[PATCH 0/7] perf, x86: Haswell LBR call stack support

[PATCH 0/7] perf, x86: Haswell LBR call stack support

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

Haswell has a new feature that utilizes the existing Last Branch Record
facility to record call chains. When the feature is enabled, function
call will be collected as normal, but as return instructions are executed
the last captured branch record is popped from the on-chip LBR registers.
The LBR call stack facility can help perf to get call chains of progam 
without frame pointer. When perf tool requests PERF_SAMPLE_CALLCHAIN +
PERF_SAMPLE_BRANCH_USER, this feature is dynamically enabled by default.
This feature can be disabled/enabled through an attribute file in the cpu
pmu sysfs directory.

The LBR call stack has following known limitations
 1. Zero length calls are not filtered out by hardware
 2. Exception handing such as setjmp/longjmp will have calls/returns not
    match
 3. Pushing different return address onto the stack will have calls/returns
    not match

The patch series depends on Andi's "Basic perf PMU support for Haswell v1" 

Available from
    https://github.com/ukernel/linux.git hws/pmu4

Regards
Yan, Zheng

[PATCH 1/7] perf, x86: Reduce lbr_sel_map size

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

The index of lbr_sel_map is bit value of perf branch_sample_type.
By using bit shift as index, we can reduce lbr_sel_map size.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.h           |  4 +++
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 50 ++++++++++++++----------------
 include/uapi/linux/perf_event.h            | 42 +++++++++++++++++--------
 3 files changed, 56 insertions(+), 40 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index ba5d043..71e87e0 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -414,6 +414,10 @@ struct x86_pmu {
 	struct perf_guest_switch_msr *(*guest_get_msrs)(int *nr);
 };
 
+enum {
+	PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE = PERF_SAMPLE_BRANCH_MAX_SHIFT,
+};
+
 #define x86_add_quirk(func_)						\
 do {									\
 	static struct x86_pmu_quirk __quirk __initdata = {		\
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 5455a00..2bac8ac 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -61,10 +61,6 @@ enum {
 #define LBR_FROM_FLAG_INTX     (1ULL << 62)
 #define LBR_FROM_FLAG_ABORT    (1ULL << 61)
 
-#define for_each_branch_sample_type(x) \
-	for ((x) = PERF_SAMPLE_BRANCH_USER; \
-	     (x) < PERF_SAMPLE_BRANCH_MAX; (x) <<= 1)
-
 /*
  * x86control flow change classification
  * x86control flow changes include branches, interrupts, traps, faults
@@ -378,14 +374,14 @@ static int intel_pmu_setup_hw_lbr_filter(struct perf_event *event)
 {
 	struct hw_perf_event_extra *reg;
 	u64 br_type = event->attr.branch_sample_type;
-	u64 mask = 0, m;
-	u64 v;
+	u64 mask = 0, v;
+	int i;
 
-	for_each_branch_sample_type(m) {
-		if (!(br_type & m))
+	for (i = 0; i < PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE; i++) {
+		if (!(br_type & (1U << i)))
 			continue;
 
-		v = x86_pmu.lbr_sel_map[m];
+		v = x86_pmu.lbr_sel_map[i];
 		if (v == LBR_NOT_SUPP)
 			return -EOPNOTSUPP;
 
@@ -631,33 +627,33 @@ intel_pmu_lbr_filter(struct cpu_hw_events *cpuc)
 /*
  * Map interface branch filters onto LBR filters
  */
-static const int nhm_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX] = {
-	[PERF_SAMPLE_BRANCH_ANY]	= LBR_ANY,
-	[PERF_SAMPLE_BRANCH_USER]	= LBR_USER,
-	[PERF_SAMPLE_BRANCH_KERNEL]	= LBR_KERNEL,
-	[PERF_SAMPLE_BRANCH_HV]		= LBR_IGN,
-	[PERF_SAMPLE_BRANCH_ANY_RETURN]	= LBR_RETURN | LBR_REL_JMP
-					| LBR_IND_JMP | LBR_FAR,
+static const int nhm_lbr_sel_map[PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE] = {
+	[PERF_SAMPLE_BRANCH_ANY_SHIFT]		= LBR_ANY,
+	[PERF_SAMPLE_BRANCH_USER_SHIFT]		= LBR_USER,
+	[PERF_SAMPLE_BRANCH_KERNEL_SHIFT]	= LBR_KERNEL,
+	[PERF_SAMPLE_BRANCH_HV_SHIFT]		= LBR_IGN,
+	[PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT]	= LBR_RETURN | LBR_REL_JMP
+						| LBR_IND_JMP | LBR_FAR,
 	/*
 	 * NHM/WSM erratum: must include REL_JMP+IND_JMP to get CALL branches
 	 */
-	[PERF_SAMPLE_BRANCH_ANY_CALL] =
+	[PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT] =
 	 LBR_REL_CALL | LBR_IND_CALL | LBR_REL_JMP | LBR_IND_JMP | LBR_FAR,
 	/*
 	 * NHM/WSM erratum: must include IND_JMP to capture IND_CALL
 	 */
-	[PERF_SAMPLE_BRANCH_IND_CALL] = LBR_IND_CALL | LBR_IND_JMP,
+	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT] = LBR_IND_CALL | LBR_IND_JMP,
 };
 
-static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX] = {
-	[PERF_SAMPLE_BRANCH_ANY]	= LBR_ANY,
-	[PERF_SAMPLE_BRANCH_USER]	= LBR_USER,
-	[PERF_SAMPLE_BRANCH_KERNEL]	= LBR_KERNEL,
-	[PERF_SAMPLE_BRANCH_HV]		= LBR_IGN,
-	[PERF_SAMPLE_BRANCH_ANY_RETURN]	= LBR_RETURN | LBR_FAR,
-	[PERF_SAMPLE_BRANCH_ANY_CALL]	= LBR_REL_CALL | LBR_IND_CALL
-					| LBR_FAR,
-	[PERF_SAMPLE_BRANCH_IND_CALL]	= LBR_IND_CALL,
+static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE] = {
+	[PERF_SAMPLE_BRANCH_ANY_SHIFT]		= LBR_ANY,
+	[PERF_SAMPLE_BRANCH_USER_SHIFT]		= LBR_USER,
+	[PERF_SAMPLE_BRANCH_KERNEL_SHIFT]	= LBR_KERNEL,
+	[PERF_SAMPLE_BRANCH_HV_SHIFT]		= LBR_IGN,
+	[PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT]	= LBR_RETURN | LBR_FAR,
+	[PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT]	= LBR_REL_CALL | LBR_IND_CALL
+						| LBR_FAR,
+	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT]	= LBR_IND_CALL,
 };
 
 /* core */
diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index 6b47798..93e945a 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -148,20 +148,36 @@ enum perf_event_sample_format {
  * The branch types can be combined, however BRANCH_ANY covers all types
  * of branches and therefore it supersedes all the other types.
  */
+enum perf_branch_sample_type_shift {
+	PERF_SAMPLE_BRANCH_USER_SHIFT		= 0, /* user branches */
+	PERF_SAMPLE_BRANCH_KERNEL_SHIFT		= 1, /* kernel branches */
+	PERF_SAMPLE_BRANCH_HV_SHIFT		= 2, /* hypervisor branches */
+
+	PERF_SAMPLE_BRANCH_ANY_SHIFT		= 3, /* any branch types */
+	PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT	= 4, /* any call branch */
+	PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT	= 5, /* any return branch */
+	PERF_SAMPLE_BRANCH_IND_CALL_SHIFT	= 6, /* indirect calls */
+	PERF_SAMPLE_BRANCH_ABORTTX_SHIFT	= 7, /* transaction aborts */
+	PERF_SAMPLE_BRANCH_INTX_SHIFT		= 8, /* in transaction */
+	PERF_SAMPLE_BRANCH_NOTX_SHIFT		= 9, /* not in transaction */
+
+	PERF_SAMPLE_BRANCH_MAX,			/* non-ABI */
+};
+
 enum perf_branch_sample_type {
-	PERF_SAMPLE_BRANCH_USER		= 1U << 0, /* user branches */
-	PERF_SAMPLE_BRANCH_KERNEL	= 1U << 1, /* kernel branches */
-	PERF_SAMPLE_BRANCH_HV		= 1U << 2, /* hypervisor branches */
-
-	PERF_SAMPLE_BRANCH_ANY		= 1U << 3, /* any branch types */
-	PERF_SAMPLE_BRANCH_ANY_CALL	= 1U << 4, /* any call branch */
-	PERF_SAMPLE_BRANCH_ANY_RETURN	= 1U << 5, /* any return branch */
-	PERF_SAMPLE_BRANCH_IND_CALL	= 1U << 6, /* indirect calls */
-	PERF_SAMPLE_BRANCH_ABORTTX	= 1U << 7, /* transaction aborts */
-	PERF_SAMPLE_BRANCH_INTX		= 1U << 8, /* in transaction (flag) */
-	PERF_SAMPLE_BRANCH_NOTX		= 1U << 9, /* not in transaction (flag) */
-
-	PERF_SAMPLE_BRANCH_MAX		= 1U << 10, /* non-ABI */
+	PERF_SAMPLE_BRANCH_USER		= 1U << PERF_SAMPLE_BRANCH_USER_SHIFT,
+	PERF_SAMPLE_BRANCH_KERNEL	= 1U << PERF_SAMPLE_BRANCH_KERNEL_SHIFT,
+	PERF_SAMPLE_BRANCH_HV		= 1U << PERF_SAMPLE_BRANCH_HV_SHIFT,
+
+	PERF_SAMPLE_BRANCH_ANY		= 1U << PERF_SAMPLE_BRANCH_ANY_SHIFT,
+	PERF_SAMPLE_BRANCH_ANY_CALL	= 1U << PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT,
+	PERF_SAMPLE_BRANCH_ANY_RETURN	= 1U << PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT,
+	PERF_SAMPLE_BRANCH_IND_CALL	= 1U << PERF_SAMPLE_BRANCH_IND_CALL_SHIFT,
+	PERF_SAMPLE_BRANCH_ABORTTX	= 1U << PERF_SAMPLE_BRANCH_ABORTTX_SHIFT,
+	PERF_SAMPLE_BRANCH_INTX		= 1U << PERF_SAMPLE_BRANCH_INTX_SHIFT,
+	PERF_SAMPLE_BRANCH_NOTX		= 1U << PERF_SAMPLE_BRANCH_NOTX_SHIFT,
+
+	PERF_SAMPLE_BRANCH_MAX		= 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT,
 };
 
 #define PERF_SAMPLE_BRANCH_PLM_ALL \
-- 
1.7.11.7

[PATCH 2/7] perf, x86: Basic Haswell LBR call stack support

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

The new HSW call stack feature provides a facility such that
unfiltered call data will be collected as normal, but as return
instructions are executed the last captured branch record is
popped from the LBR stack. Thus, branch information relative to
leaf functions will not be captured, while preserving the call
stack information of the main line execution path.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.h           |  7 ++-
 arch/x86/kernel/cpu/perf_event_intel.c     |  2 +-
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 89 ++++++++++++++++++++++--------
 include/uapi/linux/perf_event.h            |  2 +-
 4 files changed, 75 insertions(+), 25 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 71e87e0..ed1e822 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -415,7 +415,10 @@ struct x86_pmu {
 };
 
 enum {
-	PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE = PERF_SAMPLE_BRANCH_MAX_SHIFT,
+	PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT = PERF_SAMPLE_BRANCH_MAX_SHIFT,
+	PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE,
+
+	PERF_SAMPLE_BRANCH_CALL_STACK = 1U << PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT,
 };
 
 #define x86_add_quirk(func_)						\
@@ -663,6 +666,8 @@ void intel_pmu_lbr_init_atom(void);
 
 void intel_pmu_lbr_init_snb(void);
 
+void intel_pmu_lbr_init_hsw(void);
+
 int intel_pmu_setup_lbr_filter(struct perf_event *event);
 
 int p4_pmu_init(void);
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index fa20a19..fbb977e 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -2263,7 +2263,7 @@ __init int intel_pmu_init(void)
 		memcpy(hw_cache_event_ids, snb_hw_cache_event_ids,
 		       sizeof(hw_cache_event_ids));
 
-		intel_pmu_lbr_init_nhm();
+		intel_pmu_lbr_init_hsw();
 
 		x86_pmu.event_constraints = intel_hsw_event_constraints;
 		x86_pmu.pebs_constraints = intel_hsw_pebs_event_constraints;
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 2bac8ac..88367a2 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -31,6 +31,7 @@ enum {
 #define LBR_IND_JMP_BIT		6 /* do not capture indirect jumps */
 #define LBR_REL_JMP_BIT		7 /* do not capture relative jumps */
 #define LBR_FAR_BIT		8 /* do not capture far branches */
+#define LBR_CALL_STACK_BIT	9 /* enable call stack */
 
 #define LBR_KERNEL	(1 << LBR_KERNEL_BIT)
 #define LBR_USER	(1 << LBR_USER_BIT)
@@ -41,6 +42,7 @@ enum {
 #define LBR_REL_JMP	(1 << LBR_REL_JMP_BIT)
 #define LBR_IND_JMP	(1 << LBR_IND_JMP_BIT)
 #define LBR_FAR		(1 << LBR_FAR_BIT)
+#define LBR_CALL_STACK	(1 << LBR_CALL_STACK_BIT)
 
 #define LBR_PLM (LBR_KERNEL | LBR_USER)
 
@@ -66,24 +68,25 @@ enum {
  * x86control flow changes include branches, interrupts, traps, faults
  */
 enum {
-	X86_BR_NONE     = 0,      /* unknown */
-
-	X86_BR_USER     = 1 << 0, /* branch target is user */
-	X86_BR_KERNEL   = 1 << 1, /* branch target is kernel */
-
-	X86_BR_CALL     = 1 << 2, /* call */
-	X86_BR_RET      = 1 << 3, /* return */
-	X86_BR_SYSCALL  = 1 << 4, /* syscall */
-	X86_BR_SYSRET   = 1 << 5, /* syscall return */
-	X86_BR_INT      = 1 << 6, /* sw interrupt */
-	X86_BR_IRET     = 1 << 7, /* return from interrupt */
-	X86_BR_JCC      = 1 << 8, /* conditional */
-	X86_BR_JMP      = 1 << 9, /* jump */
-	X86_BR_IRQ      = 1 << 10,/* hw interrupt or trap or fault */
-	X86_BR_IND_CALL = 1 << 11,/* indirect calls */
-	X86_BR_ABORT    = 1 << 12,/* transaction abort */
-	X86_BR_INTX     = 1 << 13,/* in transaction */
-	X86_BR_NOTX     = 1 << 14,/* not in transaction */
+	X86_BR_NONE		= 0,      /* unknown */
+
+	X86_BR_USER		= 1 << 0, /* branch target is user */
+	X86_BR_KERNEL		= 1 << 1, /* branch target is kernel */
+
+	X86_BR_CALL		= 1 << 2, /* call */
+	X86_BR_RET		= 1 << 3, /* return */
+	X86_BR_SYSCALL		= 1 << 4, /* syscall */
+	X86_BR_SYSRET		= 1 << 5, /* syscall return */
+	X86_BR_INT		= 1 << 6, /* sw interrupt */
+	X86_BR_IRET		= 1 << 7, /* return from interrupt */
+	X86_BR_JCC		= 1 << 8, /* conditional */
+	X86_BR_JMP		= 1 << 9, /* jump */
+	X86_BR_IRQ		= 1 << 10,/* hw interrupt or trap or fault */
+	X86_BR_IND_CALL		= 1 << 11,/* indirect calls */
+	X86_BR_ABORT		= 1 << 12,/* transaction abort */
+	X86_BR_INTX		= 1 << 13,/* in transaction */
+	X86_BR_NOTX		= 1 << 14,/* not in transaction */
+	X86_BR_CALL_STACK	= 1 << 15,/* call stack */
 };
 
 #define X86_BR_PLM (X86_BR_USER | X86_BR_KERNEL)
@@ -127,7 +130,10 @@ static void __intel_pmu_lbr_enable(void)
 		wrmsrl(MSR_LBR_SELECT, cpuc->lbr_sel->config);
 
 	rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl);
-	debugctl |= (DEBUGCTLMSR_LBR | DEBUGCTLMSR_FREEZE_LBRS_ON_PMI);
+	debugctl |= DEBUGCTLMSR_LBR;
+	/* LBR callstack does not work well with FREEZE_LBRS_ON_PMI */
+	if (!cpuc->lbr_sel || !(cpuc->lbr_sel->config & LBR_CALL_STACK))
+		debugctl |= DEBUGCTLMSR_FREEZE_LBRS_ON_PMI;
 	wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl);
 }
 
@@ -324,7 +330,7 @@ void intel_pmu_lbr_read(void)
  * - in case there is no HW filter
  * - in case the HW filter has errata or limitations
  */
-static void intel_pmu_setup_sw_lbr_filter(struct perf_event *event)
+static int intel_pmu_setup_sw_lbr_filter(struct perf_event *event)
 {
 	u64 br_type = event->attr.branch_sample_type;
 	int mask = 0;
@@ -358,11 +364,21 @@ static void intel_pmu_setup_sw_lbr_filter(struct perf_event *event)
 	if (br_type & PERF_SAMPLE_BRANCH_NOTX)
 		mask |= X86_BR_NOTX;
 
+	if (br_type & PERF_SAMPLE_BRANCH_CALL_STACK) {
+		if (!x86_pmu.lbr_sel_map)
+			return -EOPNOTSUPP;
+		if (mask & ~(X86_BR_USER | X86_BR_KERNEL))
+			return -EINVAL;
+		mask |= X86_BR_CALL | X86_BR_IND_CALL | X86_BR_RET |
+			X86_BR_CALL_STACK;
+	}
+
 	/*
 	 * stash actual user request into reg, it may
 	 * be used by fixup code for some CPU
 	 */
 	event->hw.branch_reg.reg = mask;
+	return 0;
 }
 
 /*
@@ -392,7 +408,7 @@ static int intel_pmu_setup_hw_lbr_filter(struct perf_event *event)
 	reg->idx = EXTRA_REG_LBR;
 
 	/* LBR_SELECT operates in suppress mode so invert mask */
-	reg->config = ~mask & x86_pmu.lbr_sel_mask;
+	reg->config = mask ^ x86_pmu.lbr_sel_mask;
 
 	return 0;
 }
@@ -411,7 +427,9 @@ int intel_pmu_setup_lbr_filter(struct perf_event *event)
 	/*
 	 * setup SW LBR filter
 	 */
-	intel_pmu_setup_sw_lbr_filter(event);
+	ret = intel_pmu_setup_sw_lbr_filter(event);
+	if (ret)
+		return ret;
 
 	/*
 	 * setup HW LBR filter, if any
@@ -656,6 +674,19 @@ static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE] = {
 	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT]	= LBR_IND_CALL,
 };
 
+static const int hsw_lbr_sel_map[PERF_SAMPLE_BRANCH_SELECT_MAP_SIZE] = {
+	[PERF_SAMPLE_BRANCH_ANY_SHIFT]		= LBR_ANY,
+	[PERF_SAMPLE_BRANCH_USER_SHIFT]		= LBR_USER,
+	[PERF_SAMPLE_BRANCH_KERNEL_SHIFT]	= LBR_KERNEL,
+	[PERF_SAMPLE_BRANCH_HV_SHIFT]		= LBR_IGN,
+	[PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT]	= LBR_RETURN | LBR_FAR,
+	[PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT]	= LBR_REL_CALL | LBR_IND_CALL
+						| LBR_FAR,
+	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT]	= LBR_IND_CALL,
+	[PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT]	= LBR_REL_CALL | LBR_IND_CALL
+						| LBR_RETURN | LBR_CALL_STACK,
+};
+
 /* core */
 void intel_pmu_lbr_init_core(void)
 {
@@ -712,6 +743,20 @@ void intel_pmu_lbr_init_snb(void)
 	pr_cont("16-deep LBR, ");
 }
 
+/* haswell */
+void intel_pmu_lbr_init_hsw(void)
+{
+	x86_pmu.lbr_nr	 = 16;
+	x86_pmu.lbr_tos	 = MSR_LBR_TOS;
+	x86_pmu.lbr_from = MSR_LBR_NHM_FROM;
+	x86_pmu.lbr_to   = MSR_LBR_NHM_TO;
+
+	x86_pmu.lbr_sel_mask = LBR_SEL_MASK;
+	x86_pmu.lbr_sel_map  = hsw_lbr_sel_map;
+
+	pr_cont("16-deep LBR, ");
+}
+
 /* atom */
 void intel_pmu_lbr_init_atom(void)
 {
diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index 93e945a..8f25796 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -161,7 +161,7 @@ enum perf_branch_sample_type_shift {
 	PERF_SAMPLE_BRANCH_INTX_SHIFT		= 8, /* in transaction */
 	PERF_SAMPLE_BRANCH_NOTX_SHIFT		= 9, /* not in transaction */
 
-	PERF_SAMPLE_BRANCH_MAX,			/* non-ABI */
+	PERF_SAMPLE_BRANCH_MAX_SHIFT,		/* non-ABI */
 };
 
 enum perf_branch_sample_type {
-- 
1.7.11.7

[PATCH 3/7] perf, x86: Introduce x86 special perf event context

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

The x86 special perf event context is named x86_perf_event_context,
We can enlarge it later to store PMU special data.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.c | 12 ++++++++++++
 arch/x86/kernel/cpu/perf_event.h |  4 ++++
 include/linux/perf_event.h       |  5 +++++
 kernel/events/core.c             | 28 ++++++++++++++++++----------
 4 files changed, 39 insertions(+), 10 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 375498a..7981230 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1742,6 +1742,17 @@ static int x86_pmu_event_idx(struct perf_event *event)
 	return idx + 1;
 }
 
+static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
+{
+	struct perf_event_context *ctx;
+
+	ctx = kzalloc(sizeof(struct x86_perf_event_context), GFP_KERNEL);
+	if (!ctx)
+		return ERR_PTR(-ENOMEM);
+
+	return ctx;
+}
+
 static ssize_t get_attr_rdpmc(struct device *cdev,
 			      struct device_attribute *attr,
 			      char *buf)
@@ -1829,6 +1840,7 @@ static struct pmu pmu = {
 
 	.event_idx		= x86_pmu_event_idx,
 	.flush_branch_stack	= x86_pmu_flush_branch_stack,
+	.event_context_alloc	= x86_pmu_event_context_alloc,
 };
 
 void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index ed1e822..3080e10 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -421,6 +421,10 @@ enum {
 	PERF_SAMPLE_BRANCH_CALL_STACK = 1U << PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT,
 };
 
+struct x86_perf_event_context {
+	struct perf_event_context ctx;
+};
+
 #define x86_add_quirk(func_)						\
 do {									\
 	static struct x86_pmu_quirk __quirk __initdata = {		\
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index c32fba3..6122b2f 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -264,6 +264,11 @@ struct pmu {
 	 * flush branch stack on context-switches (needed in cpu-wide mode)
 	 */
 	void (*flush_branch_stack)	(void);
+
+	/*
+	 * Allocate PMU special perf event context
+	 */
+	void *(*event_context_alloc)	(struct perf_event_context *parent_ctx);
 };
 
 /**
diff --git a/kernel/events/core.c b/kernel/events/core.c
index b4078a0..908f2ad 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2721,13 +2721,20 @@ static void __perf_event_init_context(struct perf_event_context *ctx)
 }
 
 static struct perf_event_context *
-alloc_perf_context(struct pmu *pmu, struct task_struct *task)
+alloc_perf_context(struct pmu *pmu, struct task_struct *task,
+		   struct perf_event_context *parent_ctx)
 {
 	struct perf_event_context *ctx;
 
-	ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
-	if (!ctx)
-		return NULL;
+	if (pmu->event_context_alloc) {
+		ctx = pmu->event_context_alloc(parent_ctx);
+		if (IS_ERR(ctx))
+			return ctx;
+	} else {
+		ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
+		if (!ctx)
+			return ERR_PTR(-ENOMEM);
+	}
 
 	__perf_event_init_context(ctx);
 	if (task) {
@@ -2813,10 +2820,11 @@ retry:
 		++ctx->pin_count;
 		raw_spin_unlock_irqrestore(&ctx->lock, flags);
 	} else {
-		ctx = alloc_perf_context(pmu, task);
-		err = -ENOMEM;
-		if (!ctx)
+		ctx = alloc_perf_context(pmu, task, NULL);
+		if (IS_ERR(ctx)) {
+			err = PTR_ERR(ctx);
 			goto errout;
+		}
 
 		err = 0;
 		mutex_lock(&task->perf_event_mutex);
@@ -7132,9 +7140,9 @@ inherit_task_group(struct perf_event *event, struct task_struct *parent,
 		 * child.
 		 */
 
-		child_ctx = alloc_perf_context(event->pmu, child);
-		if (!child_ctx)
-			return -ENOMEM;
+		child_ctx = alloc_perf_context(event->pmu, child, parent_ctx);
+		if (IS_ERR(child_ctx))
+			return PTR_ERR(child_ctx);
 
 		child->perf_event_ctxp[ctxn] = child_ctx;
 	}
-- 
1.7.11.7

[PATCH 4/7] perf, x86: Save/resotre LBR stack during context switch

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

When the LBR call stack is enabled, it is necessary to save/restore
the stack on context switch. The solution is saving/restoring the
stack to/from task's perf event context. If task has no perf event
context, just flush the stack on context switch.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.c           |  18 +++--
 arch/x86/kernel/cpu/perf_event.h           |  13 +++-
 arch/x86/kernel/cpu/perf_event_intel.c     |  13 ++--
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 108 ++++++++++++++++++++++++++---
 include/linux/perf_event.h                 |   6 +-
 kernel/events/core.c                       |  64 ++++++++++-------
 6 files changed, 168 insertions(+), 54 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 7981230..8fd8894 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1742,6 +1742,13 @@ static int x86_pmu_event_idx(struct perf_event *event)
 	return idx + 1;
 }
 
+static void x86_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					bool sched_in)
+{
+	if (x86_pmu.branch_stack_sched)
+		x86_pmu.branch_stack_sched(ctx, sched_in);
+}
+
 static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 {
 	struct perf_event_context *ctx;
@@ -1750,6 +1757,9 @@ static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 	if (!ctx)
 		return ERR_PTR(-ENOMEM);
 
+	if (parent_ctx)
+		intel_pmu_lbr_init_context(ctx, parent_ctx);
+
 	return ctx;
 }
 
@@ -1807,12 +1817,6 @@ static const struct attribute_group *x86_pmu_attr_groups[] = {
 	NULL,
 };
 
-static void x86_pmu_flush_branch_stack(void)
-{
-	if (x86_pmu.flush_branch_stack)
-		x86_pmu.flush_branch_stack();
-}
-
 void perf_check_microcode(void)
 {
 	if (x86_pmu.check_microcode)
@@ -1839,7 +1843,7 @@ static struct pmu pmu = {
 	.commit_txn		= x86_pmu_commit_txn,
 
 	.event_idx		= x86_pmu_event_idx,
-	.flush_branch_stack	= x86_pmu_flush_branch_stack,
+	.branch_stack_sched     = x86_pmu_branch_stack_sched,
 	.event_context_alloc	= x86_pmu_event_context_alloc,
 };
 
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 3080e10..a6547f3 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -371,7 +371,6 @@ struct x86_pmu {
 	void		(*cpu_dead)(int cpu);
 
 	void		(*check_microcode)(void);
-	void		(*flush_branch_stack)(void);
 
 	/*
 	 * Intel Arch Perfmon v2+
@@ -401,6 +400,8 @@ struct x86_pmu {
 	int		lbr_nr;			   /* hardware stack size */
 	u64		lbr_sel_mask;		   /* LBR_SELECT valid bits */
 	const int	*lbr_sel_map;		   /* lbr_select mappings */
+	void		(*branch_stack_sched)(struct perf_event_context *ctx,
+					      bool sched_in);
 
 	/*
 	 * Extra registers for events
@@ -423,6 +424,12 @@ enum {
 
 struct x86_perf_event_context {
 	struct perf_event_context ctx;
+
+	u64 lbr_from[MAX_LBR_ENTRIES];
+	u64 lbr_to[MAX_LBR_ENTRIES];
+	u64 lbr_stack_gen;
+	int lbr_callstack_users;
+	bool lbr_stack_saved;
 };
 
 #define x86_add_quirk(func_)						\
@@ -650,8 +657,12 @@ void intel_pmu_pebs_disable_all(void);
 
 void intel_ds_init(void);
 
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx);
 void intel_pmu_lbr_reset(void);
 
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in);
+
 void intel_pmu_lbr_enable(struct perf_event *event);
 
 void intel_pmu_lbr_disable(struct perf_event *event);
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index fbb977e..61c5c17 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1807,16 +1807,11 @@ static void intel_pmu_cpu_dying(int cpu)
 	fini_debug_store_on_cpu(cpu);
 }
 
-static void intel_pmu_flush_branch_stack(void)
+static void intel_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					 bool sched_in)
 {
-	/*
-	 * Intel LBR does not tag entries with the
-	 * PID of the current task, then we need to
-	 * flush it on ctxsw
-	 * For now, we simply reset it
-	 */
 	if (x86_pmu.lbr_nr)
-		intel_pmu_lbr_reset();
+		intel_pmu_lbr_sched(ctx, sched_in);
 }
 
 PMU_FORMAT_ATTR(offcore_rsp, "config1:0-63");
@@ -1882,7 +1877,7 @@ static __initconst const struct x86_pmu intel_pmu = {
 	.cpu_starting		= intel_pmu_cpu_starting,
 	.cpu_dying		= intel_pmu_cpu_dying,
 	.guest_get_msrs		= intel_guest_get_msrs,
-	.flush_branch_stack	= intel_pmu_flush_branch_stack,
+	.branch_stack_sched	= intel_pmu_branch_stack_sched,
 };
 
 static __init void intel_clovertown_quirk(void)
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 88367a2..a1fb108 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -173,6 +173,13 @@ void intel_pmu_lbr_reset(void)
 		intel_pmu_lbr_reset_32();
 	else
 		intel_pmu_lbr_reset_64();
+
+	wrmsrl(x86_pmu.lbr_tos, 0);
+}
+
+static inline bool branch_user_callstack(unsigned br_sel)
+{
+	return (br_sel & X86_BR_USER) && (br_sel & X86_BR_CALL_STACK);
 }
 
 void intel_pmu_lbr_enable(struct perf_event *event)
@@ -182,17 +189,23 @@ void intel_pmu_lbr_enable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
-	/*
-	 * Reset the LBR stack if we changed task context to
-	 * avoid data leaks.
-	 */
-	if (event->ctx->task && cpuc->lbr_context != event->ctx) {
-		intel_pmu_lbr_reset();
-		cpuc->lbr_context = event->ctx;
-	}
 	cpuc->br_sel = event->hw.branch_reg.reg;
-
 	cpuc->lbr_users++;
+
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+		/*
+		 * Reset the LBR stack if the call stack is not
+		 * continuous enabled
+		 */
+		if (task_ctx->lbr_callstack_users == 0 &&
+		    task_ctx->lbr_stack_gen + 1 < event->ctx->sched_gen)
+			intel_pmu_lbr_reset();
+
+		task_ctx->lbr_callstack_users++;
+		task_ctx->lbr_stack_gen = event->ctx->sched_gen;
+	}
 }
 
 void intel_pmu_lbr_disable(struct perf_event *event)
@@ -202,6 +215,13 @@ void intel_pmu_lbr_disable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+
+		task_ctx->lbr_callstack_users--;
+	}
+
 	cpuc->lbr_users--;
 	WARN_ON_ONCE(cpuc->lbr_users < 0);
 
@@ -325,6 +345,76 @@ void intel_pmu_lbr_read(void)
 	intel_pmu_lbr_filter(cpuc);
 }
 
+static void __intel_pmu_lbr_restore(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		wrmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		wrmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_saved = false;
+}
+
+static void __intel_pmu_lbr_save(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		rdmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		rdmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_gen = task_ctx->ctx.sched_gen;
+	task_ctx->lbr_stack_saved = true;
+}
+
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx)
+{
+	struct x86_perf_event_context *task_ctx, *parent_task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	task_ctx = (struct x86_perf_event_context *)child_ctx;
+	parent_task_ctx = (struct x86_perf_event_context *)parent_ctx;
+
+	if (parent_task_ctx->lbr_callstack_users)
+		__intel_pmu_lbr_save(task_ctx);
+	else
+		task_ctx->lbr_stack_saved = false;
+}
+
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in)
+{
+	struct x86_perf_event_context *task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	if (!ctx) {
+		if (sched_in)
+			intel_pmu_lbr_reset();
+		return;
+	}
+
+	task_ctx = (struct x86_perf_event_context *)ctx;
+	if (sched_in) {
+		if (!task_ctx->lbr_stack_saved)
+			intel_pmu_lbr_reset();
+		else
+			__intel_pmu_lbr_restore(task_ctx);
+	} else {
+		__intel_pmu_lbr_save(task_ctx);
+	}
+}
+
 /*
  * SW filter is used:
  * - in case there is no HW filter
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 6122b2f..f130c6c 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -261,9 +261,10 @@ struct pmu {
 	int (*event_idx)		(struct perf_event *event); /*optional */
 
 	/*
-	 * flush branch stack on context-switches (needed in cpu-wide mode)
+	 * Save/restore LBR stack on context-switches
 	 */
-	void (*flush_branch_stack)	(void);
+	void (*branch_stack_sched)	(struct perf_event_context *ctx,
+					 bool sched_in);
 
 	/*
 	 * Allocate PMU special perf event context
@@ -501,6 +502,7 @@ struct perf_event_context {
 	struct perf_event_context	*parent_ctx;
 	u64				parent_gen;
 	u64				generation;
+	u64				sched_gen;
 	int				pin_count;
 	int				nr_cgroups;	 /* cgroup evts */
 	int				nr_branch_stack; /* branch_stack evt */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 908f2ad..e1ea1d5 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -138,7 +138,7 @@ enum event_type_t {
  */
 struct static_key_deferred perf_sched_events __read_mostly;
 static DEFINE_PER_CPU(atomic_t, perf_cgroup_events);
-static DEFINE_PER_CPU(atomic_t, perf_branch_stack_events);
+static DEFINE_PER_CPU(int, perf_branch_stack_events);
 
 static atomic_t nr_mmap_events __read_mostly;
 static atomic_t nr_comm_events __read_mostly;
@@ -190,6 +190,9 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
 static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
 			     enum event_type_t event_type,
 			     struct task_struct *task);
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in);
 
 static void update_context_time(struct perf_event_context *ctx);
 static u64 perf_event_time(struct perf_event *event);
@@ -1044,8 +1047,11 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 			cpuctx->cgrp = NULL;
 	}
 
-	if (has_branch_stack(event))
+	if (has_branch_stack(event)) {
+		if (ctx->is_active)
+			__get_cpu_var(perf_branch_stack_events)--;
 		ctx->nr_branch_stack--;
+	}
 
 	ctx->nr_events--;
 	if (event->attr.inherit_stat)
@@ -1566,8 +1572,10 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx,
 				struct task_struct *task)
 {
 	cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task);
-	if (ctx)
+	if (ctx) {
+		ctx->sched_gen++;
 		ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
+	}
 	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
 	if (ctx)
 		ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
@@ -1870,6 +1878,9 @@ static void ctx_sched_out(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (!ctx->is_active && is_active)
+		__get_cpu_var(perf_branch_stack_events) -= ctx->nr_branch_stack;
+
 	update_context_time(ctx);
 	update_cgrp_time_from_cpuctx(cpuctx);
 	if (!ctx->nr_active)
@@ -2059,6 +2070,10 @@ void __perf_event_task_sched_out(struct task_struct *task,
 {
 	int ctxn;
 
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(task, next, false);
+
 	for_each_task_context_nr(ctxn)
 		perf_event_context_sched_out(task, ctxn, next);
 
@@ -2166,6 +2181,9 @@ ctx_sched_in(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (ctx->is_active && !is_active)
+		__get_cpu_var(perf_branch_stack_events) += ctx->nr_branch_stack;
+
 	now = perf_clock();
 	ctx->timestamp = now;
 	perf_cgroup_set_timestamp(task, ctx);
@@ -2239,15 +2257,17 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
  * layer. It is invoked ONLY when there is at least one system-wide context
  * with at least one active event using taken branch sampling.
  */
-static void perf_branch_stack_sched_in(struct task_struct *prev,
-				       struct task_struct *task)
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in)
 {
 	struct perf_cpu_context *cpuctx;
+	struct perf_event_context *task_ctx;
 	struct pmu *pmu;
 	unsigned long flags;
 
 	/* no need to flush branch stack if not changing task */
-	if (prev == task)
+	if (task1 == task2)
 		return;
 
 	local_irq_save(flags);
@@ -2256,25 +2276,26 @@ static void perf_branch_stack_sched_in(struct task_struct *prev,
 
 	list_for_each_entry_rcu(pmu, &pmus, entry) {
 		cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+		task_ctx = cpuctx->task_ctx;
 
 		/*
 		 * check if the context has at least one
 		 * event using PERF_SAMPLE_BRANCH_STACK
 		 */
-		if (cpuctx->ctx.nr_branch_stack > 0
-		    && pmu->flush_branch_stack) {
-
+		if (pmu->branch_stack_sched &&
+		    (cpuctx->ctx.nr_branch_stack > 0 ||
+		     (task_ctx && task_ctx->nr_branch_stack > 0))) {
 			pmu = cpuctx->ctx.pmu;
 
-			perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_lock(cpuctx, task_ctx);
 
 			perf_pmu_disable(pmu);
 
-			pmu->flush_branch_stack();
+			pmu->branch_stack_sched(task_ctx, sched_in);
 
 			perf_pmu_enable(pmu);
 
-			perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_unlock(cpuctx, task_ctx);
 		}
 	}
 
@@ -2315,9 +2336,9 @@ void __perf_event_task_sched_in(struct task_struct *prev,
 	if (atomic_read(&__get_cpu_var(perf_cgroup_events)))
 		perf_cgroup_sched_in(prev, task);
 
-	/* check for system-wide branch_stack events */
-	if (atomic_read(&__get_cpu_var(perf_branch_stack_events)))
-		perf_branch_stack_sched_in(prev, task);
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(prev, task, true);
 }
 
 static u64 perf_calculate_period(struct perf_event *event, u64 nsec, u64 count)
@@ -2893,13 +2914,8 @@ static void free_event(struct perf_event *event)
 			static_key_slow_dec_deferred(&perf_sched_events);
 		}
 
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_dec_deferred(&perf_sched_events);
-			/* is system-wide event */
-			if (!(event->attach_state & PERF_ATTACH_TASK))
-				atomic_dec(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-		}
 	}
 
 	if (event->rb) {
@@ -6250,12 +6266,8 @@ done:
 				return ERR_PTR(err);
 			}
 		}
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_inc(&perf_sched_events.key);
-			if (!(event->attach_state & PERF_ATTACH_TASK))
-				atomic_inc(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-		}
 	}
 
 	return event;
-- 
1.7.11.7

[PATCH 5/7] perf, core: Pass perf_sample_data to perf_callchain()

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

New Intel CPU can record call chains by using existing last branch
record facility. perf_callchain_user() can make use of the call
chains recorded by hardware in case of there is no frame pointer.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/arm/kernel/perf_event.c     | 4 ++--
 arch/powerpc/perf/callchain.c    | 4 ++--
 arch/sparc/kernel/perf_event.c   | 4 ++--
 arch/x86/kernel/cpu/perf_event.c | 4 ++--
 include/linux/perf_event.h       | 3 ++-
 kernel/events/callchain.c        | 8 +++++---
 kernel/events/core.c             | 2 +-
 kernel/events/internal.h         | 3 ++-
 8 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c
index f9e8657..5b0c49f 100644
--- a/arch/arm/kernel/perf_event.c
+++ b/arch/arm/kernel/perf_event.c
@@ -566,8 +566,8 @@ user_backtrace(struct frame_tail __user *tail,
 	return buftail.fp - 1;
 }
 
-void
-perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
+void perf_callchain_user(struct perf_callchain_entry *entry,
+			 struct pt_regs *regs, struct perf_sample_data *data)
 {
 	struct frame_tail __user *tail;
 
diff --git a/arch/powerpc/perf/callchain.c b/arch/powerpc/perf/callchain.c
index 74d1e78..b379ebc 100644
--- a/arch/powerpc/perf/callchain.c
+++ b/arch/powerpc/perf/callchain.c
@@ -482,8 +482,8 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 	}
 }
 
-void
-perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
+void perf_callchain_user(struct perf_callchain_entry *entry,
+			 struct pt_regs *regs, struct perf_sample_data *data)
 {
 	if (current_is_64bit())
 		perf_callchain_user_64(entry, regs);
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c
index b5c38fa..cba0306 100644
--- a/arch/sparc/kernel/perf_event.c
+++ b/arch/sparc/kernel/perf_event.c
@@ -1785,8 +1785,8 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 	} while (entry->nr < PERF_MAX_STACK_DEPTH);
 }
 
-void
-perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
+void perf_callchain_user(struct perf_callchain_entry *entry,
+			 struct pt_regs *regs, struct perf_sample_data *data)
 {
 	perf_callchain_store(entry, regs->tpc);
 
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 8fd8894..19e7e99 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1973,8 +1973,8 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
 }
 #endif
 
-void
-perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
+void perf_callchain_user(struct perf_callchain_entry *entry,
+			 struct pt_regs *regs, struct perf_sample_data *data)
 {
 	struct stack_frame frame;
 	const void __user *fp;
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index f130c6c..fdf0019 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -706,7 +706,8 @@ extern void perf_event_fork(struct task_struct *tsk);
 /* Callchains */
 DECLARE_PER_CPU(struct perf_callchain_entry, perf_callchain_entry);
 
-extern void perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs);
+extern void perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs,
+				struct perf_sample_data *data);
 extern void perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs);
 
 static inline void perf_callchain_store(struct perf_callchain_entry *entry, u64 ip)
diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c
index c772061..bd7138a 100644
--- a/kernel/events/callchain.c
+++ b/kernel/events/callchain.c
@@ -30,7 +30,8 @@ __weak void perf_callchain_kernel(struct perf_callchain_entry *entry,
 }
 
 __weak void perf_callchain_user(struct perf_callchain_entry *entry,
-				struct pt_regs *regs)
+				struct pt_regs *regs,
+				struct perf_sample_data *data)
 {
 }
 
@@ -154,7 +155,8 @@ put_callchain_entry(int rctx)
 }
 
 struct perf_callchain_entry *
-perf_callchain(struct perf_event *event, struct pt_regs *regs)
+perf_callchain(struct perf_event *event, struct pt_regs *regs,
+	       struct perf_sample_data *data)
 {
 	int rctx;
 	struct perf_callchain_entry *entry;
@@ -195,7 +197,7 @@ perf_callchain(struct perf_event *event, struct pt_regs *regs)
 				goto exit_put;
 
 			perf_callchain_store(entry, PERF_CONTEXT_USER);
-			perf_callchain_user(entry, regs);
+			perf_callchain_user(entry, regs, data);
 		}
 	}
 
diff --git a/kernel/events/core.c b/kernel/events/core.c
index e1ea1d5..8a0dcb6 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4228,7 +4228,7 @@ void perf_prepare_sample(struct perf_event_header *header,
 	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
 		int size = 1;
 
-		data->callchain = perf_callchain(event, regs);
+		data->callchain = perf_callchain(event, regs, data);
 
 		if (data->callchain)
 			size += data->callchain->nr;
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index d56a64c..4ce75bf 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -126,7 +126,8 @@ DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
 
 /* Callchain handling */
 extern struct perf_callchain_entry *
-perf_callchain(struct perf_event *event, struct pt_regs *regs);
+perf_callchain(struct perf_event *event, struct pt_regs *regs,
+	       struct perf_sample_data *data);
 extern int get_callchain_buffers(void);
 extern void put_callchain_buffers(void);
 
-- 
1.7.11.7

[PATCH 6/7] perf, x86: Use LBR call stack to get user callchain

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

Try enabling the LBR call stack feature if event request recording
callchain. Try utilizing the LBR call stack to get user callchain
in case of there is no frame pointer.

This patch also adds a cpu pmu attribute to enable/disable this
feature.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.c           | 127 +++++++++++++++++++++--------
 arch/x86/kernel/cpu/perf_event.h           |   7 ++
 arch/x86/kernel/cpu/perf_event_intel.c     |  20 ++---
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |   3 +
 include/linux/perf_event.h                 |   6 ++
 kernel/events/core.c                       |  11 ++-
 6 files changed, 126 insertions(+), 48 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 19e7e99..71ace4f 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -393,36 +393,49 @@ int x86_pmu_hw_config(struct perf_event *event)
 
 		if (event->attr.precise_ip > precise)
 			return -EOPNOTSUPP;
+	}
+	/*
+	 * check that PEBS LBR correction does not conflict with
+	 * whatever the user is asking with attr->branch_sample_type
+	 */
+	if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format < 2) {
+		u64 *br_type = &event->attr.branch_sample_type;
+
+		if (has_branch_stack(event)) {
+			if (!precise_br_compat(event))
+				return -EOPNOTSUPP;
+
+			/* branch_sample_type is compatible */
+
+		} else {
+			/*
+			 * user did not specify  branch_sample_type
+			 *
+			 * For PEBS fixups, we capture all
+			 * the branches at the priv level of the
+			 * event.
+			 */
+			*br_type = PERF_SAMPLE_BRANCH_ANY;
+
+			if (!event->attr.exclude_user)
+				*br_type |= PERF_SAMPLE_BRANCH_USER;
+
+			if (!event->attr.exclude_kernel)
+				*br_type |= PERF_SAMPLE_BRANCH_KERNEL;
+		}
+	} else if ((event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) &&
+		   !has_branch_stack(event) &&
+		   x86_pmu.attr_lbr_callstack &&
+		   !event->attr.exclude_user &&
+		   (event->attach_state & PERF_ATTACH_TASK)) {
 		/*
-		 * check that PEBS LBR correction does not conflict with
-		 * whatever the user is asking with attr->branch_sample_type
+		 * user did not specify branch_sample_type,
+		 * try using the LBR call stack facility to
+		 * record call chains of user program.
 		 */
-		if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format < 2) {
-			u64 *br_type = &event->attr.branch_sample_type;
-
-			if (has_branch_stack(event)) {
-				if (!precise_br_compat(event))
-					return -EOPNOTSUPP;
-
-				/* branch_sample_type is compatible */
-
-			} else {
-				/*
-				 * user did not specify  branch_sample_type
-				 *
-				 * For PEBS fixups, we capture all
-				 * the branches at the priv level of the
-				 * event.
-				 */
-				*br_type = PERF_SAMPLE_BRANCH_ANY;
-
-				if (!event->attr.exclude_user)
-					*br_type |= PERF_SAMPLE_BRANCH_USER;
-
-				if (!event->attr.exclude_kernel)
-					*br_type |= PERF_SAMPLE_BRANCH_KERNEL;
-			}
-		}
+		event->attr.branch_sample_type =
+			PERF_SAMPLE_BRANCH_USER |
+			PERF_SAMPLE_BRANCH_CALL_STACK;
 	}
 
 	/*
@@ -1799,10 +1812,34 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
 	return count;
 }
 
+static ssize_t get_attr_lbr_callstack(struct device *cdev,
+				      struct device_attribute *attr, char *buf)
+{
+	return snprintf(buf, 40, "%d\n", x86_pmu.attr_lbr_callstack);
+}
+
+static ssize_t set_attr_lbr_callstack(struct device *cdev,
+				      struct device_attribute *attr,
+				      const char *buf, size_t count)
+{
+	unsigned long val = simple_strtoul(buf, NULL, 0);
+
+	if (x86_pmu.attr_lbr_callstack != !!val) {
+		if (val && !x86_pmu_has_lbr_callstack())
+			return -EOPNOTSUPP;
+		x86_pmu.attr_lbr_callstack = !!val;
+	}
+	return count;
+}
+
 static DEVICE_ATTR(rdpmc, S_IRUSR | S_IWUSR, get_attr_rdpmc, set_attr_rdpmc);
+static DEVICE_ATTR(lbr_callstack, S_IRUSR | S_IWUSR,
+		   get_attr_lbr_callstack, set_attr_lbr_callstack);
+
 
 static struct attribute *x86_pmu_attrs[] = {
 	&dev_attr_rdpmc.attr,
+	&dev_attr_lbr_callstack.attr,
 	NULL,
 };
 
@@ -1929,12 +1966,29 @@ static unsigned long get_segment_base(unsigned int segment)
 	return get_desc_base(desc + idx);
 }
 
+static inline void
+perf_callchain_lbr_callstack(struct perf_callchain_entry *entry,
+			     struct perf_sample_data *data)
+{
+	struct perf_branch_stack *br_stack = data->br_stack;
+
+	if (br_stack && br_stack->user_callstack &&
+	    x86_pmu.attr_lbr_callstack) {
+		int i = 0;
+		while (i < br_stack->nr && entry->nr < PERF_MAX_STACK_DEPTH) {
+			perf_callchain_store(entry, br_stack->entries[i].from);
+			i++;
+		}
+	}
+}
+
 #ifdef CONFIG_COMPAT
 
 #include <asm/compat.h>
 
 static inline int
-perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
+perf_callchain_user32(struct perf_callchain_entry *entry,
+		      struct pt_regs *regs, struct perf_sample_data *data)
 {
 	/* 32-bit process in 64-bit kernel. */
 	unsigned long ss_base, cs_base;
@@ -1963,11 +2017,16 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
 		perf_callchain_store(entry, cs_base + frame.return_address);
 		fp = compat_ptr(ss_base + frame.next_frame);
 	}
+
+	if (fp == compat_ptr(regs->bp))
+		perf_callchain_lbr_callstack(entry, data);
+
 	return 1;
 }
 #else
 static inline int
-perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
+perf_callchain_user32(struct perf_callchain_entry *entry,
+		      struct pt_regs *regs, struct perf_sample_data *data)
 {
     return 0;
 }
@@ -1997,12 +2056,12 @@ void perf_callchain_user(struct perf_callchain_entry *entry,
 	if (!current->mm)
 		return;
 
-	if (perf_callchain_user32(regs, entry))
+	if (perf_callchain_user32(entry, regs, data))
 		return;
 
 	while (entry->nr < PERF_MAX_STACK_DEPTH) {
 		unsigned long bytes;
-		frame.next_frame	     = NULL;
+		frame.next_frame = NULL;
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
@@ -2015,6 +2074,10 @@ void perf_callchain_user(struct perf_callchain_entry *entry,
 		perf_callchain_store(entry, frame.return_address);
 		fp = frame.next_frame;
 	}
+
+	/* try LBR callstack if there is no frame pointer */
+	if (fp == (void __user *)regs->bp)
+		perf_callchain_lbr_callstack(entry, data);
 }
 
 /*
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index a6547f3..2fe3e77 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -357,6 +357,7 @@ struct x86_pmu {
 	 * sysfs attrs
 	 */
 	int		attr_rdpmc;
+	int		attr_lbr_callstack;
 	struct attribute **format_attrs;
 
 	ssize_t		(*events_sysfs_show)(char *page, u64 config);
@@ -469,6 +470,12 @@ struct perf_pmu_events_attr {
 
 extern struct x86_pmu x86_pmu __read_mostly;
 
+static inline bool x86_pmu_has_lbr_callstack(void)
+{
+	return  x86_pmu.lbr_sel_map &&
+		x86_pmu.lbr_sel_map[PERF_SAMPLE_BRANCH_CALL_STACK] > 0;
+}
+
 DECLARE_PER_CPU(struct cpu_hw_events, cpu_hw_events);
 
 int x86_perf_event_set_period(struct perf_event *event);
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 61c5c17..8662360 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -820,15 +820,10 @@ static __initconst const u64 atom_hw_cache_event_ids
  },
 };
 
-static inline bool intel_pmu_needs_lbr_smpl(struct perf_event *event)
+static inline bool intel_pmu_needs_lbr_callstack(struct perf_event *event)
 {
-	/* user explicitly requested branch sampling */
-	if (has_branch_stack(event))
-		return true;
-
-	/* implicit branch sampling to correct PEBS skid */
-	if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1 &&
-	    x86_pmu.intel_cap.pebs_format < 2)
+	if ((event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) &&
+	    (event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_CALL_STACK))
 		return true;
 
 	return false;
@@ -992,7 +987,7 @@ static void intel_pmu_disable_event(struct perf_event *event)
 	 * must disable before any actual event
 	 * because any event may be combined with LBR
 	 */
-	if (intel_pmu_needs_lbr_smpl(event))
+	if (needs_branch_stack(event))
 		intel_pmu_lbr_disable(event);
 
 	if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) {
@@ -1053,7 +1048,7 @@ static void intel_pmu_enable_event(struct perf_event *event)
 	 * must enabled before any actual event
 	 * because any event may be combined with LBR
 	 */
-	if (intel_pmu_needs_lbr_smpl(event))
+	if (needs_branch_stack(event))
 		intel_pmu_lbr_enable(event);
 
 	if (event->attr.exclude_host)
@@ -1185,7 +1180,8 @@ again:
 
 		perf_sample_data_init(&data, 0, event->hw.last_period);
 
-		if (has_branch_stack(event))
+		if (has_branch_stack(event) ||
+		    (event->ctx->task && intel_pmu_needs_lbr_callstack(event)))
 			data.br_stack = &cpuc->lbr_stack;
 
 		if (perf_event_overflow(event, &data, regs))
@@ -1510,7 +1506,7 @@ static int intel_pmu_hw_config(struct perf_event *event)
 	if (event->attr.precise_ip && x86_pmu.pebs_aliases)
 		x86_pmu.pebs_aliases(event);
 
-	if (intel_pmu_needs_lbr_smpl(event)) {
+	if (needs_branch_stack(event)) {
 		ret = intel_pmu_setup_lbr_filter(event);
 		if (ret)
 			return ret;
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index a1fb108..910200b 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -691,6 +691,8 @@ intel_pmu_lbr_filter(struct cpu_hw_events *cpuc)
 	int i, j, type;
 	bool compress = false;
 
+	cpuc->lbr_stack.user_callstack = branch_user_callstack(br_sel);
+
 	/* if sampling all branches, then nothing to filter */
 	if ((br_sel & X86_BR_ALL) == X86_BR_ALL)
 		return;
@@ -843,6 +845,7 @@ void intel_pmu_lbr_init_hsw(void)
 
 	x86_pmu.lbr_sel_mask = LBR_SEL_MASK;
 	x86_pmu.lbr_sel_map  = hsw_lbr_sel_map;
+	x86_pmu.attr_lbr_callstack = 1;
 
 	pr_cont("16-deep LBR, ");
 }
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index fdf0019..71c5cda 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -98,6 +98,7 @@ struct perf_branch_entry {
  * recent branch.
  */
 struct perf_branch_stack {
+	unsigned			user_callstack:1;
 	__u64				nr;
 	struct perf_branch_entry	entries[0];
 };
@@ -757,6 +758,11 @@ static inline bool has_branch_stack(struct perf_event *event)
 	return event->attr.sample_type & PERF_SAMPLE_BRANCH_STACK;
 }
 
+static inline bool needs_branch_stack(struct perf_event *event)
+{
+	return event->attr.branch_sample_type != 0;
+}
+
 extern int perf_output_begin(struct perf_output_handle *handle,
 			     struct perf_event *event, unsigned int size);
 extern void perf_output_end(struct perf_output_handle *handle);
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 8a0dcb6..4633191 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -899,7 +899,7 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx)
 	if (is_cgroup_event(event))
 		ctx->nr_cgroups++;
 
-	if (has_branch_stack(event))
+	if (needs_branch_stack(event))
 		ctx->nr_branch_stack++;
 
 	list_add_rcu(&event->event_entry, &ctx->event_list);
@@ -1047,7 +1047,7 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 			cpuctx->cgrp = NULL;
 	}
 
-	if (has_branch_stack(event)) {
+	if (needs_branch_stack(event)) {
 		if (ctx->is_active)
 			__get_cpu_var(perf_branch_stack_events)--;
 		ctx->nr_branch_stack--;
@@ -2914,7 +2914,7 @@ static void free_event(struct perf_event *event)
 			static_key_slow_dec_deferred(&perf_sched_events);
 		}
 
-		if (has_branch_stack(event))
+		if (needs_branch_stack(event))
 			static_key_slow_dec_deferred(&perf_sched_events);
 	}
 
@@ -6234,6 +6234,9 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	if (attr->inherit && (attr->read_format & PERF_FORMAT_GROUP))
 		goto done;
 
+	if (!has_branch_stack(event))
+		event->attr.branch_sample_type = 0;
+
 	pmu = perf_init_event(event);
 
 done:
@@ -6266,7 +6269,7 @@ done:
 				return ERR_PTR(err);
 			}
 		}
-		if (has_branch_stack(event))
+		if (needs_branch_stack(event))
 			static_key_slow_inc(&perf_sched_events.key);
 	}
 
-- 
1.7.11.7

[PATCH 7/7] perf, x86: Discard zero length call entries in LBR call stack

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

"Zero length call" uses the attribute of the call instruction to push
the immediate instruction pointer on to the stack and then pops off
that address into a register. This is accomplished without any matching
return instruction. It confuses the hardware and make the recorded call
stack incorrect. Try fixing the call stack by discarding zero length
call entries.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 910200b..0546689 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -86,7 +86,8 @@ enum {
 	X86_BR_ABORT		= 1 << 12,/* transaction abort */
 	X86_BR_INTX		= 1 << 13,/* in transaction */
 	X86_BR_NOTX		= 1 << 14,/* not in transaction */
-	X86_BR_CALL_STACK	= 1 << 15,/* call stack */
+	X86_BR_ZERO_CALL	= 1 << 15,/* zero length call */
+	X86_BR_CALL_STACK	= 1 << 16,/* call stack */
 };
 
 #define X86_BR_PLM (X86_BR_USER | X86_BR_KERNEL)
@@ -103,13 +104,15 @@ enum {
 	 X86_BR_JMP	 |\
 	 X86_BR_IRQ	 |\
 	 X86_BR_ABORT	 |\
-	 X86_BR_IND_CALL)
+	 X86_BR_IND_CALL |\
+	 X86_BR_ZERO_CALL)
 
 #define X86_BR_ALL (X86_BR_PLM | X86_BR_ANY)
 
 #define X86_BR_ANY_CALL		 \
 	(X86_BR_CALL		|\
 	 X86_BR_IND_CALL	|\
+	 X86_BR_ZERO_CALL	|\
 	 X86_BR_SYSCALL		|\
 	 X86_BR_IRQ		|\
 	 X86_BR_INT)
@@ -625,6 +628,12 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 		ret = X86_BR_INT;
 		break;
 	case 0xe8: /* call near rel */
+		insn_get_immediate(&insn);
+		if (insn.immediate1.value == 0) {
+			/* zero length call */
+			ret = X86_BR_ZERO_CALL;
+			break;
+		}
 	case 0x9a: /* call far absolute */
 		ret = X86_BR_CALL;
 		break;
-- 
1.7.11.7

[PATCH V3 0/7] perf, x86: Haswell LBR call stack support

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

Haswell has a new feature that utilizes the existing Last Branch Record
facility to record call chains. When the feature is enabled, function
call will be collected as normal, but as return instructions are executed
the last captured branch record is popped from the on-chip LBR registers.
The LBR call stack facility can help perf to get call chains of progam 
without frame pointer. When perf tool requests PERF_SAMPLE_CALLCHAIN +
PERF_SAMPLE_BRANCH_USER, this feature is dynamically enabled by default.
This feature can be disabled/enabled through an attribute file in the cpu
pmu sysfs directory.

The LBR call stack has following known limitations
 1. Zero length calls are not filtered out by hardware
 2. Exception handing such as setjmp/longjmp will have calls/returns not
    match
 3. Pushing different return address onto the stack will have calls/returns
    not match

These patches are based upon Andi's linux-misc hsw/pmu5

Available from
    https://github.com/ukernel/linux.git hws/pmu5

Regards
Yan, Zheng

---
Changes since v1
 - not expose PERF_SAMPLE_BRANCH_CALL_STACK to user space
 - save/restore LBR stack on context switch for all sampling branch modes
 - reduce lbr_sel_map size

Changes since v2
 - only enable LBR call stack when user requests sampling user callchain

[PATCH 4/7] perf, x86: Save/resotre LBR stack during context switch

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

When the LBR call stack is enabled, it is necessary to save/restore
the stack on context switch. The solution is saving/restoring the
stack to/from task's perf event context. If task has no perf event
context, just flush the stack on context switch.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.c           |  18 +++--
 arch/x86/kernel/cpu/perf_event.h           |  13 +++-
 arch/x86/kernel/cpu/perf_event_intel.c     |  13 ++--
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 108 ++++++++++++++++++++++++++---
 include/linux/perf_event.h                 |   6 +-
 kernel/events/core.c                       |  64 ++++++++++-------
 6 files changed, 168 insertions(+), 54 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 55ab79d..fc69269 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1737,6 +1737,13 @@ static int x86_pmu_event_idx(struct perf_event *event)
 	return idx + 1;
 }
 
+static void x86_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					bool sched_in)
+{
+	if (x86_pmu.branch_stack_sched)
+		x86_pmu.branch_stack_sched(ctx, sched_in);
+}
+
 static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 {
 	struct perf_event_context *ctx;
@@ -1745,6 +1752,9 @@ static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 	if (!ctx)
 		return ERR_PTR(-ENOMEM);
 
+	if (parent_ctx)
+		intel_pmu_lbr_init_context(ctx, parent_ctx);
+
 	return ctx;
 }
 
@@ -1802,12 +1812,6 @@ static const struct attribute_group *x86_pmu_attr_groups[] = {
 	NULL,
 };
 
-static void x86_pmu_flush_branch_stack(void)
-{
-	if (x86_pmu.flush_branch_stack)
-		x86_pmu.flush_branch_stack();
-}
-
 void perf_check_microcode(void)
 {
 	if (x86_pmu.check_microcode)
@@ -1834,7 +1838,7 @@ static struct pmu pmu = {
 	.commit_txn		= x86_pmu_commit_txn,
 
 	.event_idx		= x86_pmu_event_idx,
-	.flush_branch_stack	= x86_pmu_flush_branch_stack,
+	.branch_stack_sched     = x86_pmu_branch_stack_sched,
 	.event_context_alloc	= x86_pmu_event_context_alloc,
 };
 
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index a8c1232..0a7527d9 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -377,7 +377,6 @@ struct x86_pmu {
 	void		(*cpu_dead)(int cpu);
 
 	void		(*check_microcode)(void);
-	void		(*flush_branch_stack)(void);
 
 	/*
 	 * Intel Arch Perfmon v2+
@@ -408,6 +407,8 @@ struct x86_pmu {
 	u64		lbr_sel_mask;		   /* LBR_SELECT valid bits */
 	const int	*lbr_sel_map;		   /* lbr_select mappings */
 	bool		lbr_double_abort;	   /* duplicated abort records */
+	void		(*branch_stack_sched)(struct perf_event_context *ctx,
+					      bool sched_in);
 
 	/*
 	 * Extra registers for events
@@ -430,6 +431,12 @@ enum {
 
 struct x86_perf_event_context {
 	struct perf_event_context ctx;
+
+	u64 lbr_from[MAX_LBR_ENTRIES];
+	u64 lbr_to[MAX_LBR_ENTRIES];
+	u64 lbr_stack_gen;
+	int lbr_callstack_users;
+	bool lbr_stack_saved;
 };
 
 #define x86_add_quirk(func_)						\
@@ -627,8 +634,12 @@ void intel_pmu_pebs_disable_all(void);
 
 void intel_ds_init(void);
 
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx);
 void intel_pmu_lbr_reset(void);
 
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in);
+
 void intel_pmu_lbr_enable(struct perf_event *event);
 
 void intel_pmu_lbr_disable(struct perf_event *event);
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index b3566bb..716f7d9 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1814,16 +1814,11 @@ static void intel_pmu_cpu_dying(int cpu)
 	fini_debug_store_on_cpu(cpu);
 }
 
-static void intel_pmu_flush_branch_stack(void)
+static void intel_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					 bool sched_in)
 {
-	/*
-	 * Intel LBR does not tag entries with the
-	 * PID of the current task, then we need to
-	 * flush it on ctxsw
-	 * For now, we simply reset it
-	 */
 	if (x86_pmu.lbr_nr)
-		intel_pmu_lbr_reset();
+		intel_pmu_lbr_sched(ctx, sched_in);
 }
 
 PMU_FORMAT_ATTR(offcore_rsp, "config1:0-63");
@@ -1889,7 +1884,7 @@ static __initconst const struct x86_pmu intel_pmu = {
 	.cpu_starting		= intel_pmu_cpu_starting,
 	.cpu_dying		= intel_pmu_cpu_dying,
 	.guest_get_msrs		= intel_guest_get_msrs,
-	.flush_branch_stack	= intel_pmu_flush_branch_stack,
+	.branch_stack_sched	= intel_pmu_branch_stack_sched,
 };
 
 static __init void intel_clovertown_quirk(void)
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index ce9e50d17..392b231 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -181,6 +181,13 @@ void intel_pmu_lbr_reset(void)
 		intel_pmu_lbr_reset_32();
 	else
 		intel_pmu_lbr_reset_64();
+
+	wrmsrl(x86_pmu.lbr_tos, 0);
+}
+
+static inline bool branch_user_callstack(unsigned br_sel)
+{
+	return (br_sel & X86_BR_USER) && (br_sel & X86_BR_CALL_STACK);
 }
 
 void intel_pmu_lbr_enable(struct perf_event *event)
@@ -190,17 +197,23 @@ void intel_pmu_lbr_enable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
-	/*
-	 * Reset the LBR stack if we changed task context to
-	 * avoid data leaks.
-	 */
-	if (event->ctx->task && cpuc->lbr_context != event->ctx) {
-		intel_pmu_lbr_reset();
-		cpuc->lbr_context = event->ctx;
-	}
 	cpuc->br_sel = event->hw.branch_reg.reg;
-
 	cpuc->lbr_users++;
+
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+		/*
+		 * Reset the LBR stack if the call stack is not
+		 * continuous enabled
+		 */
+		if (task_ctx->lbr_callstack_users == 0 &&
+		    task_ctx->lbr_stack_gen + 1 < event->ctx->sched_gen)
+			intel_pmu_lbr_reset();
+
+		task_ctx->lbr_callstack_users++;
+		task_ctx->lbr_stack_gen = event->ctx->sched_gen;
+	}
 }
 
 void intel_pmu_lbr_disable(struct perf_event *event)
@@ -210,6 +223,13 @@ void intel_pmu_lbr_disable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+
+		task_ctx->lbr_callstack_users--;
+	}
+
 	cpuc->lbr_users--;
 	WARN_ON_ONCE(cpuc->lbr_users < 0);
 
@@ -347,6 +367,76 @@ void intel_pmu_lbr_read(void)
 	intel_pmu_lbr_filter(cpuc);
 }
 
+static void __intel_pmu_lbr_restore(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		wrmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		wrmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_saved = false;
+}
+
+static void __intel_pmu_lbr_save(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		rdmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		rdmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_gen = task_ctx->ctx.sched_gen;
+	task_ctx->lbr_stack_saved = true;
+}
+
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx)
+{
+	struct x86_perf_event_context *task_ctx, *parent_task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	task_ctx = (struct x86_perf_event_context *)child_ctx;
+	parent_task_ctx = (struct x86_perf_event_context *)parent_ctx;
+
+	if (parent_task_ctx->lbr_callstack_users)
+		__intel_pmu_lbr_save(task_ctx);
+	else
+		task_ctx->lbr_stack_saved = false;
+}
+
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in)
+{
+	struct x86_perf_event_context *task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	if (!ctx) {
+		if (sched_in)
+			intel_pmu_lbr_reset();
+		return;
+	}
+
+	task_ctx = (struct x86_perf_event_context *)ctx;
+	if (sched_in) {
+		if (!task_ctx->lbr_stack_saved)
+			intel_pmu_lbr_reset();
+		else
+			__intel_pmu_lbr_restore(task_ctx);
+	} else {
+		__intel_pmu_lbr_save(task_ctx);
+	}
+}
+
 /*
  * SW filter is used:
  * - in case there is no HW filter
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 1a98087..f4a1c95 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -266,9 +266,10 @@ struct pmu {
 	int (*event_idx)		(struct perf_event *event); /*optional */
 
 	/*
-	 * flush branch stack on context-switches (needed in cpu-wide mode)
+	 * Save/restore LBR stack on context-switches
 	 */
-	void (*flush_branch_stack)	(void);
+	void (*branch_stack_sched)	(struct perf_event_context *ctx,
+					 bool sched_in);
 
 	/*
 	 * Allocate PMU special perf event context
@@ -506,6 +507,7 @@ struct perf_event_context {
 	struct perf_event_context	*parent_ctx;
 	u64				parent_gen;
 	u64				generation;
+	u64				sched_gen;
 	int				pin_count;
 	int				nr_cgroups;	 /* cgroup evts */
 	int				nr_branch_stack; /* branch_stack evt */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 0110046..ba1f959 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -138,7 +138,7 @@ enum event_type_t {
  */
 struct static_key_deferred perf_sched_events __read_mostly;
 static DEFINE_PER_CPU(atomic_t, perf_cgroup_events);
-static DEFINE_PER_CPU(atomic_t, perf_branch_stack_events);
+static DEFINE_PER_CPU(int, perf_branch_stack_events);
 
 static atomic_t nr_mmap_events __read_mostly;
 static atomic_t nr_comm_events __read_mostly;
@@ -190,6 +190,9 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
 static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
 			     enum event_type_t event_type,
 			     struct task_struct *task);
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in);
 
 static void update_context_time(struct perf_event_context *ctx);
 static u64 perf_event_time(struct perf_event *event);
@@ -1044,8 +1047,11 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 			cpuctx->cgrp = NULL;
 	}
 
-	if (has_branch_stack(event))
+	if (has_branch_stack(event)) {
+		if (ctx->is_active)
+			__get_cpu_var(perf_branch_stack_events)--;
 		ctx->nr_branch_stack--;
+	}
 
 	ctx->nr_events--;
 	if (event->attr.inherit_stat)
@@ -1566,8 +1572,10 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx,
 				struct task_struct *task)
 {
 	cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task);
-	if (ctx)
+	if (ctx) {
+		ctx->sched_gen++;
 		ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
+	}
 	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
 	if (ctx)
 		ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
@@ -1870,6 +1878,9 @@ static void ctx_sched_out(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (!ctx->is_active && is_active)
+		__get_cpu_var(perf_branch_stack_events) -= ctx->nr_branch_stack;
+
 	update_context_time(ctx);
 	update_cgrp_time_from_cpuctx(cpuctx);
 	if (!ctx->nr_active)
@@ -2059,6 +2070,10 @@ void __perf_event_task_sched_out(struct task_struct *task,
 {
 	int ctxn;
 
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(task, next, false);
+
 	for_each_task_context_nr(ctxn)
 		perf_event_context_sched_out(task, ctxn, next);
 
@@ -2166,6 +2181,9 @@ ctx_sched_in(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (ctx->is_active && !is_active)
+		__get_cpu_var(perf_branch_stack_events) += ctx->nr_branch_stack;
+
 	now = perf_clock();
 	ctx->timestamp = now;
 	perf_cgroup_set_timestamp(task, ctx);
@@ -2239,15 +2257,17 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
  * layer. It is invoked ONLY when there is at least one system-wide context
  * with at least one active event using taken branch sampling.
  */
-static void perf_branch_stack_sched_in(struct task_struct *prev,
-				       struct task_struct *task)
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in)
 {
 	struct perf_cpu_context *cpuctx;
+	struct perf_event_context *task_ctx;
 	struct pmu *pmu;
 	unsigned long flags;
 
 	/* no need to flush branch stack if not changing task */
-	if (prev == task)
+	if (task1 == task2)
 		return;
 
 	local_irq_save(flags);
@@ -2256,25 +2276,26 @@ static void perf_branch_stack_sched_in(struct task_struct *prev,
 
 	list_for_each_entry_rcu(pmu, &pmus, entry) {
 		cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+		task_ctx = cpuctx->task_ctx;
 
 		/*
 		 * check if the context has at least one
 		 * event using PERF_SAMPLE_BRANCH_STACK
 		 */
-		if (cpuctx->ctx.nr_branch_stack > 0
-		    && pmu->flush_branch_stack) {
-
+		if (pmu->branch_stack_sched &&
+		    (cpuctx->ctx.nr_branch_stack > 0 ||
+		     (task_ctx && task_ctx->nr_branch_stack > 0))) {
 			pmu = cpuctx->ctx.pmu;
 
-			perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_lock(cpuctx, task_ctx);
 
 			perf_pmu_disable(pmu);
 
-			pmu->flush_branch_stack();
+			pmu->branch_stack_sched(task_ctx, sched_in);
 
 			perf_pmu_enable(pmu);
 
-			perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_unlock(cpuctx, task_ctx);
 		}
 	}
 
@@ -2315,9 +2336,9 @@ void __perf_event_task_sched_in(struct task_struct *prev,
 	if (atomic_read(&__get_cpu_var(perf_cgroup_events)))
 		perf_cgroup_sched_in(prev, task);
 
-	/* check for system-wide branch_stack events */
-	if (atomic_read(&__get_cpu_var(perf_branch_stack_events)))
-		perf_branch_stack_sched_in(prev, task);
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(prev, task, true);
 }
 
 static u64 perf_calculate_period(struct perf_event *event, u64 nsec, u64 count)
@@ -2893,13 +2914,8 @@ static void free_event(struct perf_event *event)
 			static_key_slow_dec_deferred(&perf_sched_events);
 		}
 
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_dec_deferred(&perf_sched_events);
-			/* is system-wide event */
-			if (!(event->attach_state & PERF_ATTACH_TASK))
-				atomic_dec(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-		}
 	}
 
 	if (event->rb) {
@@ -6253,12 +6269,8 @@ done:
 				return ERR_PTR(err);
 			}
 		}
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_inc(&perf_sched_events.key);
-			if (!(event->attach_state & PERF_ATTACH_TASK))
-				atomic_inc(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-		}
 	}
 
 	return event;
-- 
1.7.11.7

[PATCH 0/7] perf, x86: Haswell LBR call stack support

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

Haswell has a new feature that utilizes the existing Last Branch Record
facility to record call chains. When the feature is enabled, function
call will be collected as normal, but as return instructions are executed
the last captured branch record is popped from the on-chip LBR registers.
The LBR call stack facility can help perf to get call chains of progam
without frame pointer. When perf tool requests PERF_SAMPLE_CALLCHAIN +
PERF_SAMPLE_BRANCH_USER, this feature is dynamically enabled by default.
This feature can be disabled/enabled through an attribute file in the cpu
pmu sysfs directory.

The LBR call stack has following known limitations
 1. Zero length calls are not filtered out by hardware
 2. Exception handing such as setjmp/longjmp will have calls/returns not
    match
 3. Pushing different return address onto the stack will have calls/returns
    not match

These patches are based upon tip/perf/core

Regards
Yan, Zheng

[PATCH 4/7] perf, x86: Save/resotre LBR stack during context switch

[Yan, Zheng]

From: "Yan, Zheng" <zheng.z.yan@intel.com>

When the LBR call stack is enabled, it is necessary to save/restore
the stack on context switch. The solution is saving/restoring the
stack to/from task's perf event context. If task has no perf event
context, just flush the stack on context switch.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
 arch/x86/kernel/cpu/perf_event.c           |  18 +++--
 arch/x86/kernel/cpu/perf_event.h           |  13 +++-
 arch/x86/kernel/cpu/perf_event_intel.c     |  13 ++--
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 108 ++++++++++++++++++++++++++---
 include/linux/perf_event.h                 |   6 +-
 kernel/events/core.c                       |  65 +++++++++--------
 6 files changed, 168 insertions(+), 55 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index b2eada9..3843f80 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1768,6 +1768,13 @@ static int x86_pmu_event_idx(struct perf_event *event)
 	return idx + 1;
 }
 
+static void x86_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					bool sched_in)
+{
+	if (x86_pmu.branch_stack_sched)
+		x86_pmu.branch_stack_sched(ctx, sched_in);
+}
+
 static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 {
 	struct perf_event_context *ctx;
@@ -1776,6 +1783,9 @@ static void *x86_pmu_event_context_alloc(struct perf_event_context *parent_ctx)
 	if (!ctx)
 		return ERR_PTR(-ENOMEM);
 
+	if (parent_ctx)
+		intel_pmu_lbr_init_context(ctx, parent_ctx);
+
 	return ctx;
 }
 
@@ -1833,12 +1843,6 @@ static const struct attribute_group *x86_pmu_attr_groups[] = {
 	NULL,
 };
 
-static void x86_pmu_flush_branch_stack(void)
-{
-	if (x86_pmu.flush_branch_stack)
-		x86_pmu.flush_branch_stack();
-}
-
 void perf_check_microcode(void)
 {
 	if (x86_pmu.check_microcode)
@@ -1865,7 +1869,7 @@ static struct pmu pmu = {
 	.commit_txn		= x86_pmu_commit_txn,
 
 	.event_idx		= x86_pmu_event_idx,
-	.flush_branch_stack	= x86_pmu_flush_branch_stack,
+	.branch_stack_sched     = x86_pmu_branch_stack_sched,
 	.event_context_alloc	= x86_pmu_event_context_alloc,
 };
 
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 08469de..0116970 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -405,7 +405,6 @@ struct x86_pmu {
 	void		(*cpu_dead)(int cpu);
 
 	void		(*check_microcode)(void);
-	void		(*flush_branch_stack)(void);
 
 	/*
 	 * Intel Arch Perfmon v2+
@@ -434,6 +433,8 @@ struct x86_pmu {
 	int		lbr_nr;			   /* hardware stack size */
 	u64		lbr_sel_mask;		   /* LBR_SELECT valid bits */
 	const int	*lbr_sel_map;		   /* lbr_select mappings */
+	void		(*branch_stack_sched)(struct perf_event_context *ctx,
+					      bool sched_in);
 
 	/*
 	 * Extra registers for events
@@ -456,6 +457,12 @@ enum {
 
 struct x86_perf_event_context {
 	struct perf_event_context ctx;
+
+	u64 lbr_from[MAX_LBR_ENTRIES];
+	u64 lbr_to[MAX_LBR_ENTRIES];
+	u64 lbr_stack_gen;
+	int lbr_callstack_users;
+	bool lbr_stack_saved;
 };
 
 #define x86_add_quirk(func_)						\
@@ -668,8 +675,12 @@ void intel_pmu_pebs_disable_all(void);
 
 void intel_ds_init(void);
 
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx);
 void intel_pmu_lbr_reset(void);
 
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in);
+
 void intel_pmu_lbr_enable(struct perf_event *event);
 
 void intel_pmu_lbr_disable(struct perf_event *event);
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 3e92a68..f59b46e 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1851,16 +1851,11 @@ static void intel_pmu_cpu_dying(int cpu)
 	fini_debug_store_on_cpu(cpu);
 }
 
-static void intel_pmu_flush_branch_stack(void)
+static void intel_pmu_branch_stack_sched(struct perf_event_context *ctx,
+					 bool sched_in)
 {
-	/*
-	 * Intel LBR does not tag entries with the
-	 * PID of the current task, then we need to
-	 * flush it on ctxsw
-	 * For now, we simply reset it
-	 */
 	if (x86_pmu.lbr_nr)
-		intel_pmu_lbr_reset();
+		intel_pmu_lbr_sched(ctx, sched_in);
 }
 
 PMU_FORMAT_ATTR(offcore_rsp, "config1:0-63");
@@ -1914,7 +1909,7 @@ static __initconst const struct x86_pmu intel_pmu = {
 	.cpu_starting		= intel_pmu_cpu_starting,
 	.cpu_dying		= intel_pmu_cpu_dying,
 	.guest_get_msrs		= intel_guest_get_msrs,
-	.flush_branch_stack	= intel_pmu_flush_branch_stack,
+	.branch_stack_sched	= intel_pmu_branch_stack_sched,
 };
 
 static __init void intel_clovertown_quirk(void)
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 2136320..43b16b4 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -181,6 +181,13 @@ void intel_pmu_lbr_reset(void)
 		intel_pmu_lbr_reset_32();
 	else
 		intel_pmu_lbr_reset_64();
+
+	wrmsrl(x86_pmu.lbr_tos, 0);
+}
+
+static inline bool branch_user_callstack(unsigned br_sel)
+{
+	return (br_sel & X86_BR_USER) && (br_sel & X86_BR_CALL_STACK);
 }
 
 void intel_pmu_lbr_enable(struct perf_event *event)
@@ -190,17 +197,23 @@ void intel_pmu_lbr_enable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
-	/*
-	 * Reset the LBR stack if we changed task context to
-	 * avoid data leaks.
-	 */
-	if (event->ctx->task && cpuc->lbr_context != event->ctx) {
-		intel_pmu_lbr_reset();
-		cpuc->lbr_context = event->ctx;
-	}
 	cpuc->br_sel = event->hw.branch_reg.reg;
-
 	cpuc->lbr_users++;
+
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+		/*
+		 * Reset the LBR stack if the call stack is not
+		 * continuous enabled
+		 */
+		if (task_ctx->lbr_callstack_users == 0 &&
+		    task_ctx->lbr_stack_gen + 1 < event->ctx->sched_gen)
+			intel_pmu_lbr_reset();
+
+		task_ctx->lbr_callstack_users++;
+		task_ctx->lbr_stack_gen = event->ctx->sched_gen;
+	}
 }
 
 void intel_pmu_lbr_disable(struct perf_event *event)
@@ -210,6 +223,13 @@ void intel_pmu_lbr_disable(struct perf_event *event)
 	if (!x86_pmu.lbr_nr)
 		return;
 
+	if (event->ctx->task &&
+	    branch_user_callstack(event->hw.branch_reg.reg)) {
+		struct x86_perf_event_context *task_ctx = (void *)event->ctx;
+
+		task_ctx->lbr_callstack_users--;
+	}
+
 	cpuc->lbr_users--;
 	WARN_ON_ONCE(cpuc->lbr_users < 0);
 
@@ -334,6 +354,76 @@ void intel_pmu_lbr_read(void)
 	intel_pmu_lbr_filter(cpuc);
 }
 
+static void __intel_pmu_lbr_restore(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		wrmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		wrmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_saved = false;
+}
+
+static void __intel_pmu_lbr_save(struct x86_perf_event_context *task_ctx)
+{
+	int i;
+	unsigned lbr_idx, mask = x86_pmu.lbr_nr - 1;
+	u64 tos = intel_pmu_lbr_tos();
+
+	for (i = 0; i < x86_pmu.lbr_nr; i++) {
+		lbr_idx = (tos - i) & mask;
+		rdmsrl(x86_pmu.lbr_from + lbr_idx, task_ctx->lbr_from[i]);
+		rdmsrl(x86_pmu.lbr_to + lbr_idx, task_ctx->lbr_to[i]);
+	}
+	task_ctx->lbr_stack_gen = task_ctx->ctx.sched_gen;
+	task_ctx->lbr_stack_saved = true;
+}
+
+void intel_pmu_lbr_init_context(struct perf_event_context *child_ctx,
+				struct perf_event_context *parent_ctx)
+{
+	struct x86_perf_event_context *task_ctx, *parent_task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	task_ctx = (struct x86_perf_event_context *)child_ctx;
+	parent_task_ctx = (struct x86_perf_event_context *)parent_ctx;
+
+	if (parent_task_ctx->lbr_callstack_users)
+		__intel_pmu_lbr_save(task_ctx);
+	else
+		task_ctx->lbr_stack_saved = false;
+}
+
+void intel_pmu_lbr_sched(struct perf_event_context *ctx, bool sched_in)
+{
+	struct x86_perf_event_context *task_ctx;
+
+	if (!x86_pmu.lbr_nr)
+		return;
+
+	if (!ctx) {
+		if (sched_in)
+			intel_pmu_lbr_reset();
+		return;
+	}
+
+	task_ctx = (struct x86_perf_event_context *)ctx;
+	if (sched_in) {
+		if (!task_ctx->lbr_stack_saved)
+			intel_pmu_lbr_reset();
+		else
+			__intel_pmu_lbr_restore(task_ctx);
+	} else {
+		__intel_pmu_lbr_save(task_ctx);
+	}
+}
+
 /*
  * SW filter is used:
  * - in case there is no HW filter
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index f6d1d59..b3e4faf 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -271,9 +271,10 @@ struct pmu {
 	int (*event_idx)		(struct perf_event *event); /*optional */
 
 	/*
-	 * flush branch stack on context-switches (needed in cpu-wide mode)
+	 * Save/restore LBR stack on context-switches
 	 */
-	void (*flush_branch_stack)	(void);
+	void (*branch_stack_sched)	(struct perf_event_context *ctx,
+					 bool sched_in);
 
 	/*
 	 * Allocate PMU special perf event context
@@ -495,6 +496,7 @@ struct perf_event_context {
 	struct perf_event_context	*parent_ctx;
 	u64				parent_gen;
 	u64				generation;
+	u64				sched_gen;
 	int				pin_count;
 	int				nr_cgroups;	 /* cgroup evts */
 	int				nr_branch_stack; /* branch_stack evt */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 3aececc..1101ce8 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -140,7 +140,7 @@ enum event_type_t {
  */
 struct static_key_deferred perf_sched_events __read_mostly;
 static DEFINE_PER_CPU(atomic_t, perf_cgroup_events);
-static DEFINE_PER_CPU(atomic_t, perf_branch_stack_events);
+static DEFINE_PER_CPU(int, perf_branch_stack_events);
 
 static atomic_t nr_mmap_events __read_mostly;
 static atomic_t nr_comm_events __read_mostly;
@@ -278,6 +278,9 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
 static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
 			     enum event_type_t event_type,
 			     struct task_struct *task);
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in);
 
 static void update_context_time(struct perf_event_context *ctx);
 static u64 perf_event_time(struct perf_event *event);
@@ -1271,8 +1274,11 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 			cpuctx->cgrp = NULL;
 	}
 
-	if (has_branch_stack(event))
+	if (has_branch_stack(event)) {
+		if (ctx->is_active)
+			__get_cpu_var(perf_branch_stack_events)--;
 		ctx->nr_branch_stack--;
+	}
 
 	ctx->nr_events--;
 	if (event->attr.inherit_stat)
@@ -1796,8 +1802,10 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx,
 				struct task_struct *task)
 {
 	cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task);
-	if (ctx)
+	if (ctx) {
+		ctx->sched_gen++;
 		ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
+	}
 	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
 	if (ctx)
 		ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
@@ -2102,6 +2110,9 @@ static void ctx_sched_out(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (!ctx->is_active && is_active)
+		__get_cpu_var(perf_branch_stack_events) -= ctx->nr_branch_stack;
+
 	update_context_time(ctx);
 	update_cgrp_time_from_cpuctx(cpuctx);
 	if (!ctx->nr_active)
@@ -2291,6 +2302,10 @@ void __perf_event_task_sched_out(struct task_struct *task,
 {
 	int ctxn;
 
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(task, next, false);
+
 	for_each_task_context_nr(ctxn)
 		perf_event_context_sched_out(task, ctxn, next);
 
@@ -2398,6 +2413,9 @@ ctx_sched_in(struct perf_event_context *ctx,
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (ctx->is_active && !is_active)
+		__get_cpu_var(perf_branch_stack_events) += ctx->nr_branch_stack;
+
 	now = perf_clock();
 	ctx->timestamp = now;
 	perf_cgroup_set_timestamp(task, ctx);
@@ -2471,15 +2489,17 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
  * layer. It is invoked ONLY when there is at least one system-wide context
  * with at least one active event using taken branch sampling.
  */
-static void perf_branch_stack_sched_in(struct task_struct *prev,
-				       struct task_struct *task)
+static void perf_branch_stack_sched(struct task_struct *task1,
+				    struct task_struct *task2,
+				    bool sched_in)
 {
 	struct perf_cpu_context *cpuctx;
+	struct perf_event_context *task_ctx;
 	struct pmu *pmu;
 	unsigned long flags;
 
 	/* no need to flush branch stack if not changing task */
-	if (prev == task)
+	if (task1 == task2)
 		return;
 
 	local_irq_save(flags);
@@ -2488,25 +2508,26 @@ static void perf_branch_stack_sched_in(struct task_struct *prev,
 
 	list_for_each_entry_rcu(pmu, &pmus, entry) {
 		cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+		task_ctx = cpuctx->task_ctx;
 
 		/*
 		 * check if the context has at least one
 		 * event using PERF_SAMPLE_BRANCH_STACK
 		 */
-		if (cpuctx->ctx.nr_branch_stack > 0
-		    && pmu->flush_branch_stack) {
-
+		if (pmu->branch_stack_sched &&
+		    (cpuctx->ctx.nr_branch_stack > 0 ||
+		     (task_ctx && task_ctx->nr_branch_stack > 0))) {
 			pmu = cpuctx->ctx.pmu;
 
-			perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_lock(cpuctx, task_ctx);
 
 			perf_pmu_disable(pmu);
 
-			pmu->flush_branch_stack();
+			pmu->branch_stack_sched(task_ctx, sched_in);
 
 			perf_pmu_enable(pmu);
 
-			perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
+			perf_ctx_unlock(cpuctx, task_ctx);
 		}
 	}
 
@@ -2547,9 +2568,9 @@ void __perf_event_task_sched_in(struct task_struct *prev,
 	if (atomic_read(&__get_cpu_var(perf_cgroup_events)))
 		perf_cgroup_sched_in(prev, task);
 
-	/* check for system-wide branch_stack events */
-	if (atomic_read(&__get_cpu_var(perf_branch_stack_events)))
-		perf_branch_stack_sched_in(prev, task);
+	/* check for branch_stack events running on this cpu */
+	if (__get_cpu_var(perf_branch_stack_events))
+		perf_branch_stack_sched(prev, task, true);
 }
 
 static u64 perf_calculate_period(struct perf_event *event, u64 nsec, u64 count)
@@ -3134,14 +3155,8 @@ static void free_event(struct perf_event *event)
 			static_key_slow_dec_deferred(&perf_sched_events);
 		}
 
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_dec_deferred(&perf_sched_events);
-			/* is system-wide event */
-			if (!(event->attach_state & PERF_ATTACH_TASK)) {
-				atomic_dec(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-			}
-		}
 	}
 
 	if (event->rb) {
@@ -6562,12 +6577,8 @@ done:
 				return ERR_PTR(err);
 			}
 		}
-		if (has_branch_stack(event)) {
+		if (has_branch_stack(event))
 			static_key_slow_inc(&perf_sched_events.key);
-			if (!(event->attach_state & PERF_ATTACH_TASK))
-				atomic_inc(&per_cpu(perf_branch_stack_events,
-						    event->cpu));
-		}
 	}
 
 	return event;
-- 
1.8.1.4

Re: [PATCH 4/7] perf, x86: Save/resotre LBR stack during context switch

[Peter Zijlstra]

On Tue, Jun 25, 2013 at 04:47:16PM +0800, Yan, Zheng wrote:
> From: "Yan, Zheng" <zheng.z.yan@intel.com>
> 
> When the LBR call stack is enabled, it is necessary to save/restore
> the stack on context switch. The solution is saving/restoring the
> stack to/from task's perf event context. If task has no perf event
> context, just flush the stack on context switch.

So I have some problems with this; suppose the LBR is shared between a
cpu and a task event, when the task switches back in, it will over-write
the LBR content with old stuff.

In that case the CPU event's LBR stack can be actively wrong, as opposed
to being empty.