Skip to contentSkip to conversation

perf: add indirect jump filtering support

PATCH In review Stephane Eranian 3 participants 8 messages May 14, 2015 – Jun 7, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover perf: add indirect jump filtering support

This short patch series adds a new filter to the branch stack sampling feature. The new filter enables sampling only indirect jumps. For each the source and target are captured.

The new branch_sample_type flag is: PERF_SAMPLE_BRANCH_IND_JUMP

The series provides full support in the perf record tool:

  $ perf record -j ind_jmp ...

On Intel x86 processors, the filter is implemented using the LBR hw filter.

This filter is useful to study targets of indirect jumps.

Stephane Eranian (3):
perf: add new PERF_SAMPLE_BRANCH_IND_JUMP branch sample type perf/x86: add support for PERF_SAMPLE_BRANCH_IND_JUMP perf record: add support for sampling indirect jumps

 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 10 +++++++++-
 include/uapi/linux/perf_event.h            |  2 ++
 tools/perf/builtin-record.c                |  1 +
 3 files changed, 12 insertions(+), 1 deletion(-)
Signature · 1 line
1.9.1
#1 lore raw

PATCH 1/3 perf: add new PERF_SAMPLE_BRANCH_IND_JUMP branch sample type

This patch adds a new branch_sample_type flag to enable filtering branch sampling to indirect jumps. The support is subject to hardware or kernel software support on each architecture.

Filtering on indirect jump is useful to study the targets of the jump.

---
 include/uapi/linux/perf_event.h | 2 ++
 1 file changed, 2 insertions(+)
include/uapi/linux/perf_event.h +2 -0
index 309211b..c4622f1 100644
@@ -167,6 +167,7 @@ enum perf_branch_sample_type_shift {
 	PERF_SAMPLE_BRANCH_COND_SHIFT		= 10, /* conditional branches */
 
 	PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT	= 11, /* call/ret stack */
+	PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT	= 12, /* indirect jumps */
 
 	PERF_SAMPLE_BRANCH_MAX_SHIFT		/* non-ABI */
 };
@@ -186,6 +187,7 @@ enum perf_branch_sample_type {
 	PERF_SAMPLE_BRANCH_COND		= 1U << PERF_SAMPLE_BRANCH_COND_SHIFT,
 
 	PERF_SAMPLE_BRANCH_CALL_STACK	= 1U << PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT,
+	PERF_SAMPLE_BRANCH_IND_JUMP	= 1U << PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT,
 
 	PERF_SAMPLE_BRANCH_MAX		= 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT,
 };
Signature · 1 line
1.9.1
#2 lore raw

[tip:perf/core] perf: add new PERF_SAMPLE_BRANCH_IND_JUMP branch sample type

perf: add new PERF_SAMPLE_BRANCH_IND_JUMP branch sample type
Commit-ID:  c9fdfa14c3792c0160849c484e83aa57afd80ccc
Gitweb:     http://git.kernel.org/tip/c9fdfa14c3792c0160849c484e83aa57afd80ccc
Author:     Stephane Eranian <[email protected]>
AuthorDate: Thu, 14 May 2015 23:09:58 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sun, 7 Jun 2015 16:08:22 +0200

perf: add new PERF_SAMPLE_BRANCH_IND_JUMP branch sample type

This patch adds a new branch_sample_type flag to enable filtering branch sampling to indirect jumps. The support is subject to hardware or kernel software support on each architecture.

Filtering on indirect jump is useful to study the targets of the jump.

---
 include/uapi/linux/perf_event.h | 2 ++
 1 file changed, 2 insertions(+)
include/uapi/linux/perf_event.h +2 -0
index 309211b..c4622f1 100644
@@ -167,6 +167,7 @@ enum perf_branch_sample_type_shift {
 	PERF_SAMPLE_BRANCH_COND_SHIFT		= 10, /* conditional branches */
 
 	PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT	= 11, /* call/ret stack */
+	PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT	= 12, /* indirect jumps */
 
 	PERF_SAMPLE_BRANCH_MAX_SHIFT		/* non-ABI */
 };
@@ -186,6 +187,7 @@ enum perf_branch_sample_type {
 	PERF_SAMPLE_BRANCH_COND		= 1U << PERF_SAMPLE_BRANCH_COND_SHIFT,
 
 	PERF_SAMPLE_BRANCH_CALL_STACK	= 1U << PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT,
+	PERF_SAMPLE_BRANCH_IND_JUMP	= 1U << PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT,
 
 	PERF_SAMPLE_BRANCH_MAX		= 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT,
 };
#3 lore raw

PATCH 2/3 perf/x86: add support for PERF_SAMPLE_BRANCH_IND_JUMP

In reply to Stephane Eranian

This patch enables support for branch sampling filter for indirect jumps (IND_JUMP). It enables LBR IND_JMP filtering where available. There is also software filtering support.

---
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
arch/x86/kernel/cpu/perf_event_intel_lbr.c +9 -1
index 94e5b50..201e16f 100644
@@ -96,6 +96,7 @@ enum {
 	X86_BR_NO_TX		= 1 << 14,/* not in transaction */
 	X86_BR_ZERO_CALL	= 1 << 15,/* zero length call */
 	X86_BR_CALL_STACK	= 1 << 16,/* call stack */
+	X86_BR_IND_JMP		= 1 << 17,/* indirect jump */
 };
 
 #define X86_BR_PLM (X86_BR_USER | X86_BR_KERNEL)
@@ -113,6 +114,7 @@ enum {
 	 X86_BR_IRQ	 |\
 	 X86_BR_ABORT	 |\
 	 X86_BR_IND_CALL |\
+	 X86_BR_IND_JMP  |\
 	 X86_BR_ZERO_CALL)
 
 #define X86_BR_ALL (X86_BR_PLM | X86_BR_ANY)
@@ -523,6 +525,9 @@ static int intel_pmu_setup_sw_lbr_filter(struct perf_event *event)
 			X86_BR_CALL_STACK;
 	}
 
+	if (br_type & PERF_SAMPLE_BRANCH_IND_JUMP)
+		mask |= X86_BR_IND_JMP;
+
 	/*
 	 * stash actual user request into reg, it may
 	 * be used by fixup code for some CPU
@@ -736,7 +741,7 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 			break;
 		case 4:
 		case 5:
-			ret = X86_BR_JMP;
+			ret = X86_BR_IND_JMP;
 			break;
 		}
 		break;
@@ -844,6 +849,7 @@ static const int nhm_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 	 */
 	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT] = LBR_IND_CALL | LBR_IND_JMP,
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]     = LBR_JCC,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT] = LBR_IND_JMP,
 };
 
 static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
@@ -856,6 +862,7 @@ static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 						| LBR_FAR,
 	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT]	= LBR_IND_CALL,
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]		= LBR_JCC,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT]	= LBR_IND_JMP,
 };
 
 static const int hsw_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
@@ -870,6 +877,7 @@ static const int hsw_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]		= LBR_JCC,
 	[PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT]	= LBR_REL_CALL | LBR_IND_CALL
 						| LBR_RETURN | LBR_CALL_STACK,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT]	= LBR_IND_JMP,
 };
 
 /* core */
Signature · 1 line
1.9.1
#4 lore raw

[tip:perf/core] perf/x86/intel: add support for PERF_SAMPLE_BRANCH_IND_JUMP

perf/x86/intel: add support for PERF_SAMPLE_BRANCH_IND_JUMP
Commit-ID:  7b74cfb2ecb4d56a25c89cdb561e4926db85feb1
Gitweb:     http://git.kernel.org/tip/7b74cfb2ecb4d56a25c89cdb561e4926db85feb1
Author:     Stephane Eranian <[email protected]>
AuthorDate: Thu, 14 May 2015 23:09:59 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sun, 7 Jun 2015 16:08:27 +0200

perf/x86/intel: add support for PERF_SAMPLE_BRANCH_IND_JUMP

This patch enables support for branch sampling filter for indirect jumps (IND_JUMP). It enables LBR IND_JMP filtering where available. There is also software filtering support.

---
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
arch/x86/kernel/cpu/perf_event_intel_lbr.c +9 -1
index 94e5b50..201e16f 100644
@@ -96,6 +96,7 @@ enum {
 	X86_BR_NO_TX		= 1 << 14,/* not in transaction */
 	X86_BR_ZERO_CALL	= 1 << 15,/* zero length call */
 	X86_BR_CALL_STACK	= 1 << 16,/* call stack */
+	X86_BR_IND_JMP		= 1 << 17,/* indirect jump */
 };
 
 #define X86_BR_PLM (X86_BR_USER | X86_BR_KERNEL)
@@ -113,6 +114,7 @@ enum {
 	 X86_BR_IRQ	 |\
 	 X86_BR_ABORT	 |\
 	 X86_BR_IND_CALL |\
+	 X86_BR_IND_JMP  |\
 	 X86_BR_ZERO_CALL)
 
 #define X86_BR_ALL (X86_BR_PLM | X86_BR_ANY)
@@ -523,6 +525,9 @@ static int intel_pmu_setup_sw_lbr_filter(struct perf_event *event)
 			X86_BR_CALL_STACK;
 	}
 
+	if (br_type & PERF_SAMPLE_BRANCH_IND_JUMP)
+		mask |= X86_BR_IND_JMP;
+
 	/*
 	 * stash actual user request into reg, it may
 	 * be used by fixup code for some CPU
@@ -736,7 +741,7 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 			break;
 		case 4:
 		case 5:
-			ret = X86_BR_JMP;
+			ret = X86_BR_IND_JMP;
 			break;
 		}
 		break;
@@ -844,6 +849,7 @@ static const int nhm_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 	 */
 	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT] = LBR_IND_CALL | LBR_IND_JMP,
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]     = LBR_JCC,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT] = LBR_IND_JMP,
 };
 
 static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
@@ -856,6 +862,7 @@ static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 						| LBR_FAR,
 	[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT]	= LBR_IND_CALL,
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]		= LBR_JCC,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT]	= LBR_IND_JMP,
 };
 
 static const int hsw_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
@@ -870,6 +877,7 @@ static const int hsw_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX_SHIFT] = {
 	[PERF_SAMPLE_BRANCH_COND_SHIFT]		= LBR_JCC,
 	[PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT]	= LBR_REL_CALL | LBR_IND_CALL
 						| LBR_RETURN | LBR_CALL_STACK,
+	[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT]	= LBR_IND_JMP,
 };
 
 /* core */
#5 lore raw

PATCH 3/3 perf record: add support for sampling indirect jumps

In reply to Stephane Eranian

This patch adds a new branch sampling type support for indirect jumps:

perf record -j ind_jmp .......

It enables analysis of indirect jumps targets. It requires kernel and possibly hardware support to operate correctly.

---
 tools/perf/builtin-record.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/builtin-record.c +1 -0
index 5dfe913..860c133 100644
@@ -773,6 +773,7 @@ static const struct branch_mode branch_modes[] = {
 	BRANCH_OPT("in_tx", PERF_SAMPLE_BRANCH_IN_TX),
 	BRANCH_OPT("no_tx", PERF_SAMPLE_BRANCH_NO_TX),
 	BRANCH_OPT("cond", PERF_SAMPLE_BRANCH_COND),
+	BRANCH_OPT("ind_jmp", PERF_SAMPLE_BRANCH_IND_JUMP),
 	BRANCH_END
 };
 
Signature · 1 line
1.9.1
#6 lore raw

[tip:perf/core] perf record: Add support for sampling indirect jumps

perf record: Add support for sampling indirect jumps
Commit-ID:  5b68164d6a1fdbe02b30bd777d1f686c6d901f28
Gitweb:     http://git.kernel.org/tip/5b68164d6a1fdbe02b30bd777d1f686c6d901f28
Author:     Stephane Eranian <[email protected]>
AuthorDate: Thu, 14 May 2015 23:10:00 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sun, 7 Jun 2015 16:08:31 +0200

perf record: Add support for sampling indirect jumps

This patch adds a new branch sampling type support for indirect jumps:

perf record -j ind_jmp .......

It enables analysis of indirect jumps targets. It requires kernel and possibly hardware support to operate correctly.

[ Fixup against: f00898f4e20b (perf tools: Move branch option parsing to own file) ]
Cc Andrew Morton <[email protected]>
Cc H. Peter Anvin <[email protected]>
Cc Linus Torvalds <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Thomas Gleixner <[email protected]>
---
 tools/perf/util/parse-branch-options.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/util/parse-branch-options.c +1 -0
index 9d99943..a3b1e13 100644
@@ -26,6 +26,7 @@ static const struct branch_mode branch_modes[] = {
 	BRANCH_OPT("in_tx", PERF_SAMPLE_BRANCH_IN_TX),
 	BRANCH_OPT("no_tx", PERF_SAMPLE_BRANCH_NO_TX),
 	BRANCH_OPT("cond", PERF_SAMPLE_BRANCH_COND),
+	BRANCH_OPT("ind_jmp", PERF_SAMPLE_BRANCH_IND_JUMP),
 	BRANCH_END
 };
 
Reviewed-by on the v1 series
#7 lore raw
In reply to Stephane Eranian
Quoted 6 lines · Stephane Eranian

This short patch series adds a new filter to the branch stack sampling feature. The new filter enables sampling only indirect jumps. For each the source and target are captured.

The new branch_sample_type flag is: PERF_SAMPLE_BRANCH_IND_JUMP

Quoted 8 lines

The series provides full support in the perf record tool:

  $ perf record -j ind_jmp ...

On Intel x86 processors, the filter is implemented using the LBR hw filter.

This filter is useful to study targets of indirect jumps.

Patches look all good to me.

-Andi