/*
***************************************************************************
* Workaround for CPU counters read from /proc/stat: Dyn-tick kernels
* have a race issue that can make those counters go backward.
***************************************************************************
*/
double ll_sp_value(unsigned long long value1, unsigned long long value2,
unsigned long long itv)
{
if (value2 < value1)
return (double) 0;
else
return SP_VALUE(value1, value2, itv);
}
/*
***************************************************************************
* Handle overflow conditions properly for counters which are read as
* unsigned long long, but which can be unsigned long long or
* unsigned long only depending on the kernel version used.
* @value1 and @value2 being two values successively read for this
* counter, if @value2 < @value1 and @value1 <= 0xffffffff, then we can
* assume that the counter's type was unsigned long and has overflown, and
* so the difference @value2 - @value1 must be casted to this type.
* NOTE: These functions should no longer be necessary to handle a particular
* stat counter when we can assume that everybody is using a recent kernel
* (defining this counter as unsigned long long).
***************************************************************************
*/
double ll_sp_value(unsigned long long value1, unsigned long long value2,
unsigned long long itv)
{
if ((value2 < value1) && (value1 <= 0xffffffff))
/* Counter's type was unsigned long and has overflown */
return ((double) ((value2 - value1) & 0xffffffff)) / itv * 100;
else
return SP_VALUE(value1, value2, itv);
}
