copy_to_user、copy_from_user 分析

在内核的学习中会遇到很多挺有意思的函数,而且能沿着一个函数扯出来很多个相关的函数。copy_to_user和copy_from_user就是在进行驱动相关程序设计的时候,要经常遇到的两个函数。由于内核空间与用户空间的内存不能直接互访,因此借助函数copy_to_user()完成用户空间到内核空间的复制,函数copy_from_user()完成内核空间到用户空间的复制。下面我们来仔细的理一下这两个函数的来龙去脉。

首先,我们来看一下这两个函数的在源码文件中是如何定义的:

~/arch/i386/lib/usercopy.c

unsigned long

copy_to_user(void __user *to, const void *from, unsigned long n)

{

<wbr><wbr><wbr><wbr><wbr><wbr> might_sleep();</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> BUG_ON((long) n &lt; 0);</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> if (access_ok(VERIFY_WRITE, to, n))</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> n = __copy_to_user(to, from, n);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> return n;</wbr></wbr></wbr></wbr></wbr></wbr>

}

EXPORT_SYMBOL(copy_to_user);

从注释中就可以看出,这个函数的主要作用就是从内核空间拷贝一块儿数据到用户空间,由于这个函数有可能睡眠,所以只能用于用户空间。它有如下三个参数,

<wbr><wbr><wbr><wbr><wbr><wbr> To 目标地址,这个地址是用户空间的地址;</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> From 源地址,这个地址是内核空间的地址;</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> N 将要拷贝的数据的字节数。</wbr></wbr></wbr></wbr></wbr></wbr>

如果数据拷贝成功,则返回零;否则,返回没有拷贝成功的数据字节数。

以上是对函数的一些说明,接下来让我们看看这个函数的内部面目:

参数to的时候有个__user限定,这个在~/include/linux/compiler.h中有如下定义:

# define __user<wbr><wbr><wbr><wbr> __attribute__((noderef, address_space(1)))</wbr></wbr></wbr></wbr>

表示这是一个用户空间的地址,即其指向的为用户空间的内存

大家可能对这个__attribute__感到比较迷惑,不过没关系,google一下嘛

__attribute__是gnu c编译器的一个功能,它用来让开发者使用此功能给所声明的函数或者变量附加一个属性,以方便编译器进行错误检查,其实就是一个内核检查器。

具体可以参考如下:

http://unixwiz.net/techtips/gnu-c-attributes.html

接下来我们看一下

might_sleep();它有两个实现版本,debug版本和非debug版本:

在debug版本中,在有可能引起sleep的函数中会给出相应的提示,如果是在原子的上下文中执行,则会打印出栈跟踪的信息,这是通过__might_sleep(__FILE__, __LINE__);函数来实现的,并且接着调用might_resched()函数进行重新调度。

在非debug版本中直接调用might_resched()函数进行重新调度。

其实现方式为,在~/ include/linux/kernel.h中:

#ifdef CONFIG_DEBUG_SPINLOCK_SLEEP

void __might_sleep(char *file, int line);

# define might_sleep() \

do { __might_sleep(__FILE__, __LINE__); might_resched(); } while (0)

#else

# define might_sleep() do { might_resched(); } while (0)

#endif

接下来是一个检查参数合法性的宏:

BUG_ON((long) n < 0);

其实现为如下(在~/include/asm-generic/bug.h):

它通过检查条件,根据结果来决定是否打印相应的提示信息;

#ifdef CONFIG_BUG

#ifndef HAVE_ARCH_BUG

#define BUG() do { \

<wbr><wbr><wbr> printk("BUG: failure at %s:%d/%s()!\n", __FILE__, __LINE__, __FUNCTION__); \</wbr></wbr></wbr>

<wbr><wbr><wbr> panic("BUG!"); \</wbr></wbr></wbr>

} while (0)

#endif

#ifndef HAVE_ARCH_BUG_ON

#define BUG_ON(condition) do { if (unlikely((condition)!=0)) BUG(); } while(0)

#endif

<wbr><wbr><wbr> 接下来是一个宏</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> access_ok(VERIFY_WRITE, to, n)</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

它是用来检查参数中一个指向用户空间数据块的指针是否有效,如果有效返回非零,否则返回零。其实现如下(在/include/asm-i386/uaccess.h中):

#define access_ok(type,addr,size) (likely(__range_ok(addr,size) == 0))

其中__range_ok(addr,size)的实现是通过内嵌汇编来实现的,内容如下(在/include/asm-i386/uaccess.h中):

#define __range_ok(addr,size) ({ \

<wbr><wbr><wbr> unsigned long flag,sum; \</wbr></wbr></wbr>

<wbr><wbr><wbr> __chk_user_ptr(addr); \</wbr></wbr></wbr>

<wbr><wbr><wbr> asm("addl %3,%1 ; sbbl %0,%0; cmpl %1,%4; sbbl $0,%0" \</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> :"=&amp;r" (flag), "=r" (sum) \</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> :"1" (addr),"g" ((int)(size)),"g" (current_thread_info()-&gt;addr_limit.seg)); \</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

flag; })

其实现的功能为:

(u33)addr + (u33)size >= (u33)current->addr_limit.seg

<wbr><wbr><wbr> 判断上式是否成立,若不成立则表示地址有效,返回零;否则返回非零</wbr></wbr></wbr>

接下来的这个函数才是最重要的函数,它实现了拷贝的工作:

<wbr><wbr><wbr> __copy_to_user(to, from, n)</wbr></wbr></wbr>

其实现方式如下(在/include/asm-i386/uaccess.h中):

static __always_inline unsigned long __must_check

__copy_to_user(void __user *to, const void *from, unsigned long n)

{

<wbr><wbr><wbr><wbr><wbr><wbr> might_sleep();</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> return __copy_to_user_inatomic(to, from, n);</wbr></wbr></wbr></wbr></wbr></wbr>

}

有一个__always_inline宏,其内容就是inline,一个__must_check,其内容是在gcc3和gcc4版本里为__attribute__((warn_unused_result))

其中might_sleep同上面__user时候的注释。

最终调用的是__copy_to_user_inatomic(to, from, n)来完成拷贝工作的,此函数的实现如下(在/include/asm-i386/uaccess.h中):

static __always_inline unsigned long __must_check

__copy_to_user_inatomic(void __user *to, const void *from, unsigned long n)

{

<wbr><wbr><wbr> if (__builtin_constant_p(n)) {</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> unsigned long ret;</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> switch (n) {</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> case 1:</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> __put_user_size(*(u8 *)from, (u8 __user *)to, 1, ret, 1);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> return ret;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> case 2:</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> __put_user_size(*(u16 *)from, (u16 __user *)to, 2, ret, 2);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> return ret;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> case 4:</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> __put_user_size(*(u32 *)from, (u32 __user *)to, 4, ret, 4);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> return ret;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> }</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> }</wbr></wbr></wbr>

<wbr><wbr><wbr> return __copy_to_user_ll(to, from, n);</wbr></wbr></wbr>

}

其中__builtin_constant_p(n)为gcc的内建函数,__builtin_constant_p用于判断一个值是否为编译时常熟,如果参数n的值为常数,函数返回1,否则返回0。很多计算或操作在参数为常数时有更优化的实现,在 GNU C 中用上面的方法可以根据参数是否为常数,只编译常数版本或非常数版本,这样既不失通用性,又能在参数是常数时编译出最优化的代码。

如果n为常数1、2或者4,就会选择某个swith来执行拷贝动作,拷贝是通过如下函数来实现的(在/include/asm-i386/uaccess.h中):

#ifdef CONFIG_X86_WP_WORKS_OK

#define __put_user_size(x,ptr,size,retval,errret)<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

do {<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> retval = 0;<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> __chk_user_ptr(ptr);<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> switch (size) {<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> case 1: __put_user_asm(x,ptr,retval,"b","b","iq",errret);break; \</wbr></wbr></wbr>

<wbr><wbr><wbr> case 2: __put_user_asm(x,ptr,retval,"w","w","ir",errret);break; \</wbr></wbr></wbr>

<wbr><wbr><wbr> case 4: __put_user_asm(x,ptr,retval,"l","","ir",errret); break; \</wbr></wbr></wbr>

<wbr><wbr><wbr> case 8: __put_user_u64((__typeof__(*ptr))(x),ptr,retval); break;\</wbr></wbr></wbr>

<wbr><wbr><wbr> default: __put_user_bad();<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> }<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

} while (0)

#else

#define __put_user_size(x,ptr,size,retval,errret)<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

do {<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> __typeof__(*(ptr)) __pus_tmp = x;<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> retval = 0;<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> if(unlikely(__copy_to_user_ll(ptr, &amp;__pus_tmp, size) != 0)) \</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> retval = errret;<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

} while (0)

#endif

其中__put_user_asm为一个宏,拷贝工作是通过如下的内联汇编来实现的(在/include/asm-i386/uaccess.h中):

#define __put_user_asm(x, addr, err, itype, rtype, ltype, errret)<wbr><wbr> \</wbr></wbr>

<wbr><wbr><wbr> __asm__ __volatile__(<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "1: mov"itype" %"rtype"1,%2\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "2:\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> ".section .fixup,\"ax\"\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "3: movl %3,%0\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "<wbr><wbr> jmp 2b\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> ".previous\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> ".section __ex_table,\"a\"\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "<wbr><wbr> .align 4\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> "<wbr><wbr> .long 1b,3b\n"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> ".previous"<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> : "=r"(err)<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> \</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> : ltype (x), "m"(__m(addr)), "i"(errret), "0"(err))</wbr></wbr></wbr>

<wbr></wbr>


以上这两个函数是为了在拷贝小字节数据比如char/int等数据的时候考虑到效率来实现小数据拷贝。

而若n不是如上所说的常数,则进行数据块区域拷贝,其实现如下(~/arch/i386/lib/usercopy.c):

unsigned long __copy_to_user_ll(void __user *to, const void *from, unsigned long n)

{

<wbr><wbr><wbr> BUG_ON((long) n &lt; 0);</wbr></wbr></wbr>

#ifndef CONFIG_X86_WP_WORKS_OK

<wbr><wbr><wbr> if (unlikely(boot_cpu_data.wp_works_ok == 0) &amp;&amp;</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> ((unsigned long )to) &lt; TASK_SIZE) {</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> while (n) {</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> unsigned long offset = ((unsigned long)to)%PAGE_SIZE;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> unsigned long len = PAGE_SIZE - offset;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> int retval;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> struct page *pg;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> void *maddr;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> if (len &gt; n)</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> len = n;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

survive:

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> down_read(&amp;current-&gt;mm-&gt;mmap_sem);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> retval = get_user_pages(current, current-&gt;mm,</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> (unsigned long )to, 1, 1, 0, &amp;pg, NULL);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> if (retval == -ENOMEM &amp;&amp; current-&gt;pid == 1) {</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> up_read(&amp;current-&gt;mm-&gt;mmap_sem);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> blk_congestion_wait(WRITE, HZ/50);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> goto survive;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> }</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> if (retval != 1) {</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> up_read(&amp;current-&gt;mm-&gt;mmap_sem);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> break;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> }</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> maddr = kmap_atomic(pg, KM_USER0);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> memcpy(maddr + offset, from, len);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> kunmap_atomic(maddr, KM_USER0);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> set_page_dirty_lock(pg);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> put_page(pg);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> up_read(&amp;current-&gt;mm-&gt;mmap_sem);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> from += len;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> to += len;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> n -= len;</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> }</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> return n;</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> }</wbr></wbr></wbr>

#endif

<wbr><wbr><wbr> if (movsl_is_ok(to, from, n))</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> __copy_user(to, from, n);</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> else</wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr> n = __copy_user_intel(to, from, n);</wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr> return n;</wbr></wbr></wbr>

}

EXPORT_SYMBOL(__copy_to_user_ll);

<wbr></wbr>

下面是copy_from_user函数的实现:

unsigned long

copy_from_user(void *to, const void __user *from, unsigned long n)

{

<wbr><wbr><wbr><wbr><wbr><wbr> might_sleep();</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> BUG_ON((long) n &lt; 0);</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> if (access_ok(VERIFY_READ, from, n))</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> n = __copy_from_user(to, from, n);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> else</wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr><wbr> memset(to, 0, n);</wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr></wbr>

<wbr><wbr><wbr><wbr><wbr><wbr> return n;</wbr></wbr></wbr></wbr></wbr></wbr>

}

EXPORT_SYMBOL(copy_from_user);

其实现方式与copy_to_user函数的实现方式类似:就不再累述了。

如上就是copy_to_user和copy_from_user两个函数的工作方式,这些进行简单的分析与跟踪。细节的部分还有待于进一步研究。

<wbr></wbr>

copy_to_user与mmap的工作原理

copy_to_user在每次拷贝时需要检测指针的合法性,也就是用户空间的指针所指向的地址的确是一段该进程本身的地址,而不是指向了不属于它的地方,而且每次都会拷贝一次数据,频繁访问内存,由于虚拟地址连续,物理地址不一定会连续,从而造成CPU的CACHE频繁失效,从而使速度降低<wbr><wbr><br><wbr> mmap仅在第一次使用时为进程建立页表,也就是将一段物理地址映射到一段虚拟地址上,以后操作时不再检测其地址的合法性(合法性交由CPU页保护异常来做),另一方面是内核下直接操作mmap地址,可以不用频繁拷贝,也就是说在内核下直接可用指针向该地址操作,而不再在内核中专门开一个缓冲区,然后将缓冲区中的数据拷贝一次进来,mmap一般是将一段连续的物理地址映射成一段虚拟地址,当然,也可以将每段连续,但各段不连续的物理地址映射成一段连续的虚拟地址,无论如何,其物理地址在每段之中是连续的,这样一来,就不会造成CPU的CACHE频繁失效,从而大大节约时间。</wbr></wbr></wbr>

  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值