Asynchronism. Asynchronism. Asynchronism. Interrupts. Asynchronism. Interrupts. Asynchronism close to hardware

Transcription

1 474 Real-Time & Embedded 5 Systems 2015 Uwe R. Zimmer - The Australian National University 476 close to hardware Interrupts Required mechanisms for interrupt driven programming: Interrupt control: Grouping, encoding, prioritising, and en-/disabling interrupt sources switching: Mechanisms for cpu-state saving and restoring + task-switching Interrupt identification: Interrupt vectors, interrupt states Hardware supported Implemented in software 2015 Uwe R. Zimmer, The Australian National University page 476 of 962 (chapter 5: up to page 620) 475 References [ Barnes2006 ] Barnes, John Programming in Ada 2005 Addison-Wesley, Pearson education, ISBN , Harlow, England 2006 [ Burns2007 ] Burns, A & Wellings, A Concurrent and Real-Time Programming in Ada, edition Cambridge University Press 2007 [ Dibble2009 ] Dibble, Peter RTSJ JSR-282 Early Draft Review version pp [ NN2004 ] The Real-Time Java Platform A Technical White Paper June pp [AdaRM2012] Ada Reference Manual - Language and Standard Libraries; ISO/IEC 8652:201x (E) 2015 Uwe R. Zimmer, The Australian National University page 475 of 962 (chapter 5: up to page 620) 477 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level 2015 Uwe R. Zimmer, The Australian National University page 477 of 962 (chapter 5: up to page 620)

2 478 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level 2015 Uwe R. Zimmer, The Australian National University page 478 of 962 (chapter 5: up to page 620) 480 LM12L458 Register bank A4 A3 A2 A1 Purpose Type D15 D14 D13 D12 D11 D10 D9 D8 D7 D6 D5 D4 D3 D2 D1 D Instruction RAM R/W Acquisition Watch- 0 to (RAM Pointer = 00) Time dog 8/12 Timer Sync V VIN+ Pause Loop Instruction RAM R/W 0 to (RAM Pointer = 0 1) Don t Car e >/< S ign Limit # Instruction RAM R/W 0 to (RAM Pointer = 1 0) Don t Car e >/< S ign Limit # Configuration R/W Test RAM I/O Auto Chan Stand- Full Auto- Reset Start Don t Care DIAG Register = 0 Pointer Sel Zero ec Mask by CAL Zero Interrupt Enable R/W Number of Conversions Sequencer INT7 Don t INT5 INT4 INT3 INT2 INT1 INT Register in Conversion FIFO Address to Care to Generate INT2 Generate INT1 Address R Actual Number of of INST7 0 INST5 INST4 INST3 INST2 INST1 INST Interrupt Status Conversion Results Sequencer Register in Conversion FIFO Instruction being Executed Timer R/W Timer Preset High Byte Timer Preset Low Byte Register Conversion R Address Sign Conversion Conversion Data: LSBs FIFO or Sign Data: MSBs Limit Status R Limit # 2 Status Limit #1: Status Register 2015 Uwe R. Zimmer, The Australian National University page 480 of 962 (chapter 5: up to page 620) 479 A/D, D/A & Interfaces: Examples LM12L458 (National Semiconductors) 2015 Uwe R. Zimmer, The Australian National University page 479 of 962 (chapter 5: up to page 620) 481 LM12L458 Data transfer options A4 A3 A2 A1 Purpose Type D15 D14 D13 D12 D11 D10 D9 D8 D7 D6 D5 D4 D3 D2 D1 D0 Interrupt Enable R/W Number of Conversions Sequencer INT7 Don t INT5 INT4 INT3 INT2 INT1 INT Register in Conversion FIFO Address to Care to Generate INT2 Generate INT1 Address R Actual Number of of INST7 0 INST5 INST4 INST3 INST2 INST1 INST Interrupt Status Conversion Results Sequencer Register in Conversion FIFO Instruction being Executed Select interrupt source (interrupt enable register, 15 bits): Watchdog (INT0): limit conditions are fulfilled. Instruction (INT1): the instruction pointer equals a pre-programmed value (bits 8-10). Conversions (INT2): a specified number of conversions (bits 11-15) have been performed. Auto-Zero (INT3): short calibration has been performed. Full-Calibration (INT4): long calibration has been performed. Pause (INT5): Sequencer arrived at a pause state. Active (INT7): Controller returned from power-down to active mode Uwe R. Zimmer, The Australian National University page 481 of 962 (chapter 5: up to page 620) :

3 482 LM12L458 Data transfer options A4 A3 A2 A1 Purpose Type D15 D14 D13 D12 D11 D10 D9 D8 D7 D6 D5 D4 D3 D2 D1 D0 Interrupt Enable R/W Number of Conversions Sequencer INT7 Don t INT5 INT4 INT3 INT2 INT1 INT Register in Conversion FIFO Address to Care to Generate INT2 Generate INT1 Address R Actual Number of of INST7 0 INST5 INST4 INST3 INST2 INST1 INST Interrupt Status Conversion Results Sequencer Register in Conversion FIFO Instruction being Executed Read interrupt status (interrupt status register, 15 bits): Indicates the currently active interrupts Shows the actual number of conversions and the currently processed instruction. Reading this register resets all status bits (bits 0-5, 7) Uwe R. Zimmer, The Australian National University page 482 of 962 (chapter 5: up to page 620) 484 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level 2015 Uwe R. Zimmer, The Australian National University page 484 of 962 (chapter 5: up to page 620) 483 A/D, D/A & Interfaces: Examples LM12L458 (National Semiconductors) Single interrupt line Identifying the actual source requires an additional read cycle 2015 Uwe R. Zimmer, The Australian National University page 483 of 962 (chapter 5: up to page 620) 485 close to hardware Interrupts System interrupt controller: Collects the interrupt signal lines from all managed devices. [Identifies the device status], if not given already as an interrupt vector index. Encodes all interrupt signals into a common interrupt vector or status scheme. Orders and masks all interrupts according to priority levels. Triggers an actual CPU interrupt. Three common forms: Embedded interrupt controller as an intrinsic part of a complex micro-controller. Dedicated interrupt controller for a set of similar or identical devices (e.g. a hard disk array). Universal interrupt controller (usually unable to fetch interrupt status information) Uwe R. Zimmer, The Australian National University page 485 of 962 (chapter 5: up to page 620)

4 486 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level Come on three levels: Interrupt service routines: Does not involve processes. Communicating interrupts to processes: Synchronize to events or data structures. Transforming interrupts to signals: Notify the scheduler that a process is to be activated Uwe R. Zimmer, The Australian National University page 486 of 962 (chapter 5: up to page 620) 488 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level Come on three levels: Interrupt service routines: Does not involve processes. Communicating interrupts to processes: Synchronize to events or data structures. Transforming interrupts to signals: Notify the scheduler that a process is to be activated Uwe R. Zimmer, The Australian National University page 488 of 962 (chapter 5: up to page 620) 487 close to hardware Interrupts (at the operating system level) Come on three levels: Interrupt service routines: Does not involve processes. All activities under strict real-time constraints will be addressed here. Communicating interrupts to processes: Synchronize to events or data structures. After the reactive part of the response has been addressed, processes might need to be notified about the current state. Transforming interrupts to signals: Notify a process via the scheduler. For activities which are not under real-time constraints, the interrupt can be translated into a signal (scheduler-level event) Uwe R. Zimmer, The Australian National University page 487 of 962 (chapter 5: up to page 620) 489 close to hardware Interrupt service routines (available in real-time operating systems and languages only) Purpose: Allow full access to the interrupt controller (interrupt vectors, priorities) and device. Activate a routine in a predictable amount of time. Cannot operate on the level of threads or tasks! Limitations regarding the accessibility of some OS-facilities (task level system calls) Uwe R. Zimmer, The Australian National University page 489 of 962 (chapter 5: up to page 620)

5 490 close to hardware Interrupt service routines (available in real-time operating systems and languages only) Some VxWorks OS entries: intconnect Connect a routine to an interrupt vector intlevelset Set the interrupt mask level IntLock Disable interrupts (besides NMI) intunlock Enable interrupts intvecset Set the interrupt vector base address intvecget Get the interrupt vector base address intvecset Set an interrupt vector intvecget Get an interrupt vector 2015 Uwe R. Zimmer, The Australian National University page 490 of 962 (chapter 5: up to page 620) 492 close to hardware Interrupt service routines (available in real-time operating systems and languages only) Initial response to interrupt (executed by CPU without any software involved): save (push) essential CPU registers (IP, condition flags, ()) jump to the (vectorized) interrupt service routine Further detail: Interrupt masks are typically raised to the current level by hardware, yet can be readjusted inside the wrapper code or service routine. Optional/common software wrapper (added by the os / runtime environment): save (push) remaining/needed CPU registers save (push) stack-frame Execute interrupt service routine code Reentrant interrupt service routines. restore (pop) stack-frame restore (pop) all remaining/touched CPU registers Special CPU instruction (also added by the os / runtime environment): restore (RTE/IRET) IP, condition flags, () 2015 Uwe R. Zimmer, The Australian National University page 492 of 962 (chapter 5: up to page 620) 491 close to hardware Interrupt service routines (available in real-time operating systems and languages only) Initial response to interrupt (executed by CPU without any software involved): save (push) essential CPU registers (IP, condition flags, ()) jump to the (vectorized) interrupt service routine Optional/common software wrapper (added by the os / runtime environment): save (push) remaining/needed CPU registers save (push) stack-frame Execute interrupt service routine code restore (pop) stack-frame restore (pop) all remaining/touched CPU registers Special CPU instruction (also added by the os / runtime environment): restore (RTE/IRET) IP, condition flags, () 2015 Uwe R. Zimmer, The Australian National University page 491 of 962 (chapter 5: up to page 620) 493 Code management Processor Architectures Interrupts Int. IP Decoder Sequencer One or multiple lines wired directly into the sequencer Required for: Pre-emptive scheduling, Timer driven actions, Transient hardware interactions, Usually preceded by an external logic ( interrupt controller ) which accumulates and encodes all external requests. Memory ALU Data management On interrupt (if unmasked): CPU stops normal sequencer flow. Lookup of interrupt handler s address Current IP and state pushed onto stack. IP set to interrupt handler Uwe R. Zimmer, The Australian National University page 493 of 705 (chapter 5: up to page 620)

6 494 Program Interrupt processing Interrupt handler Stack Code FP 2015 Uwe R. Zimmer, The Australian National University page 494 of 962 (chapter 5: up to page 620) 496 Program Stack Interrupt processing Interrupt handler Push registers Declare local variables Code Local variables 2015 Uwe R. Zimmer, The Australian National University page 496 of 962 (chapter 5: up to page 620) 495 Interrupt handler Program Interrupt processing Stack Code FP 2015 Uwe R. Zimmer, The Australian National University page 495 of 962 (chapter 5: up to page 620) 497 Program Code Stack Local variables Interrupt processing Interrupt handler Push registers Declare local variables Run handler code.. do some I/O.... or run some time critical code Uwe R. Zimmer, The Australian National University page 497 of 962 (chapter 5: up to page 620)

7 498 Program Code Stack Interrupt processing Interrupt handler Push registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables 2015 Uwe R. Zimmer, The Australian National University page 498 of 962 (chapter 5: up to page 620) 500 Program Code Stack FP Interrupt processing Interrupt handler Push registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables Pop registers Return from interrupt 2015 Uwe R. Zimmer, The Australian National University page 500 of 962 (chapter 5: up to page 620) 499 Program Code Stack FP Interrupt processing Interrupt handler Push registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables Pop registers 2015 Uwe R. Zimmer, The Australian National University page 499 of 962 (chapter 5: up to page 620) 501 Program Interrupt processing Interrupt handler Stack Code FP 2015 Uwe R. Zimmer, The Australian National University page 501 of 962 (chapter 5: up to page 620)

8 502 Interrupt handler Program Interrupt processing Stack Code FP Scratch registers LR is loaded with a special value 2015 Uwe R. Zimmer, The Australian National University page 502 of 962 (chapter 5: up to page 620) 504 Program Code Stack Local variables Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Push other registers Declare local variables Scratch registers 2015 Uwe R. Zimmer, The Australian National University page 504 of 962 (chapter 5: up to page 620) 503 Program Stack Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Code FP Scratch registers 2015 Uwe R. Zimmer, The Australian National University page 503 of 962 (chapter 5: up to page 620) 505 Program Code Stack Local variables Scratch registers Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Push other registers Declare local variables Run handler code.. do some I/O.... or run some time critical code Uwe R. Zimmer, The Australian National University page 505 of 962 (chapter 5: up to page 620)

9 506 Program Code Stack Scratch registers Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Push other registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables Pop other registers 2015 Uwe R. Zimmer, The Australian National University page 506 of 962 (chapter 5: up to page 620) 508 Program Code Stack FP Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Push other registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables Pop other registers Return ("bx lr") 2015 Uwe R. Zimmer, The Australian National University page 508 of 962 (chapter 5: up to page 620) 507 Program Code Stack Scratch registers Interrupt processing Interrupt handler Clear interrupt flag (Adjust priorities) (Re-enable interrupt) Push other registers Declare local variables Run handler code.. do some I/O.... or run some time critical code.. Remove local variables Pop other registers Return ("bx lr") 2015 Uwe R. Zimmer, The Australian National University page 507 of 962 (chapter 5: up to page 620) 509 Interrupt handler Things to consider Interrupt handler code can be interrupted as well. Are you allowing to interrupt an interrupt handler with an interrupt on the same priority level (e.g. the same interrupt)? Can you overrun a stack with interrupt handlers? 2015 Uwe R. Zimmer, The Australian National University page 509 of 962 (chapter 5: up to page 620)

10 510 Multiple programs If we can execute interrupt handler code concurrently to our main program: Can we then also have multiple main programs? 2015 Uwe R. Zimmer, The Australian National University page 510 of 962 (chapter 5: up to page 620) 512 Process 1 B PID switch Dispatcher Process 2 B PID Code Stack Code Stack FP switchvariables 2015 Uwe R. Zimmer, The Australian National University page 512 of 962 (chapter 5: up to page 620) 511 switch Dispatcher Process 1 Process 2 B PID B PID Code Stack Code Stack FP switchvariables switchvariables 2015 Uwe R. Zimmer, The Australian National University page 511 of 962 (chapter 5: up to page 620) 513 Process 1 B PID switch Dispatcher Push registers Declare local variables Process 2 B PID Code Stack Code Stack switchvariables 2015 Uwe R. Zimmer, The Australian National University page 513 of 962 (chapter 5: up to page 620)

11 514 Process 1 B PID switch Dispatcher Push registers Declare local variables Store to B 1 Process 2 B PID Code Stack Code Stack switchvariables switchvariables switchvariables 2015 Uwe R. Zimmer, The Australian National University page 514 of 962 (chapter 5: up to page 620) 516 Process 1 B PID switch Code Stack Code Stack switchvariables Dispatcher Push registers Declare local variables Store to B 1 Scheduler Load from B 2 Process 2 B PID 2015 Uwe R. Zimmer, The Australian National University page 516 of 962 (chapter 5: up to page 620) 515 Process 1 B PID switch Code Stack Code Stack switchvariables switchvariables Dispatcher Push registers Declare local variables Store to B 1 Scheduler Process 2 B PID switchvariables 2015 Uwe R. Zimmer, The Australian National University page 515 of 962 (chapter 5: up to page 620) 517 Process 1 B PID switch Code Stack Code Stack Dispatcher Push registers Declare local variables Store to B 1 Scheduler Load from B 2 Remove local variables Process 2 B PID 2015 Uwe R. Zimmer, The Australian National University page 517 of 962 (chapter 5: up to page 620)

12 518 Process 1 B PID switch Code Stack Code Stack switchvariables Dispatcher Push registers Declare local variables Store to B 1 Scheduler Load from B 2 Remove local variables Pop registers Process 2 B PID FP 2015 Uwe R. Zimmer, The Australian National University page 518 of 962 (chapter 5: up to page 620) 520 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level Come on three levels: Interrupt service routines: Does not involve processes. Communicating interrupts to processes: Synchronize to events or data structures. Transforming interrupts to signals: Notify the scheduler that a process is to be activated Uwe R. Zimmer, The Australian National University page 520 of 962 (chapter 5: up to page 620) 519 Process 1 B PID switch Code Stack Code Stack switchvariables Dispatcher Push registers Declare local variables Store to B 1 Scheduler Load from B 2 Remove local variables Pop registers Return from interrupt Process 2 B PID FP 2015 Uwe R. Zimmer, The Australian National University page 519 of 962 (chapter 5: up to page 620) 521 close to hardware Communicating interrupts to processes Interrupt service routine to task communication methods: Shared memory and ring buffers: most low level communication scheme. Semaphore: trigger a semaphore, where a task has been blocked before. Monitors: free a task, which is blocked at a monitor entry (standard Ada-method: protected object). Message queues: Send asynchronous messages to a task (if queue is not full). Pipes: Write to a pipe (if pipe is not full). Signals: indicate an asynchronous task switch to the scheduler. In all cases: The interrupt service routine cannot block! 2015 Uwe R. Zimmer, The Australian National University page 521 of 962 (chapter 5: up to page 620)

13 522 close to hardware Interrupts Interrupt control at the individual device level at the system interrupt controller level at the operating system level Come on three levels: Interrupt service routines: Does not involve processes. Communicating interrupts to processes: Synchronize to events or data structures. Transforming interrupts to signals: Notify the scheduler that a process is to be activated Uwe R. Zimmer, The Australian National University page 522 of 962 (chapter 5: up to page 620) 524 close to hardware Interrupts Signals Signals are originally a process-level synchronization / termination method ( kill ) yet are currently been used for everything from hardware-interrupts and timers to asynchronous task messaging. Signals are passed through the system wide scheduler. Signals carry little information as to what happened. Unpredictable work-around for missing direct hardware interrupt propagation. Make sure that you understand the attached strings, before employing any signals Uwe R. Zimmer, The Australian National University page 524 of 962 (chapter 5: up to page 620) 523 close to hardware Interrupts Signals POSIX b BSD-UNIX signal signal Specify the handler associated with a signal sigaction sigvec Examine or set the signal handler for a signal kill kill Send a signal (overwrite all other pending signals) sigqueue Send a queued signal sigsuspend pause Wait for a signal sigwaitinfo, sigtimedwait Wait for a signal, but do not invoke the handler sigemptyset sigsetmask Manipulate and set the mask of blocked signals sigprocmask sigblock Add to a set of blocked signals 2015 Uwe R. Zimmer, The Australian National University page 523 of 962 (chapter 5: up to page 620) 525 in general program flows Exceptions Wish list for exception handling in real-time programming languages: 1. Exception facilities should not obscure the understanding of the normal, exception-free control flow. 2. Exception facilities should produce no or minimal runtime overhead, until an exception actually occurs. 3. Exceptions should produce predictable run-time overhead when an exception occurs. 4. Exceptions indicated by the run-time environment and by the program itself should be treated uniformly. 5. The exception mechanism should be applicable to asynchronous and synchronous exceptions ( might be hard to achieve with respect to wish 3). 6. The exception mechanism should allow for appropriate recoveries (supply sufficient information and appropriate re-entry possibilities) Uwe R. Zimmer, The Australian National University page 525 of 962 (chapter 5: up to page 620)

14 526 in general program flows Exceptions (Historic exception handling methods) Use an unusual return value and a global variable convention: if (function_call (parameters) < NORMAL_RETURN_VALUE) { if errno == SOME_KNOWN_ERROR { /* react to the known error condition */ else { /* try to improvise something */ else { normal control flow ; Inflexible (exceptions from the environment cannot be detected). Error-prone (lots of chances to forget checking or to use the wrong constants). Obstructive (all fragments are commingled) Uwe R. Zimmer, The Australian National University page 526 of 962 (chapter 5: up to page 620) 528 in general program flows Exceptions (Assembler level) Provide a jump table and manipulate the return address on the stack: jsr pc, PRINT_CHAR jmp IO_ERROR jmp DEVICE_NOT_ENABLED # normal processing Caller: Subroutine: % indicate an exception: % increment the return address on the stack % by the exception number to employ the caller-provided exception handling % indicate normal operation: % increment the return address on the stack by the max. exception number Uwe R. Zimmer, The Australian National University page 528 of 962 (chapter 5: up to page 620) 527 in general program flows Exceptions (Historic exception handling methods) Use an unusual return value and a global variable convention: if (function_call (parameters) < NORMAL_RETURN_VALUE) UE { if errno == SOME_KNOWN_ERROR OR { /* react to the known error ror condition */ else { /* try to improvise something */ ; else { normal control ol flow Inflexible (exceptions from the environment cannot be detected). Error-prone (lots of chances to forget checking or to use the wrong constants). Obstructive (all fragments are commingled) Uwe R. Zimmer, The Australian National University page 527 of 962 (chapter 5: up to page 620) 529 in general program flows Exceptions (Historic languages) Emulate exception handling methods by returning a: Jump label for unrecoverable exceptions. Procedure variable for recoverable exceptions. Outdated as all current real-time programming languages provide some intrinsic means of exception handling. All exception handling methods up to here operate on subroutine level, while all current day exception handling methods operate on block level Uwe R. Zimmer, The Australian National University page 529 of 962 (chapter 5: up to page 620)

15 530 in general program flows Exceptions (Early high level constructs in not-fully-scoped languages) PL/I (IBM, 1960 s): Exception handling routines are registered globally by running over an On code section: On <exception-name> ; ( statements ) [<long jump to continuation code>] end; Exception handling routines are replaced by running over a new On code section for the same exception-name. Exception handling is concluded by either: Running over the end statement and resume execution with the statement which follows immediately after the one which raised the exception. Executing a long-jump to anywhere. Continuing in the code-scope of/after the exception handler was not considered Uwe R. Zimmer, The Australian National University page 530 of 962 (chapter 5: up to page 620) 532 in general program flows Exception forms raised: from: Run-time environment Application Synchronously Asynchronously 2015 Uwe R. Zimmer, The Australian National University page 532 of 962 (chapter 5: up to page 620) 531 in general program flows Exceptions (Early high level constructs in not-fully-scoped languages) PL/I (IBM, 1960 s): Exception handling routines are registered globally by running over an On code section: On <exception-name> ; ( statements ) [<long jump to continuation code>] end; Exception handling routines are replaced by running over a new On code section for the same exception-name. Exception handling is concluded by either: Running over the end statement and resume execution with the statement which follows immediately after the one which raised the exception. Executing a long-jump to anywhere. Continuing in the code-scope of/after the exception handler was not considered Uwe R. Zimmer, The Australian National University page 531 of 962 (chapter 5: up to page 620) 533 in general program flows Exception forms Ada raised: from: Run-time environment Application Synchronously Exceptions Asynchronously Interrupt handler Asynchronous transfer of control 2015 Uwe R. Zimmer, The Australian National University page 533 of 962 (chapter 5: up to page 620). Exception handling is messy/hard/impossible in: Weakly/not scoped languages Macro-assembler level languages Strict functional languages

16 534 in general program flows Exception forms Real-time Java raised: from: Run-time environment Application Synchronously Exceptions Asynchronously Asynchronous exceptions 2015 Uwe R. Zimmer, The Australian National University page 534 of 962 (chapter 5: up to page 620) 536 in general program flows Exception granularity (block level) Ada: -- do something dangerous exception when E: Constraint_Error => Deal with it; Handle a specific exception at the end of a defined code block end; or in Real-time Java try { // do something dangerous catch (ExceptionType e) { // handle exception e 2015 Uwe R. Zimmer, The Australian National University page 536 of 962 (chapter 5: up to page 620) 535 in general program flows Exception forms POSIX raised: from: Run-time environment Application Synchronously N/A () Asynchronously (Signals) 2015 Uwe R. Zimmer, The Australian National University page 535 of 962 (chapter 5: up to page 620) 537 in general program flows Exception granularity (block level) Ada: -- do something dangerous exception when E: others => Deal with it; Handle any exception at the end of a defined code block end; Can be combined: handle some specific exceptions first yet also provide code for other exceptions. or in Real-time Java try { // do something dangerous catch (Exception e) { // handle exception e 2015 Uwe R. Zimmer, The Australian National University page 537 of 962 (chapter 5: up to page 620)

17 538 in general program flows Exception granularity (block size large) declare subtype Temperature is Integer range ; subtype Pressure is Integer range ; subtype Flow is Integer range ; -- read temperature sensor and calculate its value -- read pressure sensor and calculate its value -- read flow sensor and calculate its value -- adjust temperature, pressure and flow -- according to requirements exception -- handler for Constraint_Error end; Might be too unspecific? 2015 Uwe R. Zimmer, The Australian National University page 538 of 962 (chapter 5: up to page 620) 540 in general program flows Exception granularity (statement level) e.g. supported in CHILL All exceptions in CHILL need to be handled A : temperature; B, C : integer; A := B + C on Exceptions are not propagated end; (overflow ): ; (rangefail): ; else ; end; Handlers are determined at compile time 2015 Uwe R. Zimmer, The Australian National University page 540 of 962 (chapter 5: up to page 620) 539 in general program flows Exception granularity (block size small) -- read temperature sensor and calculate its value exception -- handler for Constraint_Error end; -- read pressure sensor and calculate its value exception -- handler for Constraint_Error end; Might become too cluttered? exception -- handler for Constraint_Error end; 2015 Uwe R. Zimmer, The Australian National University page 539 of 962 (chapter 5: up to page 620) 541 in general program flows Exception attributes Instead of catching exceptions after each statement: Exceptions can carry further attributes about their source. Thus exception handlers can be further away from the source (and fewer in numbers) yet still be able to identify the fault. Real-time Java: Exception can carry any number of user defined parameters. Ada: The environment automatically attaches additional information to exceptions, which are indicating the position of the exception-occurrence and other observed conditions (implementation dependent, inflexible, useful for debugging) Uwe R. Zimmer, The Australian National University page 541 of 962 (chapter 5: up to page 620)

18 542 in general program flows Exception scope and propagation 1. All procedures and functions declare every potentially raised exception (requested in CHILL, requested in Real-time Java for user defined exceptions) If an appropriate exception handler can not be determined at compile-time: Either treat it as a programmer error and stop compilation ( CHILL) or propagate the exception at run-time outside its static scope ( Real-time Java) 2. Exceptions are declared for whole modules not specifically per method (Ada): The handler is determined at run-time in any case (either by propagation or in the static scope) Uwe R. Zimmer, The Australian National University page 542 of 962 (chapter 5: up to page 620) 544 in general program flows Exception recovery Resumption-model (offered in: Pearl, Mesa): 1. Find an exception handler. 2. Execute the exception handler (and potentially raise another exception). 3. After completion of the exception handler (and possibly other handlers): return to the invoker and try resume processing as if nothing happened. Feature: In case of an asynchronous exception, there is little impact on the current control flow. Problems: Some errors cannot be repaired (especially all timing related errors) infinite recursion. Exceptions can be raised in the middle of evaluations, which will be hard to restore Uwe R. Zimmer, The Australian National University page 544 of 962 (chapter 5: up to page 620) 543 in general program flows (Uncaught) exception propagation If the handler can neither be found by propagation nor in the static context of a task: or Stop the specific task exceptions are not propagated to the parent-process (Ada). Propagate the exception to the parent-task (and potentially stall the whole system). Common safety-line for each process: Use a catch all -handler (possible in Ada and Real-time Java) at the highest task level. common reaction: PANIC! 2015 Uwe R. Zimmer, The Australian National University page 543 of 962 (chapter 5: up to page 620) 545 in general program flows Exception recovery Block-Resumption-model (offered in: Eiffel): 1. Find an exception handler. 2. Execute the exception handler (and potentially raise another exception). 3. After completion of the exception handler (and possibly other handlers): return to the ning of the code block in which the exception occurred. Feature: Contracts are enforced. Problems: must not be re-initialized (potentially confusing state of the system). The code needs to be aware of all combinations of half-evaluated processing states. Trying the same method again (and again) is usually not a suitable way for real-time systems Uwe R. Zimmer, The Australian National University page 545 of 962 (chapter 5: up to page 620)

19 546 in general program flows Exception recovery Termination-model (only model in: Ada, Real-time Java; offered in: Pearl, Mesa): 1. Find an exception handler. 2. Execute the exception handler (and potentially raise another exception). 3. After completion of the exception handler (and possibly other handlers): return to the scope around the final exception handler. Feature: The method of choice, if exceptions imply that the operation was not successful and something else need to be done now real-time systems. Problems: There is no way to continue execution as if nothing happened (might be useful in case that the exception could be identified as of minor impact) Uwe R. Zimmer, The Australian National University page 546 of 962 (chapter 5: up to page 620) 548 in general program flows Multi-stage exception handling Assuming a block is holding a number of local resources, yet occurring exceptions need to be handled at the caller level: procedure Allocate (Number : Devices) is -- request each device be allocated in turn -- noting which requests are granted exception when others => -- e.g. deallocate those devices allocated raise; -- re-raise the exception end Allocate; Helpful to keep a consistent system-state and to avoid dead-locks (all-or-nothing allocation). in Real-time Java: the finally clause takes care of block consistent finalization Uwe R. Zimmer, The Australian National University page 548 of 962 (chapter 5: up to page 620) 547 in general program flows Exception recovery Hybrid-model (offered in: Mesa): 1. Find an exception handler. 2. Execute the exception handler (and potentially raise another exception). 3. After completion of the exception handler (and possibly other handlers): the exception handler can decide whether to terminate or to resume Uwe R. Zimmer, The Australian National University page 547 of 962 (chapter 5: up to page 620) 549 in general program flows Exception handling issues (Ada & Real-time Java) (Ada) Exceptions are declared at package level, i.e. it is unclear, which functions may raise which exceptions! (Ada) Exceptions may be propagated outside the scope of their declaration, i.e. only when others can handle them. (Ada, Real-time Java) Exception in task bodies are not propagated to the parent task i.e. if there could not be any handler identified in the task, the task will die silently. (Ada, Real-time Java) Exception in task declarations are always propagated to the parent task. (Ada) Exceptions in task rendezvous, which are not handled in the accept statement, are propagated to both involved tasks. Most expensive uncaught exception up to now: Half a billion dollars (maiden crash of Ariane 5, 96) 2015 Uwe R. Zimmer, The Australian National University page 549 of 962 (chapter 5: up to page 620)

20 550 in general program flows Exception handling in Real-time Java Errors ThreadDeath AssertionError VirtualMachineError Unchecked Checked to be declared per method LinkageError Throwable RunTimeException Exception User defined User defined User defined InterruptedException User defined Asynchronously- InterruptedException User defined User defined 2015 Uwe R. Zimmer, The Australian National University page 550 of 962 (chapter 5: up to page 620) 552 in general program flows Exception handling in C / POSIX There is none Possible workarounds by using POSIX long jumps or signals. see also macro-assembler or old language exception handling methods Uwe R. Zimmer, The Australian National University page 552 of 962 (chapter 5: up to page 620) 551 in general program flows Exception handling in Real-time Java Exceptions are objects in Real-time Java (Exceptions have a hierarchical relation). Exceptions handlers can catch: One individual exception. Exceptions out of a finite list of exceptions. Exceptions of a class (missing in Ada). All exceptions. The kinds of exceptions which are to be handled can be denoted precisely, completely and safely Uwe R. Zimmer, The Australian National University page 551 of 962 (chapter 5: up to page 620) 553 in general program flows Exception handling overview Real-Time Java Termination Recovery Handler Declared per: Declare as: Scope Ada Termination Dynamical / Propagating Dynamical / Propagating CHILL Termination Static Via handler Mesa Pearl Resumption or termination Resumption or termination Eiffel Class retry Dynamical / Propagating Method Class Block Package Static name Block Procedure Static handler Static procedure Statement / Block / Task Block Static Process Static name Task Dynamical / Propagating C / POSIX None Contract violation Static name Contract

21 554 Atomic actions (Definitions) An action is atomic if the processes performing it are not aware of the existence of any other active process, and no other active process is aware of the activity of the processes during the time the processes are performing the action. do not communicate with other processes while the action is being performed. cannot detect any outside state change and do not reveal their own state changes until the action is complete. can be considered to be indivisible and instantaneous Uwe R. Zimmer, The Australian National University page 554 of 962 (chapter 5: up to page 620) 556 Atomic actions Atomic Actions Atomic Indivisible phases Flow of tasks Task 4 Task 3 Task 2 Task time Commitment times 2015 Uwe R. Zimmer, The Australian National University page 556 of 962 (chapter 5: up to page 620) 555 Atomic actions (Implications) An atomic action is either performed fully, or not at all. is declared as failed, if any part of the action fails. Thus all parts of an atomic action need to be prepared: to be interrupted (due to the failure of one of them). to reset to their initial state at any Uwe R. Zimmer, The Australian National University page 555 of 962 (chapter 5: up to page 620) 557 Nested atomic actions Action actions can be nested, iff all processes involved in the nested atomic actions are a true subset of the processes involved in the enclosing atomic action Uwe R. Zimmer, The Australian National University page 557 of 962 (chapter 5: up to page 620)

22 558 Nested atomic actions Nested Atomic Actions Atomic Task 4 Task 3 Task 2 Task time 2015 Uwe R. Zimmer, The Australian National University page 558 of 962 (chapter 5: up to page 620) 560 Atomic actions failure Failure in one of the tasks Cleaning up in all tasks Action declared failed Atomic Task 4 Task 3 Task 2 Task time 2015 Uwe R. Zimmer, The Australian National University page 560 of 962 (chapter 5: up to page 620) 559 Atomic actions (Requirements for real-time environments) Well-defined boundaries A start, end and a side boundary: Define clear entry and exit points for all processes involved in the atomic action. Processes can enter at different time, but need to be released at once. Separate the involved processes from the rest of the system ( side boundary ). Indivisibility (Isolation) Prohibit or restrict communications to outside processes and resources. Employing results from one atomic action in another one implies strict serialisation. Nesting Atomic actions may be nested, if they form a true enclosing relation. Concurrency Independent atomic actions may be executed in any order and concurrently Uwe R. Zimmer, The Australian National University page 559 of 962 (chapter 5: up to page 620) 561 Failure in outer atomic action. Preserve full atomicity. Atomic Atomic actions failure Failure in outer action Inner action ready to commit Action declared failed Task 4 Task 3 Task 2 Task time Outer action rolling back Inner action rolling back 2015 Uwe R. Zimmer, The Australian National University page 561 of 962 (chapter 5: up to page 620)

23 562 Atomic actions failure Failure in outer atomic action. Action declared failed Break atomicity and propagate failure immediately. Atomic Task 4 Task 3 Task 2 Task time Failure in outer action All parts rolling back 2015 Uwe R. Zimmer, The Australian National University page 562 of 962 (chapter 5: up to page 620) 564 Atomic actions Failure in inner atomic action. Break atomicity and propagate failure immediately. Atomic Task 4 Task 3 Task 2 Task time All parts rolling back immediately 2015 Uwe R. Zimmer, The Australian National University page 564 of 962 (chapter 5: up to page 620) 563 Atomic actions failure Failure in inner atomic action. Preserve full atomicity. Atomic Task 4 Task 3 Task 2 Task time Outer action only notified after inner action rolled back and committed to failure 2015 Uwe R. Zimmer, The Australian National University page 563 of 962 (chapter 5: up to page 620) 565 Atomic actions Construct does not exist in any mainstream language action A with (P2, P3, ) do -- communication is restricted to P2, P3, -- exceptions and timing constraints violations i are propagated -- to all involved processes exception when exception_a => -- recover locally when exception_b => -- recover locally when others => raise atomic_action_failure; -- and fail the action end A; 2015 Uwe R. Zimmer, The Australian National University page 565 of 962 (chapter 5: up to page 620)

24 566 Atomic actions implementations Implemented by dedicated tasks. Client task is blocked during the atomic action. Atomic Task 4 Task 3 Task 2 Task time 2015 Uwe R. Zimmer, The Australian National University page 566 of 962 (chapter 5: up to page 620) 568 Atomic actions in Ada (specification) type Action_Part_Time_Scope is record Start_Delay_Min : Time_Span := Time_Span_Zero; Start_Delay_Max : Time_Span := Time_Span_Last; Max_Elapse : Time_Span := Time_Span_Last; Deadline : Time := Time_Last; end record; type Action_Part_Proc is access procedure; type Action_Part_Procs is record Action, Cleanup : Action_Part_Proc; Scope : Action_Part_Time_Scope; end record; type Action_Parts is array (Positive range <>) of Action_Part_Procs; The Cleanup procedure is meant to restore the initial state! ( undoing all effects of Action) 2015 Uwe R. Zimmer, The Australian National University page 568 of 962 (chapter 5: up to page 620) 567 Atomic actions in Ada with Atomic_Action_Types; use Atomic_Action_Types; generic Actions : in Action_Parts; package Generic_Atomic_Action is procedure Perform; Failure_State, Time_Out_State, Late_Activation_State : exception; end Generic_Atomic_Action; Scope mechanism is employed to limit communication possibilities of atomic action parts. The perform-call is atomic for the caller and invisible for others. Failures in one part are automatically propagated in the whole atomic action Uwe R. Zimmer, The Australian National University page 567 of 962 (chapter 5: up to page 620) 569 Atomic actions in Ada (usage) Actions : Action_Parts (Tasks_Index) := (1 => (Action => Action_Task_1 Access, Cleanup => Cleanup_Task_1 Access, Scope => (Start_Delay_Min => Milliseconds (33), Start_Delay_Max => Milliseconds (133), Max_Elapse => Time_Span_Last, Deadline => Time_Last) ), 2 => package Atomic_Action is new Generic_Atomic_Action (Actions); procedure Perform is Atomic_Action.Perform; exception when end Perform; 2015 Uwe R. Zimmer, The Australian National University page 569 of 962 (chapter 5: up to page 620)

25 570 Atomic actions in Ada (implementation) task body Action_Task is Monitor.Check_In (Task_Id); Monitor.Failed (Task_Id); Actions (Task_Id).Cleanup.all; Atomic_Action.Monitor.Check_Out (Task_Id); delay To_Duration (Actions (Task_Id).Scope.Start_Delay_Max); raise Late_Activation_State; delay To_Duration (Actions (Task_Id).Scope.Start_Delay_Min); end ; 2015 Uwe R. Zimmer, The Australian National University page 570 of 962 (chapter 5: up to page 620) 572 Atomic actions in Ada (implementation) protected Monitor is entry Check_In (Task_Id : in Task_Ids); entry Fail (Condition : in Atomic_Condition); entry Failed (Task_Id : in Task_Ids); -- blocking until Fail is called entry Check_Out (Task_Id : in Task_Ids); -- blocking until all parts are completed or all are cleaned up entry Action_Result (Condition : out Atomic_Condition); private Check_List : Task_List := Check_List_Out; State : Atomic_State := Checking_In; Final_Condition : Atomic_Condition := Succeeded; end Monitor; Check the course page for the complete sources Uwe R. Zimmer, The Australian National University page 572 of 962 (chapter 5: up to page 620) 571 Atomic actions in Ada (implementation) delay until Real_Deadline; -- based on Max_Elapse & Deadline raise Time_Out_State; Actions (Task_Id).Action.all; end ; Atomic_Action.Monitor.Check_Out (Task_Id); exception when Time_Out_State => Monitor.Fail (Time_Out); when Late_Activation_State => Monitor.Fail (Late_Activation); when others => Monitor.Fail (Other_Exception); end; end ; end Action_Task; 2015 Uwe R. Zimmer, The Australian National University page 571 of 962 (chapter 5: up to page 620) 573 Atomic actions implementations Implemented allows for tasks to dedicate themselves. Nested actions are possible, client task not blocked. Atomic Task 4 Task 3 Task 2 Task time Laboratories Uwe R. Zimmer, The Australian National University page 573 of 962 (chapter 5: up to page 620)

26 574 Atomic actions failure recovery Backward error recovery in real-time environments Since once some parts of the action might have failed: Some action-parts might re-execute the same method again, Or execute alternative methods, even if the original method was locally valid. In real-time environments, failed atomic actions often imply irreversible failures: Tracking back and re-trying the same atomic action with modified parameters or methods is rarely useful, considering timing constraints. Backtracking is often physically impossible Uwe R. Zimmer, The Australian National University page 574 of 962 (chapter 5: up to page 620) 576 Asynchronous events and transfer of control Interrupts vs. From interrupts (sub-process level) employed in: communication with slow / asynchronous / sporadic devices. sampling / control loops. closely coupled reflective systems. To asynchronous transfer of control (process level) used in: Error recovery supporting atomic actions and forward recovery. Mode changes sudden changes from normal operations to emergency measures. Partial / imprecise computations whenever timeliness is more important than precision. Operator intervention User triggered mode changes Uwe R. Zimmer, The Australian National University page 576 of 962 (chapter 5: up to page 620) 575 Atomic actions failure recovery Forward error recovery in real-time environments A mode change and a different set of (atomic) actions (and goals!) will be more useful in many cases. Forward error recovery is more common in real-time systems. (more on forward error recovery in chapter 9 ) 2015 Uwe R. Zimmer, The Australian National University page 575 of 962 (chapter 5: up to page 620) 577 Asynchronous events and transfer of control Real-time Java Supplies three basic features: 1. Binding of handlers to internal and external asynchronous events. 1a. Specifying asynchronous events 1b. Specifying adequate handlers 2. Asynchronous transfer of control. 3. Asynchronous termination of threads Uwe R. Zimmer, The Australian National University page 577 of 962 (chapter 5: up to page 620)

27 578 Asynchronous events Real-time Java: Asynchronous events Whenever an instance of AsyncEvent occurs: all.run() methods of all instances of AsyncEventHandler, which are bound to this AsyncEvent are scheduled for execution. Multiple AsyncEvents may have different impacts on the scheduling. An event counter (FireCount) is supplied. AsyncEvents and AsyncEventHandlers may be created and used by any program logic. More than one handler can be attached to one event. More than one event can be attached to one handler. Flexible, but handled as a schedulable object 2015 Uwe R. Zimmer, The Australian National University page 578 of 962 (chapter 5: up to page 620) 580 Asynchronous events Real-time Java: Asynchronous events public abstract class AsyncEventHandler implements Schedulable { public AsyncEventHandler (Scheduling scheduling, Release release, Memory memory, MemoryArea area, ProcessingGroup group); public void addtofeasibility (); public void removefromfeasibility (); protected int getandclearpendingfirecount (); public abstract void handleasyncevent (); // called by run while FireCount /= 0 public final void run (); 2015 Uwe R. Zimmer, The Australian National University page 580 of 962 (chapter 5: up to page 620) 579 Asynchronous events Real-time Java: Asynchronous events public class AsyncEvent { public AsyncEvent (); public synchronized void addhandler (AsyncEventHandler handler); public synchronized void removehandler (AsyncEventHandler handler); public void sethandler (AsyncEventHandler handler); public void bindto (java.lang.string happening); public Release createrelease (); public void fire (); 2015 Uwe R. Zimmer, The Australian National University page 579 of 962 (chapter 5: up to page 620) 581 Asynchronous events Real-time Java: Asynchronous events public class BoundAsyncEventHandler extends AsyncEventHandler { public BoundAsyncEventHandler (); // inherit modes from the current thread public BoundAsyncEventHandler (Scheduling scheduling, Release release, Memory memory, MemoryArea area, ProcessingGroup group, boolean nonheap, java.lang.runnable logic); bind the current thread to an AsyncEventHandler (otherwise the AsyncEventHandler will execute in a thread of its own) Uwe R. Zimmer, The Australian National University page 581 of 962 (chapter 5: up to page 620)

28 582 Asynchronous events Java: Interrupting exception While an AsyncEvent just schedules a handler for execution, other event classes are needed to alter the control flow more directly: Standard Java: The InterruptedException indicates the wish to interrupt a thread, which itself need to poll the isinterrupted method to find out, whether it is supposed to be interrupted. If the thread is currently executing and ignoring the flag: No effect on the actual controlflow! If the thread, which is to be interrupted is currently blocking: It is activated and receives an InterruptedException. Too weak to be employed in an asynchronous transfer of control! 2015 Uwe R. Zimmer, The Australian National University page 582 of 962 (chapter 5: up to page 620) 584 Asynchronous events Real-time Java: Asynchronously interrupting exception Cases of operations if a AsynchronouslyInterruptedException (AIE) occurs: 1. If control is in a synchronized section (within an interruptible section): AIE is put into a pending state. 2. If control is in an interruptible section: Control is transferred to the nearest dynamically enclosing catch clause, which is in a non-interruptible section. 3. If control is in wait, sleep, or join: Thread is activated and 1. or 2. is followed. 4. If control is in a non-interruptible section: Nothing happens until an interruptible section is reached. 5. If an interruptible section is reached while propagating an exception: Original exception is discarded (!) and replaced by the AIE Uwe R. Zimmer, The Australian National University page 584 of 962 (chapter 5: up to page 620) 583 Asynchronous events Real-time Java: Asynchronously interrupting exception While an AsyncEvent just schedules a handler for execution, other event classes are needed to alter the control flow more directly: Real-time Java: the AsynchronouslyInterruptedException might intercept a control flow directly, while: The code regions, which are interruptible need to be indicated. Synchronized blocks, task creations and finalizations are not interruptible. The response time of a thread must be within a bounded number of bytecodes (which itself need to be documented). If the interruptible thread is currently blocked (in a java.io.* operation) then the thread is either released or kept in the block (definition of a reasonable blocking time or situation is implementation dependent) Uwe R. Zimmer, The Australian National University page 583 of 962 (chapter 5: up to page 620) 585 Asynchronous events Real-time Java: Asynchronously interrupting exception Need to be part of a RealtimeThread public class AsynchronouslyInterruptedException extends java.lang.interruptedexception { public synchronized void disable(); public synchronized boolean enable(); public synchronized boolean fire(); Will propagate any other exception immediately public boolean dointerruptible (Interruptible logic); public boolean happened (boolean propagate); public static AsynchronouslyInterruptedException n getgeneric(); // returns the AsynchronouslyInterruptedException ption which // is generated when RealtimeThread.interrupt t () is invoked public void propagate(); 2015 Uwe R. Zimmer, The Australian National University page 585 of 962 (chapter 5: up to page 620)

29 586 Asynchronous events Real-time Java: Asynchronously interrupting exception import NonInterruptibleServices.*; public class InterruptibleService Interruptible code section { public AIE stopnow = AIE.getGeneric(); public boolean Service() throws AIE { try { // code interdispersed with calls to NonInterruptibleServices catch AIE AI { if (stopnow.happened (true)) { // handle the ATC Handle the stopnow propagate anything else 2015 Uwe R. Zimmer, The Australian National University page 586 of 962 (chapter 5: up to page 620) 588 Asynchronous events Real-time Java: Asynchronously interrupting exception While asynchronously interrupting event handling is part of the standard exception handling mechanism, there are nevertheless differences in exception propagation: A standard exception is not propagated by default when caught by any catch statement (an explicit re-raising of the exception is necessary to pass it on). A AsynchronouslyInterruptedException is propagated by default, even if caught (an explicit propagation stop is necessary to avoid its further propagation). realized in the AIE.happened method Rationale: Local catch-all clauses might not be aware of a potential asynchronous interrupting exception. Practical compromise, but destroying some integrity of the Java-exception concept Uwe R. Zimmer, The Australian National University page 588 of 962 (chapter 5: up to page 620) 587 Asynchronous events Real-time Java: Asynchronously interrupting exception import NonInterruptibleServices.*; public class InterruptibleService Interruptible code section { public AIE stopnow = AIE.getGeneric(); public boolean Service() throws AIE { try { // code interdispersed with calls to NonInterruptibleServices catch AIE AI { if (stopnow.happened (false)) { // handle the ATC else { // cleanup AI.propagate Handle the stopnow otherwise: clean up and propagate anything else 2015 Uwe R. Zimmer, The Australian National University page 587 of 962 (chapter 5: up to page 620) 589 Asynchronous events Real-time Java: Interruptible interface public interface Interruptible { public void interruptaction (AsynchronouslyInterruptedException exception); public void run (AsynchronouslyInterruptedException exception) throws AsynchronouslyInterruptedException; An object may declare an interruptible method explicitly by implementing the above interface. Sole purpose: pass this implementation to a dointerruptible method of a AIE class. The AIE.doInterruptible can only be called by one thread at a time! 2015 Uwe R. Zimmer, The Australian National University page 589 of 962 (chapter 5: up to page 620)

30 590 Asynchronous events Real-time Java: Timeout on actions public class Timed extends AsynchronouslyInterruptedException implements java.io.serializable { public Timed (HighResolutionTime time) throws IllegalArgumentException; public boolean dointerruptible (Interruptible logic); public void resettime (HighResolutionTime time); The timer is started sometime between the invocation of the dointerruptible itself and the.run method of the interruptible interface. A generic interrupt() is thrown at the expiration of the timer. time can be absolute or relative Uwe R. Zimmer, The Australian National University page 590 of 962 (chapter 5: up to page 620) 592 Interrupts Ada: Interrupt handlers package Ada.Interrupts is type Interrupt_ID is implementation-defined; type Parameterless_Handler is access protected procedure; function Is_Reserved (Interrupt : Interrupt_ID) return Boolean; function Is_Attached (Interrupt : Interrupt_ID) return Boolean; function Current_Handler (Interrupt : Interrupt_ID) return Parameterless_Handler; procedure Attach_Handler (New_Handler : in Parameterless_Handler; Interrupt : in Interrupt_ID); procedure Exchange_Handler (Old_Handler : out Parameterless_Handler; New_Handler : in Parameterless_Handler; Interrupt : in Interrupt_ID); procedure Detach_Handler (Interrupt : in Interrupt_ID); function Reference (Interrupt : Interrupt_ID) return System.Address; end Ada.Interrupts; 2015 Uwe R. Zimmer, The Australian National University page 592 of 962 (chapter 5: up to page 620) 591 Forms of asynchronism Ada Provides: Exception handling (synchronous only). Asynchronous transfer of control. Task aborts. Interrupt handling (close to hardware). Set of different methods to handle different kinds of events! 2015 Uwe R. Zimmer, The Australian National University page 591 of 962 (chapter 5: up to page 620) 593 Interrupts Ada: Interrupt handlers package Ada.Interrupts is type Interrupt_ID is implementation-defined; type Parameterless_Handler is access protected procedure; function Is_Reserved (Interrupt rupt : Interrupt_ID) return Boolean; function Is_Attached (Interrupt rupt : Interrupt_ID) ID) return Boolean; function Current_Handler (Interrupt : Interrupt_ID) return Parameterless_Handler; procedure Attach_Handler (New_Handler er : in Parameterless_Handler; Interrupt : in Interrupt_ID); procedure Exchange_Handler (Old_Handler : out Parameterless_Handler; rless_handler; New_Handler er : in Parameterless_Handler; Protected procedures need to qualify as an interrupt handler: use pragma Interrupt_Handler let the compiler evaluate the suitability. Interrupt : in Interrupt_ID); procedure Detach_Handler (Interrupt : in Interrupt_ID); function Reference (Interrupt : Interrupt_ID) return System.Address; end Ada.Interrupts; 2015 Uwe R. Zimmer, The Australian National University page 593 of 962 (chapter 5: up to page 620)