JP2569129B2 - Exclusive control method of auxiliary storage device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a resource management method for a computer system,
In particular, the present invention relates to an exclusive control method suitable for sharing a directly accessible auxiliary storage device such as a magnetic disk device among a plurality of central processing units.

[Conventional technology]

A directly accessible auxiliary storage controller represented by a conventional magnetic disk controller is the Hitachi Manual H-8538
-CI disk controller, H-8598 disk controller ", a reserve (RESERVE) command to secure a lock on the auxiliary storage device where the data to be accessed exists, and a lock command. A release (RELEASE) command is provided to release. When accessing data on an auxiliary storage device shared by a plurality of central processing units (hereinafter abbreviated as CPU), first issue a RESERVE command to the auxiliary storage device, and
Prohibit access from. After that, a data transfer command is issued. This realizes exclusive control of data on the auxiliary storage device. This is called a physical exclusion control method.

As a method for exclusive control of data on the auxiliary storage device, this exclusive control unit was made to be a logical unit, and the minutes of the International Conference on Sigmod held in December 1985 by AC M Pages 122 to 130 (Proceeding
of ACM-SIGMOD International conference on Managem
ent of Data.DEC.1985 PP122-130), and this is called a logical exclusive control method.

If CPUs sharing an auxiliary storage device can communicate with each other, there is a method described in Japanese Patent Application Laid-Open No. 57-68461. In this method, a CPU having a global lock manager (a manager for managing access rights to shared data) and a local lock manager (a manager for receiving a lock request of a user program) exist. By transmitting and receiving information about locks between CPUs, exclusive control of data shared between CPUs becomes possible. This is called an inter-CPU communication exclusive control method.

[Problems to be solved by the invention]

In general, data on a directly accessible auxiliary storage device represented by a magnetic disk device is managed in a logical unit such as a data set, and a plurality of logical storage devices are stored on one auxiliary storage device. There is a group of data summarized in FIG. Exclusive control is required for a logical unit such as a data set, but in the above physical exclusive control method, lock contention occurs frequently because the lock unit is a physical unit that is the same as the entire auxiliary storage device. , Which is a cause of a decrease in the performance of the computer system. In addition, in the logical exclusive control method in which exclusive control units are logically performed, all systems sharing this control device must perform the logical exclusive control management, and compatibility with the physical exclusive control method is required. Not considered. Furthermore, in the inter-CPU communication exclusive control method,
Two times of communication (input / output processing) are required, two times of lock information transfer processing from the CPU where the local lock manager exists and two times of lock information transfer processing from the CPU where the global lock manager exists. .

A first object of the present invention is to solve such a conventional problem and to reduce the data lock shared by a plurality of CPUs.
An object of the present invention is to provide an auxiliary storage control device having a lock mechanism capable of designating a logical lock range and reducing the overhead of lock processing while maintaining compatibility with a physical exclusive control method. is there. Further, a second object of the present invention is to mix a CPU in which an OS not supported by the auxiliary storage control device having the above-mentioned mechanism is running and a CPU in which a supported OS is running. An object of the present invention is to enable exclusive control of data without contradiction even when connected to an auxiliary storage control device.

[Means for solving the problem]

In addition to the RESERVE / RELEASE command provided in a directly accessible auxiliary storage controller represented by a conventional magnetic disk device, a RESERVE / RE with an exclusive control identifier added
Create a new LEASE command. When the RESERVE command is issued from the CPU, a memory means for storing the exclusive control identifier and an exclusive control identifier management table for storing the channel route for which the input / output is requested is provided in the auxiliary storage control device. Thereby, the first object is achieved. A second object is to provide a lock counter in the auxiliary storage control device for each auxiliary storage device connected to the auxiliary storage control device, issue a RESERVE command specifying an exclusive control identifier from the CPU, and secure the lock. Succeeds,
This is achieved by providing a means for adding 1 to the lock counter and subtracting 1 from the exclusive control counter when a RELEASE command corresponding to the RESERVE command is issued.

[Action]

A directly accessible auxiliary storage device typified by a magnetic disk device connected to a plurality of CPUs can be accessed from only one CPU at a time. Therefore, the user program issues an exclusive control request to the OS to obtain the right to use the data set on the auxiliary storage device. The OS that receives the exclusive control request issues a RESERVE command in which the data set name is specified as an exclusive control identifier at the beginning of a CCW (channel command word) for an I / O request to the auxiliary storage device issued first thereafter. And execute the I / O request. The auxiliary storage control device that has received the RESERVE command first determines whether or not the entire auxiliary storage device is locked. If the entire auxiliary storage device is not locked, the exclusive control identifier specified in the RESERVE command is used as a key to refer to the exclusive control identifier management table. If the exclusive control identifier is already registered in the exclusive control identifier management table,
The channel route that registered the exclusive control identifier is read, and the channel / channel of the RESERVE command issued this time is read.
It is determined whether the route matches a channel route for which a lock has already been secured or belongs to the same channel path group. With this, when it is determined from another CPU that the data set or the entire auxiliary storage device in which the data set exists has not been locked by the other CPU, the exclusive control identifier management table stores the exclusive control identifier in the exclusive control identifier management table. Register the channel route and add 1 to the lock counter. Thus, the securing of the lock for each data set is completed. In addition, when securing the lock fails (when the lock is already secured in another CPU)
Reports to the CPU that the device is in use.

When the OS running on the CPU issues a RESERVE command without specifying an exclusive control identifier, the auxiliary storage control device that has received the RESERVE command first checks whether the entire auxiliary storage device is locked by another CPU. Make a decision. Thereafter, the lock counter is referred to, and if the value of the lock counter is 0, a lock securing flag is set.
Thus, the securing of the lock of the entire auxiliary storage device is completed. If the exclusive control counter is other than 0, the entire auxiliary storage device cannot be locked, and the CPU is informed that the device is in use.

Accordingly, the auxiliary storage control device of the present invention can perform exclusive control of data in finer units without inconsistency with the locking mechanism of the conventional auxiliary storage control device.

〔Example〕

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram of a computer system to which the present invention is applied. In FIG. 2, three CPUs 10, 11, and 12 are provided. On each CPU, independent OSs 13, 14, and 1 are provided.
5 is working. The OSs 13 and 14 are OSs that support exclusive control in units of data sets, and the OS 15 is an OS that supports only exclusive control of the entire directly accessible auxiliary storage device represented by a conventional magnetic disk device. each
The CPU is connected to channels 16, 17, and 18, respectively, and each of the channels 16, 17, and 18 is connected to a plurality of auxiliary storage devices 21 via an auxiliary storage control device 19.

Here, the system is provided with three CPUs and three channels. However, in general, the system can be expanded to two or more n units. In the embodiments of the present invention, a magnetic disk device will be described as an example of the auxiliary storage device, but a magnetic drum device or a semiconductor storage device may be used.

By the way, access to data on a magnetic disk device shared between CPUs can be performed only by one CPU at a time. Therefore, the user must issue an exclusive control request to the OS to obtain the right to use the magnetic disk device. The OS that has received the exclusive control request does not immediately issue the exclusive control request to the magnetic disk device, but instead issues a RESERVE command for exclusive control at the beginning of the CCW corresponding to the first data transfer I / O request after the exclusive control request. To make the input / output overhead related to the lock processing almost zero.

However, in the conventional magnetic disk control device, since the lock unit is the entire magnetic disk device, when a plurality of data sets exist on one magnetic disk device, lock competition occurs. In particular, as the recording density of a magnetic disk device increases and its capacity increases, a large number of data sets are arranged on one magnetic disk device, and the performance degradation of a computer system due to the lock competition becomes a problem. I have.

Therefore, in this embodiment, RESERVE / RELEAS for exclusive control is used.
An exclusive control mechanism is realized by adding an exclusive control identifier to the E / Unconditional (UNCONDITIONAL) RESERVE command and adding a new command with the unit of exclusive control for each exclusive control identifier.

FIGS. 3 and 4 show RESERVs newly added in this embodiment.
Shows the specifications of the E / RELEASE // UNCONDITIONAL / RESERVE command. In the HITAC M series processor (Hitachi computer), there are two types of CCWs that specify these commands.

The first format shown in FIG. 3 is called format 0CCW, and the data address indicated in the command is represented by 24 bits. The data address of the format 0CCW30 indicates an area 31 storing an exclusive control identifier of 44 bytes. This area 31
Must be addressable by 24 bits.

The second format shown in FIG. 4 is called format 1CCW, and the data address indicated in the command is represented by 31 bits. The data address of this format 1CCW32 also indicates the area 31 in which the exclusive control identifier is stored, similarly to the format 0CCW. Here, the types of the commands described above are indicated by 1-byte command codes. In this embodiment, the RESERVE command is X'24 ', the RELEASE command is X'44', and the UNCONDITIONAL
RESERVE is represented by X'54 '. The specifications of these commands will be described below.

The RESERVE command uses the exclusive control identifier as a key
Secure the lock of the magnetic disk device and prohibit use from other CPUs. The RELEASE command is used to release the lock using the exclusive control identifier as a key.
The IONAL RESERVE command is a command for forcibly releasing the lock of another CPU using the exclusive control identifier as a key and securing the lock using the exclusive control identifier as a key in its own CPU.

Next, the operation of the magnetic disk control device receiving these commands will be described. FIG. 5 shows a control table for implementing exclusive control using an exclusive control identifier as a key. This control table is stored in the control device 20 in the magnetic disk control device. The table for implementing exclusive control includes a channel route management table 4
0, a device management table 50, and an exclusive control identifier management table 60.

The channel route management table manages the connection status with the CPU.
Whether it is connected to U is managed by a channel path group identifier specified from each channel route.

The device management table 50 manages the status of the magnetic disk device, and includes a lock counter 5 in this table.
3. An area 52 for storing a channel route that locks the entire magnetic disk device, a pointer 55 to an exclusive control identifier management table, and the like are arranged. The exclusive control identifier management table stores the exclusive control identifier 65 specified by the OS.
The table has a plurality of entries including an area 64 for storing a channel route to which an exclusive control request has been issued and a synonym pointer 61 for performing hashing using an exhaust control identifier as a key. Here, the number of entries is 64, but the number can be expanded to 2 or more and n.

The process of securing the lock of the magnetic disk control device 19 when the CPU issues a RESERVE command with an exclusive control identifier will be described with reference to the block diagram of FIG. 1 and the flowchart of FIG.

First, the judgment 100 reads the flag 54 stored in the device management table 50 and checks whether or not the entire magnetic disk device is locked.

If it is determined that the entire magnetic disk device is not locked by the determination 100, the process 101 first proceeds to the RES
The exclusive control identifier specified in the ERVE command is set in the exclusive control identifier register 70, and this is set in the hashing circuit 71 to calculate the entry number of the exclusive control identifier management table corresponding to the exclusive control identifier. It is stored in the number register 72. Then, the exclusive control identifier pointer 55 stored in the device management table 50
Then, the value obtained by shifting the entry number stored in the entry number register 72 by 6 bits is input to the adder 73 (6 bits shift is because the entry size of the exclusive control identifier management table is 64 bytes). The output of the adder 73 is stored in the exclusive control identifier entry pointer register 74, so that the entry of the exclusive control identifier management table can be accessed.

Subsequent to this, the judgment 102 judges whether the entry indicated by the exclusive control identifier entry pointer register 74 is in use. This determination process is performed by referring to the flag 62. When it is determined by the determination 102 that the entry is not used, the process 103 adds 1 to the lock counter 53 arranged in the device management table. The lock counter 53 is a counter that indicates the number of locks specifying different exclusive control identifiers for the magnetic disk device.

The process 104 stores the content of the exclusive control identifier register 70 in the exclusive control identifier 65, and further stores the channel number of the channel route to which the RESERVE command has been issued in the lock securing channel route 64. Thus, the securing of the lock by the designated exclusive control identifier is completed.
If the lock is successfully secured, the command processing is completed by reporting the channel end + device end DSB to the CPU that issued the RESERVE command in step 105.

When the entire magnetic disk device is already locked by the determination 100, the CPU that issued the RESERVE command in the determination 106 and the CPU that locked the entire magnetic disk device.
It is determined whether or not matches. This determination is made by referring to the lock securing channel route stored in the area 52 arranged in the device management table 50 and the channel / pal group identifier stored in the channel route management table 40.

If the entire magnetic disk device is already locked by a different CPU in the determination 106, the RES
The channel number of the channel route to which the ERVE command has been issued is set as the busy report channel route in the device management table, and the process 108 reports the DBS of the device busy to the channel route. As a result, the RESER
Execution of commands chained after the VE command is prohibited. If the entire magnetic disk is locked by the same CPU in the determination 106, the lock securing process using the exclusive control identifier as a key after the process 101 is performed.

If the exclusive control identifier entry has already been allocated in the determination 102, it is determined in a determination 109 whether the exclusive control identifiers match. This determination is made by inputting the exclusive control identifier 65 stored in the exclusive control identifier entry and the exclusive control identifier stored in the exclusive control identifier register 70 to the comparator 75. The result of this comparator is reflected in the flip-flop 76, which is performed by checking the contents. If it is determined that the exclusive control identifiers match in the determination 109, the
The channel number of the channel route to which the RESERVE command has been issued is stored in the busy report channel route 64 of the exclusive control identifier management table entry, and the process 108 reports the device busy DSB to the channel route. If the exclusive control identifiers do not match in the judgment 109, the synonym pointer is checked in the judgment 111. If the value is not set in the synonym pointer, a free exclusive control identifier management table entry is secured in steps 112, 113, and 114, this entry is chained in the synonym chain, and the flag 62, the lock securing channel route of the entry is set. By setting an exclusive control identifier, a lock using the exclusive control identifier as a key is secured. At this time, when it is determined that the empty entry does not exist by the determination 112, the channel number of the channel root where the RESERVE command is issued is stored in the area 56, which is arranged in the device management table 50 by the processing 116, and the processing is performed. 108 reports the device busy DSB to the CPU.
When it is determined that the value is set in the synonym pointer by the determination 111, the synonym pointer is set to this value in the entry number register 72, and a process 115 for following the synonym chain is performed.

Upon receiving the RESERVE command, the magnetic disk control device performs the following processing in accordance with the flowchart shown in FIG. The process 200 calculates the address of the exclusive control identifier management table entry from the exclusive control identifier specified in the RELEASE command, as in the process 101 of the RESERVE command process. In subsequent determinations 201 and 202, it is determined whether or not the exclusive control identifier designated by the RELEASE command is registered in the exclusive control identifier table entry selected by the process 200. At this time, when the synonym chain exists in the exclusive control identifier table entry, the determination 202 and the processing 2
Search for synonymuchiein by 03. If it is determined by the determination 202 that the exclusive control identifier specified in the RELEASE command is registered, the determination 206 determines whether or not there is a channel route that has reported the device-busy DSB. Entry area
Determine with reference to 63. If it is determined by the determination 206 that there is a channel route reporting the device busy, a DSB (Busy to Fres) indicating that the lock has been released to the channel route reporting the device busy exists.
The report of e) is made in processing 207. If there are a plurality of channel routes that have reported the device busy, the DSB of Busy to Free is reported for all the channel routes. Processes 208 and 209 are actually lock release processes, in which the lock counter arranged in the device management table is decremented by 1 (process 208), and the invalid bit of the entry is set in the flag portion of the exclusive control identifier management table entry. (Step 209). If the exclusive control identifier management table entry constitutes a synonym chain, the entry is separated from this chain. In the judgment 210, since there is no free space in the exclusive control identifier management table entry in the magnetic disk device, it is determined by referring to the area 56 of the device management table whether or not the busy-reported channel route exists. When it is determined by this determination 210 that a busy report channel route exists, a DSB of Busy to Free is reported to the channel route (process 211). The determination 212 is a process for determining whether or not the entire magnetic disk device is in a free state. It is determined whether there is a channel route waiting for a lock request. As a result of this determination, a Busy to Free DSB is reported by the process 214 to the channel route that has issued the busy report. As a result, the processing of the RELEASE command ends, and the channel / channel where this command is issued is
Channel end + device end DS for route
Report B.

The UNCONDITIONAL RESERVE command is a command for forcibly securing a lock using the exclusive control identifier specified in the command as a key, and performs the same processing as the RESERVE command except that the lock securing state is not determined.

The operation of the newly provided command according to the present invention has been described above. Next, the relationship between the functions of the conventional magnetic disk control device and the commands newly provided in the present invention will be described.

FIG. 8 is a flowchart showing the flow of processing of a conventional RESERVE command. In the judgments 300 and 301, it is judged whether or not the entire magnetic disk device can be clocked.
This determination is made by referring to the flag 54 and the lock counter 53 of the device management table 50. If it is determined that locking is possible, the process 302
Thus, the lock securing bit is set in the flag 54, and the lock is secured. When it is determined that the lock cannot be secured, the DSB (device busy) of the lock securing failure is reported to the channel route in steps 304 and 305.

When the user program issues a data set exclusive control request, the OS issues a RESERVE command using the data set name as an exclusive control identifier.
When the release of the data set is instructed by the user program, RELE using the data set name as the exclusive control identifier
Issuing the ASE command enables exclusive control on a data set basis.

〔The invention's effect〕

According to the present invention, it is possible to specify an exclusive control unit of an auxiliary storage device such as a magnetic disk device and a logical identifier (exclusive control identifier), thereby making it possible to reduce collision of a lock. The overhead when the storage device is shared by a plurality of CPUs can be reduced. Furthermore, the exclusive control method of the present invention
Since it is realized without contradiction with the exclusive control mechanism of the conventional auxiliary storage device, it is possible to mix an OS that supports only the exclusive control mechanism of the conventional auxiliary storage device and an OS that supports the exclusive control mechanism of the present invention. Therefore, the computer system can be easily migrated.

[Brief description of the drawings]

FIG. 1 is a block diagram showing functions relating to exclusive control in an auxiliary storage device controller, showing one embodiment of the present invention. Second
FIG. 3 is a schematic block diagram of a computer system to which the present invention is applied, FIGS. 3 and 4 are explanatory diagrams showing commands newly provided by the present invention, and FIG. 5 is stored in a control memory of a magnetic disk controller. Exclusive control relation table relation diagram, sixth
FIG. 7 is a flowchart showing processing of a RESERVE command newly established according to the present invention, FIG. 7 is a flowchart showing processing of a RELEASE command, and FIG. 8 is a flowchart showing processing of a conventional RESERVE command. 10, 11, 12, ... CPU, 16, 17, 18 ... channel, 19 ... magnetic disk control device, 20 ... control memory, 21 ... magnetic disk device, 40 ... channel route management table, 50 ...
... Device management table, 60 ... Exclusive control identifier management table, 71 ... Hashing circuit, 73 ... Adder, 75 ...
Comparator.

Claims (1)

(57) [Claims] An exclusive control method for an auxiliary storage device is provided by a storage controller connected to a plurality of central processing units via a channel and controlling a direct access type auxiliary storage device in response to a request from the central processing unit. A first exclusive control command issued from the central processing unit to exclusively use the auxiliary storage device includes a designation of an exclusive control identifier for locking the auxiliary storage device in data set units. Exclusive control command and a second exclusive control command not including the designation of the exclusive control identifier, and the exclusive control command issued from the central processing unit is the first exclusive control command.
Is the second exclusive control command issued from another central processing unit or the first exclusive control command
The first exclusive control command that specifies the same exclusive control identifier as the exclusive control identifier specified by the exclusive control command determines whether the auxiliary storage device is locked, and as a result of the determination, Is locked, a response indicating that the auxiliary storage device is in use is returned. If the result of the determination is that the auxiliary storage device is not locked, the data set of the auxiliary storage device is Exclusive use in units is permitted, and when the exclusive control command issued from the central processing unit is the second exclusive control command, the second exclusive control command issued from another processing unit or the first exclusive control command is issued. It is determined whether or not the auxiliary storage device is locked according to the exclusive control command. It returns a response indicating that the lock is if not the exclusive control method for an auxiliary storage device and permits the exclusive use of the entire auxiliary storage device.