JPH1196061A - Exclusive controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an exclusive controller capable of shortening time required for the retrieval of management information even when the controller has a generation number management function. SOLUTION: When block information corresponding to a lock request is not included in a block information string at the time of registering block information corresponding to the request outputted from a host computer, block information corresponding to the request is newly prepared and registered in the top of the block information string (step 109). When block information corresponding to the request exists in the string at the time of registering the block information corresponding to the request, the block information is connected to the top of the string and reregisters the information (step 110).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an exclusive control device, and more particularly to an exclusive control device for a shared resource in a loosely coupled system including a plurality of host systems.

[0002]

2. Description of the Related Art A conventional exclusive control apparatus is disclosed in, for example, Japanese Patent Application Laid-Open Nos. 2-194442 and 8-305.
In a system in which a plurality of host computers (hereinafter, referred to as a host) share resources such as an external storage device (hereinafter, referred to as a shared resource) as described in Japanese Patent No. 622, exclusive control of reference / update access between the hosts with respect to the shared resource is performed. It is provided in order to perform.

When a host accesses a shared resource, the host reads a data portion to be accessed into its own local memory and performs reference or update. When updating is performed, the updated data portion is written back to the shared resource.

If exclusive control is not performed, for example, when two hosts try to perform update access to the same data of a shared resource at the same time, each of them will read, perform different updates, and write back again. That is, one of the update data is overwritten. If the process proceeds without noticing this, data in the system will be destroyed.

The exclusive control aims at preventing data destruction due to the above-mentioned simultaneous update by sequentially making accesses to the same data of a shared resource by a plurality of hosts.

In the exclusive control method, when each host accesses data, an access permission request (hereinafter referred to as a lock request) for the data is issued to the exclusive control device before the access, and a permission notification is issued from the exclusive control device. Upon receipt, the data is accessed, and upon completion of the access, a release request (hereinafter referred to as an unlock request) is made to the exclusive control device.

The exclusive control device manages the lock state of the data of the shared resource based on the lock request / unlock request from the host, and upon receiving the lock request, refers to its own management information, and the other host transfers the data to another host. If not locked, a lock permission is notified, and if another host is locked, a lock disable is notified. Furthermore, when the lock is notified, the fact that there is a host waiting for the lock permission is recorded in the management information, and when the lock is unlocked, the host that is waiting for the lock permission is notified of the lock permission. I do. According to the above method, access to the same data of the shared resource is serialized.

Furthermore, some exclusive control devices have a generation management function.

When a certain host accesses the same data again after a certain time after reading the data of the shared resource in its own local memory and performing the processing, another host must update the data during that time. In this case, there is no need to read the data to the local memory again.

Therefore, the generation number is assigned to the data, and the generation number is incremented by one every time the data is updated, so that whether the data held in the local memory matches the data existing in the shared resource is determined by the generation number. to decide. The exclusive control device manages the latest generation number of the data in the shared resource. When receiving the lock request, the exclusive control device notifies the current generation number together with the lock permission notification.
When an unlock request is received, the generation number held by the request is incremented by one.

In realizing the above functions, a common exclusive control device defines a shared resource as a group of a plurality of files, further divides each file into a plurality of blocks, and defines a minimum unit for performing the exclusive control in block units. And Lock / unlock requests are made in units of tasks (also called processes) operating on the host.

[0012]

The above functions are unnecessary functions in a single host system. However, it is indispensable to use shared resources by a plurality of hosts, and it is necessary to issue a lock / unlock request to the exclusive control device every time data of the shared resources is accessed. The performance for the access to the device is reduced. Therefore, the time required for processing by the exclusive control device needs to be reduced as much as possible.

A bottleneck in the processing performance of the exclusive control device is a management information search process. The more information to manage, the more time it takes to retrieve the desired management information for each request.

Particularly, in an exclusive control device having a generation number management function, since only a large number of block information remains in the exclusive control management table even if it is not locked for generation number management, There is a disadvantage that the possibility that it takes a long time to search for the management information is further increased.

The present invention has been made in view of the above points, and it is an object of the present invention to provide an exclusive control device capable of reducing the time required for searching for management information even when having a generation number management function.

[0016]

According to a first aspect of the present invention, there is provided an exclusive control apparatus wherein a plurality of host computers define a shared resource as a set of a plurality of files in a system in which a shared resource such as an external storage device is shared. In an exclusive control device that divides each file into a plurality of blocks and performs exclusive control on a block-by-block basis in a task unit of the host computer, management of generation numbers indicating update information of the blocks and lock information indicating a lock state are performed. A management table for connecting and registering information of a file being managed for one file as a file information string for management with a pointer in a row and connecting and registering information of a block being managed for each file with a pointer as a block information string for each file in a row; When the lock information for the lock request from the host computer is registered, the lock information corresponding to the request is included in the block information sequence. When there is no block information, block information corresponding to the request is newly created and registered at the head of the block information sequence, and when the lock information for the lock request is registered, the block corresponding to the request is included in the block information sequence. And lock processing means for re-registering the block information at the head of the block information sequence when the information exists.

The exclusive control device according to a second aspect of the present invention is the exclusive control device according to the first aspect, wherein the lock processing means includes a file corresponding to the request in a file information sequence when registering lock information for a lock request from a host computer. If there is no such information, a new file information corresponding to the request is created and registered at the head of the file information sequence.

The exclusive control device according to a third aspect of the present invention is the exclusive control device according to the first or second aspect, wherein when registration of lock information for an unlock request from a host computer is deleted, block information corresponding to the request is generated. When the block information is a block number to be managed, the generation number of the block information is updated by +1 and stored in a block information sequence. When the block information corresponding to the request is block information not to be managed by the generation number, the block number is updated. When there is no other lock information that locks the block information, there is provided an unlock processing means for deleting this block information.

The exclusive control device according to a fourth aspect of the present invention is the exclusive control device according to the third aspect of the present invention, wherein the unlock processing means deletes block information corresponding to the request when deleting the registration of the lock information in response to the unlock request from the host computer. And if there is no other block information in the block information sequence of the file information corresponding to the request, the file information is deleted.

[0020]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 4 is a block diagram showing an embodiment of the exclusive control device according to the present invention. FIG. 4A shows a management table structure of file entries and task entries in the exclusive control device, and FIG. 3 shows a management table structure.

The management table has entries indicating file information (file entries 10-1 to 10-n) and entries indicating task information (task entries 20-1 to 20- 20).
-N), entries indicating block information (block entries 30-1 to 30-n), and entries indicating lock information (lock entries 40-1 to 40-n), and any of the above. Unused entries that are not used as entries (empty entries 50-1 to 50-5)
0-n). The number of entries is all n, but the present invention is not limited to this.

The empty entries are linked by a pointer (empty entry column), and are managed by an empty entry management area 50 (FIG. 4B) indicating the first entry and the last entry.

The currently registered file entries are connected by a pointer (file entry sequence), and are managed by a file entry management area 10 (FIG. 4 (a)) indicating the head entry and the tail entry.

The currently registered task entries are connected by a pointer (task entry sequence), and the task entry management area 20 indicating the first entry and the last entry.
(FIG. 4A).

The currently registered block entry is linked by a pointer to a file entry corresponding to the file having this block. When a plurality of blocks exist under the same file, a plurality of block entries are connected by a pointer (block entry sequence), and the file entry indicates the first entry and the last entry.

A locked block entry and a task entry are connected by a pointer via a lock entry, and manages which task locks which block. When one task locks a plurality of blocks, each lock entry is connected by a pointer (lock entry sequence), and the task entry indicates the first entry and the last entry.

FIG. 5 shows the structure of each entry and each entry management area.

The file entry 10-n includes a file identifier area, a head block entry pointer, a tail block entry pointer, a next entry pointer and a previous entry pointer for forming a file entry sequence (FIG. 5 ( a)).

The task entry 20-n has a host identifier area, a task identifier area, a head lock entry pointer, a tail lock entry pointer, and a next entry pointer and a previous entry pointer for forming a task entry sequence. (FIG. 5 (b)).

The block entry 30-n includes a block identifier area, a generation number area, a lock entry pointer, a lock waiting entry pointer, and a next entry pointer and a previous entry pointer for forming a block entry sequence ( FIG. 5 (c)).

The lock entry 40-n has a block entry pointer, a task entry pointer, and a next entry pointer and a previous entry pointer for forming a lock entry sequence (FIG. 5D).

The empty entry 50-n has a next entry pointer and a previous entry pointer for forming an empty entry string (FIG. 5 (e)).

The empty entry management area 50 has a head entry pointer and an end entry pointer of the empty entry row (FIG. 5 (h)).

The file entry management area 10 has a head entry pointer and a tail entry pointer of the file entry string (FIG. 5 (f)).

The task entry management area 20 has a head entry pointer and a tail entry pointer of the task entry sequence (FIG. 5 (g)).

FIG. 6 shows the transition of the management state of the block entry by the file entry in FIG. 4 for explaining the embodiment of the present invention. FIG. 4 shows only the management state of the locked block entry, but FIG. 6 also shows the management state of the unlocked block entry. The need for management in unlocked blocks is for generation number management.

(A) shows a management state of 10 entries, (b) shows a state where one entry is newly added from (a), and (c) shows a state where an existing entry is re-registered. It is a figure showing a state. For illustrative purposes, the numbers above the block entries indicate block identifiers. “L” below each block entry indicates that the block entry has a lock entry.

FIGS. 7 and 8 show a lock management state in a certain block.

FIG. 7 shows a state in which the block entry 30-1 under the file entry 10-1 is locked by the task entry 20-1, and both the block entry 30-1 and the task entry 20-1 are locked. -1
Is pointing to.

FIG. 8 shows file entry 1 similar to FIG.
Although the task entry 20-1 locks the block entry 30-1 under 0-1, the block entry 30-1 and the task entry 20- are also locked with the task entry 20-2 waiting for the lock of this block. 2 both point to lock entry 40-2. The block entry 30-1 points to the lock entry 40-1 by its own lock entry pointer and the lock entry 40-2 by the entry pointer having lock.

Registration or deletion of each entry is caused by a lock request and an unlock request.
At that time, the identifiers of the host, task, file, and block corresponding to the lock and unlock are included in the request content and supplied.

First, a procedure for using an empty entry string will be described. Each entry shares and uses an empty entry.
One entry is acquired from the empty entry sequence at the time of registration, and is returned to an empty entry at the time of deletion.

The acquisition of an empty entry is performed in the following procedure. 1. The address of the first empty entry is obtained from the empty entry management area 50. 2. The address of the next free entry is obtained from the free entry obtained in “1.” and stored in the first free entry pointer of the free entry management area 50. 3. If there is no next empty entry in “2.” (next entry pointer is 0), 0 is stored in both the leading and trailing empty entry pointers of the empty entry management area 50. With the above procedure, empty entries are extracted from the head of the empty entry string.

The entry is returned to the empty entry column in the following procedure. 1. 0 is stored in the next entry pointer of the entry to be returned. 2. The head entry pointer of the empty entry management area 50 is checked, and if the value is 0 (there is no empty entry), “3.” is executed. If the value is other than 0, “4.” is executed. 3. The address of the entry to be returned to both the leading empty entry pointer and the trailing empty entry pointer of the empty entry management area 50 is stored. 4. The address of the last free entry is obtained from the free entry management area 50, and the address of the entry returned to both the next entry pointer of this entry and the last free entry pointer of the free entry management area 50 is stored. With the above procedure, the entry is returned to the end of the empty entry sequence.

Next, the registration / deletion procedure of each entry will be described.

Registration of a file entry is performed in the following procedure. 1. One empty entry is obtained from the empty entry column, the file identifier is stored in the file identifier area, and 0 is stored in the next entry pointer, previous entry pointer, first block entry pointer, and last block entry pointer. 2. The first file entry pointer in the file entry management area 10 is checked, and the value is 0 (the file entry is 1).
If no, execute "3.", otherwise execute "4." 3. The address of the file entry created in "1." is stored in both the first file entry pointer and the last file entry pointer of the file entry management area 10. 4. The address of the last file entry is acquired from the file entry management area 10, the address of the last file entry is stored in the entry pointer before the file entry created in "1.", the next entry pointer of the last file entry and the file entry management area. The address of the file entry created in "1." is stored in both of the last file entry pointers of "10." According to the above procedure, the entry is registered at the end of the file entry sequence.

Deletion of a file entry is performed in the following procedure. 1. The previous entry pointer and the next entry pointer of the file entry to be deleted are checked. If both are 0, "2."; if only the previous entry pointer is 0, "3." 4. If both are not 0, execute 5 .. 2. In this case, there is no other file entry. In this case, 0 is stored in both the leading file entry pointer and the trailing file entry pointer of the file entry management area 10, and the file entry to be deleted is returned to the empty entry column. 3. In this case, there are a plurality of file entries, and the file entry to be deleted is located at the top. In this case, the address of the next file entry is obtained from the file entry to be deleted, 0 is stored in the previous entry pointer of the next file entry, and the address of the next file entry is stored in the first file entry pointer of the file entry management area 10. Thereafter, the file entry to be deleted is returned to the empty entry column. 4. In this case, there are a plurality of file entries, and the file entry to be deleted is located at the end. In this case, the address of the previous file entry is obtained from the file entry to be deleted, 0 is stored in the next entry pointer of the previous file entry, and the address of the previous file entry is stored in the last file entry pointer of the file entry management area 10. Thereafter, the file entry to be deleted is returned to the empty entry column. 5. In this case, there are a plurality of file entries, and the file entry to be deleted is located in the middle. In this case, the address of the previous file entry and the address of the next file entry are obtained from the file entry to be deleted, the address of the next file entry is stored in the next entry pointer of the previous file entry, and the previous entry pointer of the next file entry is stored. After storing the address of the file entry, the file entry to be deleted is returned to the empty entry column.

Registration of a task entry is performed in the following procedure. 1. One empty entry is obtained from the empty entry column, the host identifier is stored in the host identifier area, the task identifier is stored in the task identifier area, and 0 is stored in the next entry pointer, the previous entry pointer, the first lock entry pointer, and the last lock entry pointer. Store. 2. The first task entry pointer in the task entry management area 20 is checked, and if the value is 0 (there is no task entry), "3." 3. The address of the task entry created in "1." is stored in both the first task entry pointer and the last task entry pointer of the task entry management area 20. 4. The address of the last task entry is obtained from the task entry management area 20, the address of the last task entry is stored in the previous entry pointer of the task entry created in "1.", the next entry pointer of the last task entry and the task entry management area. The address of the task entry created in “1.” is stored in both of the last task entry pointers of No. 20. With the above procedure, the entry is registered at the end of the task entry sequence.

The deletion of a task entry is performed in the following procedure. 1. The previous entry pointer and the next entry pointer of the task entry to be deleted are examined. If both are 0, “2.” is entered. If only the previous entry pointer is 0, “3.” is entered. 4. If both are not 0, execute 5 .. 2. In this case, there is no other task entry. In this case, 0 is stored in both the leading task entry pointer and the ending task entry pointer of the task entry management area 20, and the task entry to be deleted is returned to the empty entry column. 3. In this case, there are a plurality of task entries, and the task entry to be deleted is located at the top. In this case, the address of the next task entry is acquired from the task entry to be deleted, 0 is stored in the previous entry pointer of the next task entry, and the task entry management area 20
After the address of the next task entry is stored in the first task entry pointer, the task entry to be deleted is returned to the empty entry column. 4. In this case, there are a plurality of task entries, and the task entry to be deleted is located at the end. In this case, the address of the previous task entry is obtained from the task entry to be deleted, 0 is stored in the next entry pointer of the previous task entry, and the task entry management area 20 is stored.
After the address of the previous task entry is stored in the last task entry pointer, the task entry to be deleted is returned to the empty entry column. 5. In this case, there are a plurality of task entries, and the task entry to be deleted is located in the middle. In this case, obtain the address of the previous task entry and the address of the next task entry from the task entry to be deleted,
The address of the next task entry is stored in the next entry pointer of the previous task entry, the address of the previous task entry is stored in the previous entry pointer of the next task entry, and the task entry to be deleted is returned to the empty entry column.

The registration of a block entry is performed in the following procedure. 1. One empty entry is obtained from the empty entry column, the block identifier is stored in the block identifier area, the initial value is stored in the generation number area, and 0 is stored in the next entry pointer, previous entry pointer, lock entry pointer, and lock waiting entry pointer. Is stored. 2. The head block entry pointer of the file entry having this block is checked. If the value is 0 (there is no block entry), "3."
Execute 3. The address of the block entry created in “1.” is stored in the first block entry pointer and the last block entry pointer of the file entry. 4. The address of the block entry created in "1." is stored in the entry pointer before the block entry indicated by the first block entry pointer of the file entry, and the first block of the file entry is stored in the next entry pointer of the block entry created in "1." The value of the entry pointer is stored, and the address of the block entry created in "1." is stored in the first block entry pointer of the file entry. By the above procedure, the block entry is registered at the head of the block entry sequence.

The deletion of a block entry is performed in the following procedure. 1. The previous entry pointer and the next entry pointer of the block entry to be deleted are checked, and if both are 0, “2.”; if the previous entry pointer is only 0, “3.”; if only the next entry pointer is 0, “2.” 4. If both are not 0, execute 5 .. 2. In this case, there is no other block entry. In this case, 0 is stored in both the head block entry pointer and the end block entry pointer of the file entry having this block, and the block entry to be deleted is returned to the empty entry column. 3. In this case, there are a plurality of block entries, and the block entry to be deleted is located at the top.
In this case, the address of the next entry is obtained from the block entry to be deleted, 0 is stored in the previous entry pointer of the next block entry, the address of the next block entry is stored in the first block entry pointer of the file entry, and then the block to be deleted is deleted. Returns an entry to the empty entry column. 4. In this case, there are a plurality of block entries, and the block entry to be deleted is located at the end.
In this case, the address of the previous entry is obtained from the block entry to be deleted, 0 is stored in the next entry pointer of the previous block entry, the address of the previous block entry is stored in the last block entry pointer of the file entry, and then the block to be deleted Returns an entry to the empty entry column. 5. In this case, there are a plurality of block entries, and the block entry to be deleted is located in the middle. In this case, the address of the previous block entry and the address of the next block entry are obtained from the block entry to be deleted, the address of the next block entry is stored in the next entry pointer of the previous block entry, and the previous address is stored in the previous entry pointer of the next block entry. After storing the address of the block entry, the block entry to be deleted is returned to the empty entry column.

The registration of a lock entry is performed in the following procedure. 1. One entry is obtained from the empty entry string. 2. When registering as a lock state, the address of the lock entry created in "1." is stored in the lock entry pointer of the block entry. When registering as a lock waiting entry, the address of the lock entry created in "1." is stored in the lock waiting entry pointer of the block entry. The address of the block entry connected to the block entry pointer of the lock entry created in “1.” is stored. 3. The address of the task entry connected to the task entry pointer of the lock entry created in “1.” is stored. Next, the head lock entry pointer of the task entry is checked. If the value is 0 (there is no lock entry), “4.” is executed. If the value is not 0, “5.” is executed. 4. The address of the lock entry created in “1.” is stored in both the head lock entry pointer and the tail lock entry pointer of the task entry. 5. The address of the tail lock entry is obtained from the task entry, the address of the tail lock entry is stored in the entry pointer before the lock entry created in “1.”, the next entry pointer of the tail lock entry and the tail lock entry pointer of the task entry Store the address of the lock entry created in "1." By the above procedure, the lock entry is registered at the end of the lock entry sequence.

The lock entry is deleted in the following procedure. 1. Check whether the lock waiting entry pointer of the block entry connecting this lock entry is 0, if 0, store 0 in the lock entry pointer, otherwise copy the address of the lock waiting entry pointer to the lock entry pointer, and wait for lock. Store 0 in the entry pointer (lock permission waiting for lock). 2. The previous entry pointer and the next entry pointer of the lock entry to be deleted are checked, and if both are 0, “3.”; if the previous entry pointer is only 0, “4.”; if only the next entry pointer is 0, “3.” 5. If both are not 0, execute 6 .. 3. In this case, there is no other lock entry. In this case, 0 is stored in both the head lock entry pointer and the tail lock entry pointer of the task entry, and the lock entry to be deleted is returned to the empty entry column. 4. In this case, there are a plurality of lock entries, and this lock entry is located at the top. In this case, the address of the next lock entry is acquired from the lock entry to be deleted, 0 is stored in the previous entry pointer of the next lock entry, the address of the next lock entry is stored in the head lock entry pointer of the task entry, and then deleted. Returns the lock entry to the empty entry column. 5. In this case, there are a plurality of lock entries, and this lock entry is located at the end. In this case, the address of the previous lock entry is obtained from the lock entry to be deleted, 0 is stored in the next entry pointer of the previous lock entry, the address of the previous lock entry is stored in the tail lock entry pointer of the task entry, and then deleted. Returns the lock entry to the empty entry column. 6. In this case, there are a plurality of lock entries, and this lock entry is located in the middle. In this case, the address of the previous lock entry and the address of the next lock entry are acquired from the lock entry to be deleted, the address of the next lock entry is stored in the next entry pointer of the previous lock entry, and the previous address is stored in the previous entry pointer of the next lock entry. After storing the address of the lock entry, the lock entry to be deleted is returned to the empty entry column.

Next, processing when a lock request and an unlock request are received from the host will be described. At the time of a lock request or an unlock request, as described above, the corresponding host, task, file, and block identifiers are included in the request and supplied.

The processing of a lock request is described below. FIG. 1 shows a first example of a lock operation of an embodiment of an exclusive control device according to the present invention.
2 is a flowchart showing a second portion other than that shown in FIG.

First, a file identifier of the lock request is obtained, and a file entry is searched. The procedure is shown below. 1. Referring to the first file entry pointer in the file entry management area 10, if the value is 0, the file entry does not exist (Y branch in step 101). If the value is not 0, “2.” is executed (N in step 101).
branch). 2. While following the address of the next entry pointer in order from the file entry indicated by the first file entry pointer, a file entry that matches the file identifier of the lock request is searched. If no matching file entry is found up to the last file entry, there is no corresponding file entry (N branch of step 102).

When the file entry is found (step 103), the block identifier of the lock request is obtained, and the block entry is searched. The procedure is shown below. 1. Referring to the head block entry pointer of the file entry, if the value is 0, the block entry does not exist (Y branch in step 104). If the value is other than 0, "2."
Execute 2. While following the address of the next entry pointer in order from the block entry indicated by the first block entry pointer, a block entry that matches the block identifier of the lock request is searched. If no matching block entry is found up to the last block entry, there is no corresponding block entry (N branch in step 105).

If a block entry is found (step 106), it is checked whether a lock entry exists. If the value of the lock entry pointer of the found block entry is 0, there is no lock entry (Y in step 107).
Branch), there is no task that locks the block. If the value is other than 0, a lock entry exists (step 111), and the block is locked by another task.

If a lock entry is found by the above processing, this lock request cannot be locked. In this case, according to the lock entry registration procedure described above, the lock entry is registered as a lock waiting entry (step 135), and a lock waiting is notified (step 13).
6). At this time, if the task entry of this lock request does not exist (Y branch of step 131, N branch of step 132)
Branch), a task entry is registered in advance in accordance with the task entry registration procedure described above (step 134).

If no file entry or block entry is found (Y branch in step 101,
This lock request can be locked (N branch of 102, Y branch of 104, N branch of 105). In this case, an entry that does not exist is registered by the above-described registration procedure of each entry (step 108 or step 109).

When a file entry and a block entry are found but a lock entry is not found (Y branch of step 107), the lock request can be locked. In this case, first, the corresponding block entry is removed from the block entry sequence according to the above-described block entry deletion procedure. However, the block entry is not returned to the empty entry column. Next, the block entry is re-registered at the head of the block entry sequence according to the above-described block entry registration procedure. However, it is not necessary to create a block entry (step 110).

The details of the re-registration of the file entry corresponding to the existing block entry to the head of the block entry string (step 110) are performed in the following procedure. 1. The previous entry pointer and the next entry pointer of the block entry to be re-registered are checked. If both are 0 or only the previous entry pointer is 0, "2." If both are not 0, “4.” is executed. 2. In this case, since this blog entry is located at the head of the block entry sequence, there is no need to re-register, and the following ends without doing anything. 3. In this case, there are a plurality of block entries, and the block entry to be reregistered is located at the end. In this case, the address of the previous entry is obtained from the block entry to be re-registered, 0 is stored in the next entry pointer of the previous block entry, the address of the previous block entry is stored in the last block entry pointer of the file entry, and then “5” .. ”and after. 4. In this case, there are a plurality of block entries, and the block entry to be re-registered is located in the middle. In this case, obtain the address of the previous block entry and the address of the next block entry from the block entry to be re-registered, store the address of the next block entry in the next entry pointer of the previous block entry, and store the address of the next block entry in the previous entry pointer of the next block entry. After the address of the previous block entry is stored, the operation from “5.” is performed. 5. 0 is stored in the entry pointer before the block entry to be reregistered. 6. The address of the block entry to be re-registered is stored in the entry pointer before the block entry indicated by the first block entry pointer of the file entry, and the value of the first block entry pointer of the file entry is stored in the next entry pointer of the block entry to be re-registered. Then, the address of the block entry to be re-registered is stored in the first block entry pointer of the file entry. The generation number of the block entry to be reregistered is not changed (the generation number is not set to the initial value by reregistration). With the above procedure, the existing block entry is re-registered at the head of the block entry sequence.

Next, the lock entry is registered as locked according to the above-described lock entry registration procedure (step 125). At this time, if the task entry of this lock request does not exist (Y branch of step 121 or N branch of step 122), the task entry is registered in advance according to the task entry registration procedure described above (step 1).
24). As a result of the above procedure, if the lock is possible, the lock request is notified to the task which has issued the lock request, and the value of the generation number area of the block entry is returned (step 126).

The host that has received the generation number thereby determines whether the data held in the local memory matches the data existing in the shared resource, and
When it is determined that they match, access to the local memory is completed, and when it is determined that they do not match, the shared resource is accessed.

Next, the processing of the unlock request will be described below. FIG. 3 is a flowchart showing an example of the unlock operation in the exclusive control device according to the present embodiment.

In the unlock request, either the block entry is deleted according to the deletion of the lock entry or the block entry is deleted or the generation number is counted up by the task issuing the request.

First, a file entry, a block entry, and a lock entry are searched by the same method as the above-described lock request procedure (steps 301 to 310). Next, the found lock entry is deleted by the above-described lock entry deletion procedure (step 311). Next, the task entry having this lock entry is checked, and if there is no lock entry due to the deletion of this lock entry (Y branch in step 312), this task entry is deleted according to the above-described task entry deletion procedure (step 312). Step 313).

Next, if deletion of the block entry is specified by the unlock request (N branch of step 314),
The block entry having this lock entry is checked, and if there is no lock entry after deleting this lock entry (Y branch of step 317), the block entry is deleted by the block entry deletion procedure (step 318). Further, if the file entry is examined and this block entry is deleted, and there is no block entry (Y branch in step 319),
This file entry is deleted by the above-described file entry deletion procedure (step 320).

If count-up of the generation number is designated by the unlock request (Y branch of step 314), the value of the generation number area of the block entry having the lock entry is counted up (step 315). If there is no corresponding file entry, block entry, lock entry or task entry in the unlock request,
It is an unlock failure, and a notification to that effect is given (step 316).

Here, the search of the block entry, which is a feature of the present invention, will be specifically described with reference to FIG.

First, as shown in (a), it is assumed that ten block entries having block identifiers 1 to 10 are registered under a certain file entry. Here, the lock entries have block identifiers 1, 4, 6,
8, 9, and 10 entries. The remaining four entries of block identifiers 2, 3, 5, and 7 do not have lock entries, and only generation numbers are registered.

In this state, the state where the lock request for the block with the block identifier 11 is received is shown in FIG.
According to the present invention, the block entry with the block identifier 11 is registered at the head of the block entry sequence.

Further, the state where the lock request for the block identifier 3 is received in the state (b) is (c). According to the present invention, since the block entry with the block identifier 3 already exists in the block entry sequence, the entry is temporarily removed from the block entry sequence and is re-registered at the head of the block entry sequence.

Normally, a task locks a block, and after processing data, unlocks quickly. The interval between this lock and unlock is very short. Therefore, there is a high possibility that the entry of the block identifier 3 is unlocked in the state of (c). In this case, unless the present invention is applied,
In the state of (b), the block entry 3 is searched, and the number of searched entries is nine. But,
According to the present invention, the block entry 3 is searched in the state of (c), and only one entry needs to be searched.

As described above, when the lock information is registered in response to the lock request, if there is no corresponding block information, the exclusive control device according to the present embodiment adds the newly created block information to the head of the existing block entry sequence. When there is corresponding block information, the block information is temporarily removed from the block entry sequence and re-registered at the beginning of the block entry sequence, thereby greatly reducing the time required for searching for management information. Can be.

In the above description, when the lock information corresponding to the lock request is registered, if there is no corresponding block information, the newly created block information is registered at the head of the existing block entry sequence. If there is no corresponding file information at the time of registering the lock information in response to the lock request, if newly created file information is registered at the head of the existing file entry column, further management information search is performed. Required time can be shortened.

The procedure for registering the file entry in this case is as follows. 1. One empty entry is obtained from the empty entry column, the file identifier is stored in the file identifier area, and 0 is stored in the next entry pointer, the previous entry pointer, the first block entry pointer, and the last block entry pointer. 2. The first file entry pointer in the file entry management area 10 is checked, and the value is 0 (the file entry is 1).
If no, execute "3.", otherwise execute "4." 3. The address of the file entry created in "1." is stored in both the first file entry pointer and the last file entry pointer of the file entry management area 10. 4. The address of the first file entry is acquired from the file entry management area 10, the address of the first file entry is stored in the next entry pointer of the file entry created in "1.", the entry pointer before the first file entry and the file entry management area. The address of the file entry created in "1." is stored in both of the first file entry pointers of "10." According to the above procedure, the entry is registered at the head of the file entry sequence.

[0079]

As described above, the exclusive control device of the present invention, when registering lock information for a lock request, if there is no corresponding block information, adds the newly created block information to the head of the existing block entry sequence. When there is corresponding block information, the block information is temporarily removed from the block entry sequence and re-registered at the beginning of the block entry sequence, thereby greatly reducing the time required for searching for management information. It has the effect that it can be done.

Further, when there is no corresponding file information at the time of registering the lock information in response to the lock request, the newly created file information is registered at the head of the existing file entry sequence, so that the time required for searching the management information is reduced. Further, there is an effect that the length can be further reduced.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a first part of a lock operation of an embodiment of an exclusive control device according to the present invention.

FIG. 2 is a flowchart showing a second part of the lock operation of the exclusive control device according to the embodiment of the present invention other than that shown in FIG. 1;

FIG. 3 is a flowchart illustrating an example of an unlock operation in the exclusive control device according to the present embodiment.

FIG. 4 is a block diagram showing a management table structure of each registration information. FIG. 4A shows a management structure of a file entry, a block entry, a lock entry, and a task entry.
(B) shows the management structure of the empty entry.

FIG. 5 is a detailed diagram showing the structure of each entry and an entry management area. (A) is a file entry, (b)
Is a task entry, (c) is a block entry, and (d)
Shows a lock entry, (e) shows a free entry, (f) shows a file entry management area, (g) shows a task entry management area, and (h) shows a free entry management area.

FIG. 6 is a detailed diagram showing an example of transition of the management state of a block entry by a file entry.
FIG. 7B illustrates a management state of zero entries, FIG. 7B illustrates a state in which one entry is newly added from FIG. 7A, and FIG. 7C illustrates a state in which an existing entry is re-registered.

FIG. 7 is a detailed diagram showing an example of a lock state in the management table.

FIG. 8 is a detailed diagram showing an example of a lock state and a lock wait state in the management table.

[Explanation of symbols]

 10 File entry management area 10-1 to 10-n File entry 20 Task entry management area 20-1 to 20-n Task entry 30-1 to 30-n Block entry 40-1 to 40-n Lock entry 50 Free entry management Area 50-1 to 50-n Free entry

Claims (4)

[Claims] In a system in which a plurality of host computers share a shared resource such as an external storage device, the shared resource is defined as a set of a plurality of files, and each of the files is divided into a plurality of blocks to divide the file into a plurality of blocks. An exclusive control device that performs exclusive control in units of a task in units of a block, wherein information of a file being managed for managing a generation number indicating update information of the block and managing lock information indicating a lock state is file information. A management table for connecting and registering the information of the block being managed by connecting the pointer in a line as a column and connecting and registering the information of the block being managed in a file as a block information column for each file, and the block when registering lock information in response to a lock request from the host computer. If there is no block information corresponding to the request in the information sequence, the request When the block information corresponding to the request exists in the block information sequence at the time of registering the lock information corresponding to the lock request, the block information is registered at the top of the block information sequence. An exclusive control device comprising: a lock processing unit for re-registering connection at the head of the block information sequence. 2. The lock processing means according to claim 1, wherein when the lock information for the lock request from the host computer is registered, if there is no file information corresponding to the request in the file information sequence, the file information corresponding to the request is newly created. 2. The exclusive control device according to claim 1, wherein the file is registered at the beginning of the file information sequence. 3. When the block information corresponding to the request is the block information to be managed with the generation number when deleting the registration of the lock information in response to the unlock request from the host computer, the generation number of the block information is updated by +1. Unlock processing for saving the block information in the block information column and deleting the block information when there is no other lock information that locks the block when the block information corresponding to the request is out of the management target of the generation number 3. The exclusive control device according to claim 1, further comprising a unit. 4. The unlock processing means deletes block information corresponding to the request when the registration of the lock information for the unlock request from the host computer is deleted, and adds another block information to the block information sequence of the file information corresponding to the request. 4. The exclusive control device according to claim 3, wherein the file information is deleted when there is no block information.