perf script: export sampled callchains to database

The existing implementation of thread__resolve_callchain, under certain circumstances, can assemble callchain entries in the incorrect order.

                0x558f2a04c774
                0x558f2a04dded
                rand
                __random
                ret_from_intr
                do_IRQ
                handle_irq
                handle_edge_irq
                handle_irq_event
                handle_irq_event_percpu
                gen8_irq_handler

Signed-off-by Chris Phlipot <[email protected]>

Quoted 2 lines

The existing implementation of thread__resolve_callchain, under certain circumstances, can assemble callchain entries in the incorrect order.

SNIP

Quoted 12 lines

                0x558f2a04c774
                0x558f2a04dded
                rand
                __random
                ret_from_intr
                do_IRQ
                handle_irq
                handle_edge_irq
                handle_irq_event
                handle_irq_event_percpu
                gen8_irq_handler

Quoted 1 line

Signed-off-by Chris Phlipot <[email protected]>

for this patch

Acked-by Jiri Olsa <[email protected]>

can't really ack the rest of the patchset, but it looks good to me

thanks,
jirka

Move the call path handling code out of thread-stack.c and thread-stack.h to allow other components that are not part of thread-stack to create call paths.

Summary:

(quoted diff) +0 -2

-Create call-path.c and call-path.h and add them to the build.
-Move all call path related code out of thread-stack.c and thread-stack.h

and into call-path.c and call-path.h.
-A small subset of structures and functions are now visible through
call-path.h, which is required for thread-stack.c to continue to compile.

This change is a prerequisite for subsequent patches in this change set and by itself contains no user-visible changes.

Signed-off-by Chris Phlipot <[email protected]>

---
 tools/perf/util/Build                              |   1 +
 tools/perf/util/call-path.c                        | 122 ++++++++++++++++++++
 tools/perf/util/call-path.h                        |  77 +++++++++++++
 tools/perf/util/db-export.c                        |   1 +
 .../util/scripting-engines/trace-event-python.c    |   1 +
 tools/perf/util/thread-stack.c                     | 126 +--------------------
 tools/perf/util/thread-stack.h                     |  25 +---
 7 files changed, 204 insertions(+), 149 deletions(-)
 create mode 100644 tools/perf/util/call-path.c
 create mode 100644 tools/perf/util/call-path.h

tools/perf/util/Build +1 -0

index 90229a8..027bb2b 100644
@@ -74,6 +74,7 @@ libperf-y += srcline.o
 libperf-y += data.o
 libperf-y += tsc.o
 libperf-y += cloexec.o
+libperf-y += call-path.o
 libperf-y += thread-stack.o
 libperf-$(CONFIG_AUXTRACE) += auxtrace.o
 libperf-$(CONFIG_AUXTRACE) += intel-pt-decoder/

tools/perf/util/call-path.c new file +122 -0

new file mode 100644
index 0000000..904a170
@@ -0,0 +1,122 @@
+/*
+ * call-path.h: Manipulate a tree data structure containing function call paths
+ * Copyright (c) 2014, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <linux/rbtree.h>
+#include <linux/list.h>
+
+#include "util.h"
+#include "call-path.h"
+
+static void call_path__init(struct call_path *cp, struct call_path *parent,
+			    struct symbol *sym, u64 ip, bool in_kernel)
+{
+	cp->parent = parent;
+	cp->sym = sym;
+	cp->ip = sym ? 0 : ip;
+	cp->db_id = 0;
+	cp->in_kernel = in_kernel;
+	RB_CLEAR_NODE(&cp->rb_node);
+	cp->children = RB_ROOT;
+}
+
+struct call_path_root *call_path_root__new(void)
+{
+	struct call_path_root *cpr;
+
+	cpr = zalloc(sizeof(struct call_path_root));
+	if (!cpr)
+		return NULL;
+	call_path__init(&cpr->call_path, NULL, NULL, 0, false);
+	INIT_LIST_HEAD(&cpr->blocks);
+	return cpr;
+}
+
+void call_path_root__free(struct call_path_root *cpr)
+{
+	struct call_path_block *pos, *n;
+
+	list_for_each_entry_safe(pos, n, &cpr->blocks, node) {
+		list_del(&pos->node);
+		free(pos);
+	}
+	free(cpr);
+}
+
+static struct call_path *call_path__new(struct call_path_root *cpr,
+					struct call_path *parent,
+					struct symbol *sym, u64 ip,
+					bool in_kernel)
+{
+	struct call_path_block *cpb;
+	struct call_path *cp;
+	size_t n;
+
+	if (cpr->next < cpr->sz) {
+		cpb = list_last_entry(&cpr->blocks, struct call_path_block,
+				      node);
+	} else {
+		cpb = zalloc(sizeof(struct call_path_block));
+		if (!cpb)
+			return NULL;
+		list_add_tail(&cpb->node, &cpr->blocks);
+		cpr->sz += CALL_PATH_BLOCK_SIZE;
+	}
+
+	n = cpr->next++ & CALL_PATH_BLOCK_MASK;
+	cp = &cpb->cp[n];
+
+	call_path__init(cp, parent, sym, ip, in_kernel);
+
+	return cp;
+}
+
+struct call_path *call_path__findnew(struct call_path_root *cpr,
+				     struct call_path *parent,
+				     struct symbol *sym, u64 ip, u64 ks)
+{
+	struct rb_node **p;
+	struct rb_node *node_parent = NULL;
+	struct call_path *cp;
+	bool in_kernel = ip >= ks;
+
+	if (sym)
+		ip = 0;
+
+	if (!parent)
+		return call_path__new(cpr, parent, sym, ip, in_kernel);
+
+	p = &parent->children.rb_node;
+	while (*p != NULL) {
+		node_parent = *p;
+		cp = rb_entry(node_parent, struct call_path, rb_node);
+
+		if (cp->sym == sym && cp->ip == ip)
+			return cp;
+
+		if (sym < cp->sym || (sym == cp->sym && ip < cp->ip))
+			p = &(*p)->rb_left;
+		else
+			p = &(*p)->rb_right;
+	}
+
+	cp = call_path__new(cpr, parent, sym, ip, in_kernel);
+	if (!cp)
+		return NULL;
+
+	rb_link_node(&cp->rb_node, node_parent, p);
+	rb_insert_color(&cp->rb_node, &parent->children);
+
+	return cp;
+}

tools/perf/util/call-path.h new file +77 -0

new file mode 100644
index 0000000..477f6d0
@@ -0,0 +1,77 @@
+/*
+ * call-path.h: Manipulate a tree data structure containing function call paths
+ * Copyright (c) 2014, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#ifndef __PERF_CALL_PATH_H
+#define __PERF_CALL_PATH_H
+
+#include <sys/types.h>
+
+#include <linux/types.h>
+#include <linux/rbtree.h>
+
+/**
+ * struct call_path - node in list of calls leading to a function call.
+ * @parent: call path to the parent function call
+ * @sym: symbol of function called
+ * @ip: only if sym is null, the ip of the function
+ * @db_id: id used for db-export
+ * @in_kernel: whether function is a in the kernel
+ * @rb_node: node in parent's tree of called functions
+ * @children: tree of call paths of functions called
+ *
+ * In combination with the call_return structure, the call_path structure
+ * defines a context-sensitve call-graph.
+ */
+struct call_path {
+	struct call_path *parent;
+	struct symbol *sym;
+	u64 ip;
+	u64 db_id;
+	bool in_kernel;
+	struct rb_node rb_node;
+	struct rb_root children;
+};
+
+#define CALL_PATH_BLOCK_SHIFT 8
+#define CALL_PATH_BLOCK_SIZE (1 << CALL_PATH_BLOCK_SHIFT)
+#define CALL_PATH_BLOCK_MASK (CALL_PATH_BLOCK_SIZE - 1)
+
+struct call_path_block {
+	struct call_path cp[CALL_PATH_BLOCK_SIZE];
+	struct list_head node;
+};
+
+/**
+ * struct call_path_root - root of all call paths.
+ * @call_path: root call path
+ * @blocks: list of blocks to store call paths
+ * @next: next free space
+ * @sz: number of spaces
+ */
+struct call_path_root {
+	struct call_path call_path;
+	struct list_head blocks;
+	size_t next;
+	size_t sz;
+};
+
+struct call_path_root *call_path_root__new(void);
+void call_path_root__free(struct call_path_root *cpr);
+
+struct call_path *call_path__findnew(struct call_path_root *cpr,
+				     struct call_path *parent,
+				     struct symbol *sym, u64 ip, u64 ks);
+
+#endif

tools/perf/util/db-export.c +1 -0

index 049438d..4fc607c 100644
@@ -23,6 +23,7 @@
 #include "event.h"
 #include "util.h"
 #include "thread-stack.h"
+#include "call-path.h"
 #include "db-export.h"
 
 struct deferred_export {

tools/perf/util/scripting-engines/trace-event-python.c +1 -0

index 525eb49..7bb8592 100644
@@ -41,6 +41,7 @@
 #include "../thread-stack.h"
 #include "../trace-event.h"
 #include "../machine.h"
+#include "../call-path.h"
 #include "thread_map.h"
 #include "cpumap.h"
 #include "stat.h"

tools/perf/util/thread-stack.c +1 -125

index 679688e..fc419a5 100644
@@ -22,31 +22,9 @@
 #include "debug.h"
 #include "symbol.h"
 #include "comm.h"
+#include "call-path.h"
 #include "thread-stack.h"
 
-#define CALL_PATH_BLOCK_SHIFT 8
-#define CALL_PATH_BLOCK_SIZE (1 << CALL_PATH_BLOCK_SHIFT)
-#define CALL_PATH_BLOCK_MASK (CALL_PATH_BLOCK_SIZE - 1)
-
-struct call_path_block {
-	struct call_path cp[CALL_PATH_BLOCK_SIZE];
-	struct list_head node;
-};
-
-/**
- * struct call_path_root - root of all call paths.
- * @call_path: root call path
- * @blocks: list of blocks to store call paths
- * @next: next free space
- * @sz: number of spaces
- */
-struct call_path_root {
-	struct call_path call_path;
-	struct list_head blocks;
-	size_t next;
-	size_t sz;
-};
-
 /**
  * struct call_return_processor - provides a call-back to consume call-return
  *                                information.

@@ -335,108 +313,6 @@ void thread_stack__sample(struct thread *thread, struct ip_callchain *chain,
 		chain->ips[i] = thread->ts->stack[thread->ts->cnt - i].ret_addr;
 }
 
-static void call_path__init(struct call_path *cp, struct call_path *parent,
-			    struct symbol *sym, u64 ip, bool in_kernel)
-{
-	cp->parent = parent;
-	cp->sym = sym;
-	cp->ip = sym ? 0 : ip;
-	cp->db_id = 0;
-	cp->in_kernel = in_kernel;
-	RB_CLEAR_NODE(&cp->rb_node);
-	cp->children = RB_ROOT;
-}
-
-static struct call_path_root *call_path_root__new(void)
-{
-	struct call_path_root *cpr;
-
-	cpr = zalloc(sizeof(struct call_path_root));
-	if (!cpr)
-		return NULL;
-	call_path__init(&cpr->call_path, NULL, NULL, 0, false);
-	INIT_LIST_HEAD(&cpr->blocks);
-	return cpr;
-}
-
-static void call_path_root__free(struct call_path_root *cpr)
-{
-	struct call_path_block *pos, *n;
-
-	list_for_each_entry_safe(pos, n, &cpr->blocks, node) {
-		list_del(&pos->node);
-		free(pos);
-	}
-	free(cpr);
-}
-
-static struct call_path *call_path__new(struct call_path_root *cpr,
-					struct call_path *parent,
-					struct symbol *sym, u64 ip,
-					bool in_kernel)
-{
-	struct call_path_block *cpb;
-	struct call_path *cp;
-	size_t n;
-
-	if (cpr->next < cpr->sz) {
-		cpb = list_last_entry(&cpr->blocks, struct call_path_block,
-				      node);
-	} else {
-		cpb = zalloc(sizeof(struct call_path_block));
-		if (!cpb)
-			return NULL;
-		list_add_tail(&cpb->node, &cpr->blocks);
-		cpr->sz += CALL_PATH_BLOCK_SIZE;
-	}
-
-	n = cpr->next++ & CALL_PATH_BLOCK_MASK;
-	cp = &cpb->cp[n];
-
-	call_path__init(cp, parent, sym, ip, in_kernel);
-
-	return cp;
-}
-
-static struct call_path *call_path__findnew(struct call_path_root *cpr,
-					    struct call_path *parent,
-					    struct symbol *sym, u64 ip, u64 ks)
-{
-	struct rb_node **p;
-	struct rb_node *node_parent = NULL;
-	struct call_path *cp;
-	bool in_kernel = ip >= ks;
-
-	if (sym)
-		ip = 0;
-
-	if (!parent)
-		return call_path__new(cpr, parent, sym, ip, in_kernel);
-
-	p = &parent->children.rb_node;
-	while (*p != NULL) {
-		node_parent = *p;
-		cp = rb_entry(node_parent, struct call_path, rb_node);
-
-		if (cp->sym == sym && cp->ip == ip)
-			return cp;
-
-		if (sym < cp->sym || (sym == cp->sym && ip < cp->ip))
-			p = &(*p)->rb_left;
-		else
-			p = &(*p)->rb_right;
-	}
-
-	cp = call_path__new(cpr, parent, sym, ip, in_kernel);
-	if (!cp)
-		return NULL;
-
-	rb_link_node(&cp->rb_node, node_parent, p);
-	rb_insert_color(&cp->rb_node, &parent->children);
-
-	return cp;
-}
-
 struct call_return_processor *
 call_return_processor__new(int (*process)(struct call_return *cr, void *data),
 			   void *data)

tools/perf/util/thread-stack.h +1 -24

index e1528f1..ec9bedd 100644
@@ -19,7 +19,6 @@
 #include <sys/types.h>
 
 #include <linux/types.h>
-#include <linux/rbtree.h>
 
 struct thread;
 struct comm;

@@ -30,6 +29,7 @@ struct call_return_processor;
 struct comm;
 struct perf_sample;
 struct addr_location;
+struct call_path;
 
 /*
  * Call/Return flags.

@@ -68,29 +68,6 @@ struct call_return {
 	u32 flags;
 };
 
-/**
- * struct call_path - node in list of calls leading to a function call.
- * @parent: call path to the parent function call
- * @sym: symbol of function called
- * @ip: only if sym is null, the ip of the function
- * @db_id: id used for db-export
- * @in_kernel: whether function is a in the kernel
- * @rb_node: node in parent's tree of called functions
- * @children: tree of call paths of functions called
- *
- * In combination with the call_return structure, the call_path structure
- * defines a context-sensitve call-graph.
- */
-struct call_path {
-	struct call_path *parent;
-	struct symbol *sym;
-	u64 ip;
-	u64 db_id;
-	bool in_kernel;
-	struct rb_node rb_node;
-	struct rb_root children;
-};
-
 int thread_stack__event(struct thread *thread, u32 flags, u64 from_ip,
 			u64 to_ip, u16 insn_len, u64 trace_nr);
 void thread_stack__set_trace_nr(struct thread *thread, u64 trace_nr);

Quoted 12 lines

Move the call path handling code out of thread-stack.c and thread-stack.h to allow other components that are not part of thread-stack to create call paths.

Summary:

(quoted diff) +0 -2

-Create call-path.c and call-path.h and add them to the build.
-Move all call path related code out of thread-stack.c and thread-stack.h

and into call-path.c and call-path.h.
-A small subset of structures and functions are now visible through
call-path.h, which is required for thread-stack.c to continue to compile.

Quoted 4 lines

This change is a prerequisite for subsequent patches in this change set and by itself contains no user-visible changes.

Signed-off-by Chris Phlipot <[email protected]>

Acked-by Adrian Hunter <[email protected]>

This change enables the db export api to export callchains. This is accomplished by adding callchains obtained from samples to the call_path_root structure and exporting them via the current call path export API.

While the current API does support exporting call paths, this is not supported when sampling. This commit addresses that missing feature by allowing the export of call paths when callchains are present in samples.

Summary:
-This feature is activated by initializing the call_path_root member
inside the db_export structure to a non-null value.
-Callchains are resolved with thread__resolve_callchain() and then stored
and exported by adding a call path under call path root.
-Symbol and DSO for each callchain node are exported via db_ids_from_al()

This commit puts in place infrastructure to be used by subsequent commits, and by itself, does not introduce any user-visible changes.

Signed-off-by Chris Phlipot <[email protected]>

---
 tools/perf/util/db-export.c | 86 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/db-export.h |  2 ++
 2 files changed, 88 insertions(+)

tools/perf/util/db-export.c +0 -0

index 4fc607c..cb96591 100644

tools/perf/util/db-export.c +5 -0

+
 #include "evsel.h"
 #include "machine.h"
 #include "thread.h"

@@ -23,6 +25,7 @@
 #include "event.h"
 #include "util.h"
 #include "thread-stack.h"
+#include "callchain.h"
 #include "call-path.h"
 #include "db-export.h"
 

@@ -277,6 +280,81 @@ static int db_ids_from_al(struct db_export *dbe, struct addr_location *al,
 	return 0;
 }
 
+static struct call_path *call_path_from_sample(struct db_export *dbe,
+					       struct machine *machine,
+					       struct thread *thread,

tools/perf/util/db-export.c +8 -0

+	if (err) {
+		callchain_param.order = saved_order;
+		return NULL;
+	}
+	callchain_cursor_commit(&callchain_cursor);
+
+	while (1) {
+		struct callchain_cursor_node *node;

tools/perf/util/db-export.c +40 -0

+		 * constructing an addr_location struct and then passing it to
+		 * db_ids_from_al() to perform the export.
+		 */
+		al.sym = node->sym;
+		al.map = node->map;
+		al.machine = machine;
+		if (al.map)
+			al.addr = al.map->map_ip(al.map, node->ip);
+		else
+			al.addr = node->ip;
+
+		db_ids_from_al(dbe, &al, &dso_db_id, &sym_db_id, &offset);
+
+		/* add node to the call path tree if it doesn't exist */
+		current = call_path__findnew(dbe->cpr, current,
+					     al.sym, node->ip,
+					     kernel_start);
+
+		callchain_cursor_advance(&callchain_cursor);
+	}
+
+	/* Reset the callchain order to its prior value. */
+	callchain_param.order = saved_order;
+
+	if (current == &dbe->cpr->call_path) {
+		/* Bail because the callchain was empty. */
+		return NULL;
+	}
+
+	return current;
+}
+
 int db_export__branch_type(struct db_export *dbe, u32 branch_type,
 			   const char *name)
 {

@@ -330,6 +408,14 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,
 	if (err)
 		goto out_put;
 
+	if (dbe->cpr) {
+		struct call_path *cp = call_path_from_sample(dbe, al->machine,
+							     thread, sample,
+							     evsel);
+		if (cp)
+			db_export__call_path(dbe, cp);
+	}
+
 	if ((evsel->attr.sample_type & PERF_SAMPLE_ADDR) &&
 	    sample_addr_correlates_sym(&evsel->attr)) {
 		struct addr_location addr_al;

tools/perf/util/db-export.h +2 -0

index 25e22fd..f5daf55 100644
@@ -27,6 +27,7 @@ struct dso;
 struct perf_sample;
 struct addr_location;
 struct call_return_processor;
+struct call_path_root;
 struct call_path;
 struct call_return;
 

@@ -64,6 +65,7 @@ struct db_export {
 	int (*export_call_return)(struct db_export *dbe,
 				  struct call_return *cr);
 	struct call_return_processor *crp;
+	struct call_path_root *cpr;
 	u64 evsel_last_db_id;
 	u64 machine_last_db_id;
 	u64 thread_last_db_id;

Quoted 16 lines

This change enables the db export api to export callchains. This is accomplished by adding callchains obtained from samples to the call_path_root structure and exporting them via the current call path export API.

While the current API does support exporting call paths, this is not supported when sampling. This commit addresses that missing feature by allowing the export of call paths when callchains are present in samples.

Summary:
-This feature is activated by initializing the call_path_root member
inside the db_export structure to a non-null value.
-Callchains are resolved with thread__resolve_callchain() and then stored
and exported by adding a call path under call path root.
-Symbol and DSO for each callchain node are exported via db_ids_from_al()

Quoted 4 lines

This commit puts in place infrastructure to be used by subsequent commits, and by itself, does not introduce any user-visible changes.

Signed-off-by Chris Phlipot <[email protected]>

Looks good. A few of nitpicks below. But apart from those:

Acked-by Adrian Hunter <[email protected]>

Quoted 10 lines

---
 tools/perf/util/db-export.c | 86 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/db-export.h |  2 ++
 2 files changed, 88 insertions(+)

tools/perf/util/db-export.c +0 -0

index 4fc607c..cb96591 100644

tools/perf/util/db-export.c line 18

Quoted 4 lines

tools/perf/util/db-export.c +1 -0

@@ -15,6 +15,8 @@
 
 #include <errno.h>
 
+#include <sys/types.h>

Why?

Quoted 19 lines

tools/perf/util/db-export.c +5 -0

+
 #include "evsel.h"
 #include "machine.h"
 #include "thread.h"

@@ -23,6 +25,7 @@
 #include "event.h"
 #include "util.h"
 #include "thread-stack.h"
+#include "callchain.h"
 #include "call-path.h"
 #include "db-export.h"
 

@@ -277,6 +280,81 @@ static int db_ids_from_al(struct db_export *dbe, struct addr_location *al,
 	return 0;
 }
 
+static struct call_path *call_path_from_sample(struct db_export *dbe,
+					       struct machine *machine,
+					       struct thread *thread,

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 295

Quoted 10 lines

tools/perf/util/db-export.c +10 -0

+					       struct perf_sample *sample,
+					       struct perf_evsel *evsel
+					       )
+{
+	u64 kernel_start = machine__kernel_start(machine);
+
+	struct call_path *current = &dbe->cpr->call_path;
+	enum chain_order saved_order = callchain_param.order;
+
+	int err = 0;

err needn't be initialized. And remove blank lines in the group of local declarations.

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 300

Quoted 5 lines

tools/perf/util/db-export.c +5 -0

+
+	if (!symbol_conf.use_callchain || !sample->callchain)
+		return NULL;
+
+	/* Since the call path tree must be built starting with the root, we

Normal kernel comment style has "/*" on separate line.

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 307

Quoted 7 lines

tools/perf/util/db-export.c +7 -0

+	 * must use ORDER_CALL for call chain resolution, in order to process
+	 * the callchain starting with the root node and ending with the leaf.
+	 */
+	callchain_param.order = ORDER_CALLER;
+	err = thread__resolve_callchain(thread, &callchain_cursor, evsel,
+					sample, NULL, NULL,
+					PERF_MAX_STACK_DEPTH);

Should probably be sysctl_perf_event_max_stack not PERF_MAX_STACK_DEPTH

Quoted 8 lines

tools/perf/util/db-export.c +8 -0

+	if (err) {
+		callchain_param.order = saved_order;
+		return NULL;
+	}
+	callchain_cursor_commit(&callchain_cursor);
+
+	while (1) {
+		struct callchain_cursor_node *node;

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 325

Quoted 10 lines

tools/perf/util/db-export.c +10 -0

+		struct addr_location al;
+		u64 dso_db_id = 0, sym_db_id = 0, offset = 0;
+
+		memset(&al, 0, sizeof(al));
+
+		node = callchain_cursor_current(&callchain_cursor);
+		if (!node)
+			break;
+
+		/* handle export of symbol and dso for this node by

Normal kernel comment style has "/*" on separate line.

Quoted 71 lines

tools/perf/util/db-export.c +40 -0

+		 * constructing an addr_location struct and then passing it to
+		 * db_ids_from_al() to perform the export.
+		 */
+		al.sym = node->sym;
+		al.map = node->map;
+		al.machine = machine;
+		if (al.map)
+			al.addr = al.map->map_ip(al.map, node->ip);
+		else
+			al.addr = node->ip;
+
+		db_ids_from_al(dbe, &al, &dso_db_id, &sym_db_id, &offset);
+
+		/* add node to the call path tree if it doesn't exist */
+		current = call_path__findnew(dbe->cpr, current,
+					     al.sym, node->ip,
+					     kernel_start);
+
+		callchain_cursor_advance(&callchain_cursor);
+	}
+
+	/* Reset the callchain order to its prior value. */
+	callchain_param.order = saved_order;
+
+	if (current == &dbe->cpr->call_path) {
+		/* Bail because the callchain was empty. */
+		return NULL;
+	}
+
+	return current;
+}
+
 int db_export__branch_type(struct db_export *dbe, u32 branch_type,
 			   const char *name)
 {

@@ -330,6 +408,14 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,
 	if (err)
 		goto out_put;
 
+	if (dbe->cpr) {
+		struct call_path *cp = call_path_from_sample(dbe, al->machine,
+							     thread, sample,
+							     evsel);
+		if (cp)
+			db_export__call_path(dbe, cp);
+	}
+
 	if ((evsel->attr.sample_type & PERF_SAMPLE_ADDR) &&
 	    sample_addr_correlates_sym(&evsel->attr)) {
 		struct addr_location addr_al;

tools/perf/util/db-export.h +2 -0

index 25e22fd..f5daf55 100644
@@ -27,6 +27,7 @@ struct dso;
 struct perf_sample;
 struct addr_location;
 struct call_return_processor;
+struct call_path_root;
 struct call_path;
 struct call_return;
 

@@ -64,6 +65,7 @@ struct db_export {
 	int (*export_call_return)(struct db_export *dbe,
 				  struct call_return *cr);
 	struct call_return_processor *crp;
+	struct call_path_root *cpr;
 	u64 evsel_last_db_id;
 	u64 machine_last_db_id;
 	u64 thread_last_db_id;

Em Fri, May 06, 2016 at 02:27:49PM +0300, Adrian Hunter escreveu:

Quoted 17 lines · Chris Phlipot

Quoted 16 lines

This change enables the db export api to export callchains. This is accomplished by adding callchains obtained from samples to the call_path_root structure and exporting them via the current call path export API.

While the current API does support exporting call paths, this is not supported when sampling. This commit addresses that missing feature by allowing the export of call paths when callchains are present in samples.

Summary:
-This feature is activated by initializing the call_path_root member
inside the db_export structure to a non-null value.
-Callchains are resolved with thread__resolve_callchain() and then stored
and exported by adding a call path under call path root.
-Symbol and DSO for each callchain node are exported via db_ids_from_al()

Quoted 6 lines

Quoted 4 lines

This commit puts in place infrastructure to be used by subsequent commits, and by itself, does not introduce any user-visible changes.

Signed-off-by Chris Phlipot <[email protected]>

Looks good. A few of nitpicks below. But apart from those:

Agreed, will fix those.

Quoted 18 lines

Acked-by Adrian Hunter <[email protected]>

Quoted 10 lines

---
 tools/perf/util/db-export.c | 86 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/db-export.h |  2 ++
 2 files changed, 88 insertions(+)

tools/perf/util/db-export.c +0 -0

index 4fc607c..cb96591 100644

tools/perf/util/db-export.c line 18

Quoted 4 lines

tools/perf/util/db-export.c +1 -0

@@ -15,6 +15,8 @@
 
 #include <errno.h>
 
+#include <sys/types.h>

Why?

Quoted 16 lines

Quoted 16 lines

tools/perf/util/db-export.c +2 -0

+
 #include "evsel.h"
 #include "machine.h"
 #include "thread.h"

@@ -23,6 +25,7 @@
 #include "event.h"
 #include "util.h"
 #include "thread-stack.h"
+#include "callchain.h"
 #include "call-path.h"
 #include "db-export.h"
 

@@ -277,6 +280,81 @@ static int db_ids_from_al(struct db_export *dbe, struct addr_location *al,
 	return 0;
 }
 

Quoted 6 lines

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 288

Quoted 6 lines

tools/perf/util/db-export.c +6 -0

+static struct call_path *call_path_from_sample(struct db_export *dbe,
+					       struct machine *machine,
+					       struct thread *thread,
+					       struct perf_sample *sample,
+					       struct perf_evsel *evsel
+					       )

Quoted 10 lines

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 295

Quoted 7 lines

tools/perf/util/db-export.c +7 -0

+{
+	u64 kernel_start = machine__kernel_start(machine);
+
+	struct call_path *current = &dbe->cpr->call_path;
+	enum chain_order saved_order = callchain_param.order;
+
+	int err = 0;

err needn't be initialized. And remove blank lines in the group of local declarations.

Quoted 7 lines

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 300

Quoted 5 lines

tools/perf/util/db-export.c +5 -0

+
+	if (!symbol_conf.use_callchain || !sample->callchain)
+		return NULL;
+
+	/* Since the call path tree must be built starting with the root, we

Normal kernel comment style has "/*" on separate line.

Quoted 9 lines

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 307

Quoted 7 lines

tools/perf/util/db-export.c +7 -0

+	 * must use ORDER_CALL for call chain resolution, in order to process
+	 * the callchain starting with the root node and ending with the leaf.
+	 */
+	callchain_param.order = ORDER_CALLER;
+	err = thread__resolve_callchain(thread, &callchain_cursor, evsel,
+					sample, NULL, NULL,
+					PERF_MAX_STACK_DEPTH);

Should probably be sysctl_perf_event_max_stack not PERF_MAX_STACK_DEPTH

Quoted 12 lines

Quoted 12 lines

tools/perf/util/db-export.c +12 -0

+	if (err) {
+		callchain_param.order = saved_order;
+		return NULL;
+	}
+	callchain_cursor_commit(&callchain_cursor);
+
+	while (1) {
+		struct callchain_cursor_node *node;
+		struct addr_location al;
+		u64 dso_db_id = 0, sym_db_id = 0, offset = 0;
+
+		memset(&al, 0, sizeof(al));

Quoted 8 lines

tools/perf/util/db-export.c static int db_ids_from_al(struct db_export *dbe, struct addr_location *al, · line 325

Quoted 6 lines

tools/perf/util/db-export.c +6 -0

+
+		node = callchain_cursor_current(&callchain_cursor);
+		if (!node)
+			break;
+
+		/* handle export of symbol and dso for this node by

Normal kernel comment style has "/*" on separate line.

Quoted 71 lines

Quoted 71 lines

tools/perf/util/db-export.c +40 -0

+		 * constructing an addr_location struct and then passing it to
+		 * db_ids_from_al() to perform the export.
+		 */
+		al.sym = node->sym;
+		al.map = node->map;
+		al.machine = machine;
+		if (al.map)
+			al.addr = al.map->map_ip(al.map, node->ip);
+		else
+			al.addr = node->ip;
+
+		db_ids_from_al(dbe, &al, &dso_db_id, &sym_db_id, &offset);
+
+		/* add node to the call path tree if it doesn't exist */
+		current = call_path__findnew(dbe->cpr, current,
+					     al.sym, node->ip,
+					     kernel_start);
+
+		callchain_cursor_advance(&callchain_cursor);
+	}
+
+	/* Reset the callchain order to its prior value. */
+	callchain_param.order = saved_order;
+
+	if (current == &dbe->cpr->call_path) {
+		/* Bail because the callchain was empty. */
+		return NULL;
+	}
+
+	return current;
+}
+
 int db_export__branch_type(struct db_export *dbe, u32 branch_type,
 			   const char *name)
 {

@@ -330,6 +408,14 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,
 	if (err)
 		goto out_put;
 
+	if (dbe->cpr) {
+		struct call_path *cp = call_path_from_sample(dbe, al->machine,
+							     thread, sample,
+							     evsel);
+		if (cp)
+			db_export__call_path(dbe, cp);
+	}
+
 	if ((evsel->attr.sample_type & PERF_SAMPLE_ADDR) &&
 	    sample_addr_correlates_sym(&evsel->attr)) {
 		struct addr_location addr_al;

tools/perf/util/db-export.h +2 -0

index 25e22fd..f5daf55 100644
@@ -27,6 +27,7 @@ struct dso;
 struct perf_sample;
 struct addr_location;
 struct call_return_processor;
+struct call_path_root;
 struct call_path;
 struct call_return;
 

@@ -64,6 +65,7 @@ struct db_export {
 	int (*export_call_return)(struct db_export *dbe,
 				  struct call_return *cr);
 	struct call_return_processor *crp;
+	struct call_path_root *cpr;
 	u64 evsel_last_db_id;
 	u64 machine_last_db_id;
 	u64 thread_last_db_id;

The exported sample now contains a reference to the call_path_id that represents its callchain.

While callchains themselves are nice to have, being able to associate them with samples makes them much more useful, and can allow for such things as determining how much cumulative time is spent in a particular function. This information is normally possible to get from the call return processor. However, when doing normal sampling, call/return information is not available, thus necessitating the need for associating samples directly with call paths.

This commit include changes to db-export layer to make this information available for subsequent patches in this change set, but by itself, does not make any changes visible to the user.

Signed-off-by Chris Phlipot <[email protected]>

---
 tools/perf/util/db-export.c | 4 +++-
 tools/perf/util/db-export.h | 1 +
 2 files changed, 4 insertions(+), 1 deletion(-)

tools/perf/util/db-export.c +3 -1

index cb96591..74f6875 100644
@@ -412,8 +412,10 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,
 		struct call_path *cp = call_path_from_sample(dbe, al->machine,
 							     thread, sample,
 							     evsel);
-		if (cp)
+		if (cp) {
 			db_export__call_path(dbe, cp);
+			es.call_path_id = cp->db_id;
+		}
 	}
 
 	if ((evsel->attr.sample_type & PERF_SAMPLE_ADDR) &&

tools/perf/util/db-export.h +1 -0

index f5daf55..67bc6b8 100644
@@ -44,6 +44,7 @@ struct export_sample {
 	u64			addr_dso_db_id;
 	u64			addr_sym_db_id;
 	u64			addr_offset; /* addr offset from symbol start */
+	u64			call_path_id;
 };
 
 struct db_export {

Quoted 11 lines

The exported sample now contains a reference to the call_path_id that represents its callchain.

While callchains themselves are nice to have, being able to associate them with samples makes them much more useful, and can allow for such things as determining how much cumulative time is spent in a particular function. This information is normally possible to get from the call return processor. However, when doing normal sampling, call/return information is not available, thus necessitating the need for associating samples directly with call paths.

Quoted 5 lines

This commit include changes to db-export layer to make this information available for subsequent patches in this change set, but by itself, does not make any changes visible to the user.

Signed-off-by Chris Phlipot <[email protected]>

Acked-by Adrian Hunter <[email protected]>

This change allows python scripts to be able to utilize the recent changes to the db export api allowing the export of call_paths derived from sampled callchains. These call paths are also now associated with the samples from which they were derived.

(quoted diff) +0 -5

-This feature is enabled by setting "perf_db_export_callchains" to true
 
-When enabled, samples that have callchain information will have the
    callchains exported via call_path_table
 
-The call_path_id field is added to sample_table to enable association of
    samples with the corresponding callchain stored in the call paths
    table. A call_path_id of 0 will be exported if there is no
    corresponding callchain.
 
-When "perf_db_export_callchains" and "perf_db_export_calls" are both
    set to True, the call path root data structure will be shared. This
    prevents duplicating of data and call path ids that would result from
    building two separate call path trees in memory.
 
-The call_return_processor structure definition was relocated to the header
    file to make its contents visible to db-export.c. This enables the
    sharing of call path trees between the two features, as mentioned
    above.

This change is visible to python scripts using the python db export api. The change is backwards compatible with scripts written against the previous API, assuming that the scripts model the sample_table function after the one in export-to-postgresql.py script by allowing for additional arguments to be added in the future. ie. using *x as the final argument of the sample_table function.

Signed-off-by Chris Phlipot <[email protected]>

---
 .../util/scripting-engines/trace-event-python.c    | 35 ++++++++++++++++++++--
 tools/perf/util/thread-stack.c                     | 13 --------
 tools/perf/util/thread-stack.h                     | 14 ++++++++-
 3 files changed, 46 insertions(+), 16 deletions(-)

tools/perf/util/scripting-engines/trace-event-python.c +33 -2

index 7bb8592..091bce6 100644
@@ -682,7 +682,7 @@ static int python_export_sample(struct db_export *dbe,
 	struct tables *tables = container_of(dbe, struct tables, dbe);
 	PyObject *t;
 
-	t = tuple_new(21);
+	t = tuple_new(22);
 
 	tuple_set_u64(t, 0, es->db_id);
 	tuple_set_u64(t, 1, es->evsel->db_id);

@@ -705,6 +705,7 @@ static int python_export_sample(struct db_export *dbe,
 	tuple_set_u64(t, 18, es->sample->data_src);
 	tuple_set_s32(t, 19, es->sample->flags & PERF_BRANCH_MASK);
 	tuple_set_s32(t, 20, !!(es->sample->flags & PERF_IP_FLAG_IN_TX));
+	tuple_set_u64(t, 21, es->call_path_id);
 
 	call_object(tables->sample_handler, t, "sample_table");
 

@@ -999,8 +1000,10 @@ static void set_table_handlers(struct tables *tables)
 {
 	const char *perf_db_export_mode = "perf_db_export_mode";
 	const char *perf_db_export_calls = "perf_db_export_calls";
-	PyObject *db_export_mode, *db_export_calls;
+	const char *perf_db_export_callchains = "perf_db_export_callchains";
+	PyObject *db_export_mode, *db_export_calls, *db_export_callchains;
 	bool export_calls = false;
+	bool export_callchains = false;
 	int ret;
 
 	memset(tables, 0, sizeof(struct tables));

@@ -1017,6 +1020,7 @@ static void set_table_handlers(struct tables *tables)
 	if (!ret)
 		return;
 
+	/* handle export calls */
 	tables->dbe.crp = NULL;
 	db_export_calls = PyDict_GetItemString(main_dict, perf_db_export_calls);
 	if (db_export_calls) {

@@ -1034,6 +1038,33 @@ static void set_table_handlers(struct tables *tables)
 			Py_FatalError("failed to create calls processor");
 	}
 
+	/* handle export callchains */
+	tables->dbe.cpr = NULL;
+	db_export_callchains = PyDict_GetItemString(main_dict,
+						    perf_db_export_callchains);
+	if (db_export_callchains) {
+		ret = PyObject_IsTrue(db_export_callchains);
+		if (ret == -1)
+			handler_call_die(perf_db_export_callchains);
+		export_callchains = !!ret;
+	}
+
+	if (export_callchains) {
+		/*
+		 * Attempt to use the call path root from the call return
+		 * processor, if the call return processor is in use. Otherwise,
+		 * we allocate a new call path root. This prevents exporting
+		 * duplicate call path ids when both are in use simultaniously.
+		 */
+		if (tables->dbe.crp)
+			tables->dbe.cpr = tables->dbe.crp->cpr;
+		else
+			tables->dbe.cpr = call_path_root__new();
+
+		if (!tables->dbe.cpr)
+			Py_FatalError("failed to create calls processor");
+	}
+
 	tables->db_export_mode = true;
 	/*
 	 * Reserve per symbol space for symbol->db_id via symbol__priv()

tools/perf/util/thread-stack.c +0 -13

index fc419a5..825086a 100644
@@ -25,19 +25,6 @@
 #include "call-path.h"
 #include "thread-stack.h"
 
-/**
- * struct call_return_processor - provides a call-back to consume call-return
- *                                information.
- * @cpr: call path root
- * @process: call-back that accepts call/return information
- * @data: anonymous data for call-back
- */
-struct call_return_processor {
-	struct call_path_root *cpr;
-	int (*process)(struct call_return *cr, void *data);
-	void *data;
-};
-
 #define STACK_GROWTH 2048
 
 /**

tools/perf/util/thread-stack.h +13 -1

index ec9bedd..ad44c79 100644
@@ -25,7 +25,6 @@ struct comm;
 struct ip_callchain;
 struct symbol;
 struct dso;
-struct call_return_processor;
 struct comm;
 struct perf_sample;
 struct addr_location;

@@ -68,6 +67,19 @@ struct call_return {
 	u32 flags;
 };
 
+/**
+ * struct call_return_processor - provides a call-back to consume call-return
+ *                                information.
+ * @cpr: call path root
+ * @process: call-back that accepts call/return information
+ * @data: anonymous data for call-back
+ */
+struct call_return_processor {
+	struct call_path_root *cpr;
+	int (*process)(struct call_return *cr, void *data);
+	void *data;
+};
+
 int thread_stack__event(struct thread *thread, u32 flags, u64 from_ip,
 			u64 to_ip, u16 insn_len, u64 trace_nr);
 void thread_stack__set_trace_nr(struct thread *thread, u64 trace_nr);

Quoted 32 lines

This change allows python scripts to be able to utilize the recent changes to the db export api allowing the export of call_paths derived from sampled callchains. These call paths are also now associated with the samples from which they were derived.

(quoted diff) +0 -5

-This feature is enabled by setting "perf_db_export_callchains" to true
 
-When enabled, samples that have callchain information will have the
    callchains exported via call_path_table
 
-The call_path_id field is added to sample_table to enable association of
    samples with the corresponding callchain stored in the call paths
    table. A call_path_id of 0 will be exported if there is no
    corresponding callchain.
 
-When "perf_db_export_callchains" and "perf_db_export_calls" are both
    set to True, the call path root data structure will be shared. This
    prevents duplicating of data and call path ids that would result from
    building two separate call path trees in memory.
 
-The call_return_processor structure definition was relocated to the header
    file to make its contents visible to db-export.c. This enables the
    sharing of call path trees between the two features, as mentioned
    above.

This change is visible to python scripts using the python db export api. The change is backwards compatible with scripts written against the previous API, assuming that the scripts model the sample_table function after the one in export-to-postgresql.py script by allowing for additional arguments to be added in the future. ie. using *x as the final argument of the sample_table function.

Quoted 1 line

Signed-off-by Chris Phlipot <[email protected]>

Acked-by Adrian Hunter <[email protected]>

Update the export-to-postgresql.py to support the newly introduced callchain export.

callchains are added into the existing call_paths table and can now be associated with samples when the "callpaths" commandline option is used with the script.

ex. $perf script -s export-to-postgresql.py example_db all callchains

Includes the following changes to enable callchain export via the python export APIs:

(quoted diff) +0 -3

-Add the "callchains" commandline option, which is used to enable
    callchain export by setting the perf_db_export_callchains global
-Add perf_db_export_callchains checks for call_path table creation
     and population.
-Add call_path_id to samples_table to conform with the new API

example usage and output using a small test app:

test_app.c:

	volatile int x = 0;
	void inc_x_loop()
	{
		int i;
		for(i=0; i<100000000; i++)
			x++;
	}

	void a()
	{
		inc_x_loop();
	}

	void b()
	{
		inc_x_loop();
	}

	int main()
	{
		a();
		b();
		return 0;
	}

example usage:

$ gcc -g -O0 test_app.c
$ ./perf record --call-graph=dwarf ./a.out
[ perf record: Woken up 77 times to write data ]
[ perf record: Captured and wrote 19.373 MB perf.data (2404 samples) ]

$ ./perf script -s scripts/python/export-to-postgresql.py
	example_db all callchains

$ psql example_db

example_db=#
SELECT
(SELECT name FROM symbols WHERE id = cps.symbol_id) as symbol, (SELECT name FROM symbols WHERE id =
(SELECT symbol_id from call_paths where id = cps.parent_id)) as parent_symbol,
sum(period) as event_count
FROM samples join call_paths as cps on call_path_id = cps.id GROUP BY cps.id,evsel_id
ORDER BY event_count DESC
LIMIT 5;

symbol | parent_symbol | event_count
------------------+--------------------------+-------------

 inc_x_loop       | a                        |   734250982
 inc_x_loop       | b                        |   731028057
 unknown          | unknown                  |     1335858
 task_tick_fair   | scheduler_tick           |     1238842
 update_wall_time | tick_do_update_jiffies64 |      650373

(5 rows)

The above data shows total "self time" in cycles for each call path that was sampled. It is intended to demonstrate how it accounts separately for the two ways to reach the "inc_x_loop" function(via "a" and "b"). Recursive common table expressions can be used as well to get cumulative time spent in a function as well, but that is beyond the scope of this basic example.

Signed-off-by Chris Phlipot <[email protected]>

---
 tools/perf/scripts/python/export-to-postgresql.py | 47 +++++++++++++++--------
 1 file changed, 30 insertions(+), 17 deletions(-)

tools/perf/scripts/python/export-to-postgresql.py +30 -17

index 6f0ca68..7656ff8 100644
@@ -223,11 +223,14 @@ sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 
 perf_db_export_mode = True
 perf_db_export_calls = False
+perf_db_export_callchains = False
+
 
 def usage():
-	print >> sys.stderr, "Usage is: export-to-postgresql.py <database name> [<columns>] [<calls>]"
+	print >> sys.stderr, "Usage is: export-to-postgresql.py <database name> [<columns>] [<calls>] [<callchains>]"
 	print >> sys.stderr, "where:	columns		'all' or 'branches'"
-	print >> sys.stderr, "		calls		'calls' => create calls table"
+	print >> sys.stderr, "		calls		'calls' => create calls and call_paths table"
+	print >> sys.stderr, "		callchains	'callchains' => create call_paths table"
 	raise Exception("Too few arguments")
 
 if (len(sys.argv) < 2):

@@ -245,9 +248,11 @@ if columns not in ("all", "branches"):
 
 branches = (columns == "branches")
 
-if (len(sys.argv) >= 4):
-	if (sys.argv[3] == "calls"):
+for i in range(3,len(sys.argv)):
+	if (sys.argv[i] == "calls"):
 		perf_db_export_calls = True
+	elif (sys.argv[i] == "callchains"):
+		perf_db_export_callchains = True
 	else:
 		usage()
 

@@ -358,14 +363,16 @@ else:
 		'transaction	bigint,'
 		'data_src	bigint,'
 		'branch_type	integer,'
-		'in_tx		boolean)')
+		'in_tx		boolean,'
+		'call_path_id	bigint)')
 
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	do_query(query, 'CREATE TABLE call_paths ('
 		'id		bigint		NOT NULL,'
 		'parent_id	bigint,'
 		'symbol_id	bigint,'
 		'ip		bigint)')
+if perf_db_export_calls:
 	do_query(query, 'CREATE TABLE calls ('
 		'id		bigint		NOT NULL,'
 		'thread_id	bigint,'

@@ -427,7 +434,7 @@ do_query(query, 'CREATE VIEW comm_threads_view AS '
 		'(SELECT tid FROM threads WHERE id = thread_id) AS tid'
 	' FROM comm_threads')
 
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	do_query(query, 'CREATE VIEW call_paths_view AS '
 		'SELECT '
 			'c.id,'

@@ -443,6 +450,7 @@ if perf_db_export_calls:
 			'(SELECT dso_id FROM symbols WHERE id = p.symbol_id) AS parent_dso_id,'
 			'(SELECT dso FROM symbols_view  WHERE id = p.symbol_id) AS parent_dso_short_name'
 		' FROM call_paths c INNER JOIN call_paths p ON p.id = c.parent_id')
+if perf_db_export_calls:
 	do_query(query, 'CREATE VIEW calls_view AS '
 		'SELECT '
 			'calls.id,'

@@ -540,8 +548,9 @@ dso_file		= open_output_file("dso_table.bin")
 symbol_file		= open_output_file("symbol_table.bin")
 branch_type_file	= open_output_file("branch_type_table.bin")
 sample_file		= open_output_file("sample_table.bin")
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	call_path_file		= open_output_file("call_path_table.bin")
+if perf_db_export_calls:
 	call_file		= open_output_file("call_table.bin")
 
 def trace_begin():

@@ -553,8 +562,8 @@ def trace_begin():
 	comm_table(0, "unknown")
 	dso_table(0, 0, "unknown", "unknown", "")
 	symbol_table(0, 0, 0, 0, 0, "unknown")
-	sample_table(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
-	if perf_db_export_calls:
+	sample_table(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
+	if perf_db_export_calls or perf_db_export_callchains:
 		call_path_table(0, 0, 0, 0)
 
 unhandled_count = 0

@@ -570,8 +579,9 @@ def trace_end():
 	copy_output_file(symbol_file,		"symbols")
 	copy_output_file(branch_type_file,	"branch_types")
 	copy_output_file(sample_file,		"samples")
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		copy_output_file(call_path_file,	"call_paths")
+	if perf_db_export_calls:
 		copy_output_file(call_file,		"calls")
 
 	print datetime.datetime.today(), "Removing intermediate files..."

@@ -584,8 +594,9 @@ def trace_end():
 	remove_output_file(symbol_file)
 	remove_output_file(branch_type_file)
 	remove_output_file(sample_file)
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		remove_output_file(call_path_file)
+	if perf_db_export_calls:
 		remove_output_file(call_file)
 	os.rmdir(output_dir_name)
 	print datetime.datetime.today(), "Adding primary keys"

@@ -598,8 +609,9 @@ def trace_end():
 	do_query(query, 'ALTER TABLE symbols         ADD PRIMARY KEY (id)')
 	do_query(query, 'ALTER TABLE branch_types    ADD PRIMARY KEY (id)')
 	do_query(query, 'ALTER TABLE samples         ADD PRIMARY KEY (id)')
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		do_query(query, 'ALTER TABLE call_paths      ADD PRIMARY KEY (id)')
+	if perf_db_export_calls:
 		do_query(query, 'ALTER TABLE calls           ADD PRIMARY KEY (id)')
 
 	print datetime.datetime.today(), "Adding foreign keys"

@@ -622,10 +634,11 @@ def trace_end():
 					'ADD CONSTRAINT symbolfk   FOREIGN KEY (symbol_id)    REFERENCES symbols    (id),'
 					'ADD CONSTRAINT todsofk    FOREIGN KEY (to_dso_id)    REFERENCES dsos       (id),'
 					'ADD CONSTRAINT tosymbolfk FOREIGN KEY (to_symbol_id) REFERENCES symbols    (id)')
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		do_query(query, 'ALTER TABLE call_paths '
 					'ADD CONSTRAINT parentfk    FOREIGN KEY (parent_id)    REFERENCES call_paths (id),'
 					'ADD CONSTRAINT symbolfk    FOREIGN KEY (symbol_id)    REFERENCES symbols    (id)')
+	if perf_db_export_calls:
 		do_query(query, 'ALTER TABLE calls '
 					'ADD CONSTRAINT threadfk    FOREIGN KEY (thread_id)    REFERENCES threads    (id),'
 					'ADD CONSTRAINT commfk      FOREIGN KEY (comm_id)      REFERENCES comms      (id),'

@@ -693,11 +706,11 @@ def branch_type_table(branch_type, name, *x):
 	value = struct.pack(fmt, 2, 4, branch_type, n, name)
 	branch_type_file.write(value)
 
-def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, branch_type, in_tx, *x):
+def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, branch_type, in_tx, call_path_id, *x):
 	if branches:
-		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiiiB", 17, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 4, branch_type, 1, in_tx)
+		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiiiBiq", 18, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 4, branch_type, 1, in_tx, 8, call_path_id)
 	else:
-		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiqiiiB", 21, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src, 4, branch_type, 1, in_tx)
+		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiqiiiBiq", 22, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src, 4, branch_type, 1, in_tx, 8, call_path_id)
 	sample_file.write(value)
 
 def call_path_table(cp_id, parent_id, symbol_id, ip, *x):

Quoted 1 line

Signed-off-by Chris Phlipot <[email protected]>

Acked-by Adrian Hunter <[email protected]>

Update the export-to-postgresql.py to support the newly introduced callchain export.

callchains are added into the existing call_paths table and can now be associated with samples when the "callpaths" commandline option is used with the script.

ex. $perf script -s export-to-postgresql.py example_db all callchains

Includes the following changes to enable callchain export via the python export APIs:

(quoted diff) +0 -3

-Add the "callchains" commandline option, which is used to enable
    callchain export by setting the perf_db_export_callchains global
-Add perf_db_export_callchains checks for call_path table creation
     and population.
-Add call_path_id to samples_table to conform with the new API

example usage and output using a small test app:

test_app.c:

	volatile int x = 0;
	void inc_x_loop()
	{
		int i;
		for(i=0; i<100000000; i++)
			x++;
	}

	void a()
	{
		inc_x_loop();
	}

	void b()
	{
		inc_x_loop();
	}

	int main()
	{
		a();
		b();
		return 0;
	}

example usage:

$ gcc -g -O0 test_app.c
$ ./perf record --call-graph=dwarf ./a.out
[ perf record: Woken up 77 times to write data ]
[ perf record: Captured and wrote 19.373 MB perf.data (2404 samples) ]

$ ./perf script -s scripts/python/export-to-postgresql.py
	example_db all callchains

$ psql example_db

example_db=#
SELECT
(SELECT name FROM symbols WHERE id = cps.symbol_id) as symbol, (SELECT name FROM symbols WHERE id =
(SELECT symbol_id from call_paths where id = cps.parent_id)) as parent_symbol,
sum(period) as event_count
FROM samples join call_paths as cps on call_path_id = cps.id GROUP BY cps.id,evsel_id
ORDER BY event_count DESC
LIMIT 5;

symbol | parent_symbol | event_count
------------------+--------------------------+-------------

 inc_x_loop       | a                        |   734250982
 inc_x_loop       | b                        |   731028057
 unknown          | unknown                  |     1335858
 task_tick_fair   | scheduler_tick           |     1238842
 update_wall_time | tick_do_update_jiffies64 |      650373

(5 rows)

The above data shows total "self time" in cycles for each call path that was sampled. It is intended to demonstrate how it accounts separately for the two ways to reach the "inc_x_loop" function(via "a" and "b"). Recursive common table expressions can be used as well to get cumulative time spent in a function as well, but that is beyond the scope of this basic example.

Signed-off-by Chris Phlipot <[email protected]>

---
 tools/perf/scripts/python/export-to-postgresql.py | 47 +++++++++++++++--------
 1 file changed, 30 insertions(+), 17 deletions(-)

tools/perf/scripts/python/export-to-postgresql.py +30 -17

index 6f0ca68..7656ff8 100644
@@ -223,11 +223,14 @@ sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 
 perf_db_export_mode = True
 perf_db_export_calls = False
+perf_db_export_callchains = False
+
 
 def usage():
-	print >> sys.stderr, "Usage is: export-to-postgresql.py <database name> [<columns>] [<calls>]"
+	print >> sys.stderr, "Usage is: export-to-postgresql.py <database name> [<columns>] [<calls>] [<callchains>]"
 	print >> sys.stderr, "where:	columns		'all' or 'branches'"
-	print >> sys.stderr, "		calls		'calls' => create calls table"
+	print >> sys.stderr, "		calls		'calls' => create calls and call_paths table"
+	print >> sys.stderr, "		callchains	'callchains' => create call_paths table"
 	raise Exception("Too few arguments")
 
 if (len(sys.argv) < 2):

@@ -245,9 +248,11 @@ if columns not in ("all", "branches"):
 
 branches = (columns == "branches")
 
-if (len(sys.argv) >= 4):
-	if (sys.argv[3] == "calls"):
+for i in range(3,len(sys.argv)):
+	if (sys.argv[i] == "calls"):
 		perf_db_export_calls = True
+	elif (sys.argv[i] == "callchains"):
+		perf_db_export_callchains = True
 	else:
 		usage()
 

@@ -358,14 +363,16 @@ else:
 		'transaction	bigint,'
 		'data_src	bigint,'
 		'branch_type	integer,'
-		'in_tx		boolean)')
+		'in_tx		boolean,'
+		'call_path_id	bigint)')
 
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	do_query(query, 'CREATE TABLE call_paths ('
 		'id		bigint		NOT NULL,'
 		'parent_id	bigint,'
 		'symbol_id	bigint,'
 		'ip		bigint)')
+if perf_db_export_calls:
 	do_query(query, 'CREATE TABLE calls ('
 		'id		bigint		NOT NULL,'
 		'thread_id	bigint,'

@@ -427,7 +434,7 @@ do_query(query, 'CREATE VIEW comm_threads_view AS '
 		'(SELECT tid FROM threads WHERE id = thread_id) AS tid'
 	' FROM comm_threads')
 
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	do_query(query, 'CREATE VIEW call_paths_view AS '
 		'SELECT '
 			'c.id,'

@@ -443,6 +450,7 @@ if perf_db_export_calls:
 			'(SELECT dso_id FROM symbols WHERE id = p.symbol_id) AS parent_dso_id,'
 			'(SELECT dso FROM symbols_view  WHERE id = p.symbol_id) AS parent_dso_short_name'
 		' FROM call_paths c INNER JOIN call_paths p ON p.id = c.parent_id')
+if perf_db_export_calls:
 	do_query(query, 'CREATE VIEW calls_view AS '
 		'SELECT '
 			'calls.id,'

@@ -540,8 +548,9 @@ dso_file		= open_output_file("dso_table.bin")
 symbol_file		= open_output_file("symbol_table.bin")
 branch_type_file	= open_output_file("branch_type_table.bin")
 sample_file		= open_output_file("sample_table.bin")
-if perf_db_export_calls:
+if perf_db_export_calls or perf_db_export_callchains:
 	call_path_file		= open_output_file("call_path_table.bin")
+if perf_db_export_calls:
 	call_file		= open_output_file("call_table.bin")
 
 def trace_begin():

@@ -553,8 +562,8 @@ def trace_begin():
 	comm_table(0, "unknown")
 	dso_table(0, 0, "unknown", "unknown", "")
 	symbol_table(0, 0, 0, 0, 0, "unknown")
-	sample_table(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
-	if perf_db_export_calls:
+	sample_table(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
+	if perf_db_export_calls or perf_db_export_callchains:
 		call_path_table(0, 0, 0, 0)
 
 unhandled_count = 0

@@ -570,8 +579,9 @@ def trace_end():
 	copy_output_file(symbol_file,		"symbols")
 	copy_output_file(branch_type_file,	"branch_types")
 	copy_output_file(sample_file,		"samples")
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		copy_output_file(call_path_file,	"call_paths")
+	if perf_db_export_calls:
 		copy_output_file(call_file,		"calls")
 
 	print datetime.datetime.today(), "Removing intermediate files..."

@@ -584,8 +594,9 @@ def trace_end():
 	remove_output_file(symbol_file)
 	remove_output_file(branch_type_file)
 	remove_output_file(sample_file)
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		remove_output_file(call_path_file)
+	if perf_db_export_calls:
 		remove_output_file(call_file)
 	os.rmdir(output_dir_name)
 	print datetime.datetime.today(), "Adding primary keys"

@@ -598,8 +609,9 @@ def trace_end():
 	do_query(query, 'ALTER TABLE symbols         ADD PRIMARY KEY (id)')
 	do_query(query, 'ALTER TABLE branch_types    ADD PRIMARY KEY (id)')
 	do_query(query, 'ALTER TABLE samples         ADD PRIMARY KEY (id)')
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		do_query(query, 'ALTER TABLE call_paths      ADD PRIMARY KEY (id)')
+	if perf_db_export_calls:
 		do_query(query, 'ALTER TABLE calls           ADD PRIMARY KEY (id)')
 
 	print datetime.datetime.today(), "Adding foreign keys"

@@ -622,10 +634,11 @@ def trace_end():
 					'ADD CONSTRAINT symbolfk   FOREIGN KEY (symbol_id)    REFERENCES symbols    (id),'
 					'ADD CONSTRAINT todsofk    FOREIGN KEY (to_dso_id)    REFERENCES dsos       (id),'
 					'ADD CONSTRAINT tosymbolfk FOREIGN KEY (to_symbol_id) REFERENCES symbols    (id)')
-	if perf_db_export_calls:
+	if perf_db_export_calls or perf_db_export_callchains:
 		do_query(query, 'ALTER TABLE call_paths '
 					'ADD CONSTRAINT parentfk    FOREIGN KEY (parent_id)    REFERENCES call_paths (id),'
 					'ADD CONSTRAINT symbolfk    FOREIGN KEY (symbol_id)    REFERENCES symbols    (id)')
+	if perf_db_export_calls:
 		do_query(query, 'ALTER TABLE calls '
 					'ADD CONSTRAINT threadfk    FOREIGN KEY (thread_id)    REFERENCES threads    (id),'
 					'ADD CONSTRAINT commfk      FOREIGN KEY (comm_id)      REFERENCES comms      (id),'

@@ -693,11 +706,11 @@ def branch_type_table(branch_type, name, *x):
 	value = struct.pack(fmt, 2, 4, branch_type, n, name)
 	branch_type_file.write(value)
 
-def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, branch_type, in_tx, *x):
+def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, branch_type, in_tx, call_path_id, *x):
 	if branches:
-		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiiiB", 17, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 4, branch_type, 1, in_tx)
+		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiiiBiq", 18, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 4, branch_type, 1, in_tx, 8, call_path_id)
 	else:
-		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiqiiiB", 21, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src, 4, branch_type, 1, in_tx)
+		value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiqiiiBiq", 22, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src, 4, branch_type, 1, in_tx, 8, call_path_id)
 	sample_file.write(value)
 
 def call_path_table(cp_id, parent_id, symbol_id, ip, *x):

Signature · 1 line

2.7.4

This patch set contains a set of changes to allow the export of sampled callchains, and to associate them with samples, via the Python db export API and export-to-postgresql.py script.

Call path information is currently only available in the database when call/return info is available, but not when doing normal sampling. These changes make this information available for normal sampling runs as well.

Patches 2-6 are required to make this information available in the database.

Patch 1 is needed to fix an existing issue where callchains are processed incorrectly which can cause the other patches to export incorrect call paths for a small percentage of samples (depending on the workload).

Chris Phlipot (6):
perf tools: fix incorrect ordering of callchain entries perf tools: refractor code to move call path handling out of
thread-stack
perf script: enable db export to output sampled callchains perf script: add call path id to exported sample in db export perf script: expose usage of the callchain db export via the python
api
perf script: update export-to-postgresql to support callchain export

 tools/perf/scripts/python/export-to-postgresql.py  |  47 ++++---
 tools/perf/util/Build                              |   1 +
 tools/perf/util/call-path.c                        | 122 ++++++++++++++++++
 tools/perf/util/call-path.h                        |  77 ++++++++++++
 tools/perf/util/db-export.c                        |  89 +++++++++++++
 tools/perf/util/db-export.h                        |   3 +
 tools/perf/util/machine.c                          |  56 ++++++---
 .../util/scripting-engines/trace-event-python.c    |  36 +++++-
 tools/perf/util/thread-stack.c                     | 139 +--------------------
 tools/perf/util/thread-stack.h                     |  31 ++---
 10 files changed, 408 insertions(+), 193 deletions(-)
 create mode 100644 tools/perf/util/call-path.c
 create mode 100644 tools/perf/util/call-path.h