2003-06-03 22:50:25

by john stultz

[permalink] [raw]
Subject: [RFC][PATCH] linux-2.5.70_btime-fix_A0

All,

Since jiffies didn't necessarily start incrementing at a second
boundary, jiffies/HZ doesn't increment at the same moment as
xtime.tv_sec. This causes one second wobbles in the calculation of btime
(xtime.tv_sec - jiffies/HZ).

This fix increases the precision of the calculation so the usec
component of xtime is used as well. Additionally it fixes some of the
non-atomic reading of time values.


This is a fix for bugme bug #764.
http://bugme.osdl.org/show_bug.cgi?id=764


Let me know if you have any comments

thanks
-john

--- 1.77/fs/proc/proc_misc.c Sun May 25 14:08:09 2003
+++ edited/fs/proc/proc_misc.c Tue Jun 3 15:52:41 2003
@@ -378,8 +378,22 @@
{
int i, len;
extern unsigned long total_forks;
- u64 jif = get_jiffies_64() - INITIAL_JIFFIES;
+ u64 jif;
unsigned int sum = 0, user = 0, nice = 0, system = 0, idle = 0, iowait = 0;
+ struct timeval now;
+ unsigned long seq;
+
+ /* Atomically read jiffies and time of day */
+ do {
+ seq = read_seqbegin(&xtime_lock);
+
+ jif = get_jiffies_64() - INITIAL_JIFFIES;
+ do_gettimeofday(&now);
+ } while (read_seqretry(&xtime_lock, seq));
+
+ /* calc # of seconds since boot time */
+ jif = ((u64)now.tv_sec * HZ) + (now.tv_usec/(1000000/HZ)) - jif;
+ do_div(jif, HZ);

for (i = 0 ; i < NR_CPUS; i++) {
int j;
@@ -419,7 +433,6 @@
len += sprintf(page + len, " %u", kstat_irqs(i));
#endif

- do_div(jif, HZ);
len += sprintf(page + len,
"\nctxt %lu\n"
"btime %lu\n"
@@ -427,7 +440,7 @@
"procs_running %lu\n"
"procs_blocked %lu\n",
nr_context_switches(),
- xtime.tv_sec - (unsigned long) jif,
+ (unsigned long)jif,
total_forks,
nr_running(),
nr_iowait());




2003-06-04 02:14:41

by jw schultz

[permalink] [raw]
Subject: Re: [RFC][PATCH] linux-2.5.70_btime-fix_A0

On Tue, Jun 03, 2003 at 04:00:59PM -0700, john stultz wrote:
> All,
>
> Since jiffies didn't necessarily start incrementing at a second
> boundary, jiffies/HZ doesn't increment at the same moment as
> xtime.tv_sec. This causes one second wobbles in the calculation of btime
> (xtime.tv_sec - jiffies/HZ).
>
> This fix increases the precision of the calculation so the usec
> component of xtime is used as well. Additionally it fixes some of the
> non-atomic reading of time values.
>
>
> This is a fix for bugme bug #764.
> http://bugme.osdl.org/show_bug.cgi?id=764
>
>
> Let me know if you have any comments

Might it not be cheaper to start jiffies at the 1 second
boundary or with a value that simulates that?

--
________________________________________________________________
J.W. Schultz Pegasystems Technologies
email address: [email protected]

Remember Cernan and Schmitt

2003-06-04 18:11:22

by john stultz

[permalink] [raw]
Subject: Re: [RFC][PATCH] linux-2.5.70_btime-fix_A0

jw schultz wrote:
> On Tue, Jun 03, 2003 at 04:00:59PM -0700, john stultz wrote:
>> Since jiffies didn't necessarily start incrementing at a second
>> boundary, jiffies/HZ doesn't increment at the same moment as
>> xtime.tv_sec. This causes one second wobbles in the calculation of btime
>> (xtime.tv_sec - jiffies/HZ).
>
> Might it not be cheaper to start jiffies at the 1 second
> boundary or with a value that simulates that?

Ehhh... Mucking with the meaning of jiffies (# of timer ticks) just to fix a
math error in /proc/stat seems like a bit much.

But I'll look to see if I can improve the math some.

thanks
-john