在邮件列表里讨论了一下writel是如何实现的,这个函数实现在操作系统层,有内存保护的情况下,往一个寄存器或者内存地址写一个数据。
在arch/alpha/kernel/io.c中有
188
void
writel(u32 b,
volatile
void
__iomem
*
addr)
189 {
190 __raw_writel(b, addr);
191 mb();
192 }
189 {
190 __raw_writel(b, addr);
191 mb();
192 }
这样一个writel函数的作用应该是向一个地址上写一个值,我想知道这个函数底下具体实现的细节,于是往下继续跟踪代码:__raw_writel(b, addr);
129
void
__raw_writel(u32 b,
volatile
void
__iomem
*
addr)
130 {
131 IO_CONCAT(__IO_PREFIX,writel)(b, addr);
132 }
130 {
131 IO_CONCAT(__IO_PREFIX,writel)(b, addr);
132 }
再往下跟踪 IO_CONCAT,在对应的io.h中的定义如下:
134
#define IO_CONCAT(a,b) _IO_CONCAT(a,b)
135 #define _IO_CONCAT(a,b) a ## _ ## b
135 #define _IO_CONCAT(a,b) a ## _ ## b
这段代码前几天问过了,是标示将两边的字符串连接起来的意思。
跟踪__IO_PREFIX 定义如下
501
#undef __IO_PREFIX
502 #define __IO_PREFIX apecs
502 #define __IO_PREFIX apecs
到这里就结束了,再往下我就晕了,有问题如下:
1、到底是怎么将数据写入地址的?我把这些单独提取出来,进行预编译,宏展开后,发现是这样的:
void
__raw_writel(u32 b,
volatile
void
__iomem
*
addr)
{
apecs_writel(b, addr);
}
{
apecs_writel(b, addr);
}
但是在内核里根本就没找到apecs_writel函数,请帮忙解释下。
For the first question,
you should refer to the file " arch\alpha\kernle\Machvec_impl.h "
" ~\Machve.h " " ~\io.c " " ~\io.h " " ~\core_**.h " .
as you have analysized before, in the file Machvec_impl.h and Machve.h,
DO_CIA_IO,IO,IO_LITE, these three macros implement the symbole
connection between ** arch and writel function, and the function
pointer initializations.
so, the details implementation to writel is to init the
alpha_machine_vector structure and the definition to the relevant
function pointer invoked to complete the low - level write operation.
.mv_writel = CAT(low,_writel), <--- IO(CIA,cia) <--> cia_writel(b, addr); <---
|
writel(b, addr) --> __raw_writel(b, addr); ---> cia_writel(b,addr) ---------------
For the second quesiton,
mb() ---> __asm__ __volatile__( " mb " : : : " memory " );
so, it is a memory barrier for alpha architecture to ensure some
operations before some actions could be occured.
and, it is similiar with the barrier() in x86 platform / arm platform.
you should refer to the file " arch\alpha\kernle\Machvec_impl.h "
" ~\Machve.h " " ~\io.c " " ~\io.h " " ~\core_**.h " .
as you have analysized before, in the file Machvec_impl.h and Machve.h,
DO_CIA_IO,IO,IO_LITE, these three macros implement the symbole
connection between ** arch and writel function, and the function
pointer initializations.
so, the details implementation to writel is to init the
alpha_machine_vector structure and the definition to the relevant
function pointer invoked to complete the low - level write operation.
.mv_writel = CAT(low,_writel), <--- IO(CIA,cia) <--> cia_writel(b, addr); <---
|
writel(b, addr) --> __raw_writel(b, addr); ---> cia_writel(b,addr) ---------------
For the second quesiton,
mb() ---> __asm__ __volatile__( " mb " : : : " memory " );
so, it is a memory barrier for alpha architecture to ensure some
operations before some actions could be occured.
and, it is similiar with the barrier() in x86 platform / arm platform.
继续阅读代码,看看定义__IO_PREFIX之后紧接着包含了哪个头文件。在哪个头文
件里面寻找答案。对于你的apsec,看看以下代码段(linux-2.6.28-rc4)
件里面寻找答案。对于你的apsec,看看以下代码段(linux-2.6.28-rc4)
arch
/
alpha
/
include
/
asm
/
core_apecs.h
------------------------------------------
#undef __IO_PREFIX
#define __IO_PREFIX apecs
#define apecs_trivial_io_bw 0
#define apecs_trivial_io_lq 0
#define apecs_trivial_rw_bw 2
#define apecs_trivial_rw_lq 1
#define apecs_trivial_iounmap 1
#include < asm / io_trivial.h >
------------------------------------------
arch / alpha / include / asm / io_trivial.h
------------------------------------------
__EXTERN_INLINE void
IO_CONCAT(__IO_PREFIX,writel)(u32 b, volatile void __iomem * a)
{
* ( volatile u32 __force * )a = b;
}
------------------------------------------
#undef __IO_PREFIX
#define __IO_PREFIX apecs
#define apecs_trivial_io_bw 0
#define apecs_trivial_io_lq 0
#define apecs_trivial_rw_bw 2
#define apecs_trivial_rw_lq 1
#define apecs_trivial_iounmap 1
#include < asm / io_trivial.h >
------------------------------------------
arch / alpha / include / asm / io_trivial.h
------------------------------------------
__EXTERN_INLINE void
IO_CONCAT(__IO_PREFIX,writel)(u32 b, volatile void __iomem * a)
{
* ( volatile u32 __force * )a = b;
}
就是最终通过*(volatile u32 __force *)a = b;
来写入数据的。
来写入数据的。
如果在没有os,没有mmu的情况下,当开发板裸跑的时候,我们只需要一句话就一切ok:
*(unsigned long *)addr = value;