Linux 读写锁
线程的读写锁函数：
1，读写锁的初始化与销毁，静态初始化的话，可以直接使用PTHREAD_RWLOCK_INITIALIZER。
#include <pthread.h>
int pthread_rwlock_destroy(pthread_rwlock_t *rwlock);
int pthread_rwlock_init(pthread_rwlock_t *restrict rwlock,
const pthread_rwlockattr_t *restrict attr);
pthread_rwlock_t rwlock = PTHREAD_RWLOCK_INITIALIZER;
2，用读的方式加锁和尝试（没锁上就立即返回）加锁。
#include <pthread.h>
int pthread_rwlock_rdlock(pthread_rwlock_t *rwlock);
int pthread_rwlock_tryrdlock(pthread_rwlock_t *rwlock);
3，用写的方式加锁和尝试（没锁上就立即返回）加锁。
#include <pthread.h>
int pthread_rwlock_wrlock(pthread_rwlock_t *rwlock);
int pthread_rwlock_trywrlock(pthread_rwlock_t *rwlock);
4，解锁
#include <pthread.h>
int pthread_rwlock_unlock(pthread_rwlock_t *rwlock);
多个进程在同时读写同一个文件，会发生什么？
例子1：用下面的例子的执行结果，观察多个进程在同时读写同一个文件，会发生什么。
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#define MAXLINE 100
#define FN "num1"
void my_lock(int fd){
return;
}
void my_unlock(int fd){
return;
}
int main(int args, char** argv){
int fd;
long i,seqno;
pid_t pid;
ssize_t n;
char line[MAXLINE + 1];
pid = getpid();
fd = open(FN, O_RDWR, 0664);
for(i = 0; i < 20; ++i){
my_lock(fd);
lseek(fd, 0L, SEEK_SET);
n = read(fd, line, MAXLINE);
line[n] = '\0';
seqno = atol(line);
printf("%s:pid = %ld, seq = %ld\n", argv[0], (long)pid, seqno);
seqno++;
snprintf(line, sizeof(line), "%ld\n", seqno);
lseek(fd, 0L, SEEK_SET);
write(fd, line, strlen(line));
my_unlock(fd);
}
return 0;
}
执行方法：同时执行上面例子的程序2次，也就是2个进程同时读写同一个文件。
ubuntu$ ./flockmain1 & ./flockmain1 &
执行结果如下，发现2个进程同时读写，在①处开始，内核切换进程时，数字乱套了。
ubuntu$ ./flockmain1 & ./flockmain1 &
[1] 4760
[2] 4761
ubuntu$ ./flockmain1:pid = 4761, seq = 1
./flockmain1:pid = 4761, seq = 2
./flockmain1:pid = 4761, seq = 3
./flockmain1:pid = 4761, seq = 4
./flockmain1:pid = 4761, seq = 5
./flockmain1:pid = 4761, seq = 6
./flockmain1:pid = 4761, seq = 7
./flockmain1:pid = 4761, seq = 8
./flockmain1:pid = 4761, seq = 9
./flockmain1:pid = 4761, seq = 10 ------------①
./flockmain1:pid = 4760, seq = 10
./flockmain1:pid = 4761, seq = 11
./flockmain1:pid = 4761, seq = 12
./flockmain1:pid = 4761, seq = 13
./flockmain1:pid = 4761, seq = 14
./flockmain1:pid = 4761, seq = 15
./flockmain1:pid = 4761, seq = 16
./flockmain1:pid = 4761, seq = 17
./flockmain1:pid = 4761, seq = 18
./flockmain1:pid = 4761, seq = 19
./flockmain1:pid = 4761, seq = 20
./flockmain1:pid = 4760, seq = 11
./flockmain1:pid = 4760, seq = 12
./flockmain1:pid = 4760, seq = 13
./flockmain1:pid = 4760, seq = 14
./flockmain1:pid = 4760, seq = 15
./flockmain1:pid = 4760, seq = 16
./flockmain1:pid = 4760, seq = 17
./flockmain1:pid = 4760, seq = 18
./flockmain1:pid = 4760, seq = 19
./flockmain1:pid = 4760, seq = 20
./flockmain1:pid = 4760, seq = 21
./flockmain1:pid = 4760, seq = 22
./flockmain1:pid = 4760, seq = 23
./flockmain1:pid = 4760, seq = 24
./flockmain1:pid = 4760, seq = 25
./flockmain1:pid = 4760, seq = 26
./flockmain1:pid = 4760, seq = 27
./flockmain1:pid = 4760, seq = 28
./flockmain1:pid = 4760, seq = 29
为了解决上面的问题，必须对文件的内容进行加锁。
如何对文件内容加锁？
使用fcntl函数，它既可以锁整文件，也可以锁文件里的某段内容。通过结构体flock来指定要锁的范围。如果 whence = SEEK_SET;l_start = 0;l_len = 0;就是锁定整个文件。
struct flock {
...
short l_type; /* Type of lock: F_RDLCK,
F_WRLCK, F_UNLCK */
short l_whence; /* How to interpret l_start:
SEEK_SET, SEEK_CUR, SEEK_END */
off_t l_start; /* Starting offset for lock */
off_t l_len; /* Number of bytes to lock */
pid_t l_pid; /* PID of process blocking our lock
(set by F_GETLK and F_OFD_GETLK) */
...
};
- F_SETLK：上锁。如果发现已经被别的进程上锁了，就直接返回-1，errno被设置成EACCES或者EAGAIN，不阻塞。
- F_SETLKW：上锁。阻塞等待。
- F_GETLK：得到锁的状态。
修改上面的函数my_lock，my_unlock。main函数不变。
例子2：
void my_lock(int fd){
struct flock lock;
lock.l_type = F_WRLCK;
wlock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLKW, lock);
}
void my_unlock(int fd){
struct flock lock;
lock.l_type = F_UNLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLK, lock);
}
执行结果如下，发现数字不乱套了。
ubuntu$ ./flockmain & ./flockmain &
[1] 4882
[2] 4883
ubuntu$ ./flockmain:pid = 4883, seq = 1
./flockmain:pid = 4883, seq = 2
./flockmain:pid = 4883, seq = 3
./flockmain:pid = 4883, seq = 4
./flockmain:pid = 4883, seq = 5
./flockmain:pid = 4883, seq = 6
./flockmain:pid = 4883, seq = 7
./flockmain:pid = 4883, seq = 8
./flockmain:pid = 4883, seq = 9
./flockmain:pid = 4883, seq = 10
./flockmain:pid = 4883, seq = 11
./flockmain:pid = 4883, seq = 12
./flockmain:pid = 4883, seq = 13
./flockmain:pid = 4883, seq = 14
./flockmain:pid = 4883, seq = 15
./flockmain:pid = 4883, seq = 16
./flockmain:pid = 4883, seq = 17
./flockmain:pid = 4883, seq = 18
./flockmain:pid = 4883, seq = 19
./flockmain:pid = 4883, seq = 20
./flockmain:pid = 4882, seq = 21
./flockmain:pid = 4882, seq = 22
./flockmain:pid = 4882, seq = 23
./flockmain:pid = 4882, seq = 24
./flockmain:pid = 4882, seq = 25
./flockmain:pid = 4882, seq = 26
./flockmain:pid = 4882, seq = 27
./flockmain:pid = 4882, seq = 28
./flockmain:pid = 4882, seq = 29
./flockmain:pid = 4882, seq = 30
./flockmain:pid = 4882, seq = 31
./flockmain:pid = 4882, seq = 32
./flockmain:pid = 4882, seq = 33
./flockmain:pid = 4882, seq = 34
./flockmain:pid = 4882, seq = 35
./flockmain:pid = 4882, seq = 36
./flockmain:pid = 4882, seq = 37
./flockmain:pid = 4882, seq = 38
./flockmain:pid = 4882, seq = 39
./flockmain:pid = 4882, seq = 40
到此为止，貌似解决了问题，但是如果同时执行例子1和例子2，结果如下，发现还是乱的。
也就是说在协作线程（cooperating processes）间，文件锁（也叫劝告性上锁）也起作用的。但是不完全不相关的进程中，文件锁也不起作用的。如何解决呢？使用强制性上锁。
ys@ys-VirtualBox:~/IPC$ ./flockmain1 & ./flockmain &
[1] 3602
[2] 3603
ys@ys-VirtualBox:~/IPC$ ./flockmain1:pid = 3602, seq = 1
./flockmain:pid = 3603, seq = 1
./flockmain:pid = 3603, seq = 2
./flockmain:pid = 3603, seq = 3
./flockmain:pid = 3603, seq = 4
./flockmain:pid = 3603, seq = 5
./flockmain:pid = 3603, seq = 6
./flockmain:pid = 3603, seq = 7
./flockmain:pid = 3603, seq = 8
./flockmain:pid = 3603, seq = 9
./flockmain:pid = 3603, seq = 10
./flockmain1:pid = 3602, seq = 2
./flockmain1:pid = 3602, seq = 3
./flockmain1:pid = 3602, seq = 4
./flockmain:pid = 3603, seq = 11
./flockmain:pid = 3603, seq = 12
./flockmain1:pid = 3602, seq = 5
./flockmain:pid = 3603, seq = 13
./flockmain1:pid = 3602, seq = 6
./flockmain1:pid = 3602, seq = 7
./flockmain1:pid = 3602, seq = 8
./flockmain:pid = 3603, seq = 14
./flockmain:pid = 3603, seq = 15
./flockmain1:pid = 3602, seq = 9
./flockmain1:pid = 3602, seq = 10
./flockmain:pid = 3603, seq = 16
./flockmain:pid = 3603, seq = 17
./flockmain1:pid = 3602, seq = 11
./flockmain:pid = 3603, seq = 18
./flockmain1:pid = 3602, seq = 12
./flockmain1:pid = 3602, seq = 13
./flockmain1:pid = 3602, seq = 14
./flockmain:pid = 3603, seq = 19
./flockmain:pid = 3603, seq = 20
./flockmain1:pid = 3602, seq = 15
./flockmain1:pid = 3602, seq = 16
./flockmain1:pid = 3602, seq = 17
./flockmain1:pid = 3602, seq = 18
./flockmain1:pid = 3602, seq = 19
./flockmain1:pid = 3602, seq = 20
第一个问题：假如一个文件被一个进程以读的方式锁定，并有另一个进程在等待读锁定解锁后，用写入的方式锁定，这时是否允许另一个进程的还以读的方式取得锁定？
用例子3来观察：
#include <time.h>
#include <sys/time.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <pthread.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/wait.h>
void gftime(char* buf){
struct timeval tv;
gettimeofday(&tv, NULL);
long usec = tv.tv_usec;
struct tm* tm = localtime(&tv.tv_sec);
sprintf(buf, "%d:%d:%d.%ld",tm->tm_hour, tm->tm_min, tm->tm_sec,usec);
}
int main(){
char buff[100] = {0};
int fd = open("test.dat", O_RDWR | O_CREAT, 0664);
struct flock lock;
lock.l_type = F_RDLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLK, &lock);
gftime(buff);
printf("%s: parent has read lock\n", buff);
//first child
if(fork() == 0){
char buf2[100] = {0};
sleep(1);
gftime(buf2);
printf("%s: first child tries to obtain write lock\n", buf2);
struct flock lock2;
lock2.l_type = F_WRLCK;
lock2.l_whence = SEEK_SET;
lock2.l_start = 0;
lock2.l_len = 0;
fcntl(fd, F_SETLKW, &lock2);
gftime(buf2);
printf("%s: first child obtains write lock\n", buf2);
sleep(2);
lock2.l_type = F_UNLCK;
lock2.l_whence = SEEK_SET;
lock2.l_start = 0;
lock2.l_len = 0;
fcntl(fd, F_SETLK, &lock2);
gftime(buf2);
printf("%s: first child releases write lock\n", buf2);
exit(0);
}
//secodn child
if(fork() == 0){
char buf1[100] = {0};
sleep(3);
gftime(buf1);
printf("%s: second child tries to obtain read lock\n", buf1);
struct flock lock1;
lock1.l_type = F_RDLCK;
lock1.l_whence = SEEK_SET;
lock1.l_start = 0;
lock1.l_len = 0;
fcntl(fd, F_SETLKW, &lock1);
gftime(buf1);
printf("%s: second child obtains read lock\n", buf1);
sleep(4);
lock1.l_type = F_UNLCK;
lock1.l_whence = SEEK_SET;
lock1.l_start = 0;
lock1.l_len = 0;
fcntl(fd, F_SETLK, &lock1);
gftime(buf1);
printf("%s: second child release read lock\n", buf1);
exit(0);
}
//parent process
sleep(5);
lock.l_type = F_UNLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLK, &lock);
gftime(buff);
printf("%s: parent releases read lock\n", buff);
wait(NULL);
wait(NULL);
exit(0);
}
在ubuntu上执行结果：
17:49:44.348946: parent has read lock
17:49:45.350191: first child tries to obtain write lock
17:49:47.350155: second child tries to obtain read lock
17:49:47.350409: second child obtains read lock
17:49:49.349442: parent releases read lock
17:49:51.351197: second child release read lock
17:49:51.351582: first child obtains write lock
17:49:53.351689: first child releases write lock
第一个问题的答案：允许另一个进程的还以读的方式取得锁定
第二个问题：假如一个文件被一个进程以写的方式锁定，这时又有2个进程在等待这个锁的释放，其中一个进程是以写锁的方式等待，其中另一个进程是以读锁的方式等待，哪一个会优先取得锁？
用例子4来观察：
#include <time.h>
#include <sys/time.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <pthread.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/wait.h>
void gftime(char* buf){
struct timeval tv;
gettimeofday(&tv, NULL);
long usec = tv.tv_usec;
struct tm* tm = localtime(&tv.tv_sec);
sprintf(buf, "%d:%d:%d.%ld",tm->tm_hour, tm->tm_min, tm->tm_sec,usec);
}
int main(){
char buff[100] = {0};
int fd = open("test.dat", O_RDWR | O_CREAT, 0664);
struct flock lock;
lock.l_type = F_WRLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLK, &lock);
gftime(buff);
printf("%s: parent has write lock\n", buff);
//first child
if(fork() == 0){
char buf2[100] = {0};
sleep(1);
gftime(buf2);
printf("%s: first child tries to obtain write lock\n", buf2);
struct flock lock2;
lock2.l_type = F_WRLCK;
lock2.l_whence = SEEK_SET;
lock2.l_start = 0;
lock2.l_len = 0;
fcntl(fd, F_SETLKW, &lock2);
gftime(buf2);
printf("%s: first child obtains write lock\n", buf2);
sleep(2);
lock2.l_type = F_UNLCK;
lock2.l_whence = SEEK_SET;
lock2.l_start = 0;
lock2.l_len = 0;
fcntl(fd, F_SETLK, &lock2);
gftime(buf2);
printf("%s: first child releases write lock\n", buf2);
exit(0);
}
//secodn child
if(fork() == 0){
char buf1[100] = {0};
sleep(3);
gftime(buf1);
printf("%s: second child tries to obtain read lock\n", buf1);
struct flock lock1;
lock1.l_type = F_RDLCK;
lock1.l_whence = SEEK_SET;
lock1.l_start = 0;
lock1.l_len = 0;
fcntl(fd, F_SETLKW, &lock1);
gftime(buf1);
printf("%s: second child obtains read lock\n", buf1);
sleep(4);
lock1.l_type = F_UNLCK;
lock1.l_whence = SEEK_SET;
lock1.l_start = 0;
lock1.l_len = 0;
fcntl(fd, F_SETLK, &lock1);
gftime(buf1);
printf("%s: second child release read lock\n", buf1);
exit(0);
}
//parent process
sleep(5);
lock.l_type = F_UNLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
fcntl(fd, F_SETLK, &lock);
gftime(buff);
printf("%s: parent releases write lock\n", buff);
wait(NULL);
wait(NULL);
exit(0);
}
在ubuntu上执行结果：
17:49:29.796599: parent has write lock
17:49:30.797099: first child tries to obtain write lock
17:49:32.796885: second child tries to obtain read lock
17:49:34.796868: parent releases write lock
17:49:34.796987: second child obtains read lock
17:49:38.797148: second child release read lock
17:49:38.797297: first child obtains write lock
17:49:40.797727: first child releases write lock
第二个问题的答案：没有准确答案。在Ubuntu上的执行结果上看，读锁优先了，但是，可能在别的环境上又是写锁优先。按道理来说应该写锁优先吧？
c/c++ 学习互助QQ群：877684253
本人微信：xiaoshitou5854
Linux 读写锁的更多相关文章
- linux读写锁
一.概述 读写锁与互斥量的功能类似,对临界区的共享资源进行保护!互斥量一次只让一个线程进入临界区, ...
- linux 读写锁应用实例
转自:http://blog.csdn.net/dsg333/article/details/22113489 /*使用读写锁实现四个线程读写一段程序的实例,共创建了四个新的线程,其中两个线程用来读取 ...
- Linux读写锁的使用
读写锁是用来解决读者写者问题的,读操作可以共享,写操作是排它的,读可以有多个在读,写只有唯一个在写,写的时候不允许读. 具有强读者同步和强写者同步两种形式: 强读者同步:当写者没有进行写操作时,读者就 ...
- Linux的线程同步对象：互斥量Mutex，读写锁，条件变量
进程是Linux资源分配的对象,Linux会为进程分配虚拟内存(4G)和文件句柄等 资源,是一个静态的概念.线程是CPU调度的对象,是一个动态的概念.一个进程之中至少包含有一个或者多个线程.这 ...
- linux中读写锁的rwlock介绍-nk_ysg-ChinaUnix博客
linux中读写锁的rwlock介绍-nk_ysg-ChinaUnix博客 linux中读写锁的rwlock介绍 2013-02-26 13:59:35 分类: C/C++ http://yaro ...
- linux 内核的另一个自旋锁 - 读写锁
除spinlock外,linux 内核还有一个自旋锁,名为arch_rwlock_t.它的头文件是qrwlock.h,包含在spinlock.h,头文件中对它全称为"Queue read/w ...
- linux线程间同步(1)读写锁
读写锁比mutex有更高的适用性,能够多个线程同一时候占用读模式的读写锁.可是仅仅能一个线程占用写模式的读写锁. 1. 当读写锁是写加锁状态时,在这个锁被解锁之前,全部试图对这个锁加锁的线程都会被堵塞 ...
- Linux系统编程 —读写锁rwlock
读写锁是另一种实现线程间同步的方式.与互斥量类似,但读写锁将操作分为读.写两种方式,可以多个线程同时占用读模式的读写锁,这样使得读写锁具有更高的并行性. 读写锁的特性为:写独占,读共享:写锁优先级高. ...
- linux kernel RCU 以及读写锁
信号量有一个很明显的缺点,没有区分临界区的读写属性,读写锁允许多个线程进程并发的访问临界区,但是写访问只限于一个线程,在多处理器系统中允许多个读者访问共享资源,但是写者有排他性,读写锁的特性如下:允许 ...
随机推荐
- [PHP] 内部接口简单加密验证方式
1. 当有内部系统之间进行调用的时候,也需要简单的进行一下调用方的验证,一种简单的内部接口加密验证方式.此加密方式需要三个参数,分别是api地址,pin码,entry标识,其中pin和entry是接口 ...
- Java面试题-基础篇二（干货）
11.是否可以从一个static方法内部发出对非static方法的调用? 不可以.因为非static方法是要与对象关联在一起的,必须创建一个对象后,才可以在该对象上进行方法调用,而static方法调用 ...
- 受限玻尔兹曼机(Restricted Boltzmann Machine)
受限玻尔兹曼机(Restricted Boltzmann Machine) 作者:凯鲁嘎吉 - 博客园 http://www.cnblogs.com/kailugaji/ 1. 生成模型 2. 参数学 ...
- 学习CNN系列一：原理篇
CNN的发展历程: 1962年,卷积神经网络的研究起源于Hubel和Wiesel研究毛脑皮层的发现局部互连网络可以有效降低反馈神经网络的复杂性. 1980年,CNN的第一个实现网络:Fukushima ...
- template might not exist or might not be accessible by any of the configured Template Resolvers at org.thymeleaf.engine.TemplateManager.resolveTemplate(TemplateManager.java:869)
org.thymeleaf.exceptions.TemplateInputException: Error resolving template [code/leading], template m ...
- git 推送本地项目到远程库
git 推送本地项目到远程库 1@DESKTOP-3H9092J MINGW64 /e/mozq/00store/01/SmartCard_MS $ git init Initialized empt ...
- Ubuntu下的录GIF神器——Peek
最近一直在找Ubuntu下面录GIF好用方便的软件一直没找到,很多都是要获取录屏四角的坐标,现在终于找到了,就是我们的录GIF神器——Peek. 1 获取Peek的ppa源 sudo add-apt- ...
- 海边拾贝-A-算法篇
收集若干算法博客地址,不定期会更新: 算法 陈浩,leetcode 的C++解法 https://github.com/haoel/leetcode 王亮,很多题目讲解的比较详细 https:/ ...
- LeetCode 206：反转链表 Reverse Linked List
反转一个单链表. Reverse a singly linked list. 示例: 输入: 1->2->3->4->5->NULL 输出: 5->4->3- ...
- jvm的组成入门
JVM的组成分为整体组成部分和运行时数据区组成部分. JVM的整体组成 JVM的整体组成可以分为4个部分:类加载器(Classloader).运行时数据区(Runtime Data Area).执行引 ...