UDP的校验和

 

UDP的校验和包括UDP首部和数据，方法是计算16位的二进制和，首先将校验  
  和字段(16位)置0，将每个16位的反码相加，若是奇数字节即最后未到16字节，  
  则以0补充，结果存在校验和字段中，因接收端包含了之前的校验和，若传输  
  中没有任何差错，则接收端计算的校验和该为全1，否则有错。

-----------------------------------------------------------------

这是lwip协议栈的校验程序

static u16_t
lwip_standard_chksum(void *dataptr, u16_t len)
{
  u32_t acc;
  u16_t src;
  u8_t *octetptr;

  acc = 0;
  /* dataptr may be at odd or even addresses */
  octetptr = (u8_t*)dataptr;
  while (len > 1)
  {
    /* declare first octet as most significant
       thus assume network order, ignoring host order */
    src = (*octetptr) << 8;
    octetptr++;
    /* declare second octet as least significant */
    src |= (*octetptr);
    octetptr++;
    acc += src;
    len -= 2;
  }
  if (len > 0)
  {
    /* accumulate remaining octet */
    src = (*octetptr) << 8;
    acc += src;
  }
  /* add deferred carry bits */
  acc = (acc >> 16) + (acc & 0x0000ffffUL);
  if ((acc & 0xffff0000) != 0) {
    acc = (acc >> 16) + (acc & 0x0000ffffUL);
  }
  /* This maybe a little confusing: reorder sum using htons()
     instead of ntohs() since it has a little less call overhead.
     The caller must invert bits for Internet sum ! */
  return htons((u16_t)acc);
}