fflush and fsync区别

2006-10-24 09:46

缓冲还是不缓冲？这是个问题

主要涉及到 setbuf, fflush, fsync,sync等函数。

首先来说输入输出库的缓冲。

The  three  types  of  buffering  available  are unbuffered, block buffered, and line buffered.  When an output  stream is unbuffered, information appears on the destination file or terminal as soon as written;  when  it  is
       block  buffered  many  characters  are saved up and written as a block; when it is line buffered characters are
       saved up until a newline is output or input is read from any stream attached to a  terminal  device  (typically
       stdin).  The function fflush(3) may be used to force the block out early.  (See fclose(3).)  Normally all files
       are block buffered.  When the first I/O operation occurs on a file,  malloc(3)  is  called,  and  a  buffer  is
       obtained.   If a stream refers to a terminal (as stdout normally does) it is line buffered.  The standard error
       stream stderr is always unbuffered by default.

一般来说，block buffered的效率高些，将多次的操作合并成一次操作。现在标准库里缓存一部分，
直到该缓冲区满了，或者程序显示的调用fflush时，将进行更新操作。

而setbuf 则可以设置该缓冲区的大小。
#include <stdio.h>
void setbuf(FILE *stream, char *buf);

这个函数应该必须在如何输出被写到该文件之前调用。一般放在main里靠前面的语句！

但是setbuf有个经典的错误，man手册上也提到了，c陷阱和缺陷上也提到了

You  must  make  sure  that both buf and the space it points to still exist by the time stream is closed, which    also happens at program termination.  For example, the following is illegal:

       #include <stdio.h>
       int main()
       {
           char buf[BUFSIZ];
           setbuf(stdin, buf);
           printf("Hello, world!\n");
           return 0;
       }
这个程序是错误的。buf缓冲区最后一次清空应该在main函数结束之后，程序交回控制给操作系统
之前C运行库所必须进行的清理工作的一部分，但是此时buf字符数组已经释放。
修改的方法是 将buf设置为static，或者全局变量； 或者调用malloc来动态申请内存。
char * malloc();
setbuf(stdout,malloc(BUFSIZE));
这里不需要判断malloc的返回值，如果malloc调用失败，将返回一个null指针，setbuf的第二个参数可以
是null,此时不进行缓冲！

对应的,fflush函数则刷新缓冲区，将缓冲区上的内容更新到文件里。

       #include <stdio.h>

       int fflush(FILE *stream);

The  function  fflush  forces a write of all user-space buffered data for the given output or update stream via  the stream underlying write function.  The open status of the stream is unaffected.
 If the stream argument is NULL, fflush flushes all open output streams.

但是fflush仅仅刷新C库里的缓冲。
其他的一些数据的刷新需要调用fsync或者sync!!
 Note  that  fflush() only flushes the user space buffers provided by the C library.  To ensure that the data is     physically stored on disk the kernel buffers must be flushed too, e.g. with sync(2) or fsync(2).

fsync和sync最终将缓冲的数据更新到文件里。
 #include <unistd.h>
int fsync(int fd);
 fsync copies all in-core parts of a file to disk, and waits until the device reports that all parts are on stable storage.  It also updates metadata stat information. It does not necessarily ensure that the entry  in  the      directory  containing the file has also reached disk.  For that an explicit fsync on the file descriptor of the directory is also needed.

NAME
       sync - commit buffer cache to disk

SYNOPSIS
       #include <unistd.h>

       void sync(void);

DESCRIPTION
       sync first commits inodes to buffers, and then buffers to disk.

ERRORS
       This function is always successful.

同步命令sync就直接调用了sync函数来更新磁盘上的缓冲！！

 

这两天工作上的新任务，需要把设备上抓包生成的pcap文件拷贝到U盘上，不复杂，写个函数完成把原始pcap文件搬运到u盘就可以了，函数如下：

static void save_pcapfile(const char* szSrcFile, const char* szDstFile)
{
FILE *Src = NULL, *Dst = NULL;
int read_size = 0;
char read_buff[1024];

if ((szSrcFile == NULL) || (szDstFile == NULL))
return;

Src = fopen(szSrcFile, “rb”);
if (Src == NULL)
return;
Dst = fopen(szDstFile, “wb”);
if (Dst == NULL)
return;

memset(read_buff, 0, sizeof(read_buff));
while ((read_size = fread(read_buff, 1, sizeof(read_buff), Src)) > 0)
{
fwrite(read_buff, 1, read_size, Dst);
}

/* maybe User unplug the USB disk immediatly */
fflush(Dst);
fsync(fileno(Dst));

fclose(Src);
fclose(Dst);
}

其中后面两个系统函数 fflush() 和 fsync() 是为了避免用户在拷贝成功之后很快拔掉U盘而导致数据丢失的问题，而且两个函数必须都调用，否则还是会出现拷贝到U盘的文件不完整甚至为空的情况。

在 man 3 fflush 中也解释的比较清楚：

Note that fflush() only flushes the user space buffers provided by the C
library. To ensure that the data is physically stored on disk the kernel buf‐
fers must be flushed too, for example, with sync(2) or fsync(2).