Skip to contentSkip to conversation

2.6.1: process start times by procps

DISCUSSION George Anzinger 1 participant 1 message Jan 28, 2004 1 message not in the archive Proc filesystem

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

George Anzinger

author
#1 lore raw
Quoted 1 line · Andrew Morton

Does this look right to you??

At first blush, no. If it is supposed to be the time of boot then it should be the same as -wall_to_monotonic.... see below.

Quoted 51 lines

Thanks.

Begin forwarded message:

Date: Tue, 27 Jan 2004 17:52:54 +0200
From: Petri Kaukasoina <[email protected]>
To: [email protected]
Subject: Re: 2.6.1: process start times by procps

Quoted 9 lines · On Sun, Jan 25, 2004 at 01:08:47PM +0200, I wrote:
Quoted 4 lines · On Fri, Jan 23, 2004 at 09:47:14PM +0200, I wrote:

For example, I started this bash process really at 21:24 (date showed 21:24 then):

kaukasoi 22108 0.0 0.2 4452 1532 pts/4 R 21:28 0:00 /bin/bash

OK, I would like to make my bug report more accurate: the problem seems to be that the value of btime in /proc/stat is not correct.

btime in /proc/stat does not stay constant but decreases at a rate of 15 secs/day. (So I thought that that's why there is that four minute error in ps output after uptime of a couple of weeks.) Maybe this has something to do with the fact that the 'timer' line in /proc/interrupts does not seem to increase at an exact rate of 1000 steps per second but about 1000.18 steps per second, instead. (The relative error is the same: 0.18 divided by 1000 is equal to 15 seconds divided by 24 hours).

I made an experiment shown below. I know nothing about kernel programming, so this is probably not correct, but at least btime seemed to stay constant. (I don't believe this fixes procps, though. If HZ if off by 180 ppm then I guess ps can't possibly get its calculations involving HZ right. But at least the bootup time reported by procinfo stays constant.)

fs/proc/proc_misc.c renamed from fs/proc/proc_misc.c.orig +0 -6
@@ -363,19 +363,13 @@
 	u64 jif;
 	unsigned int sum = 0, user = 0, nice = 0, system = 0, idle = 0, iowait = 0, irq = 0, softirq = 0;
 	struct timeval now; 
-	unsigned long seq;
-
-	/* Atomically read jiffies and time of day */ 
-	do {
-		seq = read_seqbegin(&xtime_lock);
-

fs/proc/proc_misc.c line 378

Quoted 10 lines
fs/proc/proc_misc.c renamed from fs/proc/proc_misc.c.orig +3 -5
-		jif = get_jiffies_64();
-		do_gettimeofday(&now);
-	} while (read_seqretry(&xtime_lock, seq));
+	struct timespec uptime;
 
+	do_gettimeofday(&now);
+	do_posix_clock_monotonic_gettime(&uptime);
 	/* calc # of seconds since boot time */
-	jif -= INITIAL_JIFFIES;
-	jif = ((u64)now.tv_sec * HZ) + (now.tv_usec/(1000000/HZ)) - jif;

This conversion is the problem. The time.h function timespec_to_jiffies is the correct way to do this conversion. Still, wall_to_monotonic is defined such that:
posix_clock_monotonic = wall_clock - wall_to_monotonic

This means that - wall_to_monotonic is the value you want and that is, for the
most part a constant, and all that is needed is the second part..

Try the attached.

Quoted 12 lines
fs/proc/proc_misc.c renamed from fs/proc/proc_misc.c.orig +2 -1
+	jif = ((u64)now.tv_sec * HZ) + (now.tv_usec/(1000000/HZ)) \
+            - ((u64)uptime.tv_sec * HZ) - (uptime.tv_nsec/(NSEC_PER_SEC/HZ));
 	do_div(jif, HZ);
 
 	for (i = 0; i < NR_CPUS; i++) {
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 3 lines
George Anzinger   [email protected]
High-res-timers:  http://sourceforge.net/projects/high-res-timers/
Preemption patch: http://www.kernel.org/pub/linux/kernel/people/rml
fs/proc/proc_misc.c +4 -15
@@ -360,23 +360,12 @@
 {
 	int i;
 	extern unsigned long total_forks;
-	u64 jif;
+	unsigned long jif;
 	unsigned int sum = 0, user = 0, nice = 0, system = 0, idle = 0, iowait = 0, irq = 0, softirq = 0;
-	struct timeval now; 
-	unsigned long seq;
 
-	/* Atomically read jiffies and time of day */ 
-	do {
-		seq = read_seqbegin(&xtime_lock);
-
-		jif = get_jiffies_64();
-		do_gettimeofday(&now);
-	} while (read_seqretry(&xtime_lock, seq));
-
-	/* calc # of seconds since boot time */
-	jif -= INITIAL_JIFFIES;
-	jif = ((u64)now.tv_sec * HZ) + (now.tv_usec/(1000000/HZ)) - jif;
-	do_div(jif, HZ);
+	jif = - wall_to_monotonic.tv_sec;
+	if (wall_to_monotonic.tv_nsec)
+		--jif;
 
 	for (i = 0; i < NR_CPUS; i++) {
 		int j;