CH631820A5 - Method and arrangement for dealing with interrupt requests in a multi-programmable data processing system - Google Patents

Description

The present invention relates to a method for handling interrupt requests in a multiprogrammable data processing system, and an arrangement for performing this method.

In multiprogrammable data processing systems, e.g. the IBM System / 360 or IMB System / 370, it can happen that the system alternates between an interrupting and an interrupted program several times in quick succession (due to pending interrupt requests of the same class as the one that was just being served by the interrupting program) ). This causes an excessive amount of administrative operations, in particular because of the operations that are carried out to exchange status and environmental information.

In data processing systems such as the IMB System / 360 or IBM System / 370 there are commands for examining and conditionally accepting certain interrupt requests (e.g. TEST I / O and CLEAR I / O). However, each such instruction is only used in conjunction with a particular interrupt request source associated with a particular class of interrupts (e.g. I / O interrupts). In a system that serves a large number (eg hundreds) of asynchronous I / O devices and in which there can therefore be many I / O interruption requests at the same time (still pending), it would be extremely ineffective and also practically impossible to Use source-specific »commands to handle all (or at least a large part of) all pending I / O interrupt requests. However, it was found that there is currently a need - apparently earlier, which, although important for effective operation, was not recognized or not correctly assessed - to program-link the processes for handling non-source-specific interruptions that belong to a common class (to one another to hang) without the exchange of environmental or state information.

The invention has for its object to provide an improvement here and to enable the successive execution of several requested interrupt operations of the same class without the need for multiple exchange of program status data, which is otherwise necessary when changing between programs.

The subject matter of the invention can be gathered from the claims.

The invention enables multiple interrupt requests of the same type (e.g. I / O requests) to be serviced recursively in a program that was originally only activated to handle a single specific interrupt request. The simplification is done to a certain extent by "linking" the various interruption treatments, whereby these processes cannot be interrupted as long as they run for the specified interruption class, and can only be carried out if the system is under the control of a monitoring program (supervisor).

The implementation of the invention is possible by means of a single specific command "Query for waiting interrupt request" TPI (= Test Pending Interrupt), which e.g. can: a) check a group of non-specific (ie not specified by this command) sources of interrupt requests of a certain class, b.) accept (by deletion) a pending (waiting) request, if at least one is pending , c.) setting a condition code to indicate this acceptance (or the fact that no acceptance was made), and d.) storing an interrupt code which identifies the source of a possibly accepted request, such that this code corresponds to the currently active program is accessible

This command can be executed at a point (phase) in the execution of an interrupt handler supervisor program associated with the class in question where the control of the system is effectively not able to accept unscheduled interruptions. Execution of the command in question can, however, currently remove this "blocking" (inability to accept) and thereby enable "command-controlled acceptance" of unspecified interruptions in this class, but without any exchange of status or environmental information that would otherwise occur with the normal release and acceptance of such interruptions.

The use of this TPI command (query for waiting interrupt request) can e.g. at a point in program execution where all operations related to handling an assumed break in the class concerned have been completed and where the program is "ready" to return control of the system to a program that was previously interrupted .

As a result of the command-controlled “assumption” of an interruption in the class concerned (as described above), the program which is currently in control of the system when the TPI command “query for interruption request” is executed (ie the program which has it TPI instruction contains) that handle the “accepted” interrupt request with a simple branch subroutine, which is called conditionally when the TPI instruction is executed. Such a subroutine is much shorter than the recursive program operations that would otherwise be required to accept and handle an unplanned break in the class in question.

2nd

5

10th

15

20th

25th

30th

35

40

45

50

55

60

65

Furthermore, the controlled "acceptance" of interrupt requests makes it possible to eliminate a few commands from the existing programs for the recursive handling of interruptions of a certain class, which also results in better utilization of the system.

It is also to be expected that the measures according to the invention will result in improvements which, in particular in the case of time-dependent processes, reduce the statistical frequency of process faults and in this way reduce that part of the system time which is used for processes for diagnosis, correction or restart Is "wasted".

An embodiment of the invention is described below with reference to drawings. Show it:

1 shows the sequence of operations for executing commands and accepting interruptions in known data processing systems in which the present invention can be used and which can be modified in a simple manner for this purpose,

2 shows details of the modifications carried out for the application of the invention in one of the known systems,

3 shows a flowchart to illustrate an application of the invention in a supervisor program of a DV system,

4 shows a flowchart of the recursive handling of interruptions of a certain class in the manner known hitherto, for comparison with the sequence according to the invention according to FIG. 3.

1 shows the sequence of operations for executing program instructions and for accepting any interruptions that occur in known multiprogrammed data processing systems in which the present invention can be advantageously used. These are e.g. in particular systems of the type IBM System / 360 and IBM System / 370.

Fundamentals of the structure and operation of such systems in which the present invention can be used are described in the brochures “IBM System / 360 Principles of Operation”, order number GA 22-6821, and “IBM System / 370 Principles of Operation”, Order number GA 22-7000.

Micro program sequencers for various models of such systems are e.g. described in US Pat. Nos. 3,400,371 and 3,585,599 and in the article "Microprogram Control for System / 360" by S. Tucker, IBM Systems Journal, Volume 6, No. 4, pages 222-241.

Operating system programs (supervisor programs) for the treatment of interruptions which are of interest for the present description are explained in the brochures “IBM System / 360 Operating System Introduction”, order number GC 28-5634, “OS Concepts and Facilities”, order number GC 28-6535 and “OS / VS2 I / O Supervisor Logic”, order number SY 26-3823 and “OS / VS2 System Logic Library”, order number SY 28-0716.

1 therefore shows the sequence in the execution of program instructions and in the handling of any interruptions that may occur in the specified data processing systems. A system sequence control (for example micro program control) takes care of the execution of the program commands, which effects the following processes: (1.1) calling up a command in a memory location, the address of which is usually given by a command address counter, (1.2) decoding the operating part (operation code) of the command and ( 1.3) Execution of the operations designated or called by the command. When executing certain commands, which are defined in the system structure descriptions identified above, a condition code is set, as specified under 1.3.1.

Between the completion operation (END OP) of the execution of a command (position 1.3.2 in Fig. 1) and the retrieval

631820

of the next command (position 1.1 in FIG. 1), the sequence control of the system can be enabled, regardless of the program currently active, assuming an interruption (position 1.4 in FIG. 1). The control determines whether the system is enabled to accept interruptions at this specific time (in the current program status) (item 1.4.1). If no interruptions are allowed in the system, the next command is called (transition to 1.1). However, if the system is approved for accepting interruptions, the controller checks whether there are unmasked interruption requests (item 1.4.2, Fig. 1). If there are no unmasked interrupt requests, control passes directly to the next command at 1.1. However, if there is at least one unmasked interrupt request, the controller initiates operations 1.4.3 which are related to the acceptance of an interrupt.

If there is more than one request, a single request is selected according to a priority scheme (operation 1.4.3.1, Fig. 1) and this request is deleted (1.4.3.2). (“Deleted” here means that the request for an interruption has been accepted and therefore no longer appears as an unresolved request). If there is only a single request, this request is deleted (accepted). Deletion of a request means that the advertisement representing this request is ended. An interrupt code associated with the selected request is stored (operation 1.4.3.3) and environment or status information is exchanged (operation 1.4.3.4) to control the system from the program containing the most recently executed command. to an interrupt handling program designated by the selected request. These processes are described in detail in the above handbooks and in US Patent 3,400,371 already cited above.

During this exchange, a "new" PSW (program status word), which is assigned to the program handling the interruption, is removed from the memory, and an "old" PSW, which is associated with the interrupted program (ie the program which contains the last command executed) ) is saved for later reuse. The sequence control then proceeds to process 1.1 in order to call up the first command of a supervisor program for interrupt handling (UB).

The relevant command “Query for waiting interruption request” (TPI) is intended to initiate recursive processes within the programs which carry out the handling of interruptions in the systems mentioned. As shown in Fig. 2, this command (see 2.1) contains an eight bit long operation part (OP code), which can be seen at 2.1.1, and an eight bit long field 2.1.2, which is either not used, or, as described below, can be used to identify the class of interrupt requests that are "polled" by this command. The decoding device 2.2 for decoding instructions receives an additional decoding output 2.2.1, which is uniquely assigned to this new instruction. As shown in FIG. 2, the control devices 2.2, with the exception of the line 2.2.1 (and the associated code recognition circuits), are essentially identical to the control devices 1.2 of FIG. 1. When OP codes are present, they give - except when there is a signal 2.1.1 is present - control signals are sent to the “normal” execution control devices 1.3 of FIG. 1.

When a signal occurs on TPI line 2.2.1, operations are initiated to determine the class of interrupts that the interrupt handler will use.

3rd

5

10th

15

20th

25th

30th

35

40

45

50

55

60

65

631 820 4

gram which contains the just decoded TPIgram X (shown at 3.1 in Fig. 3 and 4.1 in Fig. 4) by command. The set-up assumption of an I / O interrupt request interruptions necessary for this determination are shown schematically at 2.3 and 2.4 in FIG. 2. This is (at 3.2 in FIG. 3 and 4.2 in FIG. 4), and that corresponds -Information, which the class of the just dealt with the operations 1.4, which are shown in Fig. 1. Individual breaks are specified (e.g. I / O interruption, external 5 processes when executing program instructions are interruption, program interruption, etc.) can be indicated by horizontal lines, e.g. are used to release corresponding AND gates 3.1.1.

(2.3.1 for I / O interruptions, 2.3.2 for external interrupts.

chances, 2.3.3 for program interruptions, etc.), whereby treatment after acceptance (interruption treatment, ie one of them stimulated to emit an output signal io at the first level -1NUB) is determined by the supervisor if the line 2.2.1 is in turn with a signal for interrupt handling - as shown at 3.3 in. The class information can e.g. by the program Fig. 3 and 4.3 in Fig. 4 - the amount of interruption and grams have previously been entered in a register. Andes saves any other environmental information that, on the other hand, could e.g. the operation code 2.1.1 of the command is required for further handling of the interruption. In a differentiation of the class, in which for each advanced level of interrupt treatment (sub-class provides a different operation code. Another refraction treatment on the second level = 2 NUB), the possibility is to use field 2.1.2 of the Command to include the one shown at 3.4 in Fig. 3 and 4.4 in Fig. 4 causes the program to use class distinction information. Otherwise operations to service the interruption. In the case of one, it should be clear that if the TPI command only belongs to the I / O interrupt, these processes e.g. Ver

Is used in connection with a class (e.g. I / O shifting of the channel status word information (CSW) with reference breaks), the class difference described above on a specific I / O channel.

is not required, and that the 2.3 and 2.4

facilities shown would lapse. Command 3.4.2 in sequence 3.4 at the second level

Using the example of the class of I / O interrupts, I will initially execute later than the first instruction 3.4.1, shown in the following, how the execution of the relevant TPI-25 follows the same sequence, and provides a recursive entry point to the instruction «Query on waiting interrupt request which clearly runs towards the execution of the TPI command (after step 2.3.1). In step 2.5 the order is determined as described below. As shown at 3.5, sequence control, whether in the relevant class (here class- the execution of the TPI command (requests for waiting I / O interrupts) currently causes an interruption interrupt request) sets a condition

request is pending. This process is obviously a code of 0 or 1, as described above. The at 3.6 ne subset of the process shown at 1.4.2 in Fig. 2. the branch instruction shown queries this code.

So it could be used in a given system when the condition code is set to 0, the ProA subset of the associated logic is running. grams after the branch instruction continue to an instruction If there is no waiting request in the relevant class «Load program status word» (LPSW), which is shown at 3.7. the condition code is set to 0, as in 2.6 35 With this command the «old» program status word PSW is shown in, and the relevant execution sequence is loaded in 2.7 the registers of the system, and the control is ended with the usual Final surgery (END OP). the system back to the interrupted program (X)

The control then goes from step 1.3.2 in FIG. 1 to give. Then follows the command 3.8, which would have been the next execution of process 1.4 for general assumption if the interruption was not accepted at 3.2 interruptions, which would have become the group of interruptions with 40 men.

Normal treatment (PSW exchange etc.) belong. If the condition code is executed by executing the TPI-Be

On the other hand, if the execution of the test step failed set to 1, the branch instruction 2.5 results in an active interrupt request in the relevant 3.6 - as shown in 3.6.1 - a branch of the program class is present, the control system carries out the TPI- to it recursive entry point 3.4.2. From this point on command executes the following operations shown at 2.8: Set-45, the program 3.4 only executes operations related to the condition code reference 1 (2.8.1), selection of an active action of the interrupt request, request of relevant class (I / O interruptions) which was accepted by executing the TPI command and deleted in accordance with a predetermined priority scheme (2.8.2), as shown above. Since the TPI instruction of the selected request (2.8.3 and storage of the and the associated branch instruction at an “end position” interrupt code, which is associated with the selected request in the program handling the interrupt (2.8.4) Selection of an anau before returning the control to the interrupted Proforderang, deletion of the request and the save program X), and since the TPI command also an interruption of the interrupt code represent a subset of the instruction in the same class 1 (including the PSW-Aus previously discussed interruption, it is clear that there is no exchange), so you can do it by a similar logic. 55 There is a need for Relocation of status information or by using the best information for the operations 1.4.3 or environment information in connection with the TPI-based logic s (re-entrant) Mi command that can be used at any time. Because the TPI command specifically on such end-programs. position in the interrupt treatment program,

After execution of operations 2.8, the control goes to "white" the program in principle, that all operations that belonged to the way through the final operation 2.7 / 1.3.2 (ENDE OP) 60 treatment of the previous interruption, compared to the usual Sequence 1.4 for the conditional acceptance of locks and that the next interruptions to be dealt with that belong to the group of interruptions with a request for refraction due to the currently stored normal treatment. Interrupt code is marked (i.e. the code which

The use of the described command (TPI) was overwritten by the TPI command). For this reason, the supervisor program for handling I / O interruptions does not require a change in the program status, and a minimal one is shown in FIG. 3. A comparative representation of the male program routine used to handle the prior art recursive treatment is interrupted, which is shown by executing the TPI instruction at-in FIG. 4. It is assumed that any program has been done.

In contrast, the first level interrupt handling sequence, shown at 3.4 in Fig. 4, contains a preparatory sub-sequence, see at 4.3.1, which is not needed when the TPI instruction is provided. The preparation sequence temporarily shifts the program status information belonging to program X to a background memory, in preparation for a further program status change, which is described below and which enables the program to "skip" unnecessary program steps during the recursive interrupt handling as it does is described below.

In the interrupt handling sequence 4.4 of the second level, at point 4.4.1, which is achieved when all the operations necessary for the interruption assumed at 4.2 have been carried out, the program contains a command to set an enable bit in the program status word, which at the time is the control of the system certainly. This enables the system to recursively accept an interrupt at 4.6 in the class (I / O) that belongs to the interrupt that was assumed at 4.2. If there are no pending interrupt requests in this class (wait), the program proceeds to instruction 4.5.1 in segment 4.5 of the interrupt handling program at the second level. Command 4.5.1 resets the release bit in the program status word PSW so that a blocking condition arises, and the program at 4.7 starts to restore the environmental conditions that existed in the system prior to the preparatory sequence 4.3.1. When this process has been carried out, the command 4.8 «Load PSW» (LPSW)

5 631820

executed, and the control goes back to the interrupted program X at 4.9.

If an I / O interrupt request is accepted at 4.6, the PSW-5 exchange shown in Fig. 1 at 1.4.3.4 and other operations shown at 1.4 are performed. The «new» program status word PSW contains the address of command 4.3.2, which was used via recursive entry point 4.6.1 in the initial stage of the interrupt handling program at the first level (4.3). As a result of the preparations made in 4.3.1, the interrupt handler can perform the necessary operations, but it skips a number of steps that are normally performed during the first pass through the program from acceptance level 4.2. Finally, the program returns to release level 4.4.1 and continues to lock level 4.5.1 if no further interruptions are assumed at 4.6. Operations 4.7.1 to restore the environmental conditions bring the program into the same state as it was before preparatory operations 4.3.1. The control is then returned to the interrupted program X by executing the «Load PSW» command at 4.8. By comparing FIGS. 3 and 4, it becomes completely clear that the TPI command (query for waiting interruption request) avoids the program status change at 4.6, and that the preparatory processes 4.3.1 and the restoring processes 4.7. 1 can be eliminated, thereby shortening the interrupt handling program.

C.

1 sheet of drawings

Claims (2)

631 820 PATENT CLAIMS 1. A method for handling interrupt requests in a multiprogrammable data processing system, which runs under the control of a monitoring program, in which system status data are exchanged between the processor and main memory in order to switch from a running program to another, characterized in that before the transition from a program, which has been activated for handling an interrupt request of a certain class, a query operation is carried out on another program to determine whether there are at least one further interrupt request of the same class and, if this is the case, such an interrupt request is deleted as a waiting request and it is thus accepted that an identification feature for the device requesting the interruption is stored in a location accessible to the program in question, and that said accepted interruption request treatment without exchanging status data by means of a return in the currently active program. 2. Arrangement for performing the method according to claim 1, characterized in that an additional decoding output (2.2.1) is provided in the operations decoder (2.2) of the processor of the data processing system, which is excited when the operating part of a specific command (TPI) is decoded , and that devices for carrying out the interrogation operation, for accepting an interrupt request, for storing the identification feature and for returning to the currently active program are all activated as a function of a decoding signal on this additional decoding output.