linux中的jiffies变量

全局变量jiffies用来记录自系统启动以来产生的节拍的总数。启动时，内核将该变量初始化为0，此后，每次时钟中断处理程序都会增加该变量的值。一秒内时钟中断的次数等于Hz，所以jiffies一秒内增加的值也就是Hz。


   系统运行时间以秒为单位，等于jiffies/Hz。


注意，jiffies类型为无符号长整型(unsigned long)，其他任何类型存放它都不正确。


将以秒为单位的时间转化为jiffies：


seconds * Hz


将jiffies转化为以秒为单位的时间：


jiffies / Hz


相比之下，内核中将秒转换为jiffies用的多些。


jiffies的内部表示
   jiffies定义于文件<linux\Jiffies.h>中：


/*
* The 64-bit value is not atomic - you MUST NOT read it
* without sampling the sequence number in xtime_lock.
* get_jiffies_64() will do this for you as appropriate.
*/
extern u64 __jiffy_data jiffies_64;
extern unsigned long volatile __jiffy_data jiffies;
ld(1)脚本用于连接主内核映像（在x86上位于arch/i386/kernel/vmlinux.lds.S中），然后用jiffies_64变量的初值覆盖jiffies变量。因此jiffies取整个jiffies_64变量的低32位。


  访问jiffies的代码只会读取jiffies_64的低32位，通过get_jiffies_64()函数就可以读取整个64位的值。在64位体系结构上，jiffies_64和jiffies指的是同一个变量。


#if (BITS_PER_LONG < 64)
u64 get_jiffies_64(void);
#else
static inline u64 get_jiffies_64(void)
{
return (u64)jiffies;
}
#endif
在<Time.c(kernel)>中
#if (BITS_PER_LONG < 64)
u64 get_jiffies_64(void)
{
    unsigned long seq;
    u64 ret;
do {
        seq = read_seqbegin(&xtime_lock);
        ret = jiffies_64;
    } while (read_seqretry(&xtime_lock, seq));
return ret;
}
jiffies的回绕wrap around
  当jiffies的值超过它的最大存放范围后就会发生溢出。对于32位无符号长整型，最大取值为(2^32)-1,即429496795。如果节拍计数达到了最大值后还要继续增加，它的值就会回绕到0。


  内核提供了四个宏来帮助比较节拍计数，它们能正确的处理节拍计数回绕的问题：


/*
*  These inlines deal with timer wrapping correctly. You are
*  strongly encouraged to use them
*  1. Because people otherwise forget
*  2. Because if the timer wrap changes in future you won't have to
*     alter your driver code.
*
* time_after(a,b) returns true if the time a is after time b.
*
* Do this with "<0" and ">=0" to only test the sign of the result. A
* good compiler would generate better code (and a really good compiler
* wouldn't care). Gcc is currently neither.
*/
#define time_after(a,b)     \
    (typecheck(unsigned long, a) && \
     typecheck(unsigned long, b) && \
     ((long)(b) - (long)(a) < 0))
#define time_before(a,b)    time_after(b,a)
#define time_after_eq(a,b)  \
    (typecheck(unsigned long, a) && \
     typecheck(unsigned long, b) && \
     ((long)(a) - (long)(b) >= 0))
#define time_before_eq(a,b) time_after_eq(b,a)
/* Same as above, but does so with platform independent 64bit types.
* These must be used when utilizing jiffies_64 (i.e. return value of
* get_jiffies_64() */
#define time_after64(a,b)   \
    (typecheck(__u64, a) && \
     typecheck(__u64, b) && \
     ((__s64)(b) - (__s64)(a) < 0))
#define time_before64(a,b)  time_after64(b,a)
#define time_after_eq64(a,b)    \
    (typecheck(__u64, a) && \
     typecheck(__u64, b) && \
     ((__s64)(a) - (__s64)(b) >= 0))
#define time_before_eq64(a,b)   time_after_eq64(b,a)
用户空间和HZ
  问题提出：


  在2.6以前的内核中，如果改变内核中的HZ值会给用户空间中某些程序造成异常结果。因为内核是以节拍数/秒的形式给用户空间导出这个值的，应用程序便依赖这个特定的HZ值。如果在内核中改变了HZ的定义值，就打破了用户空间的常量关系---用户空间并不知道新的HZ值。


  解决方法：


  内核更改所有导出的jiffies值。内核定义了USER_HZ来代表用户空间看到的HZ值。在x86体系结构上，由于HZ值原来一直是100，所以USER_HZ值就定义为100。内核可以使用宏jiffies_to_clock_t()将一个有HZ表示的节拍计数转换为一个由USER_HZ表示的节拍计数。


在<Time.c(kernel)>中
/*
* Convert jiffies/jiffies_64 to clock_t and back.
*/
clock_t jiffies_to_clock_t(long x)
{
#if (TICK_NSEC % (NSEC_PER_SEC / USER_HZ)) == 0
return x / (HZ / USER_HZ);
#else
    u64 tmp = (u64)x * TICK_NSEC;
    do_div(tmp, (NSEC_PER_SEC / USER_HZ));
return (long)tmp;
#endif
}
unsigned long clock_t_to_jiffies(unsigned long x)
{
#if (HZ % USER_HZ)==0
if (x >= ~0UL / (HZ / USER_HZ))
return ~0UL;
return x * (HZ / USER_HZ);
#else
    u64 jif;
/* Don't worry about loss of precision here .. */
if (x >= ~0UL / HZ * USER_HZ)
return ~0UL;
/* .. but do try to contain it here */
    jif = x * (u64) HZ;
    do_div(jif, USER_HZ);
return jif;
#endif
}
u64 jiffies_64_to_clock_t(u64 x)
{
#if (TICK_NSEC % (NSEC_PER_SEC / USER_HZ)) == 0
    do_div(x, HZ / USER_HZ);
#else
/*
     * There are better ways that don't overflow early,
     * but even this doesn't overflow in hundreds of years
     * in 64 bits, so..
     */
    x *= TICK_NSEC;
    do_div(x, (NSEC_PER_SEC / USER_HZ));
#endif
return x;
}
在<Div64.h(include\asm-i385)>中
/*
* do_div() is NOT a C function. It wants to return
* two values (the quotient and the remainder), but
* since that doesn't work very well in C, what it
* does is:
*
* - modifies the 64-bit dividend _in_place_
* - returns the 32-bit remainder
*
* This ends up being the most efficient "calling
* convention" on x86.
*/
#define do_div(n,base) ({ \
    unsigned long __upper, __low, __high, __mod, __base; \
    __base = (base); \
    asm("":"=a" (__low), "=d" (__high):"A" (n)); \
    __upper = __high; \
if (__high) { \
        __upper = __high % (__base); \
        __high = __high / (__base); \
    } \
    asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (__base), "0" (__low), "1" (__upper)); \
    asm("":"=A" (n):"a" (__low),"d" (__high)); \
    __mod; \
})
  用户空间期望HZ=USER_HZ，但是如果它们不相等，则由宏完成转换。