JP4484436B2 - Information storage control method to storage device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a control method and program for executing processing requests stored in a queue, and a system using them.
[0002]
[Prior art]
Two types of queues are used as an interface with an issuing device that issues processing request information including a processing request and the receiving device that receives the processing request information, executes the processing request, and transmits the execution result to the issuing device. Sometimes used. In one queue, processing request information is accumulated as a queue, and is executed by the receiving apparatus in order from the processing request at the head of the queue.
[0003]
In the other queue, information indicating that the processing request is completed (processing completion information) is accumulated as a queue, and is collected by the issuing side device in order from the processing completion information at the head of the queue, and status confirmation at the time of processing completion, etc. Is done. For example, Infiniband (see Non-Patent Document 1) and VIA (Virtual Interface Architecture) (Non-Patent Document 2) proposed as next-generation input / output (I / O) technologies for connecting computers and peripheral devices In an architecture such as (see FIG. 4), processing requests are executed using the two types of queues described above.
[0004]
[Non-Patent Document 1]
Infiniband (SM) Trade Association, "Infiniband (TM) Architecture Specification Volume 1 Release 1.1", p. 80, section 3.2.1 Queuing, [online], Final, November 6,2002, [searched January 21, 2003 ] Internet <http://www.infinibandta.org/specs/>
[0005]
[Non-Patent Document 2]
Compaq Computer Corp., Intel Corporation, Microsoft Corporation, "Virtual Interface Architecture Specification", P.11, section 2.1. VI Architecture Components, [online], Draft Revision 1.0, December 4, 1997, [January 21, 2003 Search], Internet <ftp://download.intel.com/design/servers/vi/san_10.pdf>
[0006]
[Problems to be solved by the invention]
However, when processing requests are executed using the two types of queues described above as an interface, when processing request information or processing completion information is stored in each queue, the processing request information or processing completion information stored in the queue is stored. Is often not judged. Therefore, the number of storable data in each queue is exceeded, and the queue may overflow (overflow).
[0007]
When the queue overflows, the execution of the processing request is stopped, and the above-mentioned issuing side device or receiving side device is caused to perform a new process called a recovery process, which inevitably degrades the performance of the entire apparatus. As a matter of course, in order to perform the recovery process, implementation by software or hardware is required. In addition, in order to increase the number of data that can be stored, it is possible to secure a larger storage area and increase the size of each queue. Had occurred.
[0008]
In view of the above-described problems, the present invention provides an information accumulation control method for preventing each queue from overflowing when two types of queues are interposed as interfaces to execute processing requests accumulated in the queues. With the goal. Moreover, it is possible to prevent both queues from overflowing by using both in combination. In addition, it is necessary to simplify the implementation of hardware in order to reduce the implementation cost and achieve high performance. Therefore, it is desirable that hardware can be implemented by software implementation.
[0009]
[Means for Solving the Problems]
In order to solve the above problem, a step of setting a first area in which processing request information is accumulated, a step of setting a second area in which processing completion information that is information indicating completion of the processing request is stored, Setting the initial value of the semaphore variable corresponding to the first area as the maximum number of information that can be stored in the first area, and subtracting 1 from the value of the semaphore variable corresponding to the first area If the obtained value is a positive value, the processing request information is added to the first area, and a value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area The process of updating as the semaphore variable corresponding to the first area, and when the requested process is completed, the information specifying the process request and the process completion information are associated with each other and stored in the second area. Process and completion of the process in the second area It is determined whether information has been accumulated, and if accumulated, a value obtained by adding 1 to the semaphore variable corresponding to the first area is updated as the semaphore variable corresponding to the first area, and the processing completion information There is provided an information storage control method characterized by including a step of deleting one.
[0010]
In this case, it is possible to set the maximum number of information that can be stored in the second area to be equal to or greater than the maximum number of information that can be stored in the first area. It is also possible to set semaphore variables in both the first area and the second area.
[0011]
Further, the above-described problems include a step of setting a plurality of first areas in which processing request information is accumulated, and a step of setting a second area in which processing completion information that is information indicating completion of the processing request is accumulated. A step of setting an initial value of a semaphore variable corresponding to each of the plurality of first regions as a maximum number of pieces of information that can be stored in the plurality of first regions, and among the plurality of first regions If one is selected and a value obtained by subtracting 1 from the value of the semaphore variable corresponding to the selected first area is a positive value, and if it is a positive value, the selected first area Processing request information is added to the table, the information specifying the selected first area and the processing request is stored in the correspondence table, and the value of the semaphore variable corresponding to the selected first area is calculated. The value obtained by subtracting 1 corresponds to the selected first area. Updating as a semaphore variable; when the processing request is completed, storing information identifying the processing request in association with the processing completion information in the second area; and It is determined whether or not processing completion information is accumulated. If it is accumulated, one piece of the processing completion information is read, the processing request corresponding to the read processing completion information is specified, and the correspondence table is referred to The first area in which the specified processing request is stored is specified, and a value obtained by adding 1 to the semaphore variable corresponding to the specified first area corresponds to the specified first area. It is solved by providing an information storage control method characterized by comprising a step of updating as a semaphore variable to be performed and deleting the read processing completion information.
[0012]
In this case, it is also possible to set the maximum number of information that can be stored in the second area to be equal to or greater than the sum of the maximum number of information that can be stored in each first area. Furthermore, it is also possible to set semaphore variables in each of the plurality of first regions and the second region.
[0013]
In addition, the above-described problem is a first area in which processing request information is stored, and a second area in which processing completion information that is information indicating completion of the processing request is stored in association with information specifying the processing request. And a storage unit storing a semaphore variable table in which information specifying the first area and a semaphore variable are associated and stored, and a control unit that controls the storage unit A step of causing the control unit to set an initial value of a semaphore variable corresponding to the first area as a maximum number of information that can be stored in the first area; and corresponding to the first area It is determined whether the value obtained by subtracting 1 from the value of the semaphore variable is a positive value. If the value is positive, the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area is the first area. As a semaphore variable corresponding to Updating the semaphore variable table, adding the processing request information to the first area, and determining whether the processing completion information is accumulated in the second area. A program for executing a step of updating the semaphore variable table with a value obtained by adding 1 to a semaphore variable corresponding to one area as a semaphore variable corresponding to the first area and deleting one piece of the processing completion information It is solved by providing.
[0014]
In addition, the above-described problem is a first area in which processing request information is stored, and a second area in which processing completion information that is information indicating completion of the processing request is stored in association with information specifying the processing request. Are stored in the storage unit, the first device having the control unit that controls the storage unit and executes the processing request, and information that is connected to the first device and identifies the first region And a semaphore variable table in which semaphore variables are associated and stored, and a second device having a control unit that controls the storage unit and transmits the processing request information to the first device The control unit in the second device sets an initial value of a semaphore variable for the first area as a maximum number of information that can be stored in the first area, and at a specific timing, Semaphore change corresponding to the first region If the value obtained by subtracting 1 from the positive value is a positive value, the value obtained by subtracting 1 from the semaphore variable corresponding to the first area is updated as the semaphore variable corresponding to the first area. Then, new processing request information is transmitted to the first device, and the control unit in the first device receives the new processing request information and adds it to the first region at a specific timing. The processing request stored in the first area is executed, processing completion information corresponding to the processing request is created and stored in the second apparatus, and the control unit in the second apparatus is specified. It is determined whether the processing completion information is accumulated in the second area at the timing of, and if it is accumulated, a value obtained by adding 1 to the semaphore variable corresponding to the first area is stored in the first area. Update as the corresponding semaphore variable, Is solved by providing a system and deletes one said processing completion information stored in the region.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the technical scope of the present invention is not limited to such an embodiment, but covers the invention described in the claims and equivalents thereof.
[0016]
FIG. 1 is a diagram illustrating a configuration example of an information processing apparatus used in the first embodiment of the present invention. Although the application of the present invention is not limited to InfiniBand, in the embodiment of the present invention, a CPU that executes an OS command stored in a memory is a data (workpiece) including a processing request as a processing request issuing side device. A description will be given of how the present invention is applied when an InfiniBand host channel adapter in the same housing receives a work request as a receiving device.
[0017]
The memory 5 stores an operating system (OS), which is basic software for operating a computer, and application programs that operate on the OS. Based on the processing request from the OS or application program, the CPU 2 is connected to the memory 5 and various peripheral devices (hard disk not shown, storage not shown, connected to the outside via SCSI or fiber channel, LAN via the system bus 3. The data stored in the other computer (not shown) connected via the computer is accessed, and processing for controlling the devices in the computer 1 is executed.
[0018]
When the CPU 2 accesses the data stored in the memory 5, the memory controller 4 converts the virtual memory address designated by the CPU 2 into an actual physical memory address, and sends the corresponding data to the CPU 2 via the system bus 3. Send. When the CPU 2 accesses data stored in the peripheral device, the CPU 2 stores data including a processing request (hereinafter referred to as a work request) in a work queue (not shown) secured in the memory 7 of the host channel adapter (HCA) 6. To accumulate. The coprocessor 8 is a processing device that performs operations related to the control in the HCA, processes work requests accumulated in the work queue at a specific timing, and stores processing result data in a completion queue (memory 7). (Not shown). The CPU 2 checks the completion queue at a specific timing, and collects and deletes processing result data.
[0019]
The InfiniBand switch 9 connects channel adapters and enables point-to-point data transmission between the channel adapters. In FIG. 1, the HCA 6 has a network interface card (NIC) 11 for connecting the computer 1 to the network via the InfiniBand switch 9, a SCSI card 13 for connecting SCSI devices to the computer 1, and a fiber channel (Fiber channel). Channel) is connected to target channel adapters (TCAs) 11, 13, 15 provided in each of the FC cards 14 for connecting devices. It is also possible to connect directly to another HCA via the InfiniBand switch 9. For example, in a device such as a blade server (in this case, each computer 1 is referred to as a blade) in which a plurality of computers 1 exist in the same housing, the connection is made via an internal bus called a backplane that connects the blades. There is.
[0020]
FIG. 2 is a diagram showing an operation when the CPU 2 accesses data stored in the peripheral device in the first embodiment of the present invention. In the first embodiment of the present invention, one work queue 24 and one completion queue 28 are secured in the memory, and a semaphore variable indicating the number of work requests 26 that can be accumulated in the work queue 24 is set. The work queue 24 is prevented from overflowing.
[0021]
The value of the semaphore variable is decremented by 1 when one work request 26 is accumulated in the work queue 24, and is incremented by 1 when one work completion 30 is collected. By setting the initial value of the semaphore variable as the maximum number of data that can be stored in the work queue 24, if the semaphore variable in the work queue 24 is a positive value, it is possible that additional data can be stored in the work queue 24. Recognize.
[0022]
As the semaphore variable, the maximum amount of data that can be accumulated in the work queue 24 (for example, the size of the reserved area) is set, and the size of the work request 26 is subtracted every time the work request 26 is accumulated. The same effect can be obtained by adding the size of the work request 26 by collecting the completion 30. The latter is an effective means when the size of each work request 26 is not constant.
[0023]
The CPU 2 in FIG. 2 stores data including processing requests from the OS and application programs as work requests 26 in the work queue 24 with the queue number WQ1 secured in the memory 7 as an area specified by the start address and end address. To do. The CPU 2 refers to the semaphore variable table 22 in the memory 5 in which the semaphore variables set corresponding to the work queue 24 are stored, and determines whether or not the work requests 26 can be accumulated.
[0024]
If the work request 26 can be accumulated, the work request 26 is added to the end of the work queue 24. Then, the CPU 2 updates the semaphore variable table 22. Details of the work request accumulation process will be described later.
[0025]
The coprocessor 8 takes out the first work request 26 from the work queue 24 at a specific timing, and executes the processing request. The work request 26 stores processing request contents such as data transmission and reception, and information such as a data transmission destination, data to be transmitted, or storage location of received data, and the coprocessor 8 follows the processing request contents. Execute the process.
[0026]
The coprocessor 8 performs data transmission with the TCA connected via the InfiniBand switch 9 according to the processing request contents, and completes the work request 26. After processing the work request 26, the coprocessor 8 creates a work completion 30 corresponding to the processed work request 26, and the queue number secured in the memory 7 as an area specified by the start address and end address. Accumulate in the completion queue 28 of CQ1.
[0027]
The CPU 2 collects the work completion 30 from the completion queue 28 at a specific timing, confirms that the corresponding work request 26 is completed, and updates the semaphore variables in the semaphore variable table 22 of the memory 5. Details of the work completion collection process will be described later. Thus, the processing request is completed. Next, the data structure of the work request 26, the work completion 30, and the semaphore variable table 22 will be described.
[0028]
FIG. 3 is a diagram illustrating a data configuration example of the work request 26 accumulated in the work queue 24. Each entry of FIG. 3 stores an identification number (Work Request ID) 31 that identifies the work request 26 and data 32 related to the processing request. Although not shown, the data 32 related to the processing request stores information such as processing request contents such as data transmission and reception, data transmission destination and data to be transmitted, and storage location of the received data.
[0029]
FIG. 4 is a diagram illustrating a data configuration example of the work completion 30 accumulated in the completion queue 28. Each entry in FIG. 4 stores an identification number (Work Request ID) 41 that identifies the processed work request 26 and data 42 related to the processed work request 26. In the data 42 related to the processed work request 26, although not shown, for example, an end status indicating normal end or abnormal end, an error code indicating error content, and the like are stored. By referring to the work completion 30, it is possible to confirm which work request 26 is completed in which state.
[0030]
FIG. 5 is a semaphore variable table 22 stored in the memory 5. Each entry in FIG. 5 stores a queue number 51 and a semaphore variable value 52 of the queue number.
In FIG. 5, WQ1 indicates a queue number 51 for the work queue 24, and SW1 is a semaphore variable value 52 for the work queue 24.
[0031]
In the first embodiment of the present invention, a semaphore variable is set only for one work queue 24. However, when there are a plurality of work queues 24, the semaphore variable set for each queue is shown in FIG. It is added to the semaphore variable table as line data. Furthermore, a semaphore variable for the completion queue 28 may be set and added to the semaphore variable table. With reference to the semaphore variable table 22, a semaphore variable value 52 corresponding to the work queue 24 or the completion queue 28 can be obtained, and it can be determined whether or not additional work requests 26 can be accumulated in the work queue 24.
[0032]
In the first embodiment of the present invention, the semaphore variable table 22 is used, but implementation without using a table is also possible. For example, a pointer that identifies a semaphore corresponding to a work queue number may be known using a hash function.
[0033]
Subsequently, the work request accumulation process and the work completion collection process in the first embodiment of the present invention will be described.
[0034]
FIG. 6 is a flowchart for explaining work request accumulation processing in the first embodiment of the present invention. In the semaphore variable table 22, the maximum number of data that can be stored in the work queue 24 is set as an initial value of the semaphore variable corresponding to the work queue 24 with the queue number WQ1.
[0035]
First, the CPU 2 determines whether the semaphore variable in the work queue 24 is a positive value (S62). The CPU 2 refers to the semaphore variable table 22 in the memory 5 and obtains a semaphore variable value 52 corresponding to the work queue 24 having the queue number WQ1. If the obtained semaphore variable value is positive, it means that the work queue 24 is free and work requests 26 can be additionally accumulated.
[0036]
If the work queue 24 is free in step S62, the CPU 2 updates the semaphore variable table 22 with a value obtained by subtracting 1 from the semaphore variable in the work queue 24 as a new semaphore variable (S63). This is because when one work request 26 is accumulated, one empty work queue 24 is consumed.
[0037]
If it is determined in step S62 that there is no space, the process waits for a predetermined time and the process of step S62 is performed again. In this way, it waits for the work queue 24 to be free. Or give up the accumulation.
[0038]
After step S63, work requests 26 are additionally accumulated in the work queue 24 (S66). Then, it is determined whether or not the accumulation process in step S64 is successful (S67). Even if storage to the work queue 24 is possible, the work request storage process is canceled as it is when the storage process fails due to a power failure of the apparatus in the middle of the storage process or stop of the storage process due to interrupt processing.
[0039]
If the work request 26 is successfully accumulated in step S67, the work request accumulation process is terminated. If it fails in step S67, 1 is added to the semaphore variable in the work queue 24 (S69), and the work request accumulation process is terminated. In step S63, the semaphore variable is updated on the assumption that the work request is accumulated and one empty work queue 24 is consumed. However, since the work request accumulation fails in step S67, the empty work queue 24 is consumed. Not. Accordingly, in step S69, the value subtracted in step S63 is restored.
[0040]
The work request accumulation process in FIG. 6 appropriately updates the semaphore variable indicating the work queue availability, and if there is no available space, the work request 26 is not added, and overflow of the work queue 24 can be prevented. it can.
[0041]
FIG. 7 is a flowchart for explaining the work completion collection process according to the first embodiment of the present invention. First, it is determined whether or not the work completion 30 is accumulated in the completion queue 28 (S71). The CPU 2 collects the work completion 30 from the head of the completion queue 28, and if the completion queue 28 is empty, the work completion collection process is finished as it is.
[0042]
Then, the CPU 2 updates the value obtained by adding 1 to the semaphore variable of the work queue 24 as a semaphore variable (S73), and ends the work completion collection process. This is because the processing of the work request 26 has been completed, and it has been confirmed that one free space has been created in the work queue 24. The semaphore variable value 52 corresponding to the work queue 24 should be updated when the coprocessor 8 completes the processing of the work request 26, but the CPU 2 knows at what timing the coprocessor 8 processes the work request 26. I can't. Therefore, when the work completion 30 is collected, it is considered that the work queue 24 is empty, and the semaphore variable is updated.
[0043]
With the work completion collection process of FIG. 7, the semaphore variable indicating the empty state of the work queue 24 is appropriately updated, and overflow of the work queue 24 can be prevented. Further, by updating the semaphore variable of the work queue 24 when the work completion 30 is collected, the present invention can be implemented only on the OS or application side regardless of the processing of the HCA coprocessor 8 and can be easily introduced. Therefore, it is not necessary to complicate the hardware implementation.
[0044]
As described above, according to the first embodiment of the present invention, when a semaphore variable corresponding to the work queue 24 is appropriately updated and a work request is added, overflow of the work queue 24 is prevented by referring to the semaphore variable. It is possible.
[0045]
In the first embodiment, the size of the completion queue 28 is set in advance so that the number of data that can be accumulated in the completion queue 28 is equal to or larger than the maximum number of data that can be accumulated in the work queue 24 (completion). The present invention can be applied using only the semaphore variables set for the work queue 24 (without setting the semaphore variables for the queue 28).
[0046]
In other words, even if all the work requests 26 accumulated in the work queue 24 at a certain point in time are transferred to the completion queue 28, it is ensured that the completion queue 28 does not overflow. This is because it is only necessary to prevent the work queue 24 from overflowing when the work requests 26 are accumulated in the queue 24.
[0047]
In the first embodiment, the semaphore variable is set only in the work queue 24, but it is also possible to set the semaphore variable only in the completion queue 28 to prevent the completion queue 28 from overflowing. The present invention can be applied by appropriately replacing the work queue 24 in steps S62, S63, and S73 in FIGS. 6 and 7 with the completion queue 28.
[0048]
If there is no space in the completion queue, the work request 26 is not added. Since the work request 26 and the work completion 30 correspond one-to-one, the completion queue 28 does not overflow due to the added work request 26. In this way, overflow of the completion queue 28 can be prevented. In addition, by updating the semaphore variable in the completion queue 28 when the work request 26 is accumulated, the present invention can be implemented regardless of the processing of the HCA coprocessor, which is easy to introduce and implements hardware. It does not need to be complicated.
[0049]
As a modification of the first embodiment, semaphore variables can be set in both the work queue 24 and the completion queue 28 to prevent overflow of both queues. A work request accumulation process and a work completion collection process in a modification of the first embodiment will be described.
[0050]
FIG. 8 is a flowchart for explaining work request accumulation processing in a modification of the first embodiment. The same steps as those in FIG. 6 are assigned the same step numbers, and will not be described unless necessary. In the semaphore variable table 22, the maximum number of data that can be accumulated in each queue is set as initial values of semaphore variables corresponding to the work queue 24 and the completion queue 28 of queue numbers WQ1 and CQ1.
[0051]
First, the CPU 2 determines whether the semaphore variable corresponding to the work queue 24 is a positive value (S62). This is the same step as in FIG. If the work queue 24 is free in step S62, the CPU 2 updates the semaphore variable table 22 with a value obtained by subtracting 1 from the semaphore variable in the work queue 24 as a new semaphore variable (S63). This is because when one work request 26 is accumulated, one empty work queue 24 is consumed.
[0052]
If it is determined in step S62 that there is no space, the process waits for a predetermined time and performs the process of step S62 again. In this way, it waits for the work queue 24 to be free. Or give up the accumulation.
[0053]
After step S63, it is determined whether the semaphore variable corresponding to the completion queue 28 is a positive value (S64). The CPU 2 refers to the semaphore variable table 22 and obtains a semaphore variable value 52 corresponding to the completion queue 28 having the queue number CQ1. If the semaphore variable value is positive, it means that the completion queue 28 is empty and the work completion 30 can be accumulated.
[0054]
If it is determined in step S64 that there is a free space, the CPU 2 updates the semaphore variable table 22 using a value obtained by subtracting 1 from the semaphore variable in the completion queue 28 as a new semaphore variable (S65). Although the semaphore variable value corresponding to the completion queue 28 should be updated after the coprocessor 8 completes the processing of the work request 26, the CPU 2 can know when the coprocessor 8 processes the work request 26. Can not. The work requests 26 stored in the work queue 24 always have a work completion 30 corresponding to the work request 26 stored in the completion queue 28 after the processing is completed, and correspond one-to-one. Therefore, it is assumed that one completion queue 28 has been consumed when the work request 26 is accumulated in the work queue 24, and the semaphore variable in the completion queue 28 is updated.
[0055]
If it is determined in step S64 that there is no free space, the process waits for a predetermined time and performs the process of step S64 again. Thus, it waits for the completion queue 28 to be empty.
[0056]
Thus, when both the work queue 24 and the completion queue 28 are empty, the work request 26 is accumulated in the work queue 24 (S66). Then, it is determined whether or not the accumulation process in step S64 is successful (S67).
[0057]
If the work request 26 is successfully accumulated in step S67, the work request accumulation process is terminated. If it fails in step S67, 1 is added to each of the semaphore variables in the work queue 24 and the completion queue (S69), and the work request accumulation process is terminated. In steps S63 and S65, the semaphore variable is updated on the assumption that the work requests are accumulated and one of the work queue 24 and the completion queue 28 is consumed, but the accumulation of work requests failed in step S67. The empty space in the work queue 24 and the completion queue 28 is not consumed. Accordingly, in step S69, the value subtracted in steps S63 and S65 is restored.
[0058]
The semaphore variable indicating the free state of each queue is appropriately updated by the work request 26 accumulation process of FIG. 8, and the work request 26 is not added when there is no free space in the work queue. Similarly, the work request 26 is not added when there is no space in the completion queue. Since the work request 26 and the work completion 30 correspond one-to-one, the completion queue 28 does not overflow due to the added work request 26. Thus, overflow of the work queue 24 and the completion queue 28 can be prevented. In addition, by updating the semaphore variable in the completion queue 28 when the work request 26 is accumulated, the present invention can be implemented regardless of the processing of the HCA coprocessor, which is easy to introduce and implements hardware. It does not need to be complicated.
[0059]
FIG. 9 is a flowchart for explaining the work completion collection process in the modification of the first embodiment. The same steps as those in FIG. 7 are assigned the same step numbers, and will not be described unless necessary.
[0060]
First, it is determined whether or not the work completion 30 is accumulated in the completion queue 28 (S71). Then, the CPU 2 updates the value obtained by adding 1 to the semaphore variable of the work queue 24 as a semaphore variable (S73). Next, the value obtained by adding 1 to the semaphore variable in the completion queue 28 is updated as a new semaphore variable value (S74). This is because the completion queue 28 is vacated by collecting one work completion 30 in step S82.
[0061]
The semaphore variable indicating the free state of each queue is appropriately updated by the work completion 30 collection process of FIG. 9, and overflow of the work queue 24 and the completion queue 28 can be prevented. In addition, by updating the semaphore variable of the work queue 24 when the work completion 30 is collected, the present invention can be implemented regardless of the processing of the HCA coprocessor, which is easy to introduce and implements hardware. It does not need to be complicated.
[0062]
Note that the modification of the first embodiment described above can be applied to a configuration in which a plurality of work queues 24 and completion queues 28 correspond to each other as long as the work queue 24 and the completion queue 28 have a one-to-one correspondence.
[0063]
Next, a second embodiment of the present invention will be described. In the second embodiment, a plurality of work queues 24 are secured on the memory and one completion queue 28 is secured for the plurality of work queues 24 in the same configuration as in the first embodiment in FIG. It is. In the second embodiment of the present invention, a semaphore variable indicating the number of data (work request 26 and work completion 30) that can be stored in each queue is set for the work queue 24 and the completion queue 28. Thus, each queue is prevented from overflowing.
[0064]
FIG. 10 is a diagram showing an operation when the CPU 2 accesses data stored in the peripheral device in the second embodiment of the present invention. The same parts as those in FIG. 2 are given the same numbers, and the description thereof will be omitted if unnecessary.
[0065]
The CPU 2 in FIG. 10 uses data including processing requests from the OS and application programs as work requests 26, and each of the work queue numbers WQ1 to WQn secured in the memory 7 as areas specified by the start address and end address. n (n is a natural number) work queues 24 ₁ To work queue 24 _n To accumulate. The CPU 2 refers to the semaphore variable table 22 in the memory 5 in which semaphore variables set corresponding to each work queue 24 and a completion queue 28 described later are stored, and determines whether or not work requests 26 can be accumulated.
[0066]
If the work request 26 can be accumulated, the work request 26 is added to the end of the work queue 24. The CPU 2 stores information indicating in which work queue 24 the work request 26 is accumulated in the work queue 24 to work request 26 correspondence table 23 of the memory 5 and updates the semaphore variable table 22. Details of the work request 26 accumulation process will be described later.
[0067]
The coprocessor 8 retrieves the first work request 26 from the work queue 24 at a specific timing, executes a processing request, and transmits data to and from the TCA connected to the fabric network via the InfiniBand switch 9 according to the processing request content. To complete the work request 26. After processing the retrieved work request 26, the coprocessor 8 creates a work completion 30 and stores it in the completion queue 28. In general, the work queue 24 and the completion queue 28 do not need to correspond one-to-one, and a plurality of work queues 24 can be associated with one completion queue 28 as shown in FIG. is there.
[0068]
The CPU 2 collects the work completion 30 from the completion queue 28 at a specific timing, confirms that the corresponding work request 26 is completed, and completes the processing request. At the same time, the CPU 2 updates the semaphore variables in the semaphore variable table 22 in the memory 5. Details of the work completion collection process will be described later.
[0069]
In addition, although it is one in FIG. 2, it is also possible to employ a configuration having a plurality of completion queues 28. Each work queue 24 can be given a specific role. For example, WQ1 can be set to store work requests 26 for sending Infiniband messages, and WQ2 can be set to store work requests 26 for receiving Infiniband messages. The data structures of the work request 26, the work completion 30, and the semaphore variable table 22 are respectively as shown in FIGS. 3, 4, and 5, as in the first embodiment.
[0070]
FIG. 11 is a work queue to work request correspondence table 23 stored in the memory 5. Each entry in FIG. 11 stores a queue number 51 and a list 111 of work request numbers 31 accumulated in the work queue 24 specified by the queue number. With reference to this correspondence table, a list of work request 26 numbers accumulated in each work queue 24 can be obtained. Therefore, it is possible to specify in which work queue 24 a certain work request 26 is accumulated. In FIG. 11, for example, it can be seen that the work request 26 specified by the work request number (1, 12, 28) included in the list WRL1 is accumulated in the queue having the queue number WQ1.
[0071]
Next, work request accumulation processing and work completion in the second embodiment of the present invention will be described.
[0072]
FIG. 12 is a flowchart illustrating work request accumulation processing in the second embodiment of the present invention. The same steps as those in FIGS. 6 and 8 are given the same step numbers, and the description thereof will be omitted if not necessary. The maximum number of data that can be accumulated in each queue is set in advance in the semaphore variable table 22 as an initial value of the semaphore variable for each queue.
[0073]
First, the work queue 24 corresponding to the processing request from the OS or application program is determined (S61). The CPU 2 obtains an appropriate work queue 24 number of the work queue 24 according to the processing request content. For example, if the processing request is data transmission and the work queue 24 with the queue number WQ1 in FIG. 10 is assigned for data transmission, the work queue 24 with the queue number WQ1 is selected. The assignment of uses to each work queue 24 can be freely determined. Further, one work queue 24 may be arbitrarily selected unless a use is set for each work queue 24.
[0074]
Next, it is determined whether or not the semaphore variable of the work queue 24 determined in step S61 is a positive value (S62). The CPU 2 refers to the semaphore variable table 22 in the memory 5 and obtains the semaphore variable value of the work queue 24 determined in step S61.
[0075]
If the value of the semaphore variable is positive in step S62, the CPU 2 updates the semaphore variable table 22 with the value obtained by subtracting 1 from the semaphore variable in the work queue 24 determined in step S61 as a new semaphore variable (S63). If the value of the semaphore variable is positive in step S62, the work queue 24 is empty, and work requests 26 can be accumulated. Therefore, when the work request 26 is accumulated, one empty work queue is consumed.
[0076]
If it is determined in step S62 that there is no space, the process waits for a predetermined time and performs the process of step S62 again. In this way, it waits for the work queue 24 to be free.
[0077]
After step S63, it is determined whether the semaphore variable in the completion queue 28 is a positive value (S64). The CPU 2 refers to the semaphore variable table 22 and obtains the semaphore variable value of the completion queue 28. If the semaphore variable value is positive, the completion queue 28 is empty and the work completion 30 can be accumulated.
[0078]
If it is determined in step S64 that there is a free space, the CPU 2 updates the semaphore variable table 22 using a value obtained by subtracting 1 from the semaphore variable in the completion queue 28 as a new semaphore variable (S65). Although the semaphore variable value corresponding to the completion queue 28 should be updated after the coprocessor 8 completes the processing of the work request 26, the CPU 2 can know when the coprocessor 8 processes the work request 26. Can not. The work requests 26 stored in the work queue 24 always have a work completion 30 corresponding to the work request 26 stored in the completion queue 28 after the processing is completed, and correspond one-to-one. Therefore, it is assumed that one completion queue 28 has been consumed when the work request 26 is accumulated in the work queue 24, and the semaphore variable in the completion queue 28 is updated.
[0079]
If it is determined in step S64 that there is no free space, the process waits for a predetermined time and performs the process of step S64 again. Thus, it waits for the completion queue 28 to be empty.
[0080]
Thus, if both the work queue 24 and the completion queue 28 are free, the work request 26 is stored in the work queue 24 determined in step S61 (S66). Next, it is determined whether or not the accumulation process in step S66 is successful (S67). The work request 24 can be stored in the work queue 24. However, if the storage process fails due to power interruption of the apparatus or stop of the storage process due to the interrupt process during the storage process, the work request storage process is stopped as it is.
[0081]
If the work request 26 is successfully accumulated in step S67, the work request 26 accumulated in step S66 and the work queue 24 in which the work request 26 is accumulated are associated and recorded (S68), and the work request accumulation process is performed. finish. The work request number accumulated in step S64 may be added to the work request ID list corresponding to the queue number 51 of the work queue 24 in the work queue-to-work request correspondence table 23 stored in the memory 5.
[0082]
If it fails in step S67, 1 is added to each of the semaphore variables of the work queue 24 and the completion queue specified in step S61 (S69), and the work request accumulation process is terminated. In steps S63 and S65, the semaphore variable is updated on the assumption that the work requests are accumulated and one of the work queue 24 and the completion queue 28 is consumed, but the accumulation of work requests failed in step S67. The empty space in the work queue 24 and the completion queue 28 is not consumed. Accordingly, in step S69, the value subtracted in steps S63 and S65 is restored.
[0083]
The work request accumulation process in FIG. 12 appropriately updates the semaphore variable indicating the free state of each queue, and when there is no free space in the work queue, the work request 26 is not added. Similarly, the work request 26 is not added when there is no space in the completion queue. Since the work request 26 and the work completion 30 correspond one-to-one, the completion queue 28 does not overflow due to the added work request 26. Thus, overflow of the work queue 24 and the completion queue 28 can be prevented. In addition, by updating the semaphore variable in the completion queue 28 when the work request 26 is accumulated, the present invention can be implemented regardless of the processing of the HCA coprocessor, which is easy to introduce and implements hardware. It does not need to be complicated.
[0084]
FIG. 13 is a flowchart illustrating the work completion collection process according to the second embodiment of the present invention. First, it is determined whether or not the work completion 30 is accumulated in the completion queue 28 (S71). The CPU 2 takes out the work completion 30 from the head of the completion queue 28, and if the completion queue 28 is empty, the processing is ended as it is.
[0085]
Next, the work queue 24 in which the work requests 26 corresponding to the work completion 30 are stored is identified from the extracted work completion 30 (S72). The work completion 30 is created corresponding to the work request 26 that has been processed, and the processed work request number 41 is stored in the work completion 30 as shown in FIG. ing. The CPU 2 refers to the work queue-to-work request correspondence table 23 in the memory 5, searches the work request ID list 111 including the obtained work request number 41, and obtains the corresponding queue number 51.
[0086]
Then, the CPU 2 updates the value obtained by adding 1 to the semaphore variable of the work queue 24 specified in step S72 as a semaphore variable (S73). This is because the processing of the work request 26 has been completed, and it has been confirmed that one free space has been created in the work queue identified in step S72. Although the semaphore variable value corresponding to the work queue 24 should be updated when the coprocessor 8 completes the processing of the work request 26, the CPU 2 can know when the coprocessor 8 processes the work request 26. Can not. Therefore, when the work completion 30 is collected, it is considered that the work queue 24 is empty, and the semaphore variable is updated.
[0087]
Next, the value obtained by adding 1 to the semaphore variable in the completion queue 28 is updated as a new semaphore variable value (S74). This is because the completion queue 28 is freed by collecting one work completion 30 in step S72.
[0088]
With the work completion collection process of FIG. 13, the semaphore variables representing the free state of each queue are appropriately updated, and overflow of the work queue 24 and the completion queue 28 can be prevented. In addition, by updating the semaphore variable of the work queue 24 when collecting work completion, the present invention can be implemented regardless of the processing of the HCA coprocessor, which is easy to introduce and complicated to implement hardware. It is not necessary to make it.
[0089]
In the second embodiment, the size of the completion queue 28 is set in advance so that the number of data that can be stored in the completion queue 28 is equal to or greater than the sum of the maximum number of data that can be stored in each work queue 24. The present invention can be applied using only the semaphore variables set for the work queue 24 without setting the semaphore variables for the completion queue 28.
[0090]
That is, even if all the work requests 26 accumulated in each work queue 24 at a certain point in time are transferred to the completion queue 28, it is ensured that the completion queue 28 does not overflow. This is because it is only necessary to prevent overflow of the work queue 24 when the work requests 26 are accumulated in the work queue 24.
[0091]
The present invention can also be applied to inter-process communication between blades via a server-to-server, server-to-storage, blade server backplane, or inter-process communication within the same server. Further, it can be incorporated as a function of the OS or application program as a program, or provided as middleware that mediates between the OS and the application program.
[0092]
The above is summarized as an appendix.
[0093]
(Additional remark 1) The process of setting the 1st area | region where process request information is accumulate | stored, the process of setting the 2nd area | region where process completion information which is the information which shows completion of a process request is accumulated, said 1st A step of setting the initial value of the semaphore variable corresponding to the first area as the maximum number of information that can be stored in the first area, and a value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area If it is a positive value, if it is a positive value, the processing request information is added to the first area, and a value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area Updating as a semaphore variable corresponding to the area, and when the requested process is completed, associating information for identifying the process request and the process completion information in the second area, The processing completion information is accumulated in the second area. If it has been accumulated, the value obtained by adding 1 to the semaphore variable corresponding to the first area is updated as the semaphore variable corresponding to the first area, and one piece of the processing completion information is deleted An information storage control method comprising the step of:
[0094]
(Supplementary note 2) The information accumulation control method according to supplementary note 1, wherein the maximum number of pieces of information that can be stored in the second area is equal to or greater than the maximum number of pieces of information that can be stored in the first area. .
[0095]
(Additional remark 3) The process of setting the 1st area | region where process request information is accumulate | stored, the process of setting the 2nd area | region where process completion information which is the information which shows completion of a process request is accumulated, said 1st Setting an initial value of a semaphore variable corresponding to each of the first region and the second region as a maximum number of information that can be stored in the first region and the second region, respectively, and the first region And the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the second area are both positive values. If both are positive values, A value obtained by adding processing request information to the first area and subtracting 1 from the value of the semaphore variable corresponding to the first area as a semaphore variable corresponding to the first area, and the second area 1 from the value of the semaphore variable corresponding to A step of updating the subtracted value as a semaphore variable corresponding to the second area, and when the processing request is completed, information specifying the processing request is associated with the processing completion information in the second area. And the process completion information is stored in the second area, and if it is stored, a value obtained by adding 1 to the semaphore variable corresponding to the first area is determined. The value obtained by adding 1 to the value of the semaphore variable corresponding to the second area is updated as the semaphore variable corresponding to the second area, and the processing completion information is set to 1 An information storage control method comprising a step of deleting two.
[0096]
(Additional remark 4) The process of setting the some 1st area | region where process request information is accumulate | stored,
A step of setting a second area in which processing completion information, which is information indicating completion of a processing request, is stored; and an initial value of a semaphore variable corresponding to each of the plurality of first areas Setting the maximum number of pieces of information that can be stored in each of them, selecting one of the plurality of first areas, and subtracting 1 from the value of the semaphore variable corresponding to the selected first area If the selected value is a positive value, if it is a positive value, processing request information is added to the selected first area, and information for specifying the selected first area and the processing request is added. Updating the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the selected first area as the semaphore variable corresponding to the selected first area; When the processing request is completed, the processing request is A step of storing information to be determined in association with the processing completion information in the second area, and determining whether the processing completion information is stored in the second area. Read one information, identify the processing request corresponding to the read processing completion information, identify the first area where the identified processing request is stored with reference to the correspondence table, Updating a value obtained by adding 1 to the semaphore variable corresponding to the identified first area as a semaphore variable corresponding to the identified first area, and deleting the read processing completion information; An information storage control method comprising:
[0097]
(Supplementary Note 5) In Supplementary Note 4, the maximum number of information that can be stored in the second region is equal to or greater than a sum of the maximum number of information that can be stored in the plurality of first regions. Information storage control method.
[0098]
(Additional remark 6) The process of setting the some 1st area | region where process request information is accumulate | stored, the process of setting the 2nd area | region where the process completion information which is the information which shows completion of a process request is accumulated, The initial value of the semaphore variable corresponding to each of the first region and the second region of the first region is set as the maximum number of information that can be stored in each of the first region and the second region. Selecting one of the plurality of first regions, subtracting 1 from the value of the semaphore variable corresponding to the selected first region, and the semaphore corresponding to the second region. It is determined whether the value obtained by subtracting 1 from the variable value is a positive value. If both are positive values, processing request information is added to the selected first area, and the selected first area And the information specifying the processing request are stored in the correspondence table in association with each other. The value obtained by subtracting 1 from the value of the semaphore variable corresponding to the selected first area is used as the semaphore variable corresponding to the selected first area, and the value of the semaphore variable corresponding to the second area is A step of updating a value obtained by subtracting 1 from a value as a semaphore variable corresponding to the second area, and when the processing request is completed, information specifying the processing request is associated with the processing completion information A step of accumulating in the second area, and determining whether or not the process completion information is accumulated in the second area, and if it is accumulated, one of the process completion information is read and the read process completion is completed The processing request corresponding to the information is identified, the first area where the identified processing request is stored is identified with reference to the correspondence table, and the semaphore variable corresponding to the identified first area Add 1 to And the value obtained by adding 1 to the semaphore variable corresponding to the second area is updated as the semaphore variable corresponding to the second area. An information storage control method comprising a step of deleting the read processing completion information.
[0099]
(Additional remark 7) The 1st area | region where process request information is accumulate | stored, the 2nd area | region where process completion information which is the information which shows completion of a process request is matched with the information which identifies the said process request, and is accumulate | stored, A program that is executed in an apparatus having an accumulation unit that stores information specifying the first area and a semaphore variable table in which semaphore variables are associated and accumulated, and a control unit that controls the accumulation unit. And causing the control unit to set an initial value of a semaphore variable corresponding to the first area as the maximum number of information that can be stored in the first area, and a semaphore corresponding to the first area. Determine whether the value obtained by subtracting 1 from the variable value is a positive value. If the value is positive, the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area corresponds to the first area. As the semaphore variable Updating the fore variable table, adding the processing request information to the first area, and determining whether the processing completion information is stored in the second area. A program for executing a step of updating the semaphore variable table with a value obtained by adding 1 to a semaphore variable corresponding to the first area as a semaphore variable corresponding to the first area, and deleting one piece of the processing completion information.
[0100]
(Additional remark 8) The 1st area | region where process request information is accumulate | stored, the 2nd area | region where process completion information which shows completion of a process request is matched and stored with the information which specifies the said process request, said 1st A program that is executed in an apparatus having a storage unit that stores information for specifying a region and a semaphore variable table in which a semaphore variable is associated and stored, and a control unit that controls the storage unit, Causing the control unit to set an initial value of a semaphore variable corresponding to each of the first area and the second area as a maximum number of pieces of information that can be stored in the first area and the second area, respectively. And whether the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area and the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the second area are both positive values. Even if the value of For example, the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area is used as the semaphore variable corresponding to the first area, and 1 is subtracted from the value of the semaphore variable corresponding to the second area. Updating the semaphore variable table as a semaphore variable corresponding to the second area and adding the processing request information to the first area, and the processing completion information in the second area. If it is accumulated, a value obtained by adding 1 to the semaphore variable corresponding to the first area is used as the semaphore variable corresponding to the first area, and also corresponds to the second area. Executing the step of updating the semaphore variable table with a value obtained by adding 1 from the value of the semaphore variable to be performed as a semaphore variable corresponding to the second area, and deleting one piece of the processing completion information A program for causing.
[0101]
(Supplementary Note 9) A plurality of first areas in which process request information is stored, a second area in which process completion information indicating completion of a process request is stored in association with information identifying the process request, A program executed in an apparatus having an accumulation unit that stores information for specifying a first area and a semaphore variable table in which semaphore variables are associated and accumulated, and a control unit that controls the accumulation unit. The initial value of the semaphore variable corresponding to each of the first region and the second region of the plurality of first regions can be stored in the control unit in the first region and the second region, respectively. A step of setting the maximum number of pieces of information, selecting one of the plurality of first regions, and subtracting 1 from the value of the semaphore variable corresponding to the selected first region, Corresponding to the second region If the values obtained by subtracting 1 from the value of the semaphore variable are both positive values, and if both are positive values, the selected first area and the processing request are associated with each other and accumulated in the correspondence table, A value obtained by subtracting 1 from the value of the semaphore variable corresponding to the selected first area is used as a semaphore variable corresponding to the selected first area, and the value of the semaphore variable corresponding to the second area Updating the semaphore variable table with a value obtained by subtracting 1 from the semaphore variable corresponding to the second area, and adding the processing request information to the selected first area; It is determined whether the processing completion information is accumulated in the area. If it is accumulated, one processing completion information is read, the processing request corresponding to the read processing completion information is identified, and the identified First The value obtained by adding 1 to the semaphore variable corresponding to the area is used as the semaphore variable corresponding to the specified first area, and the value obtained by adding 1 from the value of the semaphore variable corresponding to the second area is the first value. A program for executing a process of updating the semaphore variable table as a semaphore variable corresponding to a second area and deleting one piece of the processing completion information.
[0102]
(Additional remark 10) The 1st area | region where process request information is accumulate | stored, and the 2nd area | region where process completion information which is the information which shows completion of a process request is matched with the information which specifies the said process request, and are accumulate | stored. A storage unit to be stored; a first device having a control unit for controlling the storage unit and executing the processing request; and information and a semaphore connected to the first device to identify the first region. A storage unit storing a semaphore variable table in which variables are associated and stored; and a second device having a control unit that controls the storage unit and transmits the processing request information to the first device. A control unit in the second device sets in advance an initial value of a semaphore variable for the first area as a maximum number of information that can be stored in the first area; From the semaphore variable corresponding to the region It is determined whether the value obtained by subtracting 1 is a positive value. If the value is positive, the value obtained by subtracting 1 from the semaphore variable corresponding to the first area is updated as the semaphore variable corresponding to the first area. , Transmitting new process request information to the first apparatus, and the control unit in the first apparatus receives the new process request information and adds it to the first area, and at a specific timing, The processing request stored in the first area is executed, processing completion information corresponding to the processing request is created and added to the second area, and the control unit in the second device is configured at a specific timing. It is determined whether the processing completion information is accumulated in the second area, and if it is accumulated, a value obtained by adding 1 to the semaphore variable corresponding to the first area is set as the semaphore corresponding to the first area. Updated as a variable and accumulated in the second area System characterized in that to remove a said processing completion information.
[0103]
(Additional remark 11) The 1st area | region where process request information is accumulate | stored, and the 2nd area | region where process completion information which shows completion of a process request are matched with the information which specifies the said process request, and are accumulate | stored are stored. A first device having a storage unit, a control unit that controls the storage unit and executes the processing request, and is connected to the first device and identifies the first region and the second region A second storage unit that stores a semaphore variable table in which information and a semaphore variable are associated and stored; and a control unit that controls the storage unit and transmits the processing request information to the first device. And the control unit in the second device sets the initial value of the semaphore variable for the first region as the maximum number of information that can be stored in the first region in advance for the second region. The initial value of the semaphore variable is Set as the maximum number of information that can be stored in the area, and subtract 1 from the semaphore variable corresponding to the first area and 1 from the semaphore variable corresponding to the second area at a specific timing If both values are positive, and if both are positive, the value obtained by subtracting 1 from the semaphore variable corresponding to the first area is used as the semaphore variable corresponding to the first area, and A value obtained by subtracting 1 from the semaphore variable corresponding to the second area is updated as a semaphore variable corresponding to the second area, and new processing request information is transmitted to the first device, and the first The control unit in the apparatus receives the new process request information, adds it to the first area, executes the process request stored in the first area at a specific timing, and responds to the process request Processing completion information Create and send to the second device, the control unit in the second device receives the processing completion information and adds it to the second region, the specific region to the second region It is determined whether processing completion information is accumulated. If accumulated, a value obtained by adding 1 to the semaphore variable corresponding to the first area is used as the semaphore variable corresponding to the first area, and the second Updating a value obtained by adding 1 to the value of the semaphore variable corresponding to the second area as a semaphore variable corresponding to the second area, and deleting one piece of the processing completion information accumulated in the second area A system characterized by
[0104]
(Additional remark 12) The some 1st area | region where process request information is accumulate | stored, and the 2nd area | region where process completion information which shows completion of a process request are matched with the information which specifies the said process request, and are accumulate | stored And a first device having a control unit for controlling the storage unit and executing the processing request, and each first of the plurality of first regions connected to the first device. And a semaphore variable table in which information specifying the second area and the semaphore variable are associated and stored, information specifying the processing request, and the processing request among the plurality of first areas are stored. A storage unit that stores a correspondence table in which information specifying the first area is associated and stored, and a control unit that controls the storage unit and transmits the processing request information to the first device A second device comprising: the second device The control unit in the information in which the initial values of the semaphore variables respectively corresponding to the first area and the second area of the plurality of first areas can be stored in the first area and the second area, respectively. Set as the maximum number of numbers, select one of the plurality of first regions at a specific timing, and subtract 1 from the semaphore variable corresponding to the selected first region and the second It is determined whether the value obtained by subtracting 1 from the semaphore variable corresponding to the area is positive, and if both are positive values, the value obtained by subtracting 1 from the semaphore variable corresponding to the selected first area is determined. Update the value obtained by subtracting 1 from the semaphore variable corresponding to the second region as the semaphore variable corresponding to the second region, and the value obtained by subtracting 1 from the semaphore variable corresponding to the second region, respectively. New to one device Processing request information is transmitted, the selected first area and the processing request are associated with each other and accumulated in a correspondence table, and the control unit in the first device receives the new processing request information. Accumulating in the first area, executing the processing request stored in the first area at a specific timing, creating processing completion information corresponding to the processing request and transmitting to the second device, The control unit in the second device receives the processing completion information and accumulates it in the second area, determines whether the processing completion information is accumulated in the second area at a specific timing, and accumulates it. If so, a value obtained by adding 1 to the semaphore variable corresponding to the first area is set as a semaphore variable corresponding to the first area, and 1 is calculated from the value of the semaphore variable corresponding to the second area. The added value corresponds to the second area That is updated each as a semaphore variable, and characterized in that to remove a stored the processing completion information to the second region system.
[0105]
(Appendix 13) In Appendix 10,
The system, wherein the first device is a channel adapter provided in a server machine, and the second device is a channel adapter provided in a network interface card.
[0106]
(Appendix 14) In Appendix 10,
A system characterized in that both the first device and the second device are host channel adapters provided in a server machine.
[0107]
【The invention's effect】
As described above, according to the present invention, when processing requests are executed by interposing two types of queues as interfaces, the data related to the processing requests can be stored in the queues without overflowing each queue. Can be prevented from being stopped or delayed due to overflow. Further, the present invention only needs to be implemented in the processing request issuing side device, and if the issuing side device is a device capable of executing a program, it can be implemented by software.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating a configuration example of an information processing apparatus used in a first embodiment of the present invention.
FIG. 2 is a diagram illustrating an operation when a CPU accesses data stored in a peripheral device in the first embodiment of the present invention.
FIG. 3 is a diagram illustrating a data configuration example of work requests accumulated in a work queue.
FIG. 4 is a diagram illustrating a data configuration example of work completion accumulated in a completion queue.
FIG. 5 is a semaphore variable table stored in a memory.
FIG. 6 is a flowchart illustrating work request accumulation processing according to the first embodiment of the present invention.
FIG. 7 is a flowchart illustrating work completion collection processing according to the first embodiment of the present invention.
FIG. 8 is a flowchart illustrating work request accumulation processing in a modification of the first embodiment.
FIG. 9 is a flowchart illustrating work completion collection processing according to a modification of the first embodiment.
FIG. 10 is a diagram showing an operation when a CPU accesses data stored in a peripheral device in the second embodiment of the present invention.
FIG. 11 is a work queue to work request correspondence table stored in a memory.
FIG. 12 is a flowchart illustrating work request accumulation processing according to the second embodiment of the present invention.
FIG. 13 is a flowchart illustrating a work completion collection process according to the second embodiment of the present invention.
[Explanation of symbols]
1 Computer, 2 CPU, 3 System bus, 4 Memory controller, 5 Memory, 6 Host channel adapter, 7 Memory, 8 Coprocessor, 9 Infiniband switch, 10 Network interface card, 12 SCSI card, 14 Fiber channel card, 11, 13, 15 Target channel adapter, 24 work queues, 26 work requests, 28 completion queues, 30 work completions

Claims (6)

Process request information is accumulated in the first area by the first processor, and process completion information indicating that the process executed in accordance with the process requests accumulated in the first area is completed by the second processor. An information storage control method for storing in a second area,
By the first processor,
The maximum number of information that can be stored in the second area is set to be greater than or equal to the maximum number of information that can be stored in the first area, and the initial value of the semaphore variable corresponding to the first area is A first step of setting as a maximum number of information that can be stored in the first area ;
After the first step, it is determined whether the semaphore variable corresponding to the first region is a positive value. If the semaphore variable is a positive value, the processing request information is added to the first region, and the first step and as the second factory to update a value obtained by subtracting 1 from the value of the semaphore variable corresponding to an area as a semaphore variable corresponding to said first region,
Furthermore, after the first step, wherein the processing completion information is determined whether it is accumulated in the second region, if it is accumulated, by adding 1 to the semaphore variable corresponding to the first region updates the value as a semaphore variable corresponding to said first region, performs a third step to delete one said processing completion information from said second region,
The process completion information is associated with the process request information when the process completion information indicating that the process executed according to the process request information is completed by the second processor is stored in the second area. information storage control method and performing as a fourth factory you accumulated in the second region Te. Process request information is accumulated in the first area by the first processor, and process completion information indicating that the process executed in accordance with the process requests accumulated in the first area is completed by the second processor. An information storage control method for storing in a second area,
By the first processor,
The maximum number of information that can be stored in the second area is set to be equal to or greater than the maximum number of information that can be stored in the first area, and corresponds to the first area and the second area, respectively. A first step of setting an initial value of a semaphore variable as the maximum number of pieces of information that can be accumulated in the first area and the second area ,
After the first step, it is determined whether the semaphore variable corresponding to the first area and the semaphore variable corresponding to the second area are both positive values. The processing request information is added to the area, and the value obtained by subtracting 1 from the value of the semaphore variable corresponding to the first area is used as the semaphore variable corresponding to the first area, and also corresponds to the second area. and as the second factory to update each a value obtained by subtracting 1 from the value of the semaphore variable as a semaphore variable corresponding to the second region,
Furthermore, after the first step, wherein the processing completion information is determined whether it is accumulated in the second region, if it is accumulated, by adding 1 to the semaphore variable corresponding to the first region The value is updated as a semaphore variable corresponding to the first area, and a value obtained by adding 1 from the value of the semaphore variable corresponding to the second area is updated as a semaphore variable corresponding to the second area, and the process completion information from the second area subjected to the third step to delete one,
The process completion information is associated with the process request information when the process completion information indicating that the process executed according to the process request information is completed by the second processor is stored in the second area. information storage control method and performing as a fourth factory you accumulated in the second region Te. The first processor accumulates the processing request information in any one of the plurality of first areas, and the processing completion information indicating that the processing executed in accordance with the processing requests stored in the first area is completed. An information accumulation control method for accumulating in a second area by two processors,
By the first processor,
The maximum number of information that can be stored in the second area is set to be equal to or greater than the sum of the maximum number of information that can be stored in the first area, and the semaphore variables respectively corresponding to the plurality of first areas A first step of setting an initial value as the maximum number of pieces of information that can be accumulated in each of the plurality of first regions ;
After the first step, select one of the plurality of first regions, determine whether the semaphore variable corresponding to the selected first region is a positive value, and if it is a positive value For example, processing request information is added to the selected first area, the selected first area and the processing request information are associated with each other and stored in a correspondence table, and the selected first area is stored in the selected first area. and as the second factory to update the obtained by subtracting 1 from the value of the corresponding semaphore variable value as a semaphore variable corresponding to the selected first region,
Furthermore, after the first step, the process completion information is determined whether the stored, if stored, reads one said processing completion information, corresponding to the read processing completion information the The processing request is specified, the first area in which the specified processing request is stored is specified with reference to the correspondence table, and 1 is added to the semaphore variable corresponding to the specified first area. updates the value as a semaphore variable corresponding to the first region the identified performs the third step to delete the read processing completion information from said second region,
The process completion information is associated with the process request information when the process completion information indicating that the process executed according to the process request information is completed by the second processor is stored in the second area. information storage control method and performing as a fourth factory you accumulated in the second region Te. The first processor accumulates the processing request information in any one of the plurality of first areas, and the processing completion information indicating that the processing executed in accordance with the processing requests stored in the first area is completed. An information accumulation control method for accumulating in a second area by two processors,
By the first processor,
The maximum number of information that can be stored in the second area is set to be equal to or greater than the sum of the maximum number of information that can be stored in the first area, and the first area and the first area of the plurality of first areas A first step of setting an initial value of a semaphore variable corresponding to each of the two regions as a maximum number of information that can be stored in each of the first region and the second region ;
After the first step, one of the plurality of first regions is selected, and a semaphore variable corresponding to the selected first region and a semaphore variable corresponding to the second region are both positive. If both are positive values, processing request information is added to the selected first area, and the selected first area is associated with the processing request information. A value obtained by subtracting 1 from the value of the semaphore variable corresponding to the selected first area as a semaphore variable corresponding to the selected first area and corresponding to the second area. and as the second factory to update each a value obtained by subtracting 1 from the value of the semaphore variable as a semaphore variable corresponding to the second region,
Furthermore, after the first step, wherein the processing completion information is determined whether it is accumulated in the second region, if stored, reads one said processing completion information was the read processing The processing request corresponding to the completion information is identified, the first area in which the identified processing request is stored is identified with reference to the correspondence table, and the semaphore corresponding to the identified first area is identified. A value obtained by adding 1 to a variable is used as a semaphore variable corresponding to the specified first area, and a value obtained by adding 1 to a semaphore variable corresponding to the second area is used as a semaphore corresponding to the second area. update each as a variable, performs the third step to delete the second the read processing completion information from the area of,
The process completion information is associated with the process request information when the process completion information indicating that the process executed according to the process request information is completed by the second processor is stored in the second area. information storage control method and performing as a fourth factory you accumulated in the second region Te. A first area in which processing request information is stored, a second area in which processing completion information indicating that processing executed in accordance with the processing request information is completed, and a semaphore corresponding to the first area An accumulator that stores variables;
A program to be executed in an apparatus having said storage unit controlling to that control section a,
Before SL control section, the second the maximum number of storable information area, is set to more than the maximum number of storable information number to the first region, corresponding to the first region A first step of setting an initial value of a semaphore variable as the maximum number of pieces of information that can be stored in the first area ;
After the first step, the semaphore variable corresponding to the first region is determined whether a positive value, if a positive value, obtained by subtracting 1 from the value of the semaphore variable corresponding to the first region wherein the value as a semaphore variable corresponding to said first region to update the semaphore variable table, as the second factory to the to the first region Ru is added to the processing request information,
Furthermore, after the first step, the second area is determined whether the process completion information is stored in, if it is accumulated, by adding 1 to the semaphore variable corresponding to the first region program for the value as a semaphore variable corresponding to said first region to update the semaphore variable table, to execute a third step of Ru is removed one the process completion information from said second region. A first storage unit that stores a first region in which processing request information is stored and a second region in which processing completion information indicating that processing executed in accordance with the processing request information has been completed is stored; A first device having a first control unit for controlling the first storage unit;
A second accumulator connected to the first device and storing a semaphore variable corresponding to the first area; and controlling the second accumulator and processing request information to the first device And a second device having a second control unit for transmitting to
The second control unit in the second unit, the first to the second maximum number of the number which can be stored information in the region, set to more than the maximum number of storable information number to the first region And the process of
After the first step, the initial value of the semaphore variable corresponding to the first area is set as the maximum number of information that can be stored in the first area, and the first area is set at a specific timing. It is determined whether the corresponding semaphore variable is a positive value, and if it is a positive value, a value obtained by subtracting 1 from the semaphore variable corresponding to the first area is updated as a semaphore variable corresponding to the first area, and as the second factory that sends a new processing request information to the first device,
Furthermore, after said first step, it is determined whether the processing completion information to the second region at a timing of specific are stored, if it is accumulated, the semaphore variable corresponding to said first region to a value obtained by adding 1 to update the semaphore variable corresponding to said first region, performs a third step to delete one said processing completion information from said second region,
First control portion before Symbol first device receives the new processing request information stored in said first region,
When the process completion information indicating that the process executed according to the process request information is completed is stored in the second area, the process request information and the process completion information are associated with each other at a specific timing. system characterized that you perform a fourth step of storing the region.

Images