Skip to contentSkip to conversation

KVM: X86: Some tracepoint enhancements

PATCH In review Peter Xu 5 participants 17 messages Jul 29, 2019 – Aug 13, 2019

Subsystems (3)
lore
Share & tools
lore

Peter Xu

author
#0 lore raw

cover KVM: X86: Some tracepoint enhancements

Each small patch explains itself. I noticed them when I'm tracing some IRQ paths and I found them helpful at least to me.

Please have a look. Thanks,

Peter Xu (3):
KVM: X86: Trace vcpu_id for vmexit
KVM: X86: Remove tailing newline for tracepoints
KVM: X86: Tune PLE Window tracepoint

 arch/x86/kvm/svm.c     |  8 ++++----
 arch/x86/kvm/trace.h   | 33 ++++++++++++++++-----------------
 arch/x86/kvm/vmx/vmx.c |  4 ++--
 3 files changed, 22 insertions(+), 23 deletions(-)
Signature · 1 line
2.21.0
author
#1 lore raw

PATCH 1/3 KVM: X86: Trace vcpu_id for vmexit

It helps to pair vmenters and vmexis with multi-core systems.

---
 arch/x86/kvm/trace.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/kvm/trace.h +4 -1
index 4d47a2631d1f..26423d2e45df 100644
@@ -232,17 +232,20 @@ TRACE_EVENT(kvm_exit,
 		__field(	u32,	        isa             )
 		__field(	u64,	        info1           )
 		__field(	u64,	        info2           )
+		__field(	int,	        vcpu_id         )
 	),
 
 	TP_fast_assign(
 		__entry->exit_reason	= exit_reason;
 		__entry->guest_rip	= kvm_rip_read(vcpu);
 		__entry->isa            = isa;
+		__entry->vcpu_id        = vcpu->vcpu_id;
 		kvm_x86_ops->get_exit_info(vcpu, &__entry->info1,
 					   &__entry->info2);
 	),
 
-	TP_printk("reason %s rip 0x%lx info %llx %llx",
+	TP_printk("vcpu %d reason %s rip 0x%lx info %llx %llx",
+		  __entry->vcpu_id,
 		 (__entry->isa == KVM_ISA_VMX) ?
 		 __print_symbolic(__entry->exit_reason, VMX_EXIT_REASONS) :
 		 __print_symbolic(__entry->exit_reason, SVM_EXIT_REASONS),
Signature · 1 line
2.21.0
#2 lore raw

Re: [PATCH 1/3] KVM: X86: Trace vcpu_id for vmexit

Quoted 1 line · Peter Xu

It helps to pair vmenters and vmexis with multi-core systems.

Typo "vmexis". The wording is also a bit funky. How about:

Tracing the ID helps to pair vmenters and vmexits for guests with multiple vCPUs.

Quoted 34 lines
---
 arch/x86/kvm/trace.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/kvm/trace.h +4 -1
index 4d47a2631d1f..26423d2e45df 100644
@@ -232,17 +232,20 @@ TRACE_EVENT(kvm_exit,
 		__field(	u32,	        isa             )
 		__field(	u64,	        info1           )
 		__field(	u64,	        info2           )
+		__field(	int,	        vcpu_id         )
 	),
 
 	TP_fast_assign(
 		__entry->exit_reason	= exit_reason;
 		__entry->guest_rip	= kvm_rip_read(vcpu);
 		__entry->isa            = isa;
+		__entry->vcpu_id        = vcpu->vcpu_id;
 		kvm_x86_ops->get_exit_info(vcpu, &__entry->info1,
 					   &__entry->info2);
 	),
 
-	TP_printk("reason %s rip 0x%lx info %llx %llx",
+	TP_printk("vcpu %d reason %s rip 0x%lx info %llx %llx",
+		  __entry->vcpu_id,
 		 (__entry->isa == KVM_ISA_VMX) ?
 		 __print_symbolic(__entry->exit_reason, VMX_EXIT_REASONS) :
 		 __print_symbolic(__entry->exit_reason, SVM_EXIT_REASONS),
Signature · 1 line
2.21.0
#3 lore raw

Re: [PATCH 1/3] KVM: X86: Trace vcpu_id for vmexit

Quoted 7 lines · Sean Christopherson
Quoted 1 line · Peter Xu

It helps to pair vmenters and vmexis with multi-core systems.

Typo "vmexis". The wording is also a bit funky. How about:

Tracing the ID helps to pair vmenters and vmexits for guests with multiple vCPUs.

Sure thing. Thanks,

Signature · 1 line
Peter Xu
Reviewed-by on v1 1/3
#4 lore raw

Re: [PATCH 1/3] KVM: X86: Trace vcpu_id for vmexit

In reply to Peter Xu
Quoted 23 lines · Peter Xu

It helps to pair vmenters and vmexis with multi-core systems.

---
  arch/x86/kvm/trace.h | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/kvm/trace.h +2 -0
index 4d47a2631d1f..26423d2e45df 100644
@@ -232,17 +232,20 @@ TRACE_EVENT(kvm_exit,
  		__field(	u32,	        isa             )
  		__field(	u64,	        info1           )
  		__field(	u64,	        info2           )
+		__field(	int,	        vcpu_id         )
  	),
 
  	TP_fast_assign(
  		__entry->exit_reason	= exit_reason;
  		__entry->guest_rip	= kvm_rip_read(vcpu);
  		__entry->isa            = isa;
+		__entry->vcpu_id        = vcpu->vcpu_id;

arch/x86/kvm/trace.h TRACE_EVENT(kvm_exit, · line 251

Quoted 10 lines
arch/x86/kvm/trace.h +2 -1
  		kvm_x86_ops->get_exit_info(vcpu, &__entry->info1,
  					   &__entry->info2);
  	),
 
-	TP_printk("reason %s rip 0x%lx info %llx %llx",
+	TP_printk("vcpu %d reason %s rip 0x%lx info %llx %llx",
+		  __entry->vcpu_id,
  		 (__entry->isa == KVM_ISA_VMX) ?
  		 __print_symbolic(__entry->exit_reason, VMX_EXIT_REASONS) :
  		 __print_symbolic(__entry->exit_reason, SVM_EXIT_REASONS),
author
#5 lore raw

PATCH 2/3 KVM: X86: Remove tailing newline for tracepoints

In reply to Peter Xu

It's done by TP_printk() already.

---
 arch/x86/kvm/trace.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/kvm/trace.h +3 -3
index 26423d2e45df..76a39bc25b95 100644
@@ -1323,7 +1323,7 @@ TRACE_EVENT(kvm_avic_incomplete_ipi,
 		__entry->index = index;
 	),
 
-	TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u\n",
+	TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u",
 		  __entry->vcpu, __entry->icrh, __entry->icrl,
 		  __entry->id, __entry->index)
 );
@@ -1348,7 +1348,7 @@ TRACE_EVENT(kvm_avic_unaccelerated_access,
 		__entry->vec = vec;
 	),
 
-	TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x\n",
+	TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x",
 		  __entry->vcpu,
 		  __entry->offset,
 		  __print_symbolic(__entry->offset, kvm_trace_symbol_apic),
@@ -1368,7 +1368,7 @@ TRACE_EVENT(kvm_hv_timer_state,
 			__entry->vcpu_id = vcpu_id;
 			__entry->hv_timer_in_use = hv_timer_in_use;
 			),
-		TP_printk("vcpu_id %x hv_timer %x\n",
+		TP_printk("vcpu_id %x hv_timer %x",
 			__entry->vcpu_id,
 			__entry->hv_timer_in_use)
 );
Signature · 1 line
2.21.0
Reviewed-by on v1 2/3
#6 lore raw

Re: [PATCH 2/3] KVM: X86: Remove tailing newline for tracepoints

Quoted 30 lines · Peter Xu

It's done by TP_printk() already.

---
  arch/x86/kvm/trace.h | 6 +++---
  1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/kvm/trace.h +2 -2
index 26423d2e45df..76a39bc25b95 100644
@@ -1323,7 +1323,7 @@ TRACE_EVENT(kvm_avic_incomplete_ipi,
  		__entry->index = index;
  	),
 
-	TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u\n",
+	TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u",
  		  __entry->vcpu, __entry->icrh, __entry->icrl,
  		  __entry->id, __entry->index)
  );
@@ -1348,7 +1348,7 @@ TRACE_EVENT(kvm_avic_unaccelerated_access,
  		__entry->vec = vec;
  	),
 
-	TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x\n",
+	TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x",
  		  __entry->vcpu,
  		  __entry->offset,
  		  __print_symbolic(__entry->offset, kvm_trace_symbol_apic),

arch/x86/kvm/trace.h TRACE_EVENT(kvm_hv_timer_state, · line 1374

Quoted 8 lines
arch/x86/kvm/trace.h +1 -1
@@ -1368,7 +1368,7 @@ TRACE_EVENT(kvm_hv_timer_state,
  			__entry->vcpu_id = vcpu_id;
  			__entry->hv_timer_in_use = hv_timer_in_use;
  			),
-		TP_printk("vcpu_id %x hv_timer %x\n",
+		TP_printk("vcpu_id %x hv_timer %x",
  			__entry->vcpu_id,
  			__entry->hv_timer_in_use)
  );
#7 lore raw

Re: [PATCH 2/3] KVM: X86: Remove tailing newline for tracepoints

In reply to Peter Xu
Quoted 31 lines · Peter Xu

It's done by TP_printk() already.

---
 arch/x86/kvm/trace.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/kvm/trace.h +2 -2
index 26423d2e45df..76a39bc25b95 100644
@@ -1323,7 +1323,7 @@ TRACE_EVENT(kvm_avic_incomplete_ipi,
                __entry->index = index;
        ),
 
-       TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u\n",
+       TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u",
                  __entry->vcpu, __entry->icrh, __entry->icrl,
                  __entry->id, __entry->index)
 );
@@ -1348,7 +1348,7 @@ TRACE_EVENT(kvm_avic_unaccelerated_access,
                __entry->vec = vec;
        ),
 
-       TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x\n",
+       TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x",
                  __entry->vcpu,
                  __entry->offset,
                  __print_symbolic(__entry->offset, kvm_trace_symbol_apic),

arch/x86/kvm/trace.h TRACE_EVENT(kvm_hv_timer_state, · line 1373

Quoted 7 lines
arch/x86/kvm/trace.h +1 -1
@@ -1368,7 +1368,7 @@ TRACE_EVENT(kvm_hv_timer_state,
                        __entry->vcpu_id = vcpu_id;
                        __entry->hv_timer_in_use = hv_timer_in_use;
                        ),
-               TP_printk("vcpu_id %x hv_timer %x\n",
+               TP_printk("vcpu_id %x hv_timer %x",
                        __entry->vcpu_id,
                        __entry->hv_timer_in_use)

The last one is handled by commit 7be373b6de503 .

Regards,
Wanpeng Li

#8 lore raw

Re: [PATCH 2/3] KVM: X86: Remove tailing newline for tracepoints

Quoted 26 lines · Wanpeng Li
Quoted 25 lines · Peter Xu

It's done by TP_printk() already.

---
 arch/x86/kvm/trace.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/kvm/trace.h +1 -1
index 26423d2e45df..76a39bc25b95 100644
@@ -1323,7 +1323,7 @@ TRACE_EVENT(kvm_avic_incomplete_ipi,
                __entry->index = index;
        ),
 
-       TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u\n",
+       TP_printk("vcpu=%u, icrh:icrl=%#010x:%08x, id=%u, index=%u",
                  __entry->vcpu, __entry->icrh, __entry->icrl,
                  __entry->id, __entry->index)
 );
@@ -1348,7 +1348,7 @@ TRACE_EVENT(kvm_avic_unaccelerated_access,
                __entry->vec = vec;
        ),
 
Quoted 15 lines

arch/x86/kvm/trace.h TRACE_EVENT(kvm_hv_timer_state, · line 1373

Quoted 13 lines
arch/x86/kvm/trace.h +2 -2
-       TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x\n",
+       TP_printk("vcpu=%u, offset=%#x(%s), %s, %s, vec=%#x",
                  __entry->vcpu,
                  __entry->offset,
                  __print_symbolic(__entry->offset, kvm_trace_symbol_apic),
@@ -1368,7 +1368,7 @@ TRACE_EVENT(kvm_hv_timer_state,
                        __entry->vcpu_id = vcpu_id;
                        __entry->hv_timer_in_use = hv_timer_in_use;
                        ),
-               TP_printk("vcpu_id %x hv_timer %x\n",
+               TP_printk("vcpu_id %x hv_timer %x",
                        __entry->vcpu_id,
                        __entry->hv_timer_in_use)

The last one is handled by commit 7be373b6de503 .

Right, I'll rebase. Thanks.

Signature · 1 line
Peter Xu
author
#9 lore raw

PATCH 3/3 KVM: X86: Tune PLE Window tracepoint

In reply to Peter Xu

The PLE window tracepoint triggers easily and it can be a bit confusing too. One example line:

kvm_ple_window: vcpu 0: ple_window 4096 (shrink 4096)

It easily let people think of "the window now is 4096 which is shrinked", but the truth is the value actually didn't change (4096).

Let's only dump this message if the value really changed, and we make the message even simpler like:

kvm_ple_window: vcpu 4 (4096 -> 8192)

---
 arch/x86/kvm/svm.c     |  8 ++++----
 arch/x86/kvm/trace.h   | 22 +++++++++-------------
 arch/x86/kvm/vmx/vmx.c |  4 ++--
 3 files changed, 15 insertions(+), 19 deletions(-)
arch/x86/kvm/svm.c +4 -4
index 48c865a4e5dd..0d365b621b5a 100644
@@ -1268,8 +1268,8 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (control->pause_filter_count != old)
 		mark_dirty(svm->vmcb, VMCB_INTERCEPTS);
 
-	trace_kvm_ple_window_grow(vcpu->vcpu_id,
-				  control->pause_filter_count, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id,
+				     control->pause_filter_count, old);
 }
 
 static void shrink_ple_window(struct kvm_vcpu *vcpu)
@@ -1286,8 +1286,8 @@ static void shrink_ple_window(struct kvm_vcpu *vcpu)
 	if (control->pause_filter_count != old)
 		mark_dirty(svm->vmcb, VMCB_INTERCEPTS);
 
-	trace_kvm_ple_window_shrink(vcpu->vcpu_id,
-				    control->pause_filter_count, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id,
+				     control->pause_filter_count, old);
 }
 
 static __init int svm_hardware_setup(void)
arch/x86/kvm/trace.h +9 -13
index 76a39bc25b95..91c91f358b23 100644
@@ -891,34 +891,30 @@ TRACE_EVENT(kvm_pml_full,
 );
 
 TRACE_EVENT(kvm_ple_window,
-	TP_PROTO(bool grow, unsigned int vcpu_id, int new, int old),
-	TP_ARGS(grow, vcpu_id, new, old),
+	TP_PROTO(unsigned int vcpu_id, int new, int old),
+	TP_ARGS(vcpu_id, new, old),
 
 	TP_STRUCT__entry(
-		__field(                bool,      grow         )
 		__field(        unsigned int,   vcpu_id         )
 		__field(                 int,       new         )
 		__field(                 int,       old         )
 	),
 
 	TP_fast_assign(
-		__entry->grow           = grow;
 		__entry->vcpu_id        = vcpu_id;
 		__entry->new            = new;
 		__entry->old            = old;
 	),
 
-	TP_printk("vcpu %u: ple_window %d (%s %d)",
-	          __entry->vcpu_id,
-	          __entry->new,
-	          __entry->grow ? "grow" : "shrink",
-	          __entry->old)
+	TP_printk("vcpu %u (%d -> %d)",
+	          __entry->vcpu_id, __entry->old, __entry->new)
 );
 
-#define trace_kvm_ple_window_grow(vcpu_id, new, old) \
-	trace_kvm_ple_window(true, vcpu_id, new, old)
-#define trace_kvm_ple_window_shrink(vcpu_id, new, old) \
-	trace_kvm_ple_window(false, vcpu_id, new, old)
+#define trace_kvm_ple_window_changed(vcpu, new, old)		\
+	do {							\
+		if (old != new)					\
+			trace_kvm_ple_window(vcpu, new, old);	\
+	} while (0)
 
 TRACE_EVENT(kvm_pvclock_update,
 	TP_PROTO(unsigned int vcpu_id, struct pvclock_vcpu_time_info *pvclock),
arch/x86/kvm/vmx/vmx.c +2 -2
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);
 }
 
 static void shrink_ple_window(struct kvm_vcpu *vcpu)
@@ -5229,7 +5229,7 @@ static void shrink_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
-	trace_kvm_ple_window_shrink(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);
 }
 
 /*
Signature · 1 line
2.21.0
#10 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 5 lines · Peter Xu

The PLE window tracepoint triggers easily and it can be a bit confusing too. One example line:

kvm_ple_window: vcpu 0: ple_window 4096 (shrink 4096)

Quoted 7 lines

It easily let people think of "the window now is 4096 which is shrinked", but the truth is the value actually didn't change (4096).

Let's only dump this message if the value really changed, and we make the message even simpler like:

kvm_ple_window: vcpu 4 (4096 -> 8192)

This seems a bit too terse, e.g. requires a decent amount of effort to do relatively simple things like show only cases where the windows was shrunk, or grew/shrunk by a large amount. In this case, more is likely better, e.g.:

kvm_ple_window_changed: vcpu 4 ple_window 8192 old 4096 grow 4096

and

kvm_ple_window_changed: vcpu 4 ple_window 4096 old 8192 shrink 4096

Tangentially related, it'd be nice to settle on a standard format for printing field+val. Right now there are four different styles, e.g. "field=val", "field = val", "field: val" and "field val".

Quoted 39 lines
---
 arch/x86/kvm/svm.c     |  8 ++++----
 arch/x86/kvm/trace.h   | 22 +++++++++-------------
 arch/x86/kvm/vmx/vmx.c |  4 ++--
 3 files changed, 15 insertions(+), 19 deletions(-)
arch/x86/kvm/svm.c +4 -4
index 48c865a4e5dd..0d365b621b5a 100644
@@ -1268,8 +1268,8 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (control->pause_filter_count != old)
 		mark_dirty(svm->vmcb, VMCB_INTERCEPTS);
 
-	trace_kvm_ple_window_grow(vcpu->vcpu_id,
-				  control->pause_filter_count, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id,
+				     control->pause_filter_count, old);
 }
 
 static void shrink_ple_window(struct kvm_vcpu *vcpu)
@@ -1286,8 +1286,8 @@ static void shrink_ple_window(struct kvm_vcpu *vcpu)
 	if (control->pause_filter_count != old)
 		mark_dirty(svm->vmcb, VMCB_INTERCEPTS);
 
-	trace_kvm_ple_window_shrink(vcpu->vcpu_id,
-				    control->pause_filter_count, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id,
+				     control->pause_filter_count, old);
 }
 
 static __init int svm_hardware_setup(void)
arch/x86/kvm/trace.h +0 -0
index 76a39bc25b95..91c91f358b23 100644

arch/x86/kvm/trace.h TRACE_EVENT(kvm_pml_full, · line 898

Quoted 10 lines
arch/x86/kvm/trace.h +2 -3
@@ -891,34 +891,30 @@ TRACE_EVENT(kvm_pml_full,
 );
 
 TRACE_EVENT(kvm_ple_window,
-	TP_PROTO(bool grow, unsigned int vcpu_id, int new, int old),
-	TP_ARGS(grow, vcpu_id, new, old),
+	TP_PROTO(unsigned int vcpu_id, int new, int old),
+	TP_ARGS(vcpu_id, new, old),
 
 	TP_STRUCT__entry(
-		__field(                bool,      grow         )

Side note, if the tracepoint is invoked only on changes the "grow" field can be removed even if the tracepoint prints grow vs. shrink, i.e. there's no ambiguity since new==old will never happen.

Quoted 38 lines
arch/x86/kvm/trace.h +7 -10
 		__field(        unsigned int,   vcpu_id         )
 		__field(                 int,       new         )
 		__field(                 int,       old         )
 	),
 
 	TP_fast_assign(
-		__entry->grow           = grow;
 		__entry->vcpu_id        = vcpu_id;
 		__entry->new            = new;
 		__entry->old            = old;
 	),
 
-	TP_printk("vcpu %u: ple_window %d (%s %d)",
-	          __entry->vcpu_id,
-	          __entry->new,
-	          __entry->grow ? "grow" : "shrink",
-	          __entry->old)
+	TP_printk("vcpu %u (%d -> %d)",
+	          __entry->vcpu_id, __entry->old, __entry->new)
 );
 
-#define trace_kvm_ple_window_grow(vcpu_id, new, old) \
-	trace_kvm_ple_window(true, vcpu_id, new, old)
-#define trace_kvm_ple_window_shrink(vcpu_id, new, old) \
-	trace_kvm_ple_window(false, vcpu_id, new, old)
+#define trace_kvm_ple_window_changed(vcpu, new, old)		\
+	do {							\
+		if (old != new)					\
+			trace_kvm_ple_window(vcpu, new, old);	\
+	} while (0)
 
 TRACE_EVENT(kvm_pvclock_update,
 	TP_PROTO(unsigned int vcpu_id, struct pvclock_vcpu_time_info *pvclock),
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 5 lines
arch/x86/kvm/vmx/vmx.c +1 -1
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 15 lines
arch/x86/kvm/vmx/vmx.c +1 -1
 }
 
 static void shrink_ple_window(struct kvm_vcpu *vcpu)
@@ -5229,7 +5229,7 @@ static void shrink_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
-	trace_kvm_ple_window_shrink(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);
 }
 
 /*
Signature · 1 line
2.21.0
author
#11 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 19 lines · Sean Christopherson
Quoted 5 lines · Peter Xu

The PLE window tracepoint triggers easily and it can be a bit confusing too. One example line:

kvm_ple_window: vcpu 0: ple_window 4096 (shrink 4096)

Quoted 7 lines

It easily let people think of "the window now is 4096 which is shrinked", but the truth is the value actually didn't change (4096).

Let's only dump this message if the value really changed, and we make the message even simpler like:

kvm_ple_window: vcpu 4 (4096 -> 8192)

This seems a bit too terse, e.g. requires a decent amount of effort to do relatively simple things like show only cases where the windows was shrunk, or grew/shrunk by a large amount. In this case, more is likely better, e.g.:

Quoted 5 lines

kvm_ple_window_changed: vcpu 4 ple_window 8192 old 4096 grow 4096

and

kvm_ple_window_changed: vcpu 4 ple_window 4096 old 8192 shrink 4096

How about:

kvm_ple_window: vcpu 4 (4096 -> 8192, growed)

Or:

kvm_ple_window: vcpu 4 old 4096 new 8192 growed

I would prefer the arrow which is very clear to me to show a value change, but I'd be fine to see what's your final preference or any further reviewers. Anyway I think any of them is clearer than the original version...

Quoted 5 lines

Tangentially related, it'd be nice to settle on a standard format for printing field+val. Right now there are four different styles, e.g. "field=val", "field = val", "field: val" and "field val".

Right, I ses "field val" is used most frequently. But I didn't touch those up because they haven't yet caused any confusion to me.

[...]

Quoted 6 lines

arch/x86/kvm/trace.h TRACE_EVENT(kvm_pml_full, · line 898

Quoted 2 lines
arch/x86/kvm/trace.h +0 -1
 	TP_STRUCT__entry(
-		__field(                bool,      grow         )

Side note, if the tracepoint is invoked only on changes the "grow" field can be removed even if the tracepoint prints grow vs. shrink, i.e. there's no ambiguity since new==old will never happen.

But I do see it happen... Please see below.

Quoted 42 lines
Quoted 42 lines
arch/x86/kvm/trace.h +7 -10
 		__field(        unsigned int,   vcpu_id         )
 		__field(                 int,       new         )
 		__field(                 int,       old         )
 	),
 
 	TP_fast_assign(
-		__entry->grow           = grow;
 		__entry->vcpu_id        = vcpu_id;
 		__entry->new            = new;
 		__entry->old            = old;
 	),
 
-	TP_printk("vcpu %u: ple_window %d (%s %d)",
-	          __entry->vcpu_id,
-	          __entry->new,
-	          __entry->grow ? "grow" : "shrink",
-	          __entry->old)
+	TP_printk("vcpu %u (%d -> %d)",
+	          __entry->vcpu_id, __entry->old, __entry->new)
 );
 
-#define trace_kvm_ple_window_grow(vcpu_id, new, old) \
-	trace_kvm_ple_window(true, vcpu_id, new, old)
-#define trace_kvm_ple_window_shrink(vcpu_id, new, old) \
-	trace_kvm_ple_window(false, vcpu_id, new, old)
+#define trace_kvm_ple_window_changed(vcpu, new, old)		\
+	do {							\
+		if (old != new)					\
+			trace_kvm_ple_window(vcpu, new, old);	\
+	} while (0)
 
 TRACE_EVENT(kvm_pvclock_update,
 	TP_PROTO(unsigned int vcpu_id, struct pvclock_vcpu_time_info *pvclock),
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

It didn't return, did it? :)

Thanks,

Signature · 1 line
Peter Xu
#12 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 30 lines · Peter Xu
Quoted 19 lines · Sean Christopherson
Quoted 5 lines · Peter Xu

The PLE window tracepoint triggers easily and it can be a bit confusing too. One example line:

kvm_ple_window: vcpu 0: ple_window 4096 (shrink 4096)

Quoted 7 lines

It easily let people think of "the window now is 4096 which is shrinked", but the truth is the value actually didn't change (4096).

Let's only dump this message if the value really changed, and we make the message even simpler like:

kvm_ple_window: vcpu 4 (4096 -> 8192)

This seems a bit too terse, e.g. requires a decent amount of effort to do relatively simple things like show only cases where the windows was shrunk, or grew/shrunk by a large amount. In this case, more is likely better, e.g.:

Quoted 5 lines

kvm_ple_window_changed: vcpu 4 ple_window 8192 old 4096 grow 4096

and

kvm_ple_window_changed: vcpu 4 ple_window 4096 old 8192 shrink 4096

How about:

   kvm_ple_window: vcpu 4 (4096 -> 8192, growed)
Quoted 8 lines

Or:

   kvm_ple_window: vcpu 4 old 4096 new 8192 growed

I would prefer the arrow which is very clear to me to show a value change, but I'd be fine to see what's your final preference or any further reviewers. Anyway I think any of them is clearer than the original version...

For tracepoints, I prefer to err on the side of more info as it's easy to filter out unwanted date. But odds are I'll never use this particular tracepoint, so I'll defer to folks who are actually affected.

Quoted 8 lines
Quoted 5 lines

Tangentially related, it'd be nice to settle on a standard format for printing field+val. Right now there are four different styles, e.g. "field=val", "field = val", "field: val" and "field val".

Right, I ses "field val" is used most frequently. But I didn't touch those up because they haven't yet caused any confusion to me.

Ya, it was more of a general complaint :-)

Quoted 10 lines

[...]

Quoted 6 lines

arch/x86/kvm/trace.h TRACE_EVENT(kvm_pml_full, · line 898

Quoted 2 lines
arch/x86/kvm/trace.h +0 -1
 	TP_STRUCT__entry(
-		__field(                bool,      grow         )

Side note, if the tracepoint is invoked only on changes the "grow" field can be removed even if the tracepoint prints grow vs. shrink, i.e. there's no ambiguity since new==old will never happen.

But I do see it happen... Please see below.

...

Quoted 15 lines
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 4 lines

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

Yep.

Quoted 1 line

It didn't return, did it? :)

You lost me. What's wrong with:

	if (vmx->ple_window != old) {
		vmx->ple_window_dirty = true;
		trace_kvm_ple_window_update(vcpu->vcpu_id, vmx_ple->window, old);
	}
#13 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 22 lines · Sean Christopherson
Quoted 15 lines
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 4 lines

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

Yep.

Quoted 8 lines
Quoted 1 line

It didn't return, did it? :)

You lost me. What's wrong with:

	if (vmx->ple_window != old) {
		vmx->ple_window_dirty = true;
		trace_kvm_ple_window_update(vcpu->vcpu_id, vmx_ple->window, old);
	}

Yes this looks fine to me. I'll switch.

Regards,

Signature · 1 line
Peter Xu
#14 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 25 lines · Peter Xu
Quoted 24 lines · Sean Christopherson
Quoted 15 lines
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 4 lines

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

Yep.

Quoted 1 line

It didn't return, did it? :)

Quoted 8 lines
Quoted 6 lines

You lost me. What's wrong with:

	if (vmx->ple_window != old) {
		vmx->ple_window_dirty = true;
		trace_kvm_ple_window_update(vcpu->vcpu_id, vmx_ple->window, old);
	}

Yes this looks fine to me. I'll switch.

Btw, I noticed we have this:

  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ple_window);

Is that trying to expose the tracepoints to the outter world? Is that whole chunk of EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_*) really needed?

Regards,

Signature · 1 line
Peter Xu
#15 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 33 lines · Peter Xu
Quoted 32 lines · Peter Xu
Quoted 30 lines · Sean Christopherson
Quoted 15 lines
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 4 lines

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

Yep.

Quoted 1 line

It didn't return, did it? :)

You lost me. What's wrong with:

	if (vmx->ple_window != old) {
		vmx->ple_window_dirty = true;
		trace_kvm_ple_window_update(vcpu->vcpu_id, vmx_ple->window, old);
	}
Quoted 8 lines
Quoted 1 line

Yes this looks fine to me. I'll switch.

Btw, I noticed we have this:

  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ple_window);

Is that trying to expose the tracepoints to the outter world? Is that whole chunk of EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_*) really needed?

It's needed to invoke tracepoints from VMX/SVM as the implementations live in kvm.ko. Same reason functions in x86.c and company need to be exported if they're called by VMX/SVM code.

#16 lore raw

Re: [PATCH 3/3] KVM: X86: Tune PLE Window tracepoint

Quoted 38 lines · Sean Christopherson
Quoted 37 lines · Peter Xu
Quoted 25 lines · Peter Xu
Quoted 24 lines · Sean Christopherson
Quoted 15 lines
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +0 -0
index d98eac371c0a..cc1f98130e6a 100644
@@ -5214,7 +5214,7 @@ static void grow_ple_window(struct kvm_vcpu *vcpu)
 	if (vmx->ple_window != old)
 		vmx->ple_window_dirty = true;
 
Quoted 6 lines

arch/x86/kvm/vmx/vmx.c static void grow_ple_window(struct kvm_vcpu *vcpu) · line 5217

Quoted 2 lines
arch/x86/kvm/vmx/vmx.c +1 -1
-	trace_kvm_ple_window_grow(vcpu->vcpu_id, vmx->ple_window, old);
+	trace_kvm_ple_window_changed(vcpu->vcpu_id, vmx->ple_window, old);

No need for the macro, the snippet right about already checks 'new != old'. Though I do like the rename, i.e. rename the trace function to trace_kvm_ple_window_changed().

Quoted 4 lines

Do you mean this one?

	if (vmx->ple_window != old)
		vmx->ple_window_dirty = true;

Yep.

Quoted 1 line

It didn't return, did it? :)

Quoted 8 lines
Quoted 6 lines

You lost me. What's wrong with:

	if (vmx->ple_window != old) {
		vmx->ple_window_dirty = true;
		trace_kvm_ple_window_update(vcpu->vcpu_id, vmx_ple->window, old);
	}

Yes this looks fine to me. I'll switch.

Btw, I noticed we have this:

Quoted 8 lines
Quoted 4 lines
  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ple_window);

Is that trying to expose the tracepoints to the outter world? Is that whole chunk of EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_*) really needed?

It's needed to invoke tracepoints from VMX/SVM as the implementations live in kvm.ko. Same reason functions in x86.c and company need to be exported if they're called by VMX/SVM code.

Ah right. Then I assume it's pretty safe to change the symbol name here.

Thanks!

Signature · 1 line
Peter Xu