Skip to contentSkip to conversation

perf: add support for capturing skid IP

PATCH In review Stephane Eranian 2 participants 7 messages Nov 9, 2017 – Nov 13, 2017

Subsystems (2)
x86 perf events
lore
Share & tools
lore

Stephane Eranian

author
#0 lore raw

v4 cover perf: add support for capturing skid IP

This patchs adds a new sample record type called PERF_SAMPLE_SKID_IP. The goal is to record the unmodified interrupted instruction pointer (IP) as seen by the kernel and reflected in the machine state.

On some architectures, it is possible to avoid the IP skid using hardware support. For instance, on Intel x86, the use of PEBS helps eliminate the skid on Haswell and later processors. On older Intel processor, software, i.e., the kernel, may succeed in eliminating the skid.

Without this patch, on Haswell processors, if you set:
- attr.precise = 0, then you get the skid IP
- attr.precise = 1, then you get the skid PEBS ip (off-by-1)
- attr.precise = 2, then you get the skidless PEBS ip

The IP is captured when the event has PERF_SAMPLE_IP set in sample_type. However, there are certain measurements where you need to have BOTH the skidless IP and the skid IP. For instance, when studying branches, the skid IP usually points to the target of the branch while the skidless IP points to the branch instruction itself. Today, it is not possible to retrieve both at the same time. This patch makes this possible by specifying PERF_SAMPLE_IP|PERF_SAMPLE_SKID_IP.

As an example, consider the following code snipet:

 37.51 42c2ed    je     42c2f3
       42c2ef    add    $0x1,%rdx
       42c2f3    sub    $0x1,%rax

When using PEBS (precise=2) and sampling on BR_INST_RETIRED.CONDITIONAL, the IP always points to 0x42c2ed. With precise=1, the IP would point to 0x42c2f3. It is interesting to collect both IPs in a single run to determine how often the conditional branch is taken vs. non-taken.

Understanding the skid is also interesting for other precise events.

In V2, we rebased to 10d94ff4d558 (v4.14-rc7).

In V3, code is rebased to 4.14-rc8, LKML comments have been integrated. The new way to specify skid ip is per event:

   $ perf record -e cpu/event=0xc5,skid-ip=1/ ....

In V4, we fix document of the ski-ip event option and move a session.c change to the correct patch as per Jiri's remark.

Stephane Eranian (5):
perf/core: add PERF_RECORD_SAMPLE_SKID_IP record type perf/x86: add PERF_SAMPLE_SKID_IP support for X86 PEBS perf/tools: add support for PERF_SAMPLE_SKID_IP perf/record: add documentation for using PERF_SAMPLE_SKID_IP perf/script: add support for PERF_SAMPLE_SKID_IP

 arch/x86/events/intel/ds.c               |  7 +++++++
 include/linux/perf_event.h               |  2 ++
 include/uapi/linux/perf_event.h          |  4 +++-
 kernel/events/core.c                     | 14 ++++++++++++++
 tools/include/uapi/linux/perf_event.h    |  4 +++-
 tools/perf/Documentation/perf-record.txt |  8 ++++++++
 tools/perf/Documentation/perf-script.txt |  2 +-
 tools/perf/builtin-script.c              | 10 ++++++++--
 tools/perf/util/event.h                  |  1 +
 tools/perf/util/evsel.c                  | 11 +++++++++++
 tools/perf/util/evsel.h                  |  2 ++
 tools/perf/util/parse-events.c           |  7 +++++++
 tools/perf/util/parse-events.h           |  1 +
 tools/perf/util/parse-events.l           |  1 +
 tools/perf/util/session.c                |  3 +++
 15 files changed, 72 insertions(+), 5 deletions(-)
Signature · 1 line
2.7.4
author
#1 lore raw

PATCH v4 1/5 perf/core: add PERF_RECORD_SAMPLE_SKID_IP record type

This patchs adds a new sample record type. The goal
is to record the interrupted instruction pointer (IP)
as seen by the kernel and reflected in the machine state (pt_regs).

On some architectures, it is possible to avoid the IP skid using hardware support. For instance, on Intel x86, the use of PEBS helps eliminate the skid on Haswell and later processors.

Without this patch, on Haswell processors, if you set:
- attr.precise = 0, then you get the skid IP
- attr.precise > 0, then you get the PEBS ip corrected for skid

The IP normally comes when the event has PERF_RECORD_SAMPLE_IP set. However, there are certain measuremewnts where you need to have BOTH the corrected IP and the skid IP. For instance, when studying branches, the skid IP usually points to the target of the branch while the corrected IP point to the branch instruction itself. Today, it is not possible to retrieve both at the same time. This patch makes this possible by specifying PERF_SAMPLE_IP|PERF_SAMPLE_SKID_IP.

Signed-off-by Stephane Eranian <[email protected]>
---
 include/linux/perf_event.h      |  2 ++
 include/uapi/linux/perf_event.h |  4 +++-
 kernel/events/core.c            | 14 ++++++++++++++
 3 files changed, 19 insertions(+), 1 deletion(-)
include/linux/perf_event.h +2 -0
index 874b71a70058..772530501025 100644
@@ -917,6 +917,7 @@ struct perf_sample_data {
 	u64				stack_user_size;
 
 	u64				phys_addr;
+	u64				skid_ip;
 } ____cacheline_aligned;
 
 /* default value for data source */

@@ -937,6 +938,7 @@ static inline void perf_sample_data_init(struct perf_sample_data *data,
 	data->weight = 0;
 	data->data_src.val = PERF_MEM_NA;
 	data->txn = 0;
+	data->skid_ip = 0; /* mark as uinitialized */
 }
 
 extern void perf_output_sample(struct perf_output_handle *handle,
include/uapi/linux/perf_event.h +3 -1
index 362493a2f950..48a65a90fcab 100644
@@ -141,8 +141,9 @@ enum perf_event_sample_format {
 	PERF_SAMPLE_TRANSACTION			= 1U << 17,
 	PERF_SAMPLE_REGS_INTR			= 1U << 18,
 	PERF_SAMPLE_PHYS_ADDR			= 1U << 19,
+	PERF_SAMPLE_SKID_IP			= 1U << 20,
 
-	PERF_SAMPLE_MAX = 1U << 20,		/* non-ABI */
+	PERF_SAMPLE_MAX = 1U << 21,		/* non-ABI */
 };
 
 /*

@@ -817,6 +818,7 @@ enum perf_event_type {
 	 *	{ u64			abi; # enum perf_sample_regs_abi
 	 *	  u64			regs[weight(mask)]; } && PERF_SAMPLE_REGS_INTR
 	 *	{ u64			phys_addr;} && PERF_SAMPLE_PHYS_ADDR
+	 *	{ u64			skid_ip;  } && PERF_SAMPLE_SKID_IP
 	 * };
 	 */
 	PERF_RECORD_SAMPLE			= 9,
kernel/events/core.c +14 -0
index 0649a84204e6..40f2839c8b94 100644
@@ -1565,6 +1565,9 @@ static void __perf_event_header_size(struct perf_event *event, u64 sample_type)
 	if (sample_type & PERF_SAMPLE_PHYS_ADDR)
 		size += sizeof(data->phys_addr);
 
+	if (sample_type & PERF_SAMPLE_SKID_IP)
+		size += sizeof(data->skid_ip);
+
 	event->header_size = size;
 }
 

@@ -5934,6 +5937,9 @@ void perf_output_sample(struct perf_output_handle *handle,
 	if (sample_type & PERF_SAMPLE_PHYS_ADDR)
 		perf_output_put(handle, data->phys_addr);
 
+	if (sample_type & PERF_SAMPLE_SKID_IP)
+		perf_output_put(handle, data->skid_ip);
+
 	if (!event->attr.watermark) {
 		int wakeup_events = event->attr.wakeup_events;
 

@@ -5999,6 +6005,14 @@ void perf_prepare_sample(struct perf_event_header *header,
 	if (sample_type & PERF_SAMPLE_IP)
 		data->ip = perf_instruction_pointer(regs);
 
+	/*
+	 * if skid_ip has not been set by arch specific code, then
+	 * we initialize it to IP as interrupt-based sampling has
+	 * skid
+	 */
+	if (!data->skid_ip && sample_type & PERF_SAMPLE_SKID_IP)
+		data->skid_ip = perf_instruction_pointer(regs);
+
 	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
 		int size = 1;
Signature · 1 line
2.7.4
author
#2 lore raw

PATCH v4 2/5 perf/x86: add PERF_SAMPLE_SKID_IP support for X86 PEBS

In reply to Stephane Eranian

This atch adds support for SKID_IP to Intel x86 processors in PEBS mode.

Signed-off-by Stephane Eranian <[email protected]>
---
 arch/x86/events/intel/ds.c | 7 +++++++
 1 file changed, 7 insertions(+)
arch/x86/events/intel/ds.c +7 -0
index 3674a4b6f8bd..dd248ceda452 100644
@@ -1190,6 +1190,13 @@ static void setup_pebs_sample_data(struct perf_event *event,
 	    x86_pmu.intel_cap.pebs_format >= 1)
 		data->addr = pebs->dla;
 
+	/*
+	 * unmodified, skid IP which is guaranteed to be the next
+	 * dyanmic instruction
+	 */
+	if (sample_type & PERF_SAMPLE_SKID_IP)
+		data->skid_ip = pebs->ip;
+
 	if (x86_pmu.intel_cap.pebs_format >= 2) {
 		/* Only set the TSX weight when no memory weight. */
 		if ((sample_type & PERF_SAMPLE_WEIGHT) && !fll)
Signature · 1 line
2.7.4
author
#3 lore raw

PATCH v4 3/5 perf/tools: add support for PERF_SAMPLE_SKID_IP

In reply to Stephane Eranian

This patch adds the support code to handle the PERF_SAMPLE_SKID_IP record type. This is done as an event term and as such can be enabled per event: cpu/event=xxx,skid-ip=1/. This is a boolean term which is false by default.

Signed-off-by Stephane Eranian <[email protected]>
---
 tools/include/uapi/linux/perf_event.h |  4 +++-
 tools/perf/util/event.h               |  1 +
 tools/perf/util/evsel.c               | 11 +++++++++++
 tools/perf/util/evsel.h               |  2 ++
 tools/perf/util/parse-events.c        |  7 +++++++
 tools/perf/util/parse-events.h        |  1 +
 tools/perf/util/parse-events.l        |  1 +
 tools/perf/util/session.c             |  3 +++
 8 files changed, 29 insertions(+), 1 deletion(-)
tools/include/uapi/linux/perf_event.h +3 -1
index 362493a2f950..79655228dd9b 100644
@@ -141,8 +141,9 @@ enum perf_event_sample_format {
 	PERF_SAMPLE_TRANSACTION			= 1U << 17,
 	PERF_SAMPLE_REGS_INTR			= 1U << 18,
 	PERF_SAMPLE_PHYS_ADDR			= 1U << 19,
+	PERF_SAMPLE_SKID_IP			= 1U << 20,
 
-	PERF_SAMPLE_MAX = 1U << 20,		/* non-ABI */
+	PERF_SAMPLE_MAX = 1U << 21,		/* non-ABI */
 };
 
 /*

@@ -817,6 +818,7 @@ enum perf_event_type {
 	 *	{ u64			abi; # enum perf_sample_regs_abi
 	 *	  u64			regs[weight(mask)]; } && PERF_SAMPLE_REGS_INTR
 	 *	{ u64			phys_addr;} && PERF_SAMPLE_PHYS_ADDR
+	 *	{ u64			skid_ip; } && PERF_SAMPLE_SKID_IP
 	 * };
 	 */
 	PERF_RECORD_SAMPLE			= 9,
tools/perf/util/event.h +1 -0
index 1ae95efbfb95..41622a7ed649 100644
@@ -202,6 +202,7 @@ struct perf_sample {
 	u32 raw_size;
 	u64 data_src;
 	u64 phys_addr;
+	u64 skid_ip;
 	u32 flags;
 	u16 insn_len;
 	u8  cpumode;
tools/perf/util/evsel.c +11 -0
index f894893c203d..679954ed2201 100644
@@ -775,6 +775,10 @@ static void apply_config_terms(struct perf_evsel *evsel,
 		case PERF_EVSEL__CONFIG_TERM_OVERWRITE:
 			attr->write_backward = term->val.overwrite ? 1 : 0;
 			break;
+		case PERF_EVSEL__CONFIG_TERM_SKID_IP:
+			if (term->val.skid_ip)
+				perf_evsel__set_sample_bit(evsel, SKID_IP);
+			break;
 		default:
 			break;
 		}

@@ -1478,6 +1482,7 @@ static void __p_sample_type(char *buf, size_t size, u64 value)
 		bit_name(BRANCH_STACK), bit_name(REGS_USER), bit_name(STACK_USER),
 		bit_name(IDENTIFIER), bit_name(REGS_INTR), bit_name(DATA_SRC),
 		bit_name(WEIGHT), bit_name(PHYS_ADDR),
+		bit_name(SKID_IP),
 		{ .name = NULL, }
 	};
 #undef bit_name

@@ -2225,6 +2230,12 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		array++;
 	}
 
+	data->skid_ip = 0;
+	if (type & PERF_SAMPLE_SKID_IP) {
+		data->skid_ip = *array;
+		array++;
+	}
+
 	return 0;
 }
tools/perf/util/evsel.h +2 -0
index 9277df96ffda..8555095f0d48 100644
@@ -49,6 +49,7 @@ enum {
 	PERF_EVSEL__CONFIG_TERM_OVERWRITE,
 	PERF_EVSEL__CONFIG_TERM_DRV_CFG,
 	PERF_EVSEL__CONFIG_TERM_BRANCH,
+	PERF_EVSEL__CONFIG_TERM_SKID_IP,
 	PERF_EVSEL__CONFIG_TERM_MAX,
 };
 

@@ -66,6 +67,7 @@ struct perf_evsel_config_term {
 		bool	inherit;
 		bool	overwrite;
 		char	*branch;
+		bool	skid_ip;
 	} val;
 };
tools/perf/util/parse-events.c +7 -0
index a7fcd95961ef..1a1d9fc509bd 100644
@@ -918,6 +918,7 @@ static const char *config_term_names[__PARSE_EVENTS__TERM_TYPE_NR] = {
 	[PARSE_EVENTS__TERM_TYPE_OVERWRITE]		= "overwrite",
 	[PARSE_EVENTS__TERM_TYPE_NOOVERWRITE]		= "no-overwrite",
 	[PARSE_EVENTS__TERM_TYPE_DRV_CFG]		= "driver-config",
+	[PARSE_EVENTS__TERM_TYPE_SKID_IP]		= "skid-ip",
 };
 
 static bool config_term_shrinked;

@@ -1026,6 +1027,9 @@ do {									   \
 	case PARSE_EVENTS__TERM_TYPE_MAX_STACK:
 		CHECK_TYPE_VAL(NUM);
 		break;
+	case PARSE_EVENTS__TERM_TYPE_SKID_IP:
+		CHECK_TYPE_VAL(NUM);
+		break;
 	default:
 		err->str = strdup("unknown term");
 		err->idx = term->err_term;

@@ -1159,6 +1163,9 @@ do {								\
 		case PARSE_EVENTS__TERM_TYPE_DRV_CFG:
 			ADD_CONFIG_TERM(DRV_CFG, drv_cfg, term->val.str);
 			break;
+		case PARSE_EVENTS__TERM_TYPE_SKID_IP:
+			ADD_CONFIG_TERM(SKID_IP, skid_ip, term->val.num ? 1 : 0);
+			break;
 		default:
 			break;
 		}
tools/perf/util/parse-events.h +1 -0
index be337c266697..d331f078a389 100644
@@ -74,6 +74,7 @@ enum {
 	PARSE_EVENTS__TERM_TYPE_NOOVERWRITE,
 	PARSE_EVENTS__TERM_TYPE_OVERWRITE,
 	PARSE_EVENTS__TERM_TYPE_DRV_CFG,
+	PARSE_EVENTS__TERM_TYPE_SKID_IP,
 	__PARSE_EVENTS__TERM_TYPE_NR,
 };
tools/perf/util/parse-events.l +1 -0
index 241396cd059d..383443f4fe0d 100644
@@ -257,6 +257,7 @@ inherit			{ return term(yyscanner, PARSE_EVENTS__TERM_TYPE_INHERIT); }
 no-inherit		{ return term(yyscanner, PARSE_EVENTS__TERM_TYPE_NOINHERIT); }
 overwrite		{ return term(yyscanner, PARSE_EVENTS__TERM_TYPE_OVERWRITE); }
 no-overwrite		{ return term(yyscanner, PARSE_EVENTS__TERM_TYPE_NOOVERWRITE); }
+skid-ip			{ return term(yyscanner, PARSE_EVENTS__TERM_TYPE_SKID_IP); }
 ,			{ return ','; }
 "/"			{ BEGIN(INITIAL); return '/'; }
 {name_minus}		{ return str(yyscanner, PE_NAME); }
tools/perf/util/session.c +3 -0
index 5c412310f266..ee8fb635f694 100644
@@ -1131,6 +1131,9 @@ static void dump_sample(struct perf_evsel *evsel, union perf_event *event,
 
 	if (sample_type & PERF_SAMPLE_READ)
 		sample_read__printf(sample, evsel->attr.read_format);
+
+	if (sample_type & PERF_SAMPLE_SKID_IP)
+		printf("... skid_ip: 0x%" PRIx64 "\n", sample->skid_ip);
 }
 
 static void dump_read(struct perf_evsel *evsel, union perf_event *event)
Signature · 1 line
2.7.4
author
#4 lore raw

PATCH v4 4/5 perf/record: add documentation for using PERF_SAMPLE_SKID_IP

In reply to Stephane Eranian

This patch adds documentation to describe how to use the skid ip support with perf record. The sample type can be provided per event as follows: pmu_instance/...,skid-ip=1/

For instance on Intel X86:

$ perf record -e cpu/event=0xc5,skid-ip=1/pp

does record the precise address of retired branches and their target.

Signed-off-by Stephane Eranian <[email protected]>
---
 tools/perf/Documentation/perf-record.txt | 8 ++++++++
 1 file changed, 8 insertions(+)
tools/perf/Documentation/perf-record.txt +8 -0
index 5a626ef666c2..f0e3636dc4be 100644
@@ -57,6 +57,14 @@ OPTIONS
 			 FP mode, "dwarf" for DWARF mode, "lbr" for LBR mode and
 			 "no" for disable callgraph.
 	  - 'stack-size': user stack size for dwarf mode
+	  - 'skid-ip' : boolean, captures the unmodified interrupt instruction pointer
+			(IP) in each sample. Usually with event-based sampling, the IP
+			has skid and rarely point to the instruction which caused the
+			event to overflow. On some architectures, the hardware can eliminate
+			the skid and perf_events returns it as the IP with precise sampling is
+			enabled. But for certain measurements, it may be useful to have both
+			the correct and skid ip. This option enable capturing the skid ip in
+			additional to the corrected ip. Default is: false
 
           See the linkperf:perf-list[1] man page for more parameters.
Signature · 1 line
2.7.4
author
#5 lore raw

PATCH v4 5/5 perf/script: add support for PERF_SAMPLE_SKID_IP

In reply to Stephane Eranian

This patch adds a skid_ip field to perf script
to dump the raw value of the PERF_SAMPLE_SKID_IP
field in each sample.

$ perf script -F +ip,+skid_ip ......

The field is not enabled by default.

Signed-off-by Stephane Eranian <[email protected]>
---
 tools/perf/Documentation/perf-script.txt |  2 +-
 tools/perf/builtin-script.c              | 10 ++++++++--
 2 files changed, 9 insertions(+), 3 deletions(-)
tools/perf/Documentation/perf-script.txt +1 -1
index 2811fcf684cb..96871bd3a576 100644
@@ -117,7 +117,7 @@ OPTIONS
         Comma separated list of fields to print. Options are:
         comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff,
         srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output, brstackinsn,
-        brstackoff, callindent, insn, insnlen, synth, phys_addr.
+        brstackoff, callindent, insn, insnlen, synth, phys_addr, skid_ip.
         Field list can be prepended with the type, trace, sw or hw,
         to indicate to which event type the field list applies.
         e.g., -F sw:comm,tid,time,ip,sym  and -F trace:time,cpu,trace
tools/perf/builtin-script.c +8 -2
index 68f36dc0344f..f00fc8c50f68 100644
@@ -90,6 +90,7 @@ enum perf_output_field {
 	PERF_OUTPUT_SYNTH           = 1U << 25,
 	PERF_OUTPUT_PHYS_ADDR       = 1U << 26,
 	PERF_OUTPUT_UREGS	    = 1U << 27,
+	PERF_OUTPUT_SKID_IP	    = 1U << 28,
 };
 
 struct output_option {

@@ -124,6 +125,7 @@ struct output_option {
 	{.str = "brstackoff", .field = PERF_OUTPUT_BRSTACKOFF},
 	{.str = "synth", .field = PERF_OUTPUT_SYNTH},
 	{.str = "phys_addr", .field = PERF_OUTPUT_PHYS_ADDR},
+	{.str = "skid_ip", .field = PERF_OUTPUT_SKID_IP},
 };
 
 enum {

@@ -1563,7 +1565,11 @@ static void process_event(struct perf_script *script,
 
 	if (PRINT_FIELD(PHYS_ADDR))
 		fprintf(fp, "%16" PRIx64, sample->phys_addr);
-	fprintf(fp, "\n");
+
+	if (PRINT_FIELD(SKID_IP))
+		printf(" %"PRIx64" ", sample->skid_ip);
+
+	printf("\n");
 }
 
 static struct scripting_ops	*scripting_ops;

@@ -2915,7 +2921,7 @@ int cmd_script(int argc, const char **argv)
 		     "Valid types: hw,sw,trace,raw,synth. "
 		     "Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,"
 		     "addr,symoff,period,iregs,uregs,brstack,brstacksym,flags,"
-		     "bpf-output,callindent,insn,insnlen,brstackinsn,synth,phys_addr",
+		     "bpf-output,callindent,insn,insnlen,brstackinsn,synth,phys_addr,skid_ip",
 		     parse_output_fields),
 	OPT_BOOLEAN('a', "all-cpus", &system_wide,
 		    "system-wide collection from all CPUs"),
Signature · 1 line
2.7.4
Acked-by on v4 5/5
#6 lore raw

Re: [PATCH v4 5/5] perf/script: add support for PERF_SAMPLE_SKID_IP

Quoted 19 lines · Stephane Eranian

This patch adds a skid_ip field to perf script
to dump the raw value of the PERF_SAMPLE_SKID_IP
field in each sample.

$ perf script -F +ip,+skid_ip ......

The field is not enabled by default.

Signed-off-by Stephane Eranian <[email protected]>
---
 tools/perf/Documentation/perf-script.txt |  2 +-
 tools/perf/builtin-script.c              | 10 ++++++++--
 2 files changed, 9 insertions(+), 3 deletions(-)
tools/perf/Documentation/perf-script.txt +0 -0
index 2811fcf684cb..96871bd3a576 100644

tools/perf/Documentation/perf-script.txt OPTIONS · line 120

Quoted 5 lines
tools/perf/Documentation/perf-script.txt +1 -1
@@ -117,7 +117,7 @@ OPTIONS
         Comma separated list of fields to print. Options are:
         comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff,
         srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output, brstackinsn,
-        brstackoff, callindent, insn, insnlen, synth, phys_addr.
+        brstackoff, callindent, insn, insnlen, synth, phys_addr, skid_ip.

while we are at it, could we keep it the same as the field name in perf record? I dont mind using '-' or '_', just please keep it same

other then that it looks ok, for patchset:

Acked-by Jiri Olsa <[email protected]>

thanks,
jirka