Work queue_ queue

brief introduction
Work queues are a way of delaying work in the kernel , Delayed work can be scheduled and used repeatedly in countless scenarios .

Data structure composition
/* @data: func Parameters of
 * @entry: Pointer to connect work
 * @func: Work processing function
 */
struct work_struct {
    atomic_long_t data; 
    struct list_head entry;
    work_func_t func;
#ifdef CONFIG_LOCKDEP
    struct lockdep_map lockdep_map;
#endif
};
Usage flow
API route :kernel/kernel/workqueue.c; kernel/include/linux/workqueue.h

Create a work queue ：
a. Create a service for a single CPU The work queue of

/*  @name: Team name
 *  RETURNS: Pointer to the allocated workqueue on success, %NULL on failure.
 */
create_singlethread_workqueue(name);
b. In every one of them CPU Create a work queue on

/* @name: Team name
 * RETURNS: Pointer to the allocated workqueue on success, %NULL on failure.
 */
create_workqueue(name);
be relative to create_singlethread_workqueue, create_workqueue Will also be assigned a wq The work queue of , But the difference is , For many CPU In terms of system , For each CPU, Will create a per-CPU Of cwq structure , Corresponding to each cwq, Will generate a new worker_thread process . But when used queue_work towards cwq Submit on file work Node time , Which is CPU Call this function , Then ask the CPU Corresponding cwq Upper worklist To add work node .

Initialize work items
a. Dynamic registration

//kernel/include/linux/workqueue.h
INIT_WORK(_work, _func)
b. Static registration

//kernel/include/linux/workqueue.h
DECLARE_WORK(_work, _func)
Using static registration, you can omit the definition _work, And DECLARE_WORK It needs to be preprocessed at the head of the code .

Run the specified work item
a. Custom queue run queue_work

/* @wq: workqueue to use
 * @work: work to queue
 * Returns %false if @work was already on a queue, %true otherwise.
 */
queue_work(struct workqueue_struct *wq, struct work_struct *work)
b. The system work queue is running schedule_work

/* @work: work item
 * Returns %false if @work was already on the kernel-global workqueue and %true otherwise.
 */
schedule_work(struct work_struct *work)
queue_work, Run work items using a custom queue
schedule_work, Call the work queue of the system to run the work item .

Summary ：
In general , The work item needs to be called repeatedly in the specified situation , Select timer +queue_work. If it is specified, call once , Then use schedule_work, Use the work queue of the system to execute the required work items .

Usage flow
1. Declare variables
struct test_work_dev work_dev;
static struct workqueue_struct *  test1_workqueue = NULL;
 
2. Declare callback functions
void test1_callback(struct work_struct *work);
 
3. Initialize queue entry
#if defined (DECLARE_WORK_SUPPORT )
static DECLARE_WORK(test1_item, (void *) test1_callback);
#endif
 
#if !defined(DECLARE_WORK_SUPPORT)
static struct work_struct test1_item;
test1_workqueue = create_singlethread_workqueue("test1_wq");    
INIT_WORK(&test1_item, test1_callback);
#endif
 
4. Define callback
void test1_callback(struct work_struct *work) 
{
    printk("test1_callback!");
}
 
5. Dispatch

queue_work(test1_workqueue, &test1_item);