>>>>> On Wed, 01 Mar 2006 14:44:42 +0900 (JST), Atsushi Nemoto <anemo@xxxxxxxxxxxxx> said:
anemo> Hi. I noticed that the 'jiffies' variable has 'wall_jiffies +
anemo> 1' value in most of time. I'm using MIPS platform but I think
anemo> this is same for other platforms.
anemo> I suppose this is due to gcc does not know that jiffies_64 and
anemo> jiffies share same place.
...
anemo> Is this really expected code? If no, how it can be fixed?
anemo> Insert "barrier()" right after "jiffies_64++" ?
I suppose passing updated jiffies to update_times() would be more
efficient than barrier(). Here is a patch.
Pass updated jiffies to update_times() to avoid jiffies/jiffies_64
aliasing.
Signed-off-by: Atsushi Nemoto <anemo@xxxxxxxxxxxxx>
diff --git a/kernel/timer.c b/kernel/timer.c
index fe3a9a9..7734788 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -904,11 +904,11 @@ void run_local_timers(void)
* Called by the timer interrupt. xtime_lock must already be taken
* by the timer IRQ!
*/
-static inline void update_times(void)
+static inline void update_times(unsigned long jif)
{
unsigned long ticks;
- ticks = jiffies - wall_jiffies;
+ ticks = jif - wall_jiffies;
if (ticks) {
wall_jiffies += ticks;
update_wall_time(ticks);
@@ -924,8 +924,7 @@ static inline void update_times(void)
void do_timer(struct pt_regs *regs)
{
- jiffies_64++;
- update_times();
+ update_times(++jiffies_64);
softlockup_tick(regs);
}
---
Atsushi Nemoto
