[GIT PULL 0/6] perf/urgent fixes

[GIT PULL 0/6] perf/urgent fixes

[Arnaldo Carvalho de Melo]

Hi Ingo,

        Please consider pulling from:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux perf/urgent

Regards,

- Arnaldo

Anton Blanchard (4):
  perf symbols: Fix ppc64 SEGV in dso__load_sym with debuginfo files
  perf symbols: /proc/kallsyms does not sort module symbols
  perf symbols: Preserve symbol scope when parsing /proc/kallsyms
  perf symbols: Add some heuristics for choosing the best duplicate symbol

David Ahern (1):
  perf record: Create events initially disabled and enable after init

Masami Hiramatsu (1):
  perf probe: Fix regression of variable finder

 tools/perf/builtin-record.c    |    3 +
 tools/perf/util/evlist.c       |   13 ++++
 tools/perf/util/evlist.h       |    1 +
 tools/perf/util/probe-finder.c |    2 +-
 tools/perf/util/symbol.c       |  129 ++++++++++++++++++++++++++++++++--------
 5 files changed, 123 insertions(+), 25 deletions(-)

[PATCH 1/6] perf probe: Fix regression of variable finder

[Arnaldo Carvalho de Melo]

From: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>

Fix to call convert_variable() if previous call does not fail.

To call convert_variable, it ensures "ret" is 0. However, since
"ret" has the return value of synthesize_perf_probe_arg() which
always returns positive value if it succeeded, perf probe doesn't
call convert_variable(). This will cause a SEGV when we add an
event with arguments.

This has to be fixed as it ensures "ret" is greater than 0
(or not negative).

This regression has been introduced by my previous patch, f182e3e1.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110820053922.3286.65805.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/probe-finder.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c
index 555fc38..5d73262 100644
--- a/tools/perf/util/probe-finder.c
+++ b/tools/perf/util/probe-finder.c
@@ -659,7 +659,7 @@ static int find_variable(Dwarf_Die *sc_die, struct probe_finder *pf)
 		if (!die_find_variable_at(&pf->cu_die, pf->pvar->var, 0, &vr_die))
 			ret = -ENOENT;
 	}
-	if (ret == 0)
+	if (ret >= 0)
 		ret = convert_variable(&vr_die, pf);
 
 	if (ret < 0)
-- 
1.6.2.5

[PATCH 2/6] perf symbols: Fix ppc64 SEGV in dso__load_sym with debuginfo files

[Arnaldo Carvalho de Melo]

From: Anton Blanchard <anton@samba.org>

64bit PowerPC debuginfo files have an empty function descriptor section.
I hit a SEGV when perf tried to use this section for symbol resolution.

To fix this we need to check the section is valid and we can do this by
checking for type SHT_PROGBITS.

Cc: <stable@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Eric B Munson <emunson@mgebm.net>
Link: http://lkml.kernel.org/r/20110824065242.895239970@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/symbol.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 469c026..bb5d32f 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -1111,6 +1111,8 @@ static int dso__load_sym(struct dso *dso, struct map *map, const char *name,
 	}
 
 	opdsec = elf_section_by_name(elf, &ehdr, &opdshdr, ".opd", &opdidx);
+	if (opdshdr.sh_type != SHT_PROGBITS)
+		opdsec = NULL;
 	if (opdsec)
 		opddata = elf_rawdata(opdsec, NULL);
 
-- 
1.6.2.5

[PATCH 3/6] perf symbols: /proc/kallsyms does not sort module symbols

[Arnaldo Carvalho de Melo]

From: Anton Blanchard <anton@samba.org>

kallsyms__parse assumes that /proc/kallsyms is sorted and sets the end
of the previous symbol to the start of the current one.

Unfortunately module symbols are not sorted, eg:

ffffffffa0081f30 t e1000_clean_rx_irq   [e1000e]
ffffffffa00817a0 t e1000_alloc_rx_buffers       [e1000e]

Some symbols end up with a negative length and others have a length
larger than they should. This results in confusing perf output.

We already have a function to fixup the end of zero length symbols so
use that instead.

Cc: Eric B Munson <emunson@mgebm.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110824065242.969681349@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/symbol.c |   33 +++++++++++----------------------
 1 files changed, 11 insertions(+), 22 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index bb5d32f..f119e85 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -438,18 +438,11 @@ int kallsyms__parse(const char *filename, void *arg,
 	char *line = NULL;
 	size_t n;
 	int err = -1;
-	u64 prev_start = 0;
-	char prev_symbol_type = 0;
-	char *prev_symbol_name;
 	FILE *file = fopen(filename, "r");
 
 	if (file == NULL)
 		goto out_failure;
 
-	prev_symbol_name = malloc(KSYM_NAME_LEN);
-	if (prev_symbol_name == NULL)
-		goto out_close;
-
 	err = 0;
 
 	while (!feof(file)) {
@@ -480,24 +473,18 @@ int kallsyms__parse(const char *filename, void *arg,
 			break;
 		}
 
-		if (prev_symbol_type) {
-			u64 end = start;
-			if (end != prev_start)
-				--end;
-			err = process_symbol(arg, prev_symbol_name,
-					     prev_symbol_type, prev_start, end);
-			if (err)
-				break;
-		}
-
-		memcpy(prev_symbol_name, symbol_name, len + 1);
-		prev_symbol_type = symbol_type;
-		prev_start = start;
+		/*
+		 * module symbols are not sorted so we add all
+		 * symbols with zero length and rely on
+		 * symbols__fixup_end() to fix it up.
+		 */
+		err = process_symbol(arg, symbol_name,
+				     symbol_type, start, start);
+		if (err)
+			break;
 	}
 
-	free(prev_symbol_name);
 	free(line);
-out_close:
 	fclose(file);
 	return err;
 
@@ -703,6 +690,8 @@ int dso__load_kallsyms(struct dso *dso, const char *filename,
 	if (dso__load_all_kallsyms(dso, filename, map) < 0)
 		return -1;
 
+	symbols__fixup_end(&dso->symbols[map->type]);
+
 	if (dso->kernel == DSO_TYPE_GUEST_KERNEL)
 		dso->symtab_type = SYMTAB__GUEST_KALLSYMS;
 	else
-- 
1.6.2.5

[PATCH 4/6] perf symbols: Preserve symbol scope when parsing /proc/kallsyms

[Arnaldo Carvalho de Melo]

From: Anton Blanchard <anton@samba.org>

kallsyms__parse capitalises the symbol type, so every symbol is marked
global. Remove this and fix symbol_type__is_a to handle both local and
global symbols.

Cc: Eric B Munson <emunson@mgebm.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110824065243.077125989@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/symbol.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index f119e85..0d94ddb 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -74,11 +74,13 @@ static void dso__set_sorted_by_name(struct dso *dso, enum map_type type)
 
 bool symbol_type__is_a(char symbol_type, enum map_type map_type)
 {
+	symbol_type = toupper(symbol_type);
+
 	switch (map_type) {
 	case MAP__FUNCTION:
 		return symbol_type == 'T' || symbol_type == 'W';
 	case MAP__VARIABLE:
-		return symbol_type == 'D' || symbol_type == 'd';
+		return symbol_type == 'D';
 	default:
 		return false;
 	}
@@ -463,7 +465,7 @@ int kallsyms__parse(const char *filename, void *arg,
 		if (len + 2 >= line_len)
 			continue;
 
-		symbol_type = toupper(line[len]);
+		symbol_type = line[len];
 		len += 2;
 		symbol_name = line + len;
 		len = line_len - len;
-- 
1.6.2.5

[PATCH 5/6] perf symbols: Add some heuristics for choosing the best duplicate symbol

[Arnaldo Carvalho de Melo]

From: Anton Blanchard <anton@samba.org>

Try and pick the best symbol based on a few heuristics:

-  Prefer a non weak symbol over a weak one
-  Prefer a global symbol over a non global one
-  Prefer a symbol with less underscores (idea taken from kallsyms.c)
-  If all else fails, choose the symbol with the longest name

Cc: Eric B Munson <emunson@mgebm.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110824065243.161953371@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/symbol.c |   88 ++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 88 insertions(+), 0 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 0d94ddb..870b3b8 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -86,6 +86,92 @@ bool symbol_type__is_a(char symbol_type, enum map_type map_type)
 	}
 }
 
+static int prefix_underscores_count(const char *str)
+{
+	const char *tail = str;
+
+	while (*tail == '_')
+		tail++;
+
+	return tail - str;
+}
+
+#define SYMBOL_A 0
+#define SYMBOL_B 1
+
+static int choose_best_symbol(struct symbol *syma, struct symbol *symb)
+{
+	s64 a;
+	s64 b;
+
+	/* Prefer a symbol with non zero length */
+	a = syma->end - syma->start;
+	b = symb->end - symb->start;
+	if ((b == 0) && (a > 0))
+		return SYMBOL_A;
+	else if ((a == 0) && (b > 0))
+		return SYMBOL_B;
+
+	/* Prefer a non weak symbol over a weak one */
+	a = syma->binding == STB_WEAK;
+	b = symb->binding == STB_WEAK;
+	if (b && !a)
+		return SYMBOL_A;
+	if (a && !b)
+		return SYMBOL_B;
+
+	/* Prefer a global symbol over a non global one */
+	a = syma->binding == STB_GLOBAL;
+	b = symb->binding == STB_GLOBAL;
+	if (a && !b)
+		return SYMBOL_A;
+	if (b && !a)
+		return SYMBOL_B;
+
+	/* Prefer a symbol with less underscores */
+	a = prefix_underscores_count(syma->name);
+	b = prefix_underscores_count(symb->name);
+	if (b > a)
+		return SYMBOL_A;
+	else if (a > b)
+		return SYMBOL_B;
+
+	/* If all else fails, choose the symbol with the longest name */
+	if (strlen(syma->name) >= strlen(symb->name))
+		return SYMBOL_A;
+	else
+		return SYMBOL_B;
+}
+
+static void symbols__fixup_duplicate(struct rb_root *symbols)
+{
+	struct rb_node *nd;
+	struct symbol *curr, *next;
+
+	nd = rb_first(symbols);
+
+	while (nd) {
+		curr = rb_entry(nd, struct symbol, rb_node);
+again:
+		nd = rb_next(&curr->rb_node);
+		next = rb_entry(nd, struct symbol, rb_node);
+
+		if (!nd)
+			break;
+
+		if (curr->start != next->start)
+			continue;
+
+		if (choose_best_symbol(curr, next) == SYMBOL_A) {
+			rb_erase(&next->rb_node, symbols);
+			goto again;
+		} else {
+			nd = rb_next(&curr->rb_node);
+			rb_erase(&curr->rb_node, symbols);
+		}
+	}
+}
+
 static void symbols__fixup_end(struct rb_root *symbols)
 {
 	struct rb_node *nd, *prevnd = rb_first(symbols);
@@ -692,6 +778,7 @@ int dso__load_kallsyms(struct dso *dso, const char *filename,
 	if (dso__load_all_kallsyms(dso, filename, map) < 0)
 		return -1;
 
+	symbols__fixup_duplicate(&dso->symbols[map->type]);
 	symbols__fixup_end(&dso->symbols[map->type]);
 
 	if (dso->kernel == DSO_TYPE_GUEST_KERNEL)
@@ -1269,6 +1356,7 @@ new_symbol:
 	 * For misannotated, zeroed, ASM function sizes.
 	 */
 	if (nr > 0) {
+		symbols__fixup_duplicate(&dso->symbols[map->type]);
 		symbols__fixup_end(&dso->symbols[map->type]);
 		if (kmap) {
 			/*
-- 
1.6.2.5

[PATCH 6/6] perf record: Create events initially disabled and enable after init

[Arnaldo Carvalho de Melo]

From: David Ahern <dsahern@gmail.com>

perf-record currently creates events enabled. When doing a system wide
collection (-a arg) this causes data collection for perf's
initialization activities -- eg., perf_event__synthesize_threads().

For some events (e.g., context switch S/W event or tracepoints like
syscalls) perf's initialization causes a lot of events to be captured
frequently generating "Check IO/CPU overload!" warnings on larger
systems (e.g., 2 socket, quad core, hyperthreading).

perf's initialization phase can be skipped by creating events
disabled and then enabling them once the initialization is done.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1314289075-14706-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-record.c |    3 +++
 tools/perf/util/evlist.c    |   13 +++++++++++++
 tools/perf/util/evlist.h    |    1 +
 3 files changed, 17 insertions(+), 0 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 6b0519f..f4c3fbe 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -161,6 +161,7 @@ static void config_attr(struct perf_evsel *evsel, struct perf_evlist *evlist)
 	struct perf_event_attr *attr = &evsel->attr;
 	int track = !evsel->idx; /* only the first counter needs these */
 
+	attr->disabled		= 1;
 	attr->inherit		= !no_inherit;
 	attr->read_format	= PERF_FORMAT_TOTAL_TIME_ENABLED |
 				  PERF_FORMAT_TOTAL_TIME_RUNNING |
@@ -671,6 +672,8 @@ static int __cmd_record(int argc, const char **argv)
 		}
 	}
 
+	perf_evlist__enable(evsel_list);
+
 	/*
 	 * Let the child rip
 	 */
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index c12bd47..72e9f48 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -113,6 +113,19 @@ void perf_evlist__disable(struct perf_evlist *evlist)
 	}
 }
 
+void perf_evlist__enable(struct perf_evlist *evlist)
+{
+	int cpu, thread;
+	struct perf_evsel *pos;
+
+	for (cpu = 0; cpu < evlist->cpus->nr; cpu++) {
+		list_for_each_entry(pos, &evlist->entries, node) {
+			for (thread = 0; thread < evlist->threads->nr; thread++)
+				ioctl(FD(pos, cpu, thread), PERF_EVENT_IOC_ENABLE);
+		}
+	}
+}
+
 int perf_evlist__alloc_pollfd(struct perf_evlist *evlist)
 {
 	int nfds = evlist->cpus->nr * evlist->threads->nr * evlist->nr_entries;
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index ce85ae9..f349150 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -54,6 +54,7 @@ int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite);
 void perf_evlist__munmap(struct perf_evlist *evlist);
 
 void perf_evlist__disable(struct perf_evlist *evlist);
+void perf_evlist__enable(struct perf_evlist *evlist);
 
 static inline void perf_evlist__set_maps(struct perf_evlist *evlist,
 					 struct cpu_map *cpus,
-- 
1.6.2.5