|TASKQUEUE_CALLBACK_TYPE_INIT||This callback is called by every thread in the taskqueue, before it executes any tasks. This callback must be set before the taskqueues threads are started.|
|This callback is called by every thread in the taskqueue, after it executes its last task. This callback will always be called before the taskqueue structure is reclaimed.|
To add a task to the list of tasks queued on a taskqueue, call taskqueue_enqueue with pointers to the queue and task. If the tasks ta_pending field is non-zero, then it is simply incremented to reflect the number of times the task was enqueued, up to a cap of USHRT_MAX. Otherwise, the task is added to the list before the first task which has a lower ta_priority value or at the end of the list if no tasks have a lower priority. Enqueueing a task does not perform any memory allocation which makes it suitable for calling from an interrupt handler. This function will return EPIPE if the queue is being freed.
The function taskqueue_enqueue_fast should be used in place of taskqueue_enqueue when the enqueuing must happen from a fast interrupt handler. This method uses spin locks to avoid the possibility of sleeping in the fast interrupt context.
When a task is executed, first it is removed from the queue, the value of ta_pending is recorded and then the field is zeroed. The function ta_func from the task structure is called with the value of the field ta_context as its first argument and the value of ta_pending as its second argument. After the function ta_func returns, wakeup(9) is called on the task pointer passed to taskqueue_enqueue.
The taskqueue_enqueue_timeout is used to schedule the enqueue after the specified amount of ticks. Only non-fast task queues can be used for timeout_task scheduling. If the ticks argument is negative, the already scheduled enqueueing is not re-scheduled. Otherwise, the task is scheduled for enqueueing in the future, after the absolute value of ticks is passed.
The taskqueue_cancel function is used to cancel a task. The ta_pending count is cleared, and the old value returned in the reference parameter pendp, if it is non- NULL. If the task is currently running, EBUSY is returned, otherwise 0. To implement a blocking taskqueue_cancel that waits for a running task to finish, it could look like:
while (taskqueue_cancel(tq, task, NULL) != 0) taskqueue_drain(tq, task);
Note that, as with taskqueue_drain, the caller is responsible for ensuring that the task is not re-enqueued after being canceled.
Similarly, the taskqueue_cancel_timeout function is used to cancel the scheduled task execution.
The taskqueue_drain function is used to wait for the task to finish, and the taskqueue_drain_timeout function is used to wait for the scheduled task to finish. There is no guarantee that the task will not be enqueued after call to taskqueue_drain. If the caller wants to put the task into a known state, then before calling taskqueue_drain the caller should use out-of-band means to ensure that the task would not be enqueued. For example, if the task is enqueued by an interrupt filter, then the interrupt could be disabled.
The taskqueue_drain_all function is used to wait for all pending and running tasks that are enqueued on the taskqueue to finish. The caller must arrange that the tasks are not re-enqueued. Note that taskqueue_drain_all currently does not handle tasks with delayed enqueueing.
The taskqueue_block function blocks the taskqueue. It prevents any enqueued but not running tasks from being executed. Future calls to taskqueue_enqueue will enqueue tasks, but the tasks will not be run until taskqueue_unblock is called. Please note that taskqueue_block does not wait for any currently running tasks to finish. Thus, the taskqueue_block does not provide a guarantee that taskqueue_run is not running after taskqueue_block returns, but it does provide a guarantee that taskqueue_run will not be called again until taskqueue_unblock is called. If the caller requires a guarantee that taskqueue_run is not running, then this must be arranged by the caller. Note that if taskqueue_drain is called on a task that is enqueued on a taskqueue that is blocked by taskqueue_block, then taskqueue_drain can not return until the taskqueue is unblocked. This can result in a deadlock if the thread blocked in taskqueue_drain is the thread that is supposed to call taskqueue_unblock. Thus, use of taskqueue_drain after taskqueue_block is discouraged, because the state of the task can not be known in advance. The same caveat applies to taskqueue_drain_all.
The taskqueue_unblock function unblocks the previously blocked taskqueue. All enqueued tasks can be run after this call.
The taskqueue_member function returns 1 if the given thread td is part of the given taskqueue queue and 0 otherwise.
The taskqueue_run function will run all pending tasks in the specified queue. Normally this function is only used internally.
A convenience macro, TASK_INIT task priority func context is provided to initialise a task structure. The TASK_INITIALIZER macro generates an initializer for a task structure. A macro TIMEOUT_TASK_INIT queue timeout_task priority func context initializes the timeout_task structure. The values of priority, func, and context are simply copied into the task structure fields and the ta_pending field is cleared.
Five macros TASKQUEUE_DECLARE name, TASKQUEUE_DEFINE name enqueue context init, TASKQUEUE_FAST_DEFINE name enqueue context init, and TASKQUEUE_DEFINE_THREAD name TASKQUEUE_FAST_DEFINE_THREAD name are used to declare a reference to a global queue, to define the implementation of the queue, and declare a queue that uses its own thread. The TASKQUEUE_DEFINE macro arranges to call taskqueue_create with the values of its name, enqueue and context arguments during system initialisation. After calling taskqueue_create, the init argument to the macro is executed as a C statement, allowing any further initialisation to be performed (such as registering an interrupt handler etc.)
macro defines a new taskqueue with its own kernel thread to serve tasks.
.Vt struct taskqueue *taskqueue_name is used to enqueue tasks onto the queue.
TASKQUEUE_FAST_DEFINE and TASKQUEUE_FAST_DEFINE_THREAD act just like TASKQUEUE_DEFINE and TASKQUEUE_DEFINE_THREAD respectively but taskqueue is created with taskqueue_create_fast.
The system provides four global taskqueues, taskqueue_fast, taskqueue_swi, taskqueue_swi_giant, and taskqueue_thread. The taskqueue_fast queue is for swi handlers dispatched from fast interrupt handlers, where sleep mutexes cannot be used. The swi taskqueues are run via a software interrupt mechanism. The taskqueue_swi queue runs without the protection of the Giant kernel lock, and the taskqueue_swi_giant queue runs with the protection of the Giant kernel lock. The thread taskqueue taskqueue_thread runs in a kernel thread context, and tasks run from this thread do not run under the Giant kernel lock. If the caller wants to run under Giant, he should explicitly acquire and release Giant in his taskqueue handler routine.
To use these queues, call taskqueue_enqueue with the value of the global taskqueue variable for the queue you wish to use ( taskqueue_swi, taskqueue_swi_giant, or taskqueue_thread). Use taskqueue_enqueue_fast for the global taskqueue variable taskqueue_fast.
The software interrupt queues can be used, for instance, for implementing interrupt handlers which must perform a significant amount of processing in the handler. The hardware interrupt handler would perform minimal processing of the interrupt and then enqueue a task to finish the work. This reduces to a minimum the amount of time spent with interrupts disabled.
The thread queue can be used, for instance, by interrupt level routines that need to call kernel functions that do things that can only be done from a thread context. (e.g., call malloc with the M_WAITOK flag.)
Note that tasks queued on shared taskqueues such as taskqueue_swi may be delayed an indeterminate amount of time before execution. If queueing delays cannot be tolerated then a private taskqueue should be created with a dedicated processing thread.
This interface first appeared in
.Fx 5.0 . There is a similar facility called work_queue in the Linux kernel.
This manual page was written by
.An Doug Rabson .