Semaphore

Basic Concepts

Semaphore is a mechanism used to implement synchronization between tasks or exclusive access to shared resources.

In the semaphore data structure, there is a value indicating the number of shared resources available. The value can be:

• 0: The semaphore is unavailable. In this case, tasks waiting for the semaphore may exist.

• Positive number: The semaphore is available.

The semaphore used for exclusive access to resources is different from the semaphore used for synchronization:

• Semaphore used for exclusive access: The initial semaphore counter value (non-zero) indicates the number of shared resources available. A semaphore must be acquired before a shared resource is used, and released when the resource is no longer required. When all shared resources are used, the semaphore counter is reduced to 0 and all tasks requiring the semaphore will be blocked. This ensures exclusive access to shared resources. In addition, if the number of shared resources is 1, a binary semaphore (similar to the mutex mechanism) is recommended.

• Semaphore used for synchronization: The initial semaphore counter value is 0. A task without the semaphore will be blocked, and enters the Ready or Running state only when the semaphore is released by another task or an interrupt.

Working Principles

Semaphore Control Block

/**
  * Data structure of the semaphore control block
 */
typedef struct {
    UINT16            semStat;          /* Semaphore status */
    UINT16            semType;          /* Semaphore type */
    UINT16            semCount;         /* Semaphore count */
    UINT16            semId;            /* Semaphore ID */
    LOS_DL_LIST       semList;          /* List of blocked tasks */
} LosSemCB;

Working Principles

Semaphore allows only a specified number of tasks to access a shared resource at a time. When the number of tasks accessing the resource reaches the limit, other tasks will be blocked until the semaphore is released.

• Semaphore initialization

Allocate memory for the semaphores (the number of semaphores is specified by the LOSCFG_BASE_IPC_SEM_LIMIT macro), set all semaphores to the unused state, and add them to a linked list.

• Semaphore creation

Obtain a semaphore from the linked list of unused semaphores and assign an initial value to the semaphore.

• Semaphore request

If the counter value is greater than 0 when a semaphore is requsted, the counter is decreased by 1 and a success message is returned. Otherwise, the task is blocked and added to the end of a task queue waiting for semaphores. The wait timeout period can be set.

• Semaphore release

If no task is waiting for the semaphore, the counter is incremented by 1. Otherwise, wake up the first task in the wait queue.

• Semaphore deletion

Set a semaphore in use to the unused state and add it to the linked list of unused semaphores.

The following figure illustrates the semaphore working mechanism.

Figure 1 Semaphore working mechanism for the small system

Development Guidelines

Available APIs

Table 1 APIs for creating and deleting a semaphore

Table 2 APIs for requesting and releasing a semaphore

How to Develop

1. Call LOS_SemCreate to create a semaphore. To create a binary semaphore, call LOS_BinarySemCreate.

2. Call LOS_SemPend to request a semaphore.

3. Call LOS_SemPost to release a semaphore.

4. Call LOS_SemDelete to delete a semaphore.

NOTE
As interrupts cannot be blocked, semaphores cannot be requested in block mode for interrupts.

Development Example

Example Description

This example implements the following:

1. Create a semaphore in task ExampleSem and lock task scheduling. Create two tasks ExampleSemTask1 and ExampleSemTask2 (with higher priority). Enable the two tasks to request the same semaphore. Unlock task scheduling. Enable task ExampleSem to enter sleep mode for 400 ticks. Release the semaphore in task ExampleSem.

2. Enable ExampleSemTask2 to enter sleep mode for 20 ticks after acquiring the semaphore. (When ExampleSemTask2 is delayed, ExampleSemTask1 is woken up.)

3. Enable ExampleSemTask1 to request the semaphore in scheduled block mode, with a wait timeout period of 10 ticks. (Because the semaphore is still held by ExampleSemTask2, ExampleSemTask1 is suspended. ExampleSemTask1 is woken up after 10 ticks.) Enable ExampleSemTask1 to request the semaphore in permanent block mode after it is woken up 10 ticks later. (Because the semaphore is still held by ExampleSemTask2, ExampleSemTask1 is suspended.)

4. After 20 ticks, ExampleSemTask2 is woken up and releases the semaphore. ExampleSemTask1 acquires the semaphore and is scheduled to run. When ExampleSemTask1 is complete, it releases the semaphore.

5. Task ExampleSem is woken up after 400 ticks. After that, delete the semaphore.

Sample Code

The sample code can be compiled and verified in ./kernel/liteos_a/testsuites/kernel/src/osTest.c. The ExampleSem function is called in TestTaskEntry.

The sample code is as follows:

#include "los_sem.h"
#include "securec.h"

/* Task ID*/
static UINT32 g_testTaskId01;
static UINT32 g_testTaskId02;

/* Task priority */
#define TASK_PRIO_LOW   5
#define TASK_PRIO_HI    4

/* Semaphore structure ID */
static UINT32 g_semId;

VOID ExampleSemTask1(VOID)
{
    UINT32 ret;

    dprintf("ExampleSemTask1 try get sem g_semId, timeout 10 ticks.\n");

    /* Request the semaphore in scheduled block mode, with a wait timeout period of 10 ticks. */
    ret = LOS_SemPend(g_semId, 10);
    /* The semaphore is acquired. */
    if (ret == LOS_OK) {
         LOS_SemPost(g_semId);
         return;
    }
    /* The semaphore is not acquired when the timeout period has expired. */
    if (ret == LOS_ERRNO_SEM_TIMEOUT) {
        dprintf("ExampleSemTask1 timeout and try get sem g_semId wait forever.\n");

        /* Request the semaphore in permanent block mode. */
        ret = LOS_SemPend(g_semId, LOS_WAIT_FOREVER);
        dprintf("ExampleSemTask1 wait_forever and get sem g_semId.\n");
        if (ret == LOS_OK) {
            dprintf("ExampleSemTask1 post sem g_semId.\n");
            LOS_SemPost(g_semId);
            return;
        }
    }
}

VOID ExampleSemTask2(VOID)
{
    UINT32 ret;
    dprintf("ExampleSemTask2 try get sem g_semId wait forever.\n");

    /* Request the semaphore in permanent block mode. */
    ret = LOS_SemPend(g_semId, LOS_WAIT_FOREVER);
    if (ret == LOS_OK) {
        dprintf("ExampleSemTask2 get sem g_semId and then delay 20 ticks.\n");
    }

    /* Enable the task to enter sleep mode for 20 ticks. */
    LOS_TaskDelay(20);

    dprintf("ExampleSemTask2 post sem g_semId.\n");
    /* Release the semaphore. */
    LOS_SemPost(g_semId);
    return;
}

UINT32 ExampleSem(VOID)
{
    UINT32 ret;
    TSK_INIT_PARAM_S task1;
    TSK_INIT_PARAM_S task2;

   /* Create a semaphore. */
    LOS_SemCreate(0, &g_semId);

    /* Lock task scheduling. */
    LOS_TaskLock();

    /* Create task 1. */
    (VOID)memset_s(&task1, sizeof(TSK_INIT_PARAM_S), 0, sizeof(TSK_INIT_PARAM_S));
    task1.pfnTaskEntry = (TSK_ENTRY_FUNC)ExampleSemTask1;
    task1.pcName       = "TestTask1";
    task1.uwStackSize  = LOSCFG_BASE_CORE_TSK_DEFAULT_STACK_SIZE;
    task1.usTaskPrio   = TASK_PRIO_LOW;
    ret = LOS_TaskCreate(&g_testTaskId01, &task1);
    if (ret != LOS_OK) {
        dprintf("task1 create failed .\n");
        return LOS_NOK;
    }

    /* Create task 2. */
    (VOID)memset_s(&task2, sizeof(TSK_INIT_PARAM_S), 0, sizeof(TSK_INIT_PARAM_S));
    task2.pfnTaskEntry = (TSK_ENTRY_FUNC)ExampleSemTask2;
    task2.pcName       = "TestTask2";
    task2.uwStackSize  = LOSCFG_BASE_CORE_TSK_DEFAULT_STACK_SIZE;
    task2.usTaskPrio   = TASK_PRIO_HI;
    ret = LOS_TaskCreate(&g_testTaskId02, &task2);
    if (ret != LOS_OK) {
        dprintf("task2 create failed.\n");
        return LOS_NOK;
    }

    /* Unlock task scheduling. */
    LOS_TaskUnlock();

     /* Enable the task to enter sleep mode for 400 ticks. */
    LOS_TaskDelay(400);

    ret = LOS_SemPost(g_semId);

     /* Enable the task to enter sleep mode for 400 ticks. */
    LOS_TaskDelay(400);

    /* Delete the semaphore. */
    LOS_SemDelete(g_semId);
    return LOS_OK;
}

Verification

The development is successful if the return result is as follows:

ExampleSemTask2 try get sem g_semId wait forever.
ExampleSemTask1 try get sem g_semId, timeout 10 ticks.
ExampleSemTask1 timeout and try get sem g_semId wait forever.
ExampleSemTask2 get sem g_semId and then delay 20 ticks.
ExampleSemTask2 post sem g_semId.
ExampleSemTask1 wait_forever and get sem g_semId.
ExampleSemTask1 post sem g_semId.

你可能感兴趣的鸿蒙文章

harmony 鸿蒙Kernel

harmony 鸿蒙Kernel Coding Specification

harmony 鸿蒙Standard Libraries

harmony 鸿蒙Interrupt Management

harmony 鸿蒙Event

harmony 鸿蒙Mutex

harmony 鸿蒙Queue

harmony 鸿蒙Semaphore

harmony 鸿蒙Doubly Linked List

harmony 鸿蒙Memory Management