summaryrefslogblamecommitdiffstats
path: root/ada_user/constant_bandwidth_server_scheduler_api.rst
blob: 00733f4d85a8809da024dfae8704d7fe8610d518 (plain) (tree)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484



































































































































































































































































































































































































































































































                                                                                                                                            
Constant Bandwidth Server Scheduler API
#######################################

.. index:: cbs

Introduction
============

Unlike simple schedulers, the Constant Bandwidth Server (CBS) requires
a special API for tasks to indicate their scheduling parameters.
The directives provided by the CBS API are:

- ``rtems.cbs_initialize`` - Initialize the CBS library

- ``rtems.cbs_cleanup`` - Cleanup the CBS library

- ``rtems.cbs_create_server`` - Create a new bandwidth server

- ``rtems.cbs_attach_thread`` - Attach a thread to server

- ``rtems.cbs_detach_thread`` - Detach a thread from server

- ``rtems.cbs_destroy_server`` - Destroy a bandwidth server

- ``rtems.cbs_get_server_id`` - Get an ID of a server

- ``rtems.cbs_get_parameters`` - Get scheduling parameters of a server

- ``rtems.cbs_set_parameters`` - Set scheduling parameters of a server

- ``rtems.cbs_get_execution_time`` - Get elapsed execution time

- ``rtems.cbs_get_remaining_budget`` - Get remainig execution time

- ``rtems.cbs_get_approved_budget`` - Get scheduler approved execution time

Background
==========

Constant Bandwidth Server Definitions
-------------------------------------
.. index:: CBS parameters

.. index:: rtems_cbs_parameters

The Constant Bandwidth Server API enables tasks to communicate with
the scheduler and indicate its scheduling parameters. The scheduler
has to be set up first (by defining ``CONFIGURE_SCHEDULER_CBS`` macro).

The difference to a plain EDF is the presence of servers.
It is a budget aware extention of the EDF scheduler, therefore, tasks
attached to servers behave in a similar way as with EDF unless they
exceed their budget.

The intention of servers is reservation of a certain computation
time (budget) of the processor for all subsequent periods. The structure``rtems_cbs_parameters`` determines the behavior of
a server. It contains ``deadline`` which is equal to period,
and ``budget`` which is the time the server is allowed to
spend on CPU per each period. The ratio between those two parameters
yields the maximum percentage of the CPU the server can use
(bandwidth). Moreover, thanks to this limitation the overall
utilization of CPU is under control, and the sum of bandwidths
of all servers in the system yields the overall reserved portion
of processor. The rest is still available for ordinary tasks that
are not attached to any server.

In order to make the server effective to the executing tasks,
tasks have to be attached to the servers. The``rtems_cbs_server_id`` is a type denoting an id of a server
and ``rtems_id`` a type for id of tasks.

Handling Periodic Tasks
-----------------------
.. index:: CBS periodic tasks

Each task’s execution begins with a default background priority
(see the chapter Scheduling Concepts to understand the concept of
priorities in EDF). Once you decide the tasks should start periodic
execution, you have two possibilities. Either you use only the Rate
Monotonic manager which takes care of periodic behavior, or you declare
deadline and budget using the CBS API in which case these properties
are constant for all subsequent periods, unless you change them using
the CBS API again. Task now only has to indicate and end of
each period using ``rtems_rate_monotonic_period``.

Registering a Callback Function
-------------------------------
.. index:: CBS overrun handler

In case tasks attached to servers are not aware of their execution time
and happen to exceed it, the scheduler does not guarantee execution any
more and pulls the priority of the task to background, which would
possibly lead to immediate preemption (if there is at least one ready
task with a higher pirority). However, the task is not blocked but a
callback function is invoked. The callback function
(``rtems_cbs_budget_overrun``) might be optionally registered upon
a server creation (``rtems_cbs_create_server``).

This enables the user to define what should happen in case of budget
overrun. There is obviously no space for huge operations because the
priority is down and not real time any more, however, you still can at
least in release resources for other tasks, restart the task or log an
error information. Since the routine is called directly from kernel,
use ``printk()`` instead of ``printf()``.

The calling convention of the callback function is:

Limitations
-----------
.. index:: CBS limitations

When using this scheduler you have to keep in mind several things:

- it_limitations

- In the current implementation it is possible to attach only
  a single task to each server.

- If you have a task attached to a server and you voluntatily
  block it in the beginning of its execution, its priority will be
  probably pulled to background upon unblock, thus not guaranteed
  deadline any more. This is because you are effectively raising
  computation time of the task. When unbocking, you should be always
  sure that the ratio between remaining computation time and remaining
  deadline is not higher that the utilization you have agreed with the
  scheduler.

Operations
==========

Setting up a server
-------------------

The directive ``rtems_cbs_create_server`` is used to create a new
server that is characterized by ``rtems_cbs_parameters``. You also
might want to register the ``rtems_cbs_budget_overrun`` callback
routine. After this step tasks can be attached to the server. The directive``rtems_cbs_set_parameters`` can change the scheduling parameters
to avoid destroying and creating a new server again.

Attaching Task to a Server
--------------------------

If a task is attached to a server using ``rtems_cbs_attach_thread``,
the task’s computation time per period is limited by the server and
the deadline (period) of task is equal to deadline of the server which
means if you conclude a period using ``rate_monotonic_period``,
the length of next period is always determined by the server’s property.

The task has a guaranteed bandwidth given by the server but should not
exceed it, otherwise the priority is pulled to background until the
start of next period and the ``rtems_cbs_budget_overrun`` callback
function is invoked.

When attaching a task to server, the preemptability flag of the task
is raised, otherwise it would not be possible to control the execution
of the task.

Detaching Task from a Server
----------------------------

The directive ``rtems_cbs_detach_thread`` is just an inverse
operation to the previous one, the task continues its execution with
the initial priority.

Preemptability of the task is restored to the initial value.

Examples
--------

The following example presents a simple common use of the API.

You can see the initialization and cleanup call here, if there are
multiple tasks in the system, it is obvious that the initialization
should be called before creating the task.

Notice also that in this case we decided to register an overrun handler,
instead of which there could be ``NULL``. This handler just prints
a message to terminal, what else may be done here depends on a specific
application.

During the periodic execution, remaining budget should be watched
to avoid overrun.
.. code:: c

    void overrun_handler (
    rtems_cbs_server_id server_id
    )
    {
    printk( "Budget overrun, fixing the task\\n" );
    return;
    }
    rtems_task Tasks_Periodic(
    rtems_task_argument argument
    )
    {
    rtems_id          rmid;
    rtems_cbs_server_id server_id;
    rtems_cbs_parameters params;
    params.deadline = 10;
    params.budget = 4;
    rtems_cbs_initialize();
    rtems_cbs_create_server( &params, &overrun_handler, &server_id )
    rtems_cbs_attach_thread( server_id, SELF );
    rtems_rate_monotonic_create( argument, &rmid );
    while ( 1 ) {
    if (rtems_rate_monotonic_period(rmid, params.deadline)==RTEMS_TIMEOUT)
    break;
    /* Perform some periodic action \*/
    }
    rtems_rate_monotonic_delete( rmid );
    rtems_cbs_cleanup();
    exit( 1 );
    }

Directives
==========

This section details the Constant Bandwidth Server’s directives.
A subsection is dedicated to each of this manager’s directives
and describes the calling sequence, related constants, usage,
and status codes.

CBS_INITIALIZE - Initialize the CBS library
-------------------------------------------
.. index:: initialize the CBS library

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful initialization
``RTEMS.CBS_ERROR_NO_MEMORY`` - not enough memory for data

**DESCRIPTION:**

This routine initializes the library in terms of allocating necessary memory
for the servers. In case not enough memory is available in the system,``RTEMS.CBS_ERROR_NO_MEMORY`` is returned, otherwise``RTEMS.CBS_OK``.

**NOTES:**

Additional memory per each server is allocated upon invocation of``rtems_cbs_create_server``.

Tasks in the system are not influenced, they still keep executing
with their initial parameters.

CBS_CLEANUP - Cleanup the CBS library
-------------------------------------
.. index:: cleanup the CBS library

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - always successful

**DESCRIPTION:**

This routine detaches all tasks from their servers, destroys all servers
and returns memory back to the system.

**NOTES:**

All tasks continue executing with their initial priorities.

CBS_CREATE_SERVER - Create a new bandwidth server
-------------------------------------------------
.. index:: create a new bandwidth server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successfully created
``RTEMS.CBS_ERROR_NO_MEMORY`` - not enough memory for data
``RTEMS.CBS_ERROR_FULL`` - maximum servers exceeded
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument

**DESCRIPTION:**

This routine prepares an instance of a constant bandwidth server.
The input parameter ``rtems_cbs_parameters`` specifies scheduling
parameters of the server (period and budget). If these are not valid,``RTEMS.CBS_ERROR_INVALID_PARAMETER`` is returned.
The ``budget_overrun_callback`` is an optional callback function, which is
invoked in case the server’s budget within one period is exceeded.
Output parameter ``server_id`` becomes an id of the newly created server.
If there is not enough memory, the ``RTEMS.CBS_ERROR_NO_MEMORY``
is returned. If the maximum server count in the system is exceeded,``RTEMS.CBS_ERROR_FULL`` is returned.

**NOTES:**

No task execution is being influenced so far.

CBS_ATTACH_THREAD - Attach a thread to server
---------------------------------------------
.. index:: attach a thread to server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successfully attached
``RTEMS.CBS_ERROR_FULL`` - server maximum tasks exceeded
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

Attaches a task (``task_id``) to a server (``server_id``).
The server has to be previously created. Now, the task starts
to be scheduled according to the server parameters and not
using initial priority. This implementation allows only one task
per server, if the user tries to bind another task to the same
server, ``RTEMS.CBS_ERROR_FULL`` is returned.

**NOTES:**

Tasks attached to servers become preemptible.

CBS_DETACH_THREAD - Detach a thread from server
-----------------------------------------------
.. index:: detach a thread from server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successfully detached
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive detaches a thread from server. The task continues its
execution with initial priority.

**NOTES:**

The server can be reused for any other task.

CBS_DESTROY_SERVER - Destroy a bandwidth server
-----------------------------------------------
.. index:: destroy a bandwidth server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successfully destroyed
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive destroys a server. If any task was attached to the server,
the task is detached and continues its execution according to EDF rules
with initial properties.

**NOTES:**

This again enables one more task to be created.

CBS_GET_SERVER_ID - Get an ID of a server
-----------------------------------------
.. index:: get an ID of a server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive returns an id of server belonging to a given task.

CBS_GET_PARAMETERS - Get scheduling parameters of a server
----------------------------------------------------------
.. index:: get scheduling parameters of a server

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive returns a structure with current scheduling parameters
of a given server (period and execution time).

**NOTES:**

It makes no difference if any task is assigned or not.

CBS_SET_PARAMETERS - Set scheduling parameters
----------------------------------------------
.. index:: set scheduling parameters

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive sets new scheduling parameters to the server. This operation
can be performed regardless of whether a task is assigned or not.
If a task is assigned, the parameters become effective imediately, therefore it
is recommended to apply the change between two subsequent periods.

**NOTES:**

There is an upper limit on both period and budget equal to (2^31)-1 ticks.

CBS_GET_EXECUTION_TIME - Get elapsed execution time
---------------------------------------------------
.. index:: get elapsed execution time

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This routine returns consumed execution time (``exec_time``) of a server
during the current period.

**NOTES:**

Absolute time (``abs_time``) not supported now.

CBS_GET_REMAINING_BUDGET - Get remaining execution time
-------------------------------------------------------
.. index:: get remaining execution time

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive returns remaining execution time of a given server for
current period.

**NOTES:**

If the execution time approaches zero, the assigned task should finish
computations of the current period.

CBS_GET_APPROVED_BUDGET - Get scheduler approved execution time
---------------------------------------------------------------
.. index:: get scheduler approved execution time

**CALLING SEQUENCE:**

**DIRECTIVE STATUS CODES:**

``RTEMS.CBS_OK`` - successful
``RTEMS.CBS_ERROR_INVALID_PARAMETER`` - invalid input argument
``RTEMS.CBS_ERROR_NOSERVER`` - server is not valid

**DESCRIPTION:**

This directive returns server’s approved budget for subsequent periods.

.. COMMENT: COPYRIGHT (c) 1989-2011.

.. COMMENT: On-Line Applications Research Corporation (OAR).

.. COMMENT: All rights reserved.