7.2. Background¶
7.2.1. Task Definition¶
Many definitions of a task have been proposed in computer literature. Unfortunately, none of these definitions encompasses all facets of the concept in a manner which is operating system independent. Several of the more common definitions are provided to enable each user to select a definition which best matches their own experience and understanding of the task concept:
a “dispatchable” unit.
an entity to which the processor is allocated.
an atomic unit of a real-time, multiprocessor system.
single threads of execution which concurrently compete for resources.
a sequence of closely related computations which can execute concurrently with other computational sequences.
From RTEMS’ perspective, a task is the smallest thread of execution which can compete on its own for system resources. A task is manifested by the existence of a task control block (TCB).
7.2.2. Task Control Block¶
The Task Control Block (TCB) is an RTEMS defined data structure which contains all the information that is pertinent to the execution of a task. During system initialization, RTEMS reserves a TCB for each task configured. A TCB is allocated upon creation of the task and is returned to the TCB free list upon deletion of the task.
The TCB’s elements are modified as a result of system calls made by the application in response to external and internal stimuli. TCBs are the only RTEMS internal data structure that can be accessed by an application via user extension routines. The TCB contains a task’s name, ID, current priority, current and starting states, execution mode, TCB user extension pointer, scheduling control structures, as well as data required by a blocked task.
A task’s context is stored in the TCB when a task switch occurs. When the task regains control of the processor, its context is restored from the TCB. When a task is restarted, the initial state of the task is restored from the starting context area in the task’s TCB.
7.2.3. Task Memory¶
The system uses two separate memory areas to manage a task. One memory area is the Task Control Block. The other memory area is allocated from the stack space or provided by the user and contains
the task stack,
the thread-local storage (TLS), and
an optional architecture-specific floating-point context.
The size of the thread-local storage is determined at link time. A user-provided task stack must take the size of the thread-local storage into account.
On architectures with a dedicated floating-point context, the application configuration assumes that every task is a floating-point task, but whether or not a task is actually floating-point is determined at runtime during task creation (see Floating Point Considerations). In highly memory constrained systems this potential overestimate of the task stack space can be mitigated through the CONFIGURE_MINIMUM_TASK_STACK_SIZE configuration option and aligned task stack sizes for the tasks. A user-provided task stack must take the potential floating-point context into account.
7.2.4. Task Name¶
By default, the task name is defined by the task object name given to rtems_task_create(). The task name can be obtained with the pthread_getname_np() function. Optionally, a new task name may be set with the pthread_setname_np() function. The maximum size of a task name is defined by the application configuration option CONFIGURE_MAXIMUM_THREAD_NAME_SIZE.
7.2.5. Task States¶
A task may exist in one of the following five states:
executing - Currently scheduled to the CPU
ready - May be scheduled to the CPU
blocked - Unable to be scheduled to the CPU
dormant - Created task that is not started
non-existent - Uncreated or deleted task
An active task may occupy the executing, ready, blocked or dormant state, otherwise the task is considered non-existent. One or more tasks may be active in the system simultaneously. Multiple tasks communicate, synchronize, and compete for system resources with each other via system calls. The multiple tasks appear to execute in parallel, but actually each is dispatched to the CPU for periods of time determined by the RTEMS scheduling algorithm. The scheduling of a task is based on its current state and priority.
7.2.6. Task Priority¶
A task’s priority determines its importance in relation to the other
tasks executing on the processor set owned by a scheduler. Normally,
RTEMS supports 256 levels of priority ranging from 0 to 255. The priority
level 0 represents a special priority reserved for the operating system. The
data type rtems_task_priority
is used to store task priorities. The
maximum priority level depends on the configured scheduler, see
CONFIGURE_MAXIMUM_PRIORITY, Clustered Scheduler Configuration, and
Scheduling Concepts.
Tasks of numerically smaller priority values are more important tasks than tasks of numerically larger priority values. For example, a task at priority level 5 is of higher privilege than a task at priority level 10. There is no limit to the number of tasks assigned to the same priority.
Each task has a priority associated with it at all times. The initial value of this priority is assigned at task creation time. The priority of a task may be changed at any subsequent time.
Priorities are used by the scheduler to determine which ready task will be allowed to execute. In general, the higher the logical priority of a task, the more likely it is to receive processor execution time.
7.2.7. Task Mode¶
A task’s execution mode is a combination of the following four components:
preemption
ASR processing
timeslicing
interrupt level
It is used to modify RTEMS’ scheduling process and to alter the execution
environment of the task. The data type rtems_task_mode
is used to manage
the task execution mode.
The preemption component allows a task to determine when control of the
processor is relinquished. If preemption is disabled (RTEMS_NO_PREEMPT
),
the task will retain control of the processor as long as it is in the executing
state - even if a higher priority task is made ready. If preemption is enabled
(RTEMS_PREEMPT
) and a higher priority task is made ready, then the
processor will be taken away from the current task immediately and given to the
higher priority task.
The timeslicing component is used by the RTEMS scheduler to determine how the
processor is allocated to tasks of equal priority. If timeslicing is enabled
(RTEMS_TIMESLICE
), then RTEMS will limit the amount of time the task can
execute before the processor is allocated to another ready task of equal
priority. The length of the timeslice is application dependent and specified in
the Configuration Table. If timeslicing is disabled (RTEMS_NO_TIMESLICE
),
then the task will be allowed to execute until a task of higher priority is
made ready. If RTEMS_NO_PREEMPT
is selected, then the timeslicing component
is ignored by the scheduler.
The asynchronous signal processing component is used to determine when received
signals are to be processed by the task. If signal processing is enabled
(RTEMS_ASR
), then signals sent to the task will be processed the next time
the task executes. If signal processing is disabled (RTEMS_NO_ASR
), then
all signals received by the task will remain posted until signal processing is
enabled. This component affects only tasks which have established a routine to
process asynchronous signals.
The interrupt level component is used to determine which interrupts will be
enabled when the task is executing. RTEMS_INTERRUPT_LEVEL(n)
specifies that
the task will execute at interrupt level n.
|
enable preemption (default) |
|
disable preemption |
|
disable timeslicing (default) |
|
enable timeslicing |
|
enable ASR processing (default) |
|
disable ASR processing |
|
enable all interrupts (default) |
|
execute at interrupt level n |
The set of default modes may be selected by specifying the
RTEMS_DEFAULT_MODES
constant.
7.2.8. Task Life States¶
Independent of the task state with respect to the scheduler, the task life is determined by several orthogonal states:
protected or unprotected
deferred life changes or no deferred life changes
restarting or not restarting
terminating or not terminating
detached or not detached
While the task life is protected, asynchronous task restart and termination
requests are blocked. A task may still restart or terminate itself. All tasks
are created with an unprotected task life. The task life protection is used by
the system to prevent system resources being affected by asynchronous task
restart and termination requests. The task life protection can be enabled
(PTHREAD_CANCEL_DISABLE
) or disabled (PTHREAD_CANCEL_ENABLE
) for the
calling task through the pthread_setcancelstate()
directive.
While deferred life changes are enabled, asynchronous task restart and
termination requests are delayed until the task performs a life change itself
or calls pthread_testcancel()
. Cancellation points are not implemented in
RTEMS. Deferred task life changes can be enabled (PTHREAD_CANCEL_DEFERRED
)
or disabled (PTHREAD_CANCEL_ASYNCHRONOUS
) for the calling task through the
pthread_setcanceltype()
directive. Classic API tasks are created with
deferred life changes disabled. POSIX threads are created with deferred life
changes enabled.
A task is made restarting by issuing a task restart request through the rtems_task_restart() directive.
A task is made terminating by issuing a task termination request through the
rtems_task_exit(), rtems_task_delete(),
pthread_exit()
, and pthread_cancel()
directives.
When a detached task terminates, the termination procedure completes without
the need for another task to join with the terminated task. Classic API tasks
are created as not detached. The detached state of created POSIX threads is
determined by the thread attributes. They are created as not detached by
default. The calling task is made detached through the pthread_detach()
directive. The rtems_task_exit() directive and self deletion
though rtems_task_delete() directive make the calling task
detached. In contrast, the pthread_exit()
directive does not change the
detached state of the calling task.
7.2.9. Accessing Task Arguments¶
All RTEMS tasks are invoked with a single argument which is specified when they are started or restarted. The argument is commonly used to communicate startup information to the task. The simplest manner in which to define a task which accesses it argument is:
rtems_task user_task(
rtems_task_argument argument
);
Application tasks requiring more information may view this single argument as an index into an array of parameter blocks.
7.2.10. Floating Point Considerations¶
Please consult the RTEMS CPU Architecture Supplement if this section is relevant on your architecture. On some architectures the floating-point context is contained in the normal task context and this section does not apply.
Creating a task with the RTEMS_FLOATING_POINT
attribute flag results in
additional memory being allocated for the task to store the state of the numeric
coprocessor during task switches. This additional memory is not allocated
for RTEMS_NO_FLOATING_POINT
tasks. Saving and restoring the context of a
RTEMS_FLOATING_POINT
task takes longer than that of a
RTEMS_NO_FLOATING_POINT
task because of the relatively large amount of time
required for the numeric coprocessor to save or restore its computational state.
Since RTEMS was designed specifically for embedded military applications which
are floating point intensive, the executive is optimized to avoid unnecessarily
saving and restoring the state of the numeric coprocessor. In uniprocessor
configurations, the state of the numeric coprocessor is only saved when a
RTEMS_FLOATING_POINT
task is dispatched and that task was not the last task
to utilize the coprocessor. In a uniprocessor system with only one
RTEMS_FLOATING_POINT
task, the state of the numeric coprocessor will never
be saved or restored.
Although the overhead imposed by RTEMS_FLOATING_POINT
tasks is minimal,
some applications may wish to completely avoid the overhead associated with
RTEMS_FLOATING_POINT
tasks and still utilize a numeric coprocessor. By
preventing a task from being preempted while performing a sequence of floating
point operations, a RTEMS_NO_FLOATING_POINT
task can utilize the numeric
coprocessor without incurring the overhead of a RTEMS_FLOATING_POINT
context switch. This approach also avoids the allocation of a floating point
context area. However, if this approach is taken by the application designer,
no tasks should be created as RTEMS_FLOATING_POINT
tasks. Otherwise, the
floating point context will not be correctly maintained because RTEMS assumes
that the state of the numeric coprocessor will not be altered by
RTEMS_NO_FLOATING_POINT
tasks. Some architectures with a dedicated
floating-point context raise a processor exception if a task with
RTEMS_NO_FLOATING_POINT
issues a floating-point instruction, so this
approach may not work at all.
If the supported processor type does not have hardware floating capabilities or
a standard numeric coprocessor, RTEMS will not provide built-in support for
hardware floating point on that processor. In this case, all tasks are
considered RTEMS_NO_FLOATING_POINT
whether created as
RTEMS_FLOATING_POINT
or RTEMS_NO_FLOATING_POINT
tasks. A floating
point emulation software library must be utilized for floating point
operations.
On some processors, it is possible to disable the floating point unit
dynamically. If this capability is supported by the target processor, then
RTEMS will utilize this capability to enable the floating point unit only for
tasks which are created with the RTEMS_FLOATING_POINT
attribute. The
consequence of a RTEMS_NO_FLOATING_POINT
task attempting to access the
floating point unit is CPU dependent but will generally result in an exception
condition.
7.2.11. Building a Task Attribute Set¶
In general, an attribute set is built by a bitwise OR of the desired components. The set of valid task attribute components is listed below:
|
does not use coprocessor (default) |
|
uses numeric coprocessor |
|
local task (default) |
|
global task |
Attribute values are specifically designed to be mutually exclusive, therefore
bitwise OR and addition operations are equivalent as long as each attribute
appears exactly once in the component list. A component listed as a default is
not required to appear in the component list, although it is a good programming
practice to specify default components. If all defaults are desired, then
RTEMS_DEFAULT_ATTRIBUTES
should be used.
This example demonstrates the attribute_set parameter needed to create a local
task which utilizes the numeric coprocessor. The attribute_set parameter could
be RTEMS_FLOATING_POINT
or RTEMS_LOCAL | RTEMS_FLOATING_POINT
. The
attribute_set parameter can be set to RTEMS_FLOATING_POINT
because
RTEMS_LOCAL
is the default for all created tasks. If the task were global
and used the numeric coprocessor, then the attribute_set parameter would be
RTEMS_GLOBAL | RTEMS_FLOATING_POINT
.
7.2.12. Building a Mode and Mask¶
In general, a mode and its corresponding mask is built by a bitwise OR of the desired components. The set of valid mode constants and each mode’s corresponding mask constant is listed below:
|
is masked by |
|
is masked by |
|
is masked by |
|
is masked by |
|
is masked by |
|
is masked by |
|
is masked by |
|
is masked by |
Mode values are specifically designed to be mutually exclusive, therefore
bitwise OR and addition operations are equivalent as long as each mode appears
exactly once in the component list. A mode component listed as a default is
not required to appear in the mode component list, although it is a good
programming practice to specify default components. If all defaults are
desired, the mode RTEMS_DEFAULT_MODES
and the mask RTEMS_ALL_MODE_MASKS
should be used.
The following example demonstrates the mode and mask parameters used with the
rtems_task_mode
directive to place a task at interrupt level 3 and make it
non-preemptible. The mode should be set to RTEMS_INTERRUPT_LEVEL(3) |
RTEMS_NO_PREEMPT
to indicate the desired preemption mode and interrupt level,
while the mask parameter should be set to RTEMS_INTERRUPT_MASK |
RTEMS_NO_PREEMPT_MASK
to indicate that the calling task’s interrupt level and
preemption mode are being altered.