, new technologies such as Protocol Offloading and Remote DMA have been developed.

  Protocol Offloading is a technique in which a part of data reception processing that is conventionally performed by an operating system (OS) is mounted on hardware and mounted on a network interface card (NIC). As a result, the computer system body does not need to perform the processing that has been performed when data is received from the outside, and the burden is reduced.

However, in this Protocol Offloading,
1. 1. Hardware that can perform high-speed and complex processing that can handle the processing that the computer system has performed must be prepared. Since software is evolving year by year and OS processing is also changing year by year, there is a problem that the hardware cannot follow the change.

  On the other hand, Remote DMA provides a means to solve these problems. Remote DMA greatly changes the communication method, and allows data that could not be received without complicated processing to be exchanged directly by an operating system.

  However, in order to use Remote DMA, it is necessary to prepare hardware corresponding to Remote DMA and rewrite the program for Remote DMA. For this reason, in general, it is not so popular.

Here, Patent Document 1 shown below proposes a DMA controller that performs high-speed DMA transfer across page boundaries between pages that are consecutive on a physical address as an invention related to the present invention. .
Japanese Patent Laid-Open No. 11-184797

  Thus, in order to solve the problems of the prior art, new technologies such as Protocol Offloading and Remote DMA have been developed.

  However, Protocol Offloading has a problem that expensive hardware must be prepared. In addition, Remote DMA has a problem that it is necessary to prepare hardware corresponding to Remote DMA, and there is a problem that it is necessary to rewrite a program for Remote DMA.

  From now on, it is necessary to construct a new technology that enables a computer system to receive data at high speed without preparing expensive hardware or rewriting a program.

On the other hand, in data communication involving an OS,
Step 1: NIC packet arrives from outside Step 2: The NIC forwards the packet to the area designated in advance by the OS and notifies the OS that it has been forwarded. 3. The OS analyzes the transferred packet information and determines the final destination. Data reception processing is performed by copying the data stored in the packet to the data area where the final destination is used.

  Of these steps, the process of step 2 is an indispensable operation for receiving data from the outside. Therefore, if there is room for contrivance, the process of step 3 or step 4 is performed. Step 4 is considered to be particularly important.

  The data copy operation executed in step 4 is one of very heavy processes in the current processor. This is because no matter how fast the processor is accelerated, the speed of the memory is not so high, and operations involving memory access are very time consuming. Also, since the memory does not have an infinite speed, there is an upper limit on the bandwidth for memory access. Since the data copy operation is an operation of once reading data from the memory and writing it to another location, the load on the memory is high and the memory bandwidth is wasted. By this operation, processing for accessing other memories may be delayed.

  Also, the number of data copies increases as the data increases. Therefore, as the data communication speed is improved, the data copy process becomes heavier accordingly. The packet information analysis processing performed in step 3 can be speeded up by increasing the packet size or placing header information in a cache. However, in the process of step 4 for handling huge data, it is difficult to make such a device.

  In order to speed up the operation in step 4 as described above, it is conceivable to use a technique called page map switching.

  A page map is a table for address conversion used for realizing a virtual storage space. The page map describes a set of logical addresses and physical addresses. When there is a request for referring to the data of the logical address A from the program, the processor refers to the page map and extracts the physical address A ′ corresponding to the logical address A. Then, the data at the physical address A 'of the memory is read out and transferred to the program as data corresponding to the logical address A.

  By using this mechanism, data movement can be realized without copying. For example, it is assumed that the physical address B ′ corresponds to the logical address B, and the physical address C ′ corresponds to the logical address C. When data is moved from the logical address C to the logical address B, the page map is rewritten to change the correspondence between the logical address and the physical address from “B → C ′, C → B ′”. In this way, when the data at the logical address B is referred to, the data stored at the logical address C can be referred to so far.

  The page map is not prepared for each address, but is managed for each unit of data called a page. For example, in the widely used Intel IA32 processor, the page size is 4 KB.

  Since data movement using page map switching must be performed in units of pages, the available situations are limited.

  For example, when data of 4 KB (= 0x1000B) starting from the logical address 0x10000000 is moved to an area of 4 KB from the logical address 0x20000000, the page map corresponding to the logical address 0x20000000 is moved to the current logical address 0x10000000. What is necessary is just to rewrite so that it may associate with the corresponding physical address.

  However, for example, if the destination is the logical address 0x20000100, the destination area will span the page starting from the logical address 0x20000000 and the page starting from the logical address 0x20001000. I can't. In this case, the data must be copied.

  In other words, if the page offset of the data (in this example, the transfer source is 0x000 and the transfer destination is 0x100) does not match, the data movement by rewriting the page map cannot be used.

  From now on, in the data communication involving the OS that executes the processing of Step 1 to Step 4 as described above, it is not possible to simply use data movement by rewriting the page map in Step 4.

  Therefore, in the present invention, in the data communication involving the OS that executes the processing of step 1 to step 4 as described above, the memory actually used by the application program from the memory area in which the OS temporarily prepares the received data. When transferring data to an area, by creating many situations where data movement by rewriting the page map can be used, high-speed data transfer can be realized, thereby preparing expensive hardware and rewriting programs It is an object of the present invention to provide a new data receiving technique that enables a computer system to receive data from outside at high speed.

  In order to achieve this object, according to the present invention, a data receiving device that is attached to a computer system, receives data sent from the outside, and temporarily stores it in a memory in the computer system; A data receiving device configured with a computer system that receives data transferred from a receiving device, temporarily stores the data in a memory, and then transfers the data to an application memory area managed by an application program The computer system has the following configuration.

[1] Configuration of Data Receiving Device of the Present Invention The data receiving device of the present invention comprises (a) a predicting means for predicting a page offset of an application memory area when receiving data sent from the outside; (B) a transfer means for transferring the received data to the memory by designating a data storage start address specified by the top address of the data transfer destination page on the memory and the page offset predicted by the prediction means; Setting means for setting a predicted value of the page offset used in the next data reception based on the length of the data and the predicted page offset of the prediction means, and (d) a page offset predicted by the prediction means from the computer system When there is a notification of a prediction error for the predicted value, the predicted value of the page offset set by the setting means is compiled. And changing means for changing according to the page offset notified from the computer system.

  When adopting this configuration, a first storage unit that stores the predicted value of the page offset set by the setting unit in association with the transmission source information and the transmission destination information may be provided. In the case of providing, when receiving the data, the predicting means extracts the transmission source information and the transmission destination information of the received data, and uses the extracted transmission source information and the transmission destination information as a key for the first storage. Processing is performed so as to predict the page offset by referring to the storage information of the means.

  At this time, there may be further provided initial registration means for initially registering the predicted value of the page offset in the first storage means in accordance with the initial value of the page offset notified from the computer system.

  In addition, when adopting this configuration, whether or not to use a fixed page offset in association with the transmission source information and the transmission destination information, and a fixed page offset for those using a fixed page offset are stored. For those that do not use a fixed page offset, there may be provided a second storage means for storing the predicted value of the page offset set by the setting means, and if this second storage means is provided, the prediction means When data is received, the source information and destination information of the received data are extracted, and the stored information in the second storage means is referenced using the extracted source information and destination information as a key. To process the page offset.

  At this time, it may further comprise registration means for registering the fixed page offset value in the second storage means in accordance with the fixed page offset value notified from the computer system. There may be provided initial registration means for initially registering the predicted value of the page offset in the second storage means in accordance with the notified initial value of the page offset.

  The reason why the second storage means is provided is as follows.

  That is, when there is a loop process or the like, the computer system may receive small data many times at the same location in the application memory area.

  In such a case, if the first storage means is used, the predicted value of the same page offset is written to the first storage means every time even though the predicted page offset is the same. The process will be repeated. On the other hand, if the second storage means is used, it is not necessary to perform such inconvenient processing.

[2] Configuration of the computer system of the present invention The computer system of the present invention specifies (a) the page offset predicted by the data receiving device based on the data storage start address specified at the time of data transfer from the data receiving device. (B) second specifying means for specifying the page offset of the application memory area to which the data is transferred by analyzing the data stored in the memory; Determining means for determining whether or not the page offset specified by the first specifying means and the page offset specified by the second specifying means match; and (d) when the determining means determines a mismatch between the two page offsets. In addition, based on the page offset specified by the second specifying means and the length of the data stored in the memory, The first notification means for specifying the page offset of the application memory area and notifying the data receiving device, and (e) determining the data transfer method from the memory to the application memory area based on the determination result of the determination means And determining means for performing the above-described operation.

  When adopting this configuration, the data receiving apparatus is provided with a second notification means for notifying the initial value of the predicted value of the page offset, or the fixed value of the page offset is notified to the data receiving apparatus. Third notification means may be provided.

[3] Processing of the present invention When the data receiving apparatus of the present invention receives data sent from the outside, the data receiving apparatus predicts the page offset of the application memory area as the final data transfer destination, and stores it on the memory. Specify the data storage start address specified by the start address of the data transfer destination page and its predicted page offset, transfer the received data to the memory, and set the length of the received data and its predicted page offset. Based on this, a predicted value of the page offset to be used at the next data reception is set.

  Upon receiving this data transfer, the computer system of the present invention receives the transferred data, and stores the received data in the memory starting from the data storage start address designated at the time of the data transfer.

  Subsequently, based on the data storage start address, the page offset predicted by the data receiving apparatus of the present invention is specified, and the data stored in the memory is analyzed, so that the application memory area to which the data is transferred The page offset is specified, and it is determined whether or not the two specified page offsets match.

  Subsequently, a data transfer method from the memory to the application memory area is determined based on the determination result.

  Specifically, when determining the mismatch between two page offsets, it is determined to perform data transfer according to the data copy operation, and when determining whether the two page offsets match, the page map switchable page data is determined. Data transfer is performed by the page map switching operation, and it is determined that data transfer is performed according to the data copy operation for data of other pages.

  Subsequently, when determining the mismatch between the two page offsets, the page offset of the application memory area after the data transfer is specified based on the page offset of the specified application memory area and the length of the data stored in the memory. This is notified to the data receiving apparatus of the present invention.

  Upon receiving the notification of the page offset, the data receiving apparatus according to the present invention sets the predicted value of the set page offset (the predicted value of the page offset used at the next data reception) because the predicted page offset has an error. Is also determined to be an error, and the predicted value of the set page offset is changed according to the page offset notified from the computer system of the present invention.

[4] Specific Processing Example of the Present Invention Next, assuming that the transferred data is a packet, the processing of the present invention will be specifically described.

  For example, when the length of one page is 0x1000 and the length of a packet sent from the outside is 0x2800 (2.5 pages), the data reception device of the present invention predicts the application memory area first. Assume that the predicted value of page offset is 0 pages.

From this, the data receiving apparatus of the present invention is not notified of a page offset prediction error from the computer system of the present invention,
0 page → 0.5 page → 0 page → 0.5 page → 0 page →
Thus, the page offset of the application memory area is predicted.

  (I) In this case, it is further assumed that the actual page offset of the first page of the application memory area is 0 page. At this time, when packets are transferred in order without loss, As shown in FIG. 1, the page map switching operation is used from the beginning when the page offset of the application memory area predicted by the data receiving apparatus of the present invention matches the actual page offset of the application memory area. As a result, data can be moved from the memory to the application memory area at high speed.

  (Ii) On the other hand, in this case, it is further assumed that the actual page offset of the first page of the application memory area is 0.5 page. At this time, packets are transferred in order without loss. In this case, as shown in FIG. 2, the page offset of the application memory area predicted by the data receiving apparatus of the present invention does not match the actual page offset of the application memory area only at the beginning. After that, the page offset switching operation is used when the page offset of the application memory area predicted by the data receiving apparatus of the present invention matches the actual page offset of the application memory area. Application memory from the memory It becomes possible to move data at high speed from re region.

  (Iii) On the other hand, in this case, it is further assumed that the actual page offset of the first page of the application memory area is 0 page. At this time, if packets are lost or not transferred in order As shown in FIG. 3, at that time, the page offset of the memory predicted by the data receiving apparatus of the present invention and the actual page offset of the application memory area are inconsistent, thereby correcting the prediction of the page offset. After that, the page offset of the memory predicted by the data receiving apparatus of the present invention and the actual page offset of the memory area for application coincide with each other. Data can be moved to the memory area at high speed.

  Here, in FIG. 3, it is assumed that the packet # n + 2 has been transferred before the packet # n + 1, and the computer system of the present invention analyzes this by analyzing the header information of the packet. Will be detected.

  In FIG. 1, it is assumed that the packet length is 2.5 pages, but as shown in FIG. 4, even if the packet length is halfway, such as 2.6 pages, from the beginning. The page offset of the application memory area predicted by the data receiving apparatus of the present invention matches the actual page offset of the application memory area, and thereafter the application memory area predicted by the data receiving apparatus of the present invention Since the page offset matches the actual page offset of the application memory area, data can be moved from the memory to the application memory area at high speed by using the page map switching operation.

  In the example shown in FIG. 2, even though the actual page offset of the first page of the application memory area is 0.5 page, the data receiving apparatus of the present invention first calculates the predicted value of the page offset. The computer system of the present invention notifies the initial value of the predicted value of the page offset to the data receiving apparatus of the present invention, and in response to this, the data receiving apparatus of the present invention Such inconvenience does not occur if the first page offset is predicted according to the notified initial value of the predicted page offset value.

  According to the present invention, a data receiving device that is provided in association with a computer system, receives data sent from the outside, and temporarily stores it in a memory in the computer system, and is transferred from the data receiving device. In a system comprised of a computer system that receives incoming data, temporarily stores it in memory, and then transfers it to an application memory area managed by an application program, the data receiving device rewrites the page map. It is possible to predict the page offset that enables data movement by, and by this, it is possible to create many situations that can use data movement by rewriting the page map, so that high-speed data by rewriting the page map Transfer can be realized.

  Thus, according to the present invention, the computer system can receive data from the outside at high speed without preparing expensive hardware or rewriting the program.

  Hereinafter, the present invention will be described in detail according to embodiments.

  FIG. 5 illustrates an example of a system configuration including the network interface device 1 of the present invention and the data processing device 2 of the present invention. The network interface device 1 may be abbreviated as NIC in the drawings.

  As shown in this figure, the network interface device 1 according to the present invention receives a TCP / IP packet transmitted from Ethernet (not shown) and temporarily stores it, and a received data storage buffer 10 and received data. The DMA engine 11 that transfers the packet stored in the storage buffer 10 to the OS buffer 200 of the data processing device 2, the packet stored in the reception data storage buffer 10 is analyzed, the packet length, the IP address of the transmission source, A key extractor 12 that extracts the port number of the transmission destination, and a buffer for APL of the data processing device 2 by searching the table using the transmission source IP address and the transmission destination port number extracted by the key extractor 12 as keys. Set by the offset predictor 13 that predicts the page offset of 210 and the data processing device 2 The receiving buffer register 14 that receives and registers the page head address of the OS buffer 200, and the OS buffer by adding the page head address registered by the receiving buffer register 14 and the page head address predicted by the offset predictor 13 And an adder 15 that calculates 200 data storage start addresses and gives them to the DMA engine 11.

  On the other hand, the data processing apparatus 2 of the present invention includes an operating system 20 and an application program 21 that includes an APL buffer 210 and executes predetermined data processing.

  The operating system 20 receives a packet transferred from the network interface device 1 and temporarily stores it, and a data receiving area of the APL buffer 210 as shown in FIG. The APL buffer area information acquisition unit 201 for acquiring the information and the page head address of the OS buffer 200 as shown in FIG. 7 as the data reception area are set, and the setting information is notified to the reception buffer register 14. The OS buffer address notifying unit 202 that performs, the page offset predicted by the offset predictor 13, the offset specifying unit 203 that specifies the page offset of the APL buffer 210, and the header of the packet stored in the OS buffer 200 Analyzing information Data information analysis unit 204, page map switching unit 205 that controls data transfer from OS buffer 200 to APL buffer 210 by switching page maps, and data from OS buffer 200 to APL buffer 210 by data copying A data copy control unit 206 that controls the transfer, and an offset prediction change instruction unit 207 that instructs the offset predictor 13 to change the prediction of the page offset.

  Here, FIG. 6 illustrates a continuous page configuration as the data reception area of the APL buffer 210. However, when pages are associated with each other by a link, such a continuous page configuration is not used. 7 illustrates a continuous page configuration as a data reception area of the OS buffer 200. However, when pages are associated with each other by a link, such a continuous page configuration is not used.

  FIG. 8 illustrates an example of a device configuration of the offset predictor 13.

  As shown in this figure, the offset predictor 13 includes an offset prediction table 130 that manages the predicted value of the page offset of the APL buffer 210 in association with the transmission source IP address and the transmission destination port number, and the key extractor 12. The page offset of the offset prediction table 130 pointed to by the extracted information is updated, and a new predicted value of the page offset of the APL buffer 210 is calculated according to the calculation formula “new page offset = page offset + packet length”, A table updating unit 131 that updates a page offset managed in the offset prediction table 130, a CPU interface 132 that changes a page offset managed in the offset prediction table 130 in accordance with instruction information from the offset prediction change instruction unit 207, and a key A table reference unit 133 that reads the page offset by referring to the offset prediction table 130 using the source IP address and destination port number extracted by the output unit 12 as keys, and supplies the page offset to the adder 15 and the table update unit 131. With.

  9 and 10 show an example of processing executed by the network interface device 1 of the present invention in the form of a flowchart. FIG. 11 shows an example of flowchart executed by the operating system 20 included in the data processing device 2 of the present invention. Is illustrated.

  Next, according to these flowcharts, processing executed by the present invention configured as described above will be described in detail.

  First, processing executed by the network interface device 1 of the present invention will be described with reference to the flowcharts of FIGS. 9 and 10.

  When the network interface device 1 of the present invention receives a TCP / IP packet transmitted from an Ethernet (not shown) and stores it in the received data storage buffer 10, first, as shown in the flowchart of FIG. Then, the IP address of the transmission source, the port number for designating the transmission destination, and the packet length are extracted from the header information of the packet.

  Here, the network interface device 1 of the present invention is configured by hardware, and in this step S10, the packet header information is not analyzed in detail, and the source IP address and destination Only the process of extracting the port number for designating the packet length and the packet length is performed.

  Subsequently, in step S11, the predicted value of the page offset of the APL buffer 210 is specified by referring to the offset prediction table 130 using the extracted IP address and port number as a key.

  Subsequently, in step S12, the transfer destination address of the OS buffer 200 is calculated by adding the head address of the next page of the last page of the OS buffer 200 transferred last time and the predicted value of the specified page offset. To do.

  In step S13, the received packet is transferred to the OS buffer 200 using the calculated transfer destination address as the transfer destination.

Subsequently, in step S14, the page offset specified in step S11 is used as an update target, and the packet length extracted in step S10 and the page offset specified in step S11 are used.
A new predicted value of the page offset of the APL buffer 210 is calculated according to the calculation formula of new page offset = page offset + packet length, and the predicted value of the page offset managed in the offset prediction table 130 is updated. finish.

  As will be described later, the data processing apparatus 2 of the present invention, when transferring this packet, if there is an error in the page offset predicted by the network interface apparatus 1 of the present invention (the page offset specified in step S11). The network interface device 1 of the invention is instructed to change the prediction of the page offset by designating the predicted value of the page offset used at the next data transfer.

  From now on, when there is a page offset prediction change instruction from the data processing apparatus 1 of the present invention, the network interface apparatus 1 of the present invention, first, in step S20, as shown in the flowchart of FIG. The page offset prediction value sent from the processing device 1 is received, and in step S21, the page offset prediction value updated in step S14 in the flowchart of FIG. 9 is changed to the received page offset prediction value. The process is terminated.

  Next, processing executed by the operating system 20 included in the data processing apparatus 2 of the present invention will be described with reference to the flowchart of FIG.

  When a packet is sent from the network interface device 1 of the present invention by designating a transfer destination address, the data processing device 2 of the present invention stores the packet in the memory area of the OS buffer 200 indicated by the transfer destination address. After that, by notifying the OS buffer 200 of an interrupt, the operating system 20 is notified of packet reception.

  Upon receipt of this interrupt notification, the operating system 20 included in the data processing apparatus 2 of the present invention, first, as shown in the flowchart of FIG. 11, first of all, the header of the packet stored in the OS buffer 200 in step S30. By analyzing the information, it is determined where to store the packet in the APL buffer 210.

  That is, packets transferred through the network are not always transferred in order, and packet loss may occur. From now on, the packets stored in the OS buffer 200 cannot be transferred to the APL buffer 210 in the storage order. Therefore, in step S30, the header information of the packet stored in the OS buffer 200 is analyzed to determine where the packet is stored in the APL buffer 210.

  Subsequently, in step S31, the page offset (offset from the start address) of the storage start page of the APL buffer 210 is specified based on the determination result in step S30.

  Subsequently, in step S32, the page offset (offset from the head address) of the storage start page of the OS buffer 200 is specified according to the transfer destination address instructed from the network interface device 1 of the present invention.

  Subsequently, in step S33, it is determined whether or not the page offset specified in step S31 matches the page offset specified in step S32.

  When it is determined that the two page offsets coincide with each other according to this determination process, the process proceeds to step S34, and the page stored in the OS buffer 200 is processed and the page map can be switched for the data portion. Data is moved from the OS buffer 200 to the APL buffer 210 in accordance with the map switching, and for data portions where the page map cannot be switched, data is copied from the OS buffer 200 to the APL buffer 210 in byte units. The process is terminated.

  That is, when the predicted value of the page offset of the APL buffer 210 predicted by the network interface apparatus 1 of the present invention matches the actual page offset of the APL buffer 210 obtained by analyzing the header information of the packet, the page Since data movement using map switching is possible, control is performed so that data portions that can be switched between page maps are moved from the OS buffer 200 to the APL buffer 210 according to the page map switching. is there.

  On the other hand, when it is determined that the two page offsets do not match according to the determination process in step S33, the process proceeds to step S35, and the packet stored in the OS buffer 200 is set as a processing target from the OS buffer 200 to the APL buffer 210. Copy data in bytes.

  That is, when the predicted value of the page offset of the APL buffer 210 predicted by the network interface apparatus 1 of the present invention does not match the actual page offset of the APL buffer 210 obtained by analyzing the header information of the packet, the page Since map switching cannot be used, data is copied in bytes from the OS buffer 200 to the APL buffer 210.

  Subsequently, in step S36, the page offset of the last storage page of the packet copied to the APL buffer 210 (the page offset of the storage start page of the next OS buffer 200) is specified.

  Subsequently, in step S37, the specified page offset is notified to the network interface device 1 of the present invention (IP address and port number are also notified), thereby predicting the page offset stored in the offset prediction table 130. An instruction to change the value is issued, and the process is terminated.

  Upon receiving this page offset prediction change instruction, the network interface device 1 according to the present invention, as described with reference to the flowchart of FIG. 10, stores the page offset updated in step S <b> 14 of the flowchart of FIG. 9 stored in the offset prediction table 130. Is predicted according to the page offset notified from the operating system 20.

  Next, the processing of the present invention will be specifically described with reference to FIGS. Here, in FIGS. 12 to 17, the length of one page is set to 0x1000.

  In the processing example shown in FIGS. 12 to 14 (corresponding to the processing example shown in FIG. 1), as shown in FIG. 12, the application program 21 has a start address of 0x20000000 and a length of the operating system 20. Requests that data be received in the memory area of the APL buffer 210 of 0x5000, and the operating system 20 requests the network interface apparatus 1 to receive data in the memory area of the OS buffer 200 of 0x10000000 to 0x10006000. It is assumed that.

  In this case, as shown in FIG. 13, the network interface device 1 receives packet # 1 having a packet length of 0x2800 (2.5 pages), and at this time, the network interface device 1 sets a page offset of 0x000. Suppose you predict.

  In this case, as shown in FIG. 13, the network interface device 1 transfers 2.5 pages of data to the memory area of the OS buffer 200 starting from 0x10000000, and 0x800 as the predicted value of the next page offset. (Page 0.5) is predicted and written to the offset prediction table 130.

  Upon receiving this data transfer, as shown in FIG. 13, the operating system 20 sets the page offset predicted by the network interface device 1 and the APL for 0x20000000 in the APL buffer 210 as the transfer destination area. When it is determined that the page offset of the buffer 210 matches, the data is transferred from the OS buffer 200 to the APL buffer 210 while performing data movement by rewriting the page table.

  Subsequently, as shown in FIG. 14, it is assumed that the network interface apparatus 1 receives a packet # 2 having a packet length of 0x2800 (2.5 pages). At this time, the network interface apparatus 1 predicts a page offset of 0x800 (0.5 pages) according to the data stored in the offset prediction table 130.

  In this case, as shown in FIG. 14, the network interface device 1 transfers 2.5 pages of data to the memory area of the OS buffer 200 starting from 0x10003800 (= 0x10003000 + 0x800), and the next page 0x000 (= 0x800 + 0x800) is predicted as the predicted offset value and written to the offset prediction table 130.

  Upon receiving this data transfer, as shown in FIG. 14, the operating system 20 sets the page offset predicted by the network interface device 1 and the APL for 0x20002800 in the APL buffer 210 as the transfer destination area. When it is determined that the page offset of the buffer 210 matches, the data is transferred from the OS buffer 200 to the APL buffer 210 while performing data movement by rewriting the page table.

  Thus, in the processing examples shown in FIGS. 12 to 14, data transfer is executed from the OS buffer 200 to the APL buffer 210 in the form shown in FIG.

  In the processing example shown in FIGS. 15 to 17 (corresponding to the processing example shown in FIG. 2), as shown in FIG. 15, the application program 21 has the start address 0x20000800 and the length of the operating system 20. Requests that data be received in the memory area of the APL buffer 210 of 0x5000, and the operating system 20 requests the network interface apparatus 1 to receive data in the memory area of the OS buffer 200 of 0x10000000 to 0x10006000. It is assumed that.

  In this case, as shown in FIG. 16, the network interface device 1 receives a packet # 1 having a packet length of 0x2800 (2.5 pages). At this time, the network interface device 1 sets a page offset of 0x000. Suppose you predict.

  In this case, as shown in FIG. 16, the network interface device 1 transfers 2.5 pages of data to the memory area of the OS buffer 200 starting from 0x10000000, and 0x800 as the predicted value of the next page offset. (Page 0.5) is predicted and written to the offset prediction table 130.

  Upon receiving this data transfer, as shown in FIG. 16, the operating system 20 sets the page offset predicted by the network interface device 1 and the APL for 0x20002800 in the APL buffer 210 as the transfer destination area. It is determined that the page offset of the buffer 210 does not match, and data is transferred from the OS buffer 200 to the APL buffer 210 without performing data movement by rewriting the page table. Then, it notifies the network interface device 1 to change the predicted value of the page offset to 0x000.

  Subsequently, as shown in FIG. 17, it is assumed that the network interface apparatus 1 receives a packet # 2 having a packet length of 0x2800 (2.5 pages). At this time, the network interface device 1 predicts a page offset of 0x000 in accordance with the change instruction from the operating system 20.

  In this case, as shown in FIG. 17, the network interface device 1 transfers 2.5 pages of data to the memory area of the OS buffer 200 starting from 0x10003000 (= 0x10003000 + 0x000), and the next page As a predicted offset value, 0x800 (= 0x000 + 0x800) is predicted and written to the offset prediction table 130.

  Upon receiving this data transfer, as shown in FIG. 17, the operating system 20 sets the page offset predicted by the network interface device 1 and the APL for 0x20003000 in the APL buffer 210 as the transfer destination area. When it is determined that the page offset of the buffer 210 matches, the data is transferred from the OS buffer 200 to the APL buffer 210 while performing data movement by rewriting the page table.

  In this way, in the processing examples shown in FIGS. 15 to 17, data transfer is executed from the OS buffer 200 to the APL buffer 210 in the form shown in FIG.

  Although the present invention has been described according to the illustrated embodiment, the present invention is not limited to this.

  For example, in the exemplary embodiment, the offset prediction table 130 discloses that manages the predicted value of the page offset of the APL buffer 210 in association with the transmission source IP address and the transmission destination port number. Whether or not to use a fixed page offset in association with the destination port number, and for those that use a fixed page offset, manage a fixed page offset, and for those that do not use a fixed page offset, A configuration may be adopted in which the predicted value of the page offset of the APL buffer 210 is managed.

  When there is a loop process or the like, the data processing apparatus 2 may receive small data many times in the same memory area of the APL buffer 210. In such a case, the offset prediction table 130 is used. Indicates whether or not to use a fixed page offset in association with the source IP address and destination port number, and manages a fixed page offset for those using a fixed page offset. For those not using, a configuration is adopted in which the predicted value of the page offset of the APL buffer 210 is managed.

  The present invention is provided in association with a computer system, and includes a data receiving device that receives data transmitted from the outside and transfers the data to the computer system, and a computer system that connects the data receiving device. It is possible to predict the page offset that enables the data receiving device to move the data by rewriting the page map, which makes it possible to use the data movement by rewriting the page map. Since many can be created, high-speed data transfer by rewriting the page map can be realized.

  (Supplementary note 1) A data receiving device provided in association with a computer system for receiving data sent from the outside and temporarily storing it in a memory in the computer system, and data transferred from the data receiving device A data receiving method executed by a system comprising: a computer system that receives the data and temporarily stores the data in a memory and then transfers the data to an application memory area managed by an application program; When the receiving device receives data, it predicts the page offset of the application memory area, and sets the data storage start address defined by the start address of the data transfer destination page on the memory and the predicted page offset. Specify and transfer the received data to the memory, Based on the length of the received data and the predicted page offset, a predicted value of the page offset used for the next data reception is set, and a prediction error is notified of the predicted value of the page offset predicted from the computer system. And processing to change the predicted value of the set page offset according to the page offset notified from the computer system, and the computer system performs the data storage based on the data storage start address designated at the time of data transfer. The page offset predicted by the receiving apparatus is specified, and the data stored in the memory is analyzed to identify the page offset of the application memory area to which the data is transferred, and the two specified pages Offset matches And determining whether the two page offsets do not match, based on the specified page offset of the application memory area and the length of the data stored in the memory, for the application after the data transfer A method of transferring data from the memory to the application memory area based on a determination result of whether or not two page offsets match, specifying a page offset of the memory area and notifying the data receiving apparatus A data receiving method characterized in that processing is performed so as to determine.

  (Supplementary Note 2) A data receiving device provided with the computer system for receiving data sent from the outside and temporarily storing it in a memory in the computer system. Data storage defined by a prediction unit that predicts a page offset of an application memory area that is a data transfer destination from the memory, a head address of a data transfer destination page on the memory, and a page offset predicted by the prediction unit A transfer unit that designates a start address and transfers the received data to the memory, and a page to be used for the next data reception based on the length of the received data and the page offset predicted by the prediction unit A setting means for setting a predicted value of the offset; and A change means for changing the predicted value of the page offset set by the setting means according to the page offset notified from the computer system when there is a notification of a prediction error with respect to the measured predicted value of the page offset. Data receiving device.

  (Supplementary note 3) In the data receiving device according to supplementary note 2, the data reception device includes a storage unit that stores a predicted value of the page offset set by the setting unit in association with transmission source information and transmission destination information, and the prediction unit includes: When receiving data, the page offset is obtained by extracting the transmission source information and transmission destination information of the received data and referring to the storage information of the storage means using the extracted transmission source information and transmission destination information as a key. A data receiving device characterized by predicting.

  (Supplementary Note 4) In the data receiving device described in Supplementary Note 2, whether or not to use a fixed page offset in association with the transmission source information and the transmission destination information, and what uses a fixed page offset is fixed For those which store a page offset and do not use a fixed page offset, the storage unit stores a predicted value of the page offset set by the setting unit, and the prediction unit receives the data when receiving the data. Extracting data transmission source information and transmission destination information, and predicting a page offset by referring to the storage information of the storage means using the extracted transmission source information and transmission destination information as a key Data receiving device.

  (Supplementary note 5) The data receiving device according to supplementary note 4, further comprising registration means for registering a fixed page offset value in the storage means in accordance with a fixed page offset value notified from a computer system. , A data receiving device characterized.

  (Supplementary note 6) In the data receiving device according to any one of supplementary notes 3 to 5, the predicted value of the page offset is initially registered in the storage unit in accordance with the initial value of the page offset notified from the computer system. A data receiving apparatus characterized by comprising initial registration means.

  (Supplementary note 7) A computer system for receiving data transferred from a data receiving device, temporarily storing the data in a memory, and transferring the data to an application memory area managed by an application program, wherein the data Based on a data storage start address specified at the time of data transfer from the receiving device, the first specifying means for specifying the page offset predicted by the data receiving device, and analyzing the data stored in the memory, Second specifying means for specifying a page offset of the application memory area to which the data is transferred, a page offset specified by the first specifying means, and a page offset specified by the second specifying means A judging means for judging whether or not they match, and the judging means comprises two page offsets. Is determined based on the page offset specified by the second specifying means and the length of data stored in the memory, the page offset of the application memory area after data transfer, First notification means for notifying the data receiving device; and determination means for determining a data transfer method from the memory to the application memory area based on a determination result of the determination means; Computer system.

  (Supplementary Note 8) In the computer system according to Supplementary Note 7, when the determination unit determines that two page offsets do not match, the determination unit determines to perform data transfer according to a data copy operation, and the determination unit When determining whether the two page offsets match, it is determined that the page map switchable page data is transferred by the pagemap switching operation, and other page data is transferred by the data copy operation. A computer system characterized by

  (Supplementary note 9) The computer system according to supplementary note 7 or 8, further comprising: second notification means for notifying the data receiving device of a fixed page offset value.

  (Supplementary note 10) The computer system according to Supplementary note 7 or 8, further comprising: third notification means for notifying the data reception device of an initial value of a predicted value of a page offset.

It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is an apparatus block diagram of the system which comprises this invention. It is explanatory drawing of the information which the buffer area | region information acquisition part for APL acquires. It is explanatory drawing of the information which the buffer address notification part for OS sets. It is an apparatus block diagram of an offset predictor. It is the figure which showed the process which a network interface apparatus performs in the form of the flowchart. It is the figure which showed the process which a network interface apparatus performs in the form of the flowchart. It is a flowchart which the operating system with which a data processor is provided performs. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention. It is explanatory drawing of the process of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Network interface apparatus 2 Data processing apparatus 10 Received data storage buffer 11 DMA engine 12 Key extractor 13 Offset predictor 14 Receive buffer register 15 Adder 20 Operating system 21 Application program 200 OS buffer 200
201 APL buffer area information acquisition unit 202 OS buffer address notification unit 203 Offset specification unit 204 Header information analysis unit 205 Page map switching unit 206 Data copy control unit 207 Offset prediction change instruction unit 210 APL buffer

Claims (7)

A data receiving device provided with the computer system for receiving data sent from the outside and temporarily storing it in a memory in the computer system, and receiving data transferred from the data receiving device A data receiving method executed in a system comprising a computer system that temporarily stores it in a memory and then transfers it to an application memory area managed by an application program,
The data receiving device is
When receiving data, predict the page offset of the application memory area, specify the data storage start address defined by the start address of the data transfer destination page on the memory and the predicted page offset, The received data is transferred to the memory, and furthermore, based on the length of the received data and the predicted page offset, a predicted value of the page offset used for the next data reception is set and predicted from the computer system When there is a notification of a prediction error about the predicted value of the page offset, process the predicted value of the set page offset to be changed according to the page offset notified from the computer system,
The computer system is
Based on the data storage start address specified at the time of data transfer, the page offset predicted by the data receiving device is specified, and the data stored in the memory is analyzed, so that the data transfer destination becomes The page offset of the application memory area is specified, and it is determined whether or not the two specified page offsets match. When the mismatch between the two page offsets is determined, the specified page of the application memory area is determined. Based on the offset and the length of the data stored in the memory, the page offset of the application memory area after the data transfer is identified and notified to the data receiving device, and whether the two page offsets match Based on the determination result of whether or not To process to determine the data transfer method to the application memory area,
Characteristic data receiving method. A data receiving device that is attached to a computer system, receives data sent from the outside, and temporarily stores it in a memory in the computer system,
A predicting means for predicting a page offset of an application memory area as a data transfer destination from the memory when receiving data;
A transfer means for designating a data storage start address defined by a head address of a data transfer destination page on the memory and a page offset predicted by the prediction means, and transferring the received data to the memory;
A setting unit that sets a predicted value of a page offset to be used at the time of next data reception based on the length of the received data and the page offset predicted by the prediction unit;
Change means for changing the predicted value of the page offset set by the setting means according to the page offset notified from the computer system when there is a notification of a prediction error from the computer system for the predicted value of the page offset predicted by the prediction means; Having
A characteristic data receiving apparatus. The data receiving device according to claim 2,
A storage unit that stores the predicted value of the page offset set by the setting unit in association with the transmission source information and the transmission destination information;
When receiving the data, the prediction means extracts the transmission source information and the transmission destination information of the received data, and refers to the storage information of the storage means using the extracted transmission source information and the transmission destination information as a key. Predicting the page offset by
A characteristic data receiving apparatus. The data receiving device according to claim 2,
Whether to use a fixed page offset in association with the transmission source information and the transmission destination information, and for those using a fixed page offset, store a fixed page offset and do not use a fixed page offset For, comprising storage means for storing the predicted value of the page offset set by the setting means,
When receiving the data, the prediction means extracts the transmission source information and the transmission destination information of the received data, and refers to the storage information of the storage means using the extracted transmission source information and the transmission destination information as a key. Predicting the page offset by
A characteristic data receiving apparatus. In the data receiving device according to claim 3 or 4,
In accordance with the initial value of the page offset notified from the computer system, the storage means comprises initial registration means for initially registering the predicted value of the page offset.
A characteristic data receiving apparatus. A computer system that receives data transferred from a data receiving device, temporarily stores the data in a memory, and then transfers the data to an application memory area managed by an application program,
First specifying means for specifying a page offset predicted by the data receiving device based on a data storage start address specified at the time of data transfer from the data receiving device;
Analyzing the data stored in the memory, a second specifying means for specifying a page offset of the application memory area to which the data is transferred;
Determining means for determining whether or not the page offset specified by the first specifying means and the page offset specified by the second specifying means match;
The application memory area after data transfer based on the page offset specified by the second specifying means and the length of data stored in the memory when the determining means determines a mismatch between two page offsets Notification means for identifying the page offset and notifying the data receiving device;
Determining means for determining a data transfer method from the memory to the application memory area based on a determination result of the determination means;
A featured computer system. The computer system according to claim 6.
The determining unit determines to perform data transfer according to a data copy operation when the determining unit determines a mismatch between two page offsets, and determines a page map when the determining unit determines a match between two page offsets. For page data that can be switched, data transfer is performed by a page map switching operation, and for other page data, it is determined to perform data transfer according to a data copy operation.
A featured computer system.

Images