clarify a usage constraint for cnt32_to_63()

		unsigned long long ll;

		unsigned l[2];

	} tsc64, result;

	unsigned long tsc, tmp;

	unsigned long tmp;

	unsigned product[3]; /* 96-bit intermediate value */

	/* cnt32_to_63() is not safe with preemption */

	preempt_disable();

	/* read the TSC value

	 */

	tsc = get_cycles();

	/* expand to 64-bits.

	/* expand the tsc to 64-bits.

	 * - sched_clock() must be called once a minute or better or the

	 *   following will go horribly wrong - see cnt32_to_63()

	 */

	tsc64.ll = cnt32_to_63(tsc) & 0x7fffffffffffffffULL;

	tsc64.ll = cnt32_to_63(get_cycles()) & 0x7fffffffffffffffULL;

	preempt_enable();

 *

 * 2) this code must not be preempted for a duration longer than the

 *    32-bit counter half period minus the longest period between two

 *    calls to this code.

 *    calls to this code;

 *

 * Those requirements ensure proper update to the state bit in memory.

 * This is usually not a problem in practice, but if it is then a kernel

 * timer should be scheduled to manage for this code to be executed often

 * enough.

 *

 * And finally:

 *

 * 3) the cnt_lo argument must be seen as a globally incrementing value,

 *    meaning that it should be a direct reference to the counter data which

 *    can be evaluated according to a specific ordering within the macro,

 *    and not the result of a previous evaluation stored in a variable.

 *

 * For example, this is wrong:

 *

 *	u32 partial = get_hw_count();

 *	u64 full = cnt32_to_63(partial);

 *	return full;

 *

 * This is fine:

 *

 *	u64 full = cnt32_to_63(get_hw_count());

 *	return full;

 *

 * Note that the top bit (bit 63) in the returned value should be considered

 * as garbage.  It is not cleared here because callers are likely to use a

 * multiplier on the returned value which can get rid of the top bit