Deadlock detecting device

Claim 2

A deadlock detecting device of a multitasking system, said device detecting deadlock between plurality of tasks executed in the multitasking system, comprising; task managing means for managing parallel execution of plurality of the tasks; lock managing means for detecting that one task locks a common resource , and for detecting that another task is waiting for the common resource locked by the former task and outputting waiting relation information indicating that the latter task is waiting for the former task; a wait management table for storing a plural pieces of waiting relation information; registering means for registering said waiting relation information in said wait management table, every time said lock managing means outputs said waiting relation information; and deadlock detecting means for detecting the deadlock between plurality of the tasks, of plural pieces of waiting relation involving the tasks which are respectively indicated by any one of the waiting relation information stored in the wait management table forms a loop relation, said deadlock detecting means functioning independently and asynchronously from said lock managing means and said task managing means; whereby execution of said first task and said second task is not suspended during deadlock detection.

Claim 3

The deadlock detecting device as claimed in claim 1, wherein said registering means has a table in which said waiting relations are registered.

Claim 4

The deadlock detecting device as claimed in claim 2, wherein said deadlock detecting means registers the waiting relations in the table of said registering means and detects the presence or absence of deadlock by searching said table.

Claim 5

The deadlock detecting device as claimed in claim 2, wherein said deadlock detecting means comprises a request queue receiving unit for receiving a "waiting relation registering request" indicative of an instruction to detect the deadlock loop.

Claim 6

The deadlock detecting device as claimed in claim 2, wherein the deadlock detecting device is realized on a distribution system having a first and second system, wherein each system comprises said task managing means, said lock managing means, said wait management table and said deadlock detecting means.

Claim 7

The deadlock detecting device as claimed in claim 5, wherein said second waiting relation from said first system is communicated and registered in said wait management table of said second system, and the presence or absence of the deadlock loop is determined by looking up said wait management table in said second system.

Claim 8

The deadlock detecting device as claimed in claim 5, wherein a task x generated in a system i is defined as T(i, x), a task y generated in a system j is defined as T(j, y) and waiting of T(i, x) for T(j, y) is represented by T(i, x)→T(j, y), which is registered in said table as information of the third waiting relation to detect a loop of T(i, x)→T(j, y)→T(i, x) when both T(i, x)→T(j, y) and T(j, y)→T(i, x) hold, thereby detecting the deadlock loop.

Claim 9

The deadlock detecting device as claimed in claim 1, wherein the dead lock detecting device is realized on a distribution system having a plurality of systems and comprises a waiting time monitoring unit for use in reissuing a resource gaining request to the task of which waiting relation is continued for a predetermined time.

Claim 10

A deadlock detecting device as claimed in claim 1, wherein said task managing means manages the tasks through a two-phase lock system in which a lock is continuously made when the task starts to lock a data while the lock is continuously released when the task starts to release the lock.

Claim 11

The deadlock detecting device as claimed in claim 2, wherein said lock managing means registers the first waiting relation in said wait management table and issues a deadlock detecting instruction to the deadlock detecting means only when the first waiting relation is caused to be a resource gaining request, and wherein said lock managing means does not suspend the execution operation on the transaction for the resource gaining request accompanying an absence of a waiting relation.

Claim 12

A deadlock detecting device of a multitasking system, said device detecting deadlock between plurality of tasks executed in the multitasking system, comprising: a task manager which manages parallel execution of plurality of the tasks; a lock manager which makes one task a common resource, and which detects that another task is waiting for the common resource locked by the former task and consequently output waiting relation information indicating that the latter task is waiting for the former task; a wait management table for storing the waiting relation information; and deadlock detector which detects the deadlock between plurality of the tasks if plural pieces of the waiting relation involving the tasks which are respectively indicated by any one of the waiting relation information stored in the wait management table forms a loop relation, said deadlock detector functioning independently and asynchronously from said lock manager and said task manager; whereby execution of said first task and said second task is not suspended during the deadlock detection.

Claim 13

A deadlock detecting device of a multitasking system, said device detecting deadlock between plurality of tasks executed in the multitasking system, comprising: a task manager which manages parallel execution of plurality of the tasks; a lock manager which makes one task a common resource, and which detects that another task is waiting for the common resource locked by the former task and consequently output waiting relation information indicating that the latter task is waiting for the former task; a wait management table for storing the waiting relation information; and deadlock detector which detects the deadlock between plurality of the tasks if plural pieces of the waiting relation involving the tasks which are respectively indicated by any one of the waiting relation information stored in the wait management table forms a loop relation, said deadlock detector functioning independently and asynchronously from said lock manager and said task manager; whereby execution of said first task and said second task is not suspended during the deadlock detection.

Claim 14

The deadlock detecting device as claimed in claim 11, wherein said task manager starts, commits and aborts the first and second tasks.

Claim 15

The deadlock detecting device as claimed in claim 11, wherein said lock manager includes a lock table for locking a corresponding relation between said first task and said first resource and between said second task and said second resource.

Claim 16

A deadlock detecting device of a distribution system having a plurality of systems, said device detecting deadlock between plurality of tasks each of which is executed in one of the systems and is usable common resources, each system comprising: a task manager which manages parallel execution of plurality of the task executed in own system; a lock manager which makes one task lock a common resource and which detects that another task is waiting for the common resource locked by the former task and consequently output waiting relation information indicating that the latter task is waiting for the former task; means for sending only waiting relation information indicating that one task is waiting for any task executed in other system to this other system, said sending means further sending waiting relation information indicating that one task is waiting for another task involved in the sent waiting relation information; a wait management table for storing the waiting relation information outputted by the lock manager of own system and the waiting relation information sent from the sending means of other system; a deadlock detector which detects the deadlock between plurality of the tasks, if plural pieces of the waiting relation involving the tasks which are respectively indicated by any waiting relation information stored in the wait management table forms a loop relation, said deadlock detector functioning independently and asynchronously from said lock manager and said task manager; whereby execution of said first task and said second task is not suspended during the deadlock detection.

Claim 17

A deadlock detecting device of a distribution system having a plurality of systems, said device detecting deadlock between plurality of tasks each of which is executed in one of the systems and is usable common resources, each system comprising: a task manager which manages parallel execution of plurality of the task executed in own system; a lock manager which makes one task lock a common resource and which detects that another task is waiting for the common resource locked by the former task and consequently output waiting relation information indicating that the latter task is waiting for the former task; means for sending only waiting relation information indicating that one task is waiting for any task executed in other system to this other system, said sending means further sending waiting relation information indicating that one task is waiting for another task involved in the sent waiting relation information; a wait management table for storing the waiting relation information outputted by the lock manager of own system and the waiting relation information sent from the sending means of other system; a deadlock detector which detects the deadlock between plurality of the tasks, if plural pieces of the waiting relation involving the tasks which are respectively indicated by any waiting relation information stored in the wait management table forms a loop relation, said deadlock detector functioning independently and asynchronously from said lock manager and said task manager; whereby execution of said first task and said second task is not suspended during the deadlock detection.

Claim 18

A deadlock detecting device according to claim 14, further comprising: a watchdog timer watch which monitors, for every waiting relation information, elapsed time since the waiting relation information stored in the wait management table and, when the elapsed time exceeds a predetermined time period, makes the deadlock detector erase the waiting relation information from the wait management table to make the waiting transaction involved in the waiting relation information run again.