[PATCH] xen: sched: adjustments to some performance counters

[PATCH] xen: sched: adjustments to some performance counters

[Dario Faggioli]

More specifically:

1) rename vcpu_destroy to vcpu_remove

It seems this have had to be done as part of 7e6b926a
("cpupools: Make interface more consistent"), which
renamed the function but not the counter.

In fact, because of cpupools, vcpus are not only removed
from a scheduler when they are destroyed, but also when
domains move between pools.

Make the related statistics counter reflect that more
properly.

2) rename vcpu_init to vcpu_alloc

As it lives in *_alloc_vdata.

3) add vcpu_insert

matching vcpu_remove, and useful to quickly check
whether the number of insertions and removal matches,
or in general investigare their relationship.

Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
---
Cc: George Dunlap <george.dunlap@eu.citrix.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Meng Xu <mengxu@cis.upenn.edu>
Cc: Jan Beulich <jbeulich@suse.com>
---
This would be v2 of "xen: sched: rename vcpu_destroy perf counter to
vcpu_remove", but both the title and the content changed, so I'm treating it as
a fresh submission (and not applying the Rev-by Juergen gave to it).

The original submission was only changing the name of vcpu_destroy to
vcpu_remove. This one, also changes vcpu_init to vcpu_alloc (as suggested by
Juergen) and adds vcpu_insert, for simmetry with vcpu_remove.
---
 xen/common/sched_credit.c    |    6 ++++--
 xen/common/sched_credit2.c   |    6 ++++--
 xen/common/sched_rt.c        |    6 ++++--
 xen/include/xen/perfc_defn.h |    5 +++--
 4 files changed, 15 insertions(+), 8 deletions(-)

diff --git a/xen/common/sched_credit.c b/xen/common/sched_credit.c
index a1945ac..3bb28c0 100644
--- a/xen/common/sched_credit.c
+++ b/xen/common/sched_credit.c
@@ -896,7 +896,7 @@ csched_alloc_vdata(const struct scheduler *ops, struct vcpu *vc, void *dd)
     svc->pri = is_idle_domain(vc->domain) ?
         CSCHED_PRI_IDLE : CSCHED_PRI_TS_UNDER;
     SCHED_VCPU_STATS_RESET(svc);
-    SCHED_STAT_CRANK(vcpu_init);
+    SCHED_STAT_CRANK(vcpu_alloc);
     return svc;
 }
 
@@ -907,6 +907,8 @@ csched_vcpu_insert(const struct scheduler *ops, struct vcpu *vc)
 
     if ( !__vcpu_on_runq(svc) && vcpu_runnable(vc) && !vc->is_running )
         __runq_insert(vc->processor, svc);
+
+    SCHED_STAT_CRANK(vcpu_insert);
 }
 
 static void
@@ -927,7 +929,7 @@ csched_vcpu_remove(const struct scheduler *ops, struct vcpu *vc)
     struct csched_dom * const sdom = svc->sdom;
     unsigned long flags;
 
-    SCHED_STAT_CRANK(vcpu_destroy);
+    SCHED_STAT_CRANK(vcpu_remove);
 
     if ( test_and_clear_bit(CSCHED_FLAG_VCPU_PARKED, &svc->flags) )
     {
diff --git a/xen/common/sched_credit2.c b/xen/common/sched_credit2.c
index 75e0321..bf1fe6f 100644
--- a/xen/common/sched_credit2.c
+++ b/xen/common/sched_credit2.c
@@ -796,7 +796,7 @@ csched2_alloc_vdata(const struct scheduler *ops, struct vcpu *vc, void *dd)
         svc->weight = 0;
     }
 
-    SCHED_STAT_CRANK(vcpu_init);
+    SCHED_STAT_CRANK(vcpu_alloc);
 
     return svc;
 }
@@ -891,6 +891,8 @@ csched2_vcpu_insert(const struct scheduler *ops, struct vcpu *vc)
         vcpu_schedule_unlock_irq(lock, vc);
 
         sdom->nr_vcpus++;
+
+        SCHED_STAT_CRANK(vcpu_insert);
     }
 
     CSCHED2_VCPU_CHECK(vc);
@@ -917,7 +919,7 @@ csched2_vcpu_remove(const struct scheduler *ops, struct vcpu *vc)
     {
         spinlock_t *lock;
 
-        SCHED_STAT_CRANK(vcpu_destroy);
+        SCHED_STAT_CRANK(vcpu_remove);
 
         /* Remove from runqueue */
         lock = vcpu_schedule_lock_irq(vc);
diff --git a/xen/common/sched_rt.c b/xen/common/sched_rt.c
index 4372486..6fe3bec 100644
--- a/xen/common/sched_rt.c
+++ b/xen/common/sched_rt.c
@@ -597,7 +597,7 @@ rt_alloc_vdata(const struct scheduler *ops, struct vcpu *vc, void *dd)
     if ( !is_idle_vcpu(vc) )
         svc->budget = RTDS_DEFAULT_BUDGET;
 
-    SCHED_STAT_CRANK(vcpu_init);
+    SCHED_STAT_CRANK(vcpu_alloc);
 
     return svc;
 }
@@ -635,6 +635,8 @@ rt_vcpu_insert(const struct scheduler *ops, struct vcpu *vc)
 
     /* add rt_vcpu svc to scheduler-specific vcpu list of the dom */
     list_add_tail(&svc->sdom_elem, &svc->sdom->vcpu);
+
+    SCHED_STAT_CRANK(vcpu_insert);
 }
 
 /*
@@ -648,7 +650,7 @@ rt_vcpu_remove(const struct scheduler *ops, struct vcpu *vc)
     struct rt_dom * const sdom = svc->sdom;
     spinlock_t *lock;
 
-    SCHED_STAT_CRANK(vcpu_destroy);
+    SCHED_STAT_CRANK(vcpu_remove);
 
     BUG_ON( sdom == NULL );
 
diff --git a/xen/include/xen/perfc_defn.h b/xen/include/xen/perfc_defn.h
index 526002d..76ee803 100644
--- a/xen/include/xen/perfc_defn.h
+++ b/xen/include/xen/perfc_defn.h
@@ -19,8 +19,9 @@ PERFCOUNTER(sched_ctx,              "sched: context switches")
 PERFCOUNTER(schedule,               "sched: specific scheduler")
 PERFCOUNTER(dom_init,               "sched: dom_init")
 PERFCOUNTER(dom_destroy,            "sched: dom_destroy")
-PERFCOUNTER(vcpu_init,              "sched: vcpu_init")
-PERFCOUNTER(vcpu_destroy,           "sched: vcpu_destroy")
+PERFCOUNTER(vcpu_alloc,             "sched: vcpu_alloc")
+PERFCOUNTER(vcpu_insert,            "sched: vcpu_insert")
+PERFCOUNTER(vcpu_remove,            "sched: vcpu_remove")
 PERFCOUNTER(vcpu_sleep,             "sched: vcpu_sleep")
 PERFCOUNTER(vcpu_wake_running,      "sched: vcpu_wake_running")
 PERFCOUNTER(vcpu_wake_onrunq,       "sched: vcpu_wake_onrunq")

Re: [PATCH] xen: sched: adjustments to some performance counters

[Juergen Gross]

On 09/25/2015 11:53 AM, Dario Faggioli wrote:
> More specifically:
>
> 1) rename vcpu_destroy to vcpu_remove
>
> It seems this have had to be done as part of 7e6b926a
> ("cpupools: Make interface more consistent"), which
> renamed the function but not the counter.
>
> In fact, because of cpupools, vcpus are not only removed
> from a scheduler when they are destroyed, but also when
> domains move between pools.
>
> Make the related statistics counter reflect that more
> properly.
>
> 2) rename vcpu_init to vcpu_alloc
>
> As it lives in *_alloc_vdata.
>
> 3) add vcpu_insert
>
> matching vcpu_remove, and useful to quickly check
> whether the number of insertions and removal matches,
> or in general investigare their relationship.
>
> Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
> ---
> Cc: George Dunlap <george.dunlap@eu.citrix.com>
> Cc: Juergen Gross <jgross@suse.com>
> Cc: Meng Xu <mengxu@cis.upenn.edu>
> Cc: Jan Beulich <jbeulich@suse.com>
> ---
> This would be v2 of "xen: sched: rename vcpu_destroy perf counter to
> vcpu_remove", but both the title and the content changed, so I'm treating it as
> a fresh submission (and not applying the Rev-by Juergen gave to it).

You can have it again:

Reviewed-by: Juergen Gross <jgross@suse.com>


Juergen

Re: [PATCH] xen: sched: adjustments to some performance counters

[George Dunlap]

On 25/09/15 10:53, Dario Faggioli wrote:
> More specifically:
> 
> 1) rename vcpu_destroy to vcpu_remove
> 
> It seems this have had to be done as part of 7e6b926a
> ("cpupools: Make interface more consistent"), which
> renamed the function but not the counter.
> 
> In fact, because of cpupools, vcpus are not only removed
> from a scheduler when they are destroyed, but also when
> domains move between pools.
> 
> Make the related statistics counter reflect that more
> properly.
> 
> 2) rename vcpu_init to vcpu_alloc
> 
> As it lives in *_alloc_vdata.
> 
> 3) add vcpu_insert
> 
> matching vcpu_remove, and useful to quickly check
> whether the number of insertions and removal matches,
> or in general investigare their relationship.
> 
> Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>

The changeset summary is a bit vague, but I can't immediately think of a
better headline title.

Acked-by: George Dunlap <george.dunlap@citrix.com>