new time code problem

[George Anzinger <george@mvista.com>]

[-- Attachment #1: Type: text/plain, Size: 708 bytes --]

The 64-bit conversion routine to convert 64-bit nsec time to a time spec. 
gives an unnormalized result if the value being converted is negative.  I 
think there are two ways to go about fixing this.  Most systems will give a 
negative remainder and so need to just normalize.  On the other hand, some 
systems will use div64 to do the division and, I think, it expects unsigned 
numbers.  The attached patch uses the conservative approach of expecting the 
div to be set up for unsigned numbers.

I came accross this when one of my tests set a time near 1 Jan 1970, i.e. it 
is a real problem.
-- 
George Anzinger   george@mvista.com
HRT (High-res-timers):  http://sourceforge.net/projects/high-res-timers/

[-- Attachment #2: ktime_conversion.patch --]
[-- Type: text/plain, Size: 910 bytes --]

 kernel/time.c |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)

Index: linux-2.6.16-rc/kernel/time.c
===================================================================
--- linux-2.6.16-rc.orig/kernel/time.c
+++ linux-2.6.16-rc/kernel/time.c
@@ -702,16 +702,19 @@ void set_normalized_timespec(struct time
  *
  * Returns the timespec representation of the nsec parameter.
  */
-inline struct timespec ns_to_timespec(const nsec_t nsec)
+struct timespec ns_to_timespec(const nsec_t nsec)
 {
 	struct timespec ts;
 
-	if (nsec)
+	if (nsec) return (struct timespec){0, 0};
+
+	if (nsec < 0) {
+		ts.tv_sec = div_long_long_rem_signed(-nsec, NSEC_PER_SEC,
+						     &ts.tv_nsec);
+		set_normalized_timespec(&ts, -ts.tv_sec, -ts.tv_nsec);
+	} else
 		ts.tv_sec = div_long_long_rem_signed(nsec, NSEC_PER_SEC,
 						     &ts.tv_nsec);
-	else
-		ts.tv_sec = ts.tv_nsec = 0;
-
 	return ts;
 }