Linux内核结构体--kfifo 环状缓冲区
1、前言
最近项目中用到一个环形缓冲区（ring buffer），代码是由linux内核的kfifo改过来的。缓冲区在文件系统中经常用到，通过缓冲区缓解cpu读写内存和读写磁盘的速度。例如一个进程A产生数据发给另外一个进程B，进程B需要对进程A传的数据进行处理并写入文件，如果B没有处理完，则A要延迟发送。为了保证进程A减少等待时间，可以在A和B之间采用一个缓冲区，A每次将数据存放在缓冲区中，B每次冲缓冲区中取。这是典型的生产者和消费者模型，缓冲区中数据满足FIFO特性，因此可以采用队列进行实现。Linux内核的kfifo正好是一个环形队列，可以用来当作环形缓冲区。生产者与消费者使用缓冲区如下图所示：
环形缓冲区的详细介绍及实现方法可以参考http://en.wikipedia.org/wiki/Circular_buffer，介绍的非常详细，列举了实现环形队列的几种方法。环形队列的不便之处在于如何判断队列是空还是满。维基百科上给三种实现方法。
2、linux 内核kfifo
kfifo设计的非常巧妙，代码很精简，对于入队和出对处理的出人意料。首先看一下kfifo的数据结构：
struct kfifo {
unsigned char *buffer; /* the buffer holding the data */
unsigned int size; /* the size of the allocated buffer */
unsigned int in; /* data is added at offset (in % size) */
unsigned int out; /* data is extracted from off. (out % size) */
spinlock_t *lock; /* protects concurrent modifications */
};kfifo提供的方法有：
//根据给定buffer创建一个kfifo
struct kfifo *kfifo_init(unsigned char *buffer, unsigned int size,
gfp_t gfp_mask, spinlock_t *lock);
//给定size分配buffer和kfifo
struct kfifo *kfifo_alloc(unsigned int size, gfp_t gfp_mask,
spinlock_t *lock);
//释放kfifo空间
void kfifo_free(struct kfifo *fifo)
//向kfifo中添加数据
unsigned int kfifo_put(struct kfifo *fifo,
const unsigned char *buffer, unsigned int len)
//从kfifo中取数据
unsigned int kfifo_put(struct kfifo *fifo,
const unsigned char *buffer, unsigned int len)
//获取kfifo中有数据的buffer大小
unsigned int kfifo_len(struct kfifo *fifo)定义自旋锁的目的为了防止多进程/线程并发使用kfifo。因为in和out在每次get和out时，发生改变。初始化和创建kfifo的源代码如下：
struct kfifo *kfifo_init(unsigned char *buffer, unsigned int size,
gfp_t gfp_mask, spinlock_t *lock)
{
struct kfifo *fifo;
/* size must be a power of 2 */
BUG_ON(!is_power_of_2(size));
fifo = kmalloc(sizeof(struct kfifo), gfp_mask);
if (!fifo)
return ERR_PTR(-ENOMEM);
fifo->buffer = buffer;
fifo->size = size;
fifo->in = fifo->out = 0;
fifo->lock = lock;
return fifo;
}
struct kfifo *kfifo_alloc(unsigned int size, gfp_t gfp_mask, spinlock_t *lock)
{
unsigned char *buffer;
struct kfifo *ret;
if (!is_power_of_2(size)) {
BUG_ON(size > 0x80000000);
size = roundup_pow_of_two(size);
}
buffer = kmalloc(size, gfp_mask);
if (!buffer)
return ERR_PTR(-ENOMEM);
ret = kfifo_init(buffer, size, gfp_mask, lock);
if (IS_ERR(ret))
kfree(buffer);
return ret;
}在kfifo_init和kfifo_calloc中，kfifo->size的值总是在调用者传进来的size参数的基础上向2的幂扩展，这是内核一贯的做法。这样的好处不言而喻--对kfifo->size取模运算可以转化为与运算，如： kfifo->in % kfifo->size 可以转化为 kfifo->in & (kfifo->size – 1)
kfifo的巧妙之处在于in和out定义为无符号类型，在put和get时，in和out都是增加，当达到最大值时，产生溢出，使得从0开始，进行循环使用。put和get代码如下所示：
static inline unsigned int kfifo_put(struct kfifo *fifo,
const unsigned char *buffer, unsigned int len)
{
unsigned long flags;
unsigned int ret;
spin_lock_irqsave(fifo->lock, flags);
ret = __kfifo_put(fifo, buffer, len);
spin_unlock_irqrestore(fifo->lock, flags);
return ret;
}
static inline unsigned int kfifo_get(struct kfifo *fifo,
unsigned char *buffer, unsigned int len)
{
unsigned long flags;
unsigned int ret;
spin_lock_irqsave(fifo->lock, flags);
ret = __kfifo_get(fifo, buffer, len);
//当fifo->in == fifo->out时，buufer为空
if (fifo->in == fifo->out)
fifo->in = fifo->out = 0;
spin_unlock_irqrestore(fifo->lock, flags);
return ret;
}
unsigned int __kfifo_put(struct kfifo *fifo,
const unsigned char *buffer, unsigned int len)
{
unsigned int l;
//buffer中空的长度
len = min(len, fifo->size - fifo->in + fifo->out);
/*
* Ensure that we sample the fifo->out index -before- we
* start putting bytes into the kfifo.
*/
smp_mb();
/* first put the data starting from fifo->in to buffer end */
l = min(len, fifo->size - (fifo->in & (fifo->size - 1)));
memcpy(fifo->buffer + (fifo->in & (fifo->size - 1)), buffer, l);
/* then put the rest (if any) at the beginning of the buffer */
memcpy(fifo->buffer, buffer + l, len - l);
/*
* Ensure that we add the bytes to the kfifo -before-
* we update the fifo->in index.
*/
smp_wmb();
fifo->in += len; //每次累加，到达最大值后溢出，自动转为0
return len;
}
unsigned int __kfifo_get(struct kfifo *fifo,
unsigned char *buffer, unsigned int len)
{
unsigned int l;
//有数据的缓冲区的长度
len = min(len, fifo->in - fifo->out);
/*
* Ensure that we sample the fifo->in index -before- we
* start removing bytes from the kfifo.
*/
smp_rmb();
/* first get the data from fifo->out until the end of the buffer */
l = min(len, fifo->size - (fifo->out & (fifo->size - 1)));
memcpy(buffer, fifo->buffer + (fifo->out & (fifo->size - 1)), l);
/* then get the rest (if any) from the beginning of the buffer */
memcpy(buffer + l, fifo->buffer, len - l);
/*
* Ensure that we remove the bytes from the kfifo -before-
* we update the fifo->out index.
*/
smp_mb();
fifo->out += len; //每次累加，到达最大值后溢出，自动转为0
return len;
}put和get在调用__put和__get过程都进行加锁，防止并发。从代码中可以看出put和get都调用两次memcpy，这针对的是边界条件。例如下图：蓝色表示空闲，红色表示占用。
（1）空的kfifo，
（2）put一个buffer后
（3）get一个buffer后
（4）当此时put的buffer长度超出in到末尾长度时，则将剩下的移到头部去
3、测试程序
仿照kfifo编写一个ring_buffer，现有线程互斥量进行并发控制。设计的ring_buffer如下所示：
/**@brief 仿照linux kfifo写的ring buffer
*@atuher Anker date:2013-12-18
* ring_buffer.h
* */
#ifndef KFIFO_HEADER_H
#define KFIFO_HEADER_H
#include
#include
#include
#include
#include
#include
//判断x是否是2的次方
#define is_power_of_2(x) ((x) != 0 && (((x) & ((x) - 1)) == 0))
//取a和b中最小值
#define min(a, b) (((a) < (b)) ? (a) : (b))
struct ring_buffer
{
void *buffer; //缓冲区
uint32_t size; //大小
uint32_t in; //入口位置
uint32_t out; //出口位置
pthread_mutex_t *f_lock; //互斥锁
};
//初始化缓冲区
struct ring_buffer* ring_buffer_init(void *buffer, uint32_t size, pthread_mutex_t *f_lock)
{
assert(buffer);
struct ring_buffer *ring_buf = NULL;
if (!is_power_of_2(size))
{
fprintf(stderr,"size must be power of 2.\n");
return ring_buf;
}
ring_buf = (struct ring_buffer *)malloc(sizeof(struct ring_buffer));
if (!ring_buf)
{
fprintf(stderr,"Failed to malloc memory,errno:%u,reason:%s",
errno, strerror(errno));
return ring_buf;
}
memset(ring_buf, 0, sizeof(struct ring_buffer));
ring_buf->buffer = buffer;
ring_buf->size = size;
ring_buf->in = 0;
ring_buf->out = 0;
ring_buf->f_lock = f_lock;
return ring_buf;
}
//释放缓冲区
void ring_buffer_free(struct ring_buffer *ring_buf)
{
if (ring_buf)
{
if (ring_buf->buffer)
{
free(ring_buf->buffer);
ring_buf->buffer = NULL;
}
free(ring_buf);
ring_buf = NULL;
}
}
//缓冲区的长度
uint32_t __ring_buffer_len(const struct ring_buffer *ring_buf)
{
return (ring_buf->in - ring_buf->out);
}
//从缓冲区中取数据
uint32_t __ring_buffer_get(struct ring_buffer *ring_buf, void * buffer, uint32_t size)
{
assert(ring_buf || buffer);
uint32_t len = 0;
size = min(size, ring_buf->in - ring_buf->out);
/* first get the data from fifo->out until the end of the buffer */
len = min(size, ring_buf->size - (ring_buf->out & (ring_buf->size - 1)));
memcpy(buffer, ring_buf->buffer + (ring_buf->out & (ring_buf->size - 1)), len);
/* then get the rest (if any) from the beginning of the buffer */
memcpy(buffer + len, ring_buf->buffer, size - len);
ring_buf->out += size;
return size;
}
//向缓冲区中存放数据
uint32_t __ring_buffer_put(struct ring_buffer *ring_buf, void *buffer, uint32_t size)
{
assert(ring_buf || buffer);
uint32_t len = 0;
size = min(size, ring_buf->size - ring_buf->in + ring_buf->out);
/* first put the data starting from fifo->in to buffer end */
len = min(size, ring_buf->size - (ring_buf->in & (ring_buf->size - 1)));
memcpy(ring_buf->buffer + (ring_buf->in & (ring_buf->size - 1)), buffer, len);
/* then put the rest (if any) at the beginning of the buffer */
memcpy(ring_buf->buffer, buffer + len, size - len);
ring_buf->in += size;
return size;
}
uint32_t ring_buffer_len(const struct ring_buffer *ring_buf)
{
uint32_t len = 0;
pthread_mutex_lock(ring_buf->f_lock);
len = __ring_buffer_len(ring_buf);
pthread_mutex_unlock(ring_buf->f_lock);
return len;
}
uint32_t ring_buffer_get(struct ring_buffer *ring_buf, void *buffer, uint32_t size)
{
uint32_t ret;
pthread_mutex_lock(ring_buf->f_lock);
ret = __ring_buffer_get(ring_buf, buffer, size);
//buffer中没有数据
if (ring_buf->in == ring_buf->out)
ring_buf->in = ring_buf->out = 0;
pthread_mutex_unlock(ring_buf->f_lock);
return ret;
}
uint32_t ring_buffer_put(struct ring_buffer *ring_buf, void *buffer, uint32_t size)
{
uint32_t ret;
pthread_mutex_lock(ring_buf->f_lock);
ret = __ring_buffer_put(ring_buf, buffer, size);
pthread_mutex_unlock(ring_buf->f_lock);
return ret;
}
#endif /**@brief ring buffer测试程序，创建两个线程，一个生产者，一个消费者。
* 生产者每隔1秒向buffer中投入数据，消费者每隔2秒去取数据。
*@atuher Anker date:2013-12-18
* */
#include "ring_buffer.h"
#include
#include
#define BUFFER_SIZE 1024 * 1024
typedef struct student_info
{
uint64_t stu_id;
uint32_t age;
uint32_t score;
}student_info;
void print_student_info(const student_info *stu_info)
{
assert(stu_info);
printf("id:%lu\t",stu_info->stu_id);
printf("age:%u\t",stu_info->age);
printf("score:%u\n",stu_info->score);
}
student_info * get_student_info(time_t timer)
{
student_info *stu_info = (student_info *)malloc(sizeof(student_info));
if (!stu_info)
{
fprintf(stderr, "Failed to malloc memory.\n");
return NULL;
}
srand(timer);
stu_info->stu_id = 10000 + rand() % 9999;
stu_info->age = rand() % 30;
stu_info->score = rand() % 101;
print_student_info(stu_info);
return stu_info;
}
void * consumer_proc(void *arg)
{
struct ring_buffer *ring_buf = (struct ring_buffer *)arg;
student_info stu_info;
while(1)
{
sleep(2);
printf("------------------------------------------\n");
printf("get a student info from ring buffer.\n");
ring_buffer_get(ring_buf, (void *)&stu_info, sizeof(student_info));
printf("ring buffer length: %u\n", ring_buffer_len(ring_buf));
print_student_info(&stu_info);
printf("------------------------------------------\n");
}
return (void *)ring_buf;
}
void * producer_proc(void *arg)
{
time_t cur_time;
struct ring_buffer *ring_buf = (struct ring_buffer *)arg;
while(1)
{
time(&cur_time);
srand(cur_time);
int seed = rand() % 11111;
printf("******************************************\n");
student_info *stu_info = get_student_info(cur_time + seed);
printf("put a student info to ring buffer.\n");
ring_buffer_put(ring_buf, (void *)stu_info, sizeof(student_info));
printf("ring buffer length: %u\n", ring_buffer_len(ring_buf));
printf("******************************************\n");
sleep(1);
}
return (void *)ring_buf;
}
int consumer_thread(void *arg)
{
int err;
pthread_t tid;
err = pthread_create(&tid, NULL, consumer_proc, arg);
if (err != 0)
{
fprintf(stderr, "Failed to create consumer thread.errno:%u, reason:%s\n",
errno, strerror(errno));
return -1;
}
return tid;
}
int producer_thread(void *arg)
{
int err;
pthread_t tid;
err = pthread_create(&tid, NULL, producer_proc, arg);
if (err != 0)
{
fprintf(stderr, "Failed to create consumer thread.errno:%u, reason:%s\n",
errno, strerror(errno));
return -1;
}
return tid;
}
int main()
{
void * buffer = NULL;
uint32_t size = 0;
struct ring_buffer *ring_buf = NULL;
pthread_t consume_pid, produce_pid;
pthread_mutex_t *f_lock = (pthread_mutex_t *)malloc(sizeof(pthread_mutex_t));
if (pthread_mutex_init(f_lock, NULL) != 0)
{
fprintf(stderr, "Failed init mutex,errno:%u,reason:%s\n",
errno, strerror(errno));
return -1;
}
buffer = (void *)malloc(BUFFER_SIZE);
if (!buffer)
{
fprintf(stderr, "Failed to malloc memory.\n");
return -1;
}
size = BUFFER_SIZE;
ring_buf = ring_buffer_init(buffer, size, f_lock);
if (!ring_buf)
{
fprintf(stderr, "Failed to init ring buffer.\n");
return -1;
}
#if 0
student_info *stu_info = get_student_info(638946124);
ring_buffer_put(ring_buf, (void *)stu_info, sizeof(student_info));
stu_info = get_student_info(976686464);
ring_buffer_put(ring_buf, (void *)stu_info, sizeof(student_info));
ring_buffer_get(ring_buf, (void *)stu_info, sizeof(student_info));
print_student_info(stu_info);
#endif
printf("multi thread test.......\n");
produce_pid = producer_thread((void*)ring_buf);
consume_pid = consumer_thread((void*)ring_buf);
pthread_join(produce_pid, NULL);
pthread_join(consume_pid, NULL);
ring_buffer_free(ring_buf);
free(f_lock);
return 0;
} 总结：
len = min(len, fifo->size - fifo->in + fifo->out);
在 len 和 (fifo->size - fifo->in + fifo->out) 之间取一个较小的值赋给len。注意，当 (fifo->in == fifo->out+fifo->size) 时，表示缓冲区已满，此时得到的较小值一定是0，后面实际写入的字节数也全为0。
另一种边界情况是当 len 很大时（因为len是无符号的，负数对它来说也是一个很大的正数），这一句也能保证len取到一个较小的值，因为 fifo->in 总是大于等于 fifo->out ，所以后面的那个表达式 l = min(len, fifo->size - (fifo->in & (fifo->size - 1))); 的值不会超过fifo->size的大小。
smp_mb(); smp_wmb(); 是加内存屏障，这里不是我们讨论的范围，你可以忽略它。
l = min(len, fifo->size - (fifo->in & (fifo->size - 1))); 是把上一步决定的要写入的字节数len “切开”，这里又使用了一个技巧。注意：实际分配给 fifo->buffer 的字节数 fifo->size，必须是2的幂，否则这里就会出错。既然 fifo->size 是2的幂，那么 (fifo->size-1) 也就是一个后面几位全为1的数，也就能保证(fifo->in & (fifo->size - 1)) 总为不超过 (fifo->size - 1) 的那一部分，和 (fifo->in)% (fifo->size - 1) 的效果一样。
这样后面的代码就不难理解了，它先向 fifo->in 到缓冲区末端这一块写数据，如果还没写完，在从缓冲区头开始写入剩下的，从而实现了循环缓冲。最后，把写指针后移 len 个字节，并返回len。
从上面可以看出，fifo->in的值可以从0变化到超过fifo->size的数值，fifo->out也如此，但它们的差不会超过fifo->size。