arm 2410 写入数据发生诡异错位问题

硬件环境 arm s3c2410a-20  
交叉编译环境
CT_BINUTILS_VERSION=2.19.1   
CT_KERNEL_VERSION=2.6.29.4
CT_LIBC_VERSION=2.9
[DEBUG] CT_ARCH_TARGET_CFLAGS=' -mlittle-endian -march=armv4t -mcpu=arm9tdmi -mtune=arm920t -msoft-float'
[DEBUG] CT_ARCH_TARGET_LDFLAGS=' -EL'
[DEBUG] CT_ARCH_TUNE=arm920t
gcc-4.3.2

出现问题的代码如下

int main(int argc, char *argv[])
{
    U8 testBuf[100]= {0};
    //memset( testBuf,0,100 );
    U16 *len = (U16*)testBuf;
    U8 *ver = testBuf+2;
    U8 *msgType = testBuf+3;
    U8 *flag = testBuf+4;
    U32 *data = (U32*)(testBuf+5);

    *len = 4;
    *ver = 1;
    *msgType = 2;
    *flag = 3;

    int dataLen(0);
    bool reVal = true;
    //for (int i = 0; i < 100; ++i)
    //{
        *data = 0xFFFFFFFF;
    //}

    return 0;
}

GDB调试日志如下

Breakpoint 1, main (argc=1, argv=0xbec05db4) at test.cpp:5
5        U8 testBuf[100]= {0};
1: /x *testBuf @ 30 = {0xd8, 0xd0, 0x2, 0x40, 0xb0, 0xda, 0xf9, 0xe, 0xd8,
0xee, 0x2, 0x40, 0xe8, 0xdd, 0x1, 0x40, 0x50, 0xe5, 0x1, 0x40, 0x6c, 0xbd,
0x1c, 0x40, 0xc0, 0xd7, 0x1, 0x40, 0x0, 0x50}
7        U16 *len = (U16*)testBuf;
1: /x *testBuf @ 30 = {0x0 }
8        U8 *ver = testBuf+2;
1: /x *testBuf @ 30 = {0x0 }
9        U8 *msgType = testBuf+3;
1: /x *testBuf @ 30 = {0x0 }
10        U8 *flag = testBuf+4;
1: /x *testBuf @ 30 = {0x0 }
11        U32 *data = (U32*)(testBuf+5);
1: /x *testBuf @ 30 = {0x0 }
13        *len = 4;
1: /x *testBuf @ 30 = {0x0 }
14        *ver = 1;
1: /x *testBuf @ 30 = {0x4, 0x0 }
15        *msgType = 2;
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x0 }
16        *flag = 3;
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2, 0x0 }
18        int dataLen(0);
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2, 0x3, 0x0 }
19        bool reVal = true;
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2, 0x3, 0x0 }
22            *data = 0xFFFFFFFF;
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2, 0x3, 0x0 }
25        return 0;
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2,[color=#FF0000] 0xff, 0xff, 0xff, 0xff, [/color]
0x0 }
26    }
1: /x *testBuf @ 30 = {0x4, 0x0, 0x1, 0x2, 0xff, 0xff, 0xff, 0xff,
0x0 }
0x401d5004 in __libc_start_main () from /lib/libc.so.6
Single stepping until exit from function __libc_start_main,
which has no line number information.

Program exited normally.

问题是 为什么U32 *data = (U32*)(testBuf+5);的数值会把U8 *flag = testBuf+4;的数值3覆盖掉 此代码在x86下运行正常

问题出在ARM 架构CPU的内存对齐问题上面

arm架构CPU为了性能上的考虑，会对目标数据长度的整数倍取整对齐(有人说四字节)，所以不应该用U32 *data指针直接存储数据，由于是非专业ARM程序员再加上项目紧迫，没有深究其原因。

在GCC编译的时候加上 -Wcast-align 参数既可以发出内存对齐问题的警告

后来编写了模板方法应对此问题。

template<typename T>
static void lfb(T* val,const void * const pBuf)
{
    memcpy((U8 *)val,(U8 *)pBuf,sizeof(T));
}

template<typename T>
static void stb(T *val,const void *pBuf)
{
    memcpy((U8 *)pBuf,(U8 *)val,sizeof(T));
}