Virtualization - Aug 2007 - [PATCH 2/4] Introduce a new fields "gtime" and "cgtime" in task_struct and signal_struct

[PATCH 2/4] like for cpustat, introduce the "gtime" (guest time of the
task) and
"cgtime" (guest time of the task children) fields for the
tasks. Modify signal_struct and task_struct. Modify /proc/<pid>/stat to
display
these new fields.

Signed-off-by: Laurent Vivier <Laurent.Vivier@bull.net>
-- 
------------- Laurent.Vivier@bull.net  --------------
          "Software is hard" - Donald Knuth
-------------- next part --------------
Index: kvm/fs/proc/array.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- kvm.orig/fs/proc/array.c	2007-08-20 11:11:30.000000000 +0200
+++ kvm/fs/proc/array.c	2007-08-20 13:04:03.000000000 +0200
@@ -354,6 +354,15 @@ static clock_t task_stime(struct task_st
 	return stime;
 }
 
+#ifdef CONFIG_GUEST_ACCOUNTING
+static clock_t task_gtime(struct task_struct *p)
+{
+	clock_t gtime =3D cputime_to_clock_t(p->gtime);
+
+	return gtime;
+}
+#endif
+
 static int do_task_stat(struct task_struct *task, char *buffer, int whole)
 {
 	unsigned long vsize, eip, esp, wchan =3D ~0UL;
@@ -369,6 +378,10 @@ static int do_task_stat(struct task_stru
 	unsigned long cmin_flt =3D 0, cmaj_flt =3D 0;
 	unsigned long  min_flt =3D 0,  maj_flt =3D 0;
 	cputime_t cutime, cstime;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	cputime_t cgtime;
+	clock_t gtime;
+#endif
 	clock_t utime, stime;
 	unsigned long rsslim =3D 0;
 	char tcomm[sizeof(task->comm)];
@@ -388,6 +401,10 @@ static int do_task_stat(struct task_stru
 	sigemptyset(&sigign);
 	sigemptyset(&sigcatch);
 	cutime =3D cstime =3D cputime_zero;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	cgtime =3D cputime_zero;
+	gtime =3D 0;
+#endif
 	utime =3D stime =3D 0;
 
 	rcu_read_lock();
@@ -406,6 +423,9 @@ static int do_task_stat(struct task_stru
 		cmaj_flt =3D sig->cmaj_flt;
 		cutime =3D sig->cutime;
 		cstime =3D sig->cstime;
+#ifdef CONFIG_GUEST_ACCOUNTING
+		cgtime =3D sig->cgtime;
+#endif
 		rsslim =3D sig->rlim[RLIMIT_RSS].rlim_cur;
 
 		/* add up live thread stats at the group level */
@@ -416,6 +436,9 @@ static int do_task_stat(struct task_stru
 				maj_flt +=3D t->maj_flt;
 				utime +=3D task_utime(t);
 				stime +=3D task_stime(t);
+#ifdef CONFIG_GUEST_ACCOUNTING
+				gtime +=3D task_gtime(t);
+#endif
 				t =3D next_thread(t);
 			} while (t !=3D task);
 
@@ -423,6 +446,9 @@ static int do_task_stat(struct task_stru
 			maj_flt +=3D sig->maj_flt;
 			utime +=3D cputime_to_clock_t(sig->utime);
 			stime +=3D cputime_to_clock_t(sig->stime);
+#ifdef CONFIG_GUEST_ACCOUNTING
+			gtime +=3D cputime_to_clock_t(sig->gtime);
+#endif
 		}
 
 		sid =3D signal_session(sig);
@@ -440,6 +466,9 @@ static int do_task_stat(struct task_stru
 		maj_flt =3D task->maj_flt;
 		utime =3D task_utime(task);
 		stime =3D task_stime(task);
+#ifdef CONFIG_GUEST_ACCOUNTING
+		gtime =3D task_gtime(task);
+#endif
 	}
 
 	/* scale priority and nice values from timeslices to -20..20 */
@@ -455,9 +484,13 @@ static int do_task_stat(struct task_stru
 	/* convert nsec -> ticks */
 	start_time =3D nsec_to_clock_t(start_time);
 
-	res =3D sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu \
-%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
-%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu\n",
+	res =3D sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu "
+"%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu
"
+#ifdef CONFIG_GUEST_ACCOUNTING
+"%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
+#else
+"%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu\n",
+#endif
 		task->pid,
 		tcomm,
 		state,
@@ -502,7 +535,13 @@ static int do_task_stat(struct task_stru
 		task_cpu(task),
 		task->rt_priority,
 		task->policy,
+#ifdef CONFIG_GUEST_ACCOUNTING
+		(unsigned long long)delayacct_blkio_ticks(task),
+		gtime,
+		cputime_to_clock_t(cgtime));
+#else
 		(unsigned long long)delayacct_blkio_ticks(task));
+#endif
 	if (mm)
 		mmput(mm);
 	return res;
Index: kvm/include/linux/sched.h
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- kvm.orig/include/linux/sched.h	2007-08-20 11:11:30.000000000 +0200
+++ kvm/include/linux/sched.h	2007-08-20 13:00:02.000000000 +0200
@@ -515,6 +515,10 @@ struct signal_struct {
 	 * in __exit_signal, except for the group leader.
 	 */
 	cputime_t utime, stime, cutime, cstime;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	cputime_t gtime;
+	cputime_t cgtime;
+#endif
 	unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
 	unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
 	unsigned long inblock, oublock, cinblock, coublock;
@@ -1019,6 +1023,9 @@ struct task_struct {
 
 	unsigned int rt_priority;
 	cputime_t utime, stime;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	cputime_t gtime;
+#endif
 	unsigned long nvcsw, nivcsw; /* context switch counts */
 	struct timespec start_time; 		/* monotonic time */
 	struct timespec real_start_time;	/* boot based time */
Index: kvm/kernel/exit.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- kvm.orig/kernel/exit.c	2007-08-20 11:11:30.000000000 +0200
+++ kvm/kernel/exit.c	2007-08-20 12:32:08.000000000 +0200
@@ -120,6 +120,9 @@ static void __exit_signal(struct task_st
 		 */
 		sig->utime =3D cputime_add(sig->utime, tsk->utime);
 		sig->stime =3D cputime_add(sig->stime, tsk->stime);
+#ifdef CONFIG_GUEST_ACCOUNTING
+		sig->gtime =3D cputime_add(sig->gtime, tsk->gtime);
+#endif
 		sig->min_flt +=3D tsk->min_flt;
 		sig->maj_flt +=3D tsk->maj_flt;
 		sig->nvcsw +=3D tsk->nvcsw;
@@ -1255,6 +1258,13 @@ static int wait_task_zombie(struct task_
 			cputime_add(p->stime,
 			cputime_add(sig->stime,
 				    sig->cstime)));
+#ifdef CONFIG_GUEST_ACCOUNTING
+		psig->cgtime =3D
+			cputime_add(psig->cgtime,
+			cputime_add(p->gtime,
+			cputime_add(sig->gtime,
+				    sig->cgtime)));
+#endif
 		psig->cmin_flt +=3D
 			p->min_flt + sig->min_flt + sig->cmin_flt;
 		psig->cmaj_flt +=3D
Index: kvm/kernel/fork.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- kvm.orig/kernel/fork.c	2007-08-20 11:11:30.000000000 +0200
+++ kvm/kernel/fork.c	2007-08-20 12:38:55.000000000 +0200
@@ -877,6 +877,10 @@ static inline int copy_signal(unsigned l
 	sig->tty_old_pgrp =3D NULL;
 
 	sig->utime =3D sig->stime =3D sig->cutime =3D sig->cstime =3D
cputime_zero;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	sig->gtime =3D cputime_zero;
+	sig->cgtime =3D cputime_zero;
+#endif
 	sig->nvcsw =3D sig->nivcsw =3D sig->cnvcsw =3D sig->cnivcsw =3D 0;
 	sig->min_flt =3D sig->maj_flt =3D sig->cmin_flt =3D sig->cmaj_flt
=3D 0;
 	sig->inblock =3D sig->oublock =3D sig->cinblock =3D sig->coublock
=3D 0;
@@ -1045,6 +1049,9 @@ static struct task_struct *copy_process(
 
 	p->utime =3D cputime_zero;
 	p->stime =3D cputime_zero;
+#ifdef CONFIG_GUEST_ACCOUNTING
+	p->gtime =3D cputime_zero;
+#endif
 
 #ifdef CONFIG_TASK_XACCT
 	p->rchar =3D 0;		/* I/O counter: bytes read */

Am Montag, 20. August 2007 schrieb Laurent Vivier:
> Index: kvm/fs/proc/array.c
> ==================================================================> ---
kvm.orig/fs/proc/array.c????2007-08-20 11:11:30.000000000 +0200
> +++ kvm/fs/proc/array.c?2007-08-20 13:04:03.000000000 +0200
Just a heads up, this patch collides with this fix in mm:
http://marc.info/?l=linux-mm-commits&m=118737949222362&w=2

If Ingo accepts this fix, your patch should be adopted in array.c to use 
cputime_t for gtime as well. Lets see what Ingo thinks.


Christian

Christian Borntraeger wrote:
> Am Montag, 20. August 2007 schrieb Laurent Vivier:
>> Index: kvm/fs/proc/array.c
>>
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>> --- kvm.orig/fs/proc/array.c    2007-08-20 11:11:30.000000000 +0200
>> +++ kvm/fs/proc/array.c 2007-08-20 13:04:03.000000000 +0200
> 
> Just a heads up, this patch collides with this fix in mm:
> http://marc.info/?l=3Dlinux-mm-commits&m=3D118737949222362&w=3D2
> 
> If Ingo accepts this fix, your patch should be adopted in array.c to use 
> cputime_t for gtime as well. Lets see what Ingo thinks.
Thank you for the comment. No problem for me to rewrite the patch according the
fix in the mm.

Laurent
-- 
------------- Laurent.Vivier@bull.net  --------------
          "Software is hard" - Donald Knuth

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: OpenPGP digital signature
Url :
http://lists.linux-foundation.org/pipermail/virtualization/attachments/20070820/8ebc5e1f/signature.pgp

* Laurent Vivier <Laurent.Vivier at bull.net> wrote:
> [PATCH 2/4] like for cpustat, introduce the "gtime" (guest time
of
> the task) and "cgtime" (guest time of the task children) fields 
> for the tasks. Modify signal_struct and task_struct. Modify 
> /proc/<pid>/stat to display these new fields.
> --- kvm.orig/include/linux/sched.h	2007-08-20 11:11:30.000000000 +0200
> +++ kvm/include/linux/sched.h	2007-08-20 13:00:02.000000000 +0200
> @@ -515,6 +515,10 @@ struct signal_struct {
>  	 * in __exit_signal, except for the group leader.
>  	 */
>  	cputime_t utime, stime, cutime, cstime;
> +#ifdef CONFIG_GUEST_ACCOUNTING
> +	cputime_t gtime;
> +	cputime_t cgtime;
> +#endif
A handful of general (and less general) observations about these 
patches:

 1- The code is very ugly due to being an #ifdef fest. Please
    always try to avoid them.

 2- cputime_t is very coarse on x86: measured in jiffies. This means
    that with a default HZ of 250 we'll have units of 4 msecs. 
    That's almost useless to rely on in new instrumentation: an irq 
    can come in and out without accounting noticing it, etc. If we 
    do some new statistics then it should be a lot better than 
    jiffies granular.

 3- stime of vcpu tasks/threads already approximates 'guest time' 
    adequately. (as Jeremy observed it as well) Yes, it mixes 'true 
    guest mode' and 'host mode' system time, but then again due to 
    the jiffies granularity we have a _far_ bigger skew going on 
    already.

 4- namespace collision: 'gtime' is already used as 'group time'
in
    a few places. One of the two things needs to be renamed.

 5- tracepoints and perfcounters could be used to measure guest time 
    precisely, in a low-overhead mode.

These issues need to be addressed in a meaningful way. #2 probably 
means a revamping of cputime_t handling on x86 - of not just the 
gtime. But #3 is worth keeping in mind as well.

I think #5 is the most capable solution by a wide margin - we need 
just a single tracepoint to emit 'nsecs spent in guest mode' 
information and that's it. It would be a far smaller patch.

The tracepoint might even sample the guest RIP and hence could be 
used as a VM-exit profiler and 'perf record -e kvm:vm_exit + perf 
report' could be used to examine/profile/trace guest exit reasons.

	Ingo