cpuset: Replace instances of time_t with time64_t to solve y2038 problem

The following patchset replaces instances of time_t with time64_t to address the y2038 problem. The patchset also makes the code independent of settimeofday().

Changes in v2:
- Changed the function get_seconds() that return a 32-bit integer to ktime_get_real_ts64() that returns a struct of type timespec64 and use the tv_sec part which is 64-bit.

- Change the type of ticks from time_t to u32. We keep ticks as 32-bits as the function uses 32-bit arithmetic which would prove less expensive than 64-bit arithmetic and the function is expected to be called atleast once every 32 seconds.

Changes in v3:
- Changed commenting style.

Changes in v4:
- Changed the subject tag.

Changes in v5:
- Changed a comment and added a patch that make the code independent of settimeofday().

Heena Sirwani (2):
cpuset: Replace all instances of time_t with time64_t
cpuset: Replace ktime_get_real_ts64() with ktime_get_ts64()

 kernel/cpuset.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

1.9.1

The following patch replaces all instances of time_t with time64_t i.e. change the type used for representing time from 32-bit to 64-bit. All 32-bit kernels to date use a signed 32-bit time_t type, which can only represent time until January 2038. Since embedded systems running 32-bit Linux are going to survive beyond that date, we have to change all current uses, in a backwards compatible way.

The patch also changes the function get_seconds() that returns a 32-bit integer to ktime_get_real_ts64() that returns a struct of type timespec64 and use the tv_sec part which is 64-bit.

The patch changes the type of ticks from time_t to u32. We keep ticks as 32-bits as the function uses 32-bit arithmetic which would prove less expensive than 64-bit arithmetic and the function is expected to be called atleast once every 32 seconds.

---
 kernel/cpuset.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

index 52cb04c..c3e3aac 100644
@@ -51,6 +51,7 @@
 #include <linux/stat.h>
 #include <linux/string.h>
 #include <linux/time.h>
+#include <linux/time64.h>
 #include <linux/backing-dev.h>
 #include <linux/sort.h>
 

@@ -68,7 +69,7 @@ struct static_key cpusets_enabled_key __read_mostly = STATIC_KEY_INIT_FALSE;
 struct fmeter {
 	int cnt;		/* unprocessed events count */
 	int val;		/* most recent output value */
-	time_t time;		/* clock (secs) when val computed */
+	time64_t time;		/* clock (secs) when val computed */
 	spinlock_t lock;	/* guards read or write of above */
 };
 

@@ -1355,7 +1356,7 @@ out:
  */
 
 #define FM_COEF 933		/* coefficient for half-life of 10 secs */
-#define FM_MAXTICKS ((time_t)99) /* useless computing more ticks than this */
+#define FM_MAXTICKS ((u32)99)	/* useless computing more ticks than this */
 #define FM_MAXCNT 1000000	/* limit cnt to avoid overflow */
 #define FM_SCALE 1000		/* faux fixed point scale */
 

@@ -1371,8 +1372,14 @@ static void fmeter_init(struct fmeter *fmp)
 /* Internal meter update - process cnt events and update value */
 static void fmeter_update(struct fmeter *fmp)
 {
-	time_t now = get_seconds();
-	time_t ticks = now - fmp->time;
+	struct timespec64 timespec;
+	time64_t now;
+	u32 ticks;
+
+	/* One should probably use get_seconds64().*/
+	ktime_get_real_ts64(&timespec);		
+	now = timespec.tv_sec;
+	ticks = now - fmp->time;
 
 	if (ticks == 0)
 		return;

1.9.1

The following patch makes the kernel code independent of the setimeofday() function by replacing the function ktime_get_real_ts64() with ktime_get_ts64().

The function calculates the interval between two successive calls to fmeter_update() and between these if a call to settimeofday() is called it would lead to erroneous calculation of the interval. The function ktime_get_ts64() is the monotonic wall-time clock that does an extra calculation to remove the effects of settimeofday().

---
 kernel/cpuset.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index c3e3aac..98cf976 100644
@@ -1377,7 +1377,7 @@ static void fmeter_update(struct fmeter *fmp)
 	u32 ticks;
 
 	/* One should probably use get_seconds64().*/
-	ktime_get_real_ts64(&timespec);		
+	ktime_get_ts64(&timespec);		
 	now = timespec.tv_sec;
 	ticks = now - fmp->time;
 

1.9.1