perf/core improvements and fixes

Hi Ingo,

Please consider pulling, more to come,

- Arnaldo

The following changes since commit 107eb964d8c04417e8bce9e9ec2ed61d9261aec6:

Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2015-03-13 13:30:10 +0100)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo

for you to fetch changes up to 0c8c20779c5d56b93b8cb4cd30ba129a927ab437:

perf report: Don't allow empty argument for '-t'. (2015-03-19 13:53:28 -0300)

----------------------------------------------------------------
perf/core improvements and fixes:

User visible:

- Bash completion for subcommands (Yunlong Song)

- Allow annotating entries in callchains in the hists browser (top/report).
TODO: give some visual cue to what entries in callchains have samples and thus can be annotated and/or allow showing the source code for functions without samples (Arnaldo Carvalho de Melo)

- Don't allow empty argument for '-t' in perf report, fixing segfault (Wang Nan)

Infrastructure:

- Prep work for moving the perf feature tests build system to tools/build (Jiri Olsa)

- Fix perf-read-vdsox32 not building and lib64 install dir (H.J. Lu)

- ARM64: fix building error and eh/debug frame offset cache fixes (Wang Nan)

----------------------------------------------------------------
Arnaldo Carvalho de Melo (4):
perf hists browser: Simplify symbol annotation menu setup perf hists browser: Fix up some branch alignment perf hists: Remove hist_entry->used, not used anymore perf hists browser: Allow annotating entries in callchains

H.J. Lu (1):
perf tools: Fix perf-read-vdsox32 not building and lib64 install dir

Jiri Olsa (10):
perf build: Disable default check for libbabeltrace perf build: Move features build output under features directory perf build: Fix pthread-attr-setaffinity-np include in test-all perf build: Get rid of LIB_INCLUDE variable
perf build: Rename CORE_FEATURE_TESTS to FEATURE_TESTS perf build: Get rid of VF_FEATURE_TESTS
perf build: Rename display_lib into feature_display perf build: Rename display_vf to feature_verbose perf build: Rename PERF-FEATURES into FEATURE-DUMP perf build: Rename feature_print_var_code to print_var_code

Wang Nan (3):
perf tools: Fix building error for arm64. perf callchain: Separate eh/debug frame offset cache. perf report: Don't allow empty argument for '-t'.

Yunlong Song (13):
perf tools: Fix the bash completion for listing options of perf subcommand
perf tools: Fix the bash completion for listing subsubcommands of perf subcommand perf tools: Provide the right bash completion for listing options of perf subcommand subsubcommand perf tools: Fix the bash completion for listing events of perf subcommand record|stat|top -e perf tools: Fix the bash completion to support listing events for --event
perf tools: Fix the bash completion for listing subcommands of perf
perf tools: Add the bash completion for listing subsubcommands of perf data
perf tools: Add the bash completion for listing subsubcommands of perf help
perf tools: Add the bash completion for listing subsubcommands of perf script perf tools: Add the bash completion for listing subsubcommands of perf test
perf tools: Add the bash completion for listing subsubcommands of perf timechart perf tools: Add the bash completion for listing subsubcommands of perf trace
perf tools: Avoid confusion with preloaded bash function for perf bash completion

 tools/perf/Makefile.perf                    |   3 +-
 tools/perf/builtin-data.c                   |   6 +-
 tools/perf/builtin-help.c                   |  17 ++++-
 tools/perf/builtin-report.c                 |   2 +-
 tools/perf/builtin-script.c                 |   5 +-
 tools/perf/builtin-timechart.c              |   7 +-
 tools/perf/builtin-trace.c                  |   7 +-
 tools/perf/config/Makefile                  | 111 ++++++++++++----------------
 tools/perf/config/Makefile.arch             |   2 +-
 tools/perf/config/feature-checks/Makefile   |   7 +-
 tools/perf/config/feature-checks/test-all.c |  11 ++-
 tools/perf/perf-completion.sh               |  88 +++++++++++++++++-----
 tools/perf/tests/builtin-test.c             |   5 +-
 tools/perf/tests/make                       |   2 +-
 tools/perf/ui/browsers/annotate.c           |   8 +-
 tools/perf/ui/browsers/hists.c              |  59 +++++++--------
 tools/perf/util/dso.h                       |   3 +-
 tools/perf/util/hist.c                      |   8 +-
 tools/perf/util/hist.h                      |   9 +++
 tools/perf/util/parse-options.c             |  21 +++++-
 tools/perf/util/parse-options.h             |   2 +
 tools/perf/util/sort.h                      |   1 -
 tools/perf/util/unwind-libunwind.c          |   8 +-
 23 files changed, 237 insertions(+), 155 deletions(-)

From: Arnaldo Carvalho de Melo <[email protected]>

No need to repeat some tests, skip annotation instead.

---
 tools/perf/ui/browsers/hists.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)

index 49eddeb81458..f69371b7f49b 100644
@@ -1612,19 +1612,22 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 		if (!sort__has_sym)
 			goto add_exit_option;
 
+		if (browser->selection == NULL)
+			goto skip_annotation;
+
 		if (sort__mode == SORT_MODE__BRANCH) {
 			bi = browser->he_selection->branch_info;
-			if (browser->selection != NULL &&
-			    bi &&
-			    bi->from.sym != NULL &&
+
+			if (bi == NULL)
+				goto skip_annotation;
+
+			if (bi->from.sym != NULL &&
 			    !bi->from.map->dso->annotate_warned &&
 				asprintf(&options[nr_options], "Annotate %s",
 					 bi->from.sym->name) > 0)
 				annotate_f = nr_options++;
 
-			if (browser->selection != NULL &&
-			    bi &&
-			    bi->to.sym != NULL &&
+			if (bi->to.sym != NULL &&
 			    !bi->to.map->dso->annotate_warned &&
 			    (bi->to.sym != bi->from.sym ||
 			     bi->to.map->dso != bi->from.map->dso) &&

@@ -1632,8 +1635,7 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 					 bi->to.sym->name) > 0)
 				annotate_t = nr_options++;
 		} else {
-			if (browser->selection != NULL &&
-			    browser->selection->sym != NULL &&
+			if (browser->selection->sym != NULL &&
 			    !browser->selection->map->dso->annotate_warned) {
 				struct annotation *notes;
 

@@ -1645,7 +1647,7 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 					annotate = nr_options++;
 			}
 		}
-
+skip_annotation:
 		if (thread != NULL &&
 		    asprintf(&options[nr_options], "Zoom %s %s(%d) thread",
 			     (browser->hists->thread_filter ? "out of" : "into"),

1.9.3

From: Arnaldo Carvalho de Melo <[email protected]>

Those asprintf return checks should be aligned with the other conditionals, fix it.

Also add {} blocks to further clarify.

Cc: Adrian Hunter <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Don Zickus <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Stephane Eranian <[email protected]>
echo Link: http://lkml.kernel.org/n/tip-ranpwd -l 24@git.kernel.org

---
 tools/perf/ui/browsers/hists.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

index f69371b7f49b..ebd9d3afe2d6 100644
@@ -1623,17 +1623,17 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 
 			if (bi->from.sym != NULL &&
 			    !bi->from.map->dso->annotate_warned &&
-				asprintf(&options[nr_options], "Annotate %s",
-					 bi->from.sym->name) > 0)
+			    asprintf(&options[nr_options], "Annotate %s", bi->from.sym->name) > 0) {
 				annotate_f = nr_options++;
+			}
 
 			if (bi->to.sym != NULL &&
 			    !bi->to.map->dso->annotate_warned &&
 			    (bi->to.sym != bi->from.sym ||
 			     bi->to.map->dso != bi->from.map->dso) &&
-				asprintf(&options[nr_options], "Annotate %s",
-					 bi->to.sym->name) > 0)
+			    asprintf(&options[nr_options], "Annotate %s", bi->to.sym->name) > 0) {
 				annotate_t = nr_options++;
+			}
 		} else {
 			if (browser->selection->sym != NULL &&
 			    !browser->selection->map->dso->annotate_warned) {

@@ -1643,8 +1643,9 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 
 				if (notes->src &&
 				    asprintf(&options[nr_options], "Annotate %s",
-						 browser->selection->sym->name) > 0)
+						 browser->selection->sym->name) > 0) {
 					annotate = nr_options++;
+				}
 			}
 		}
 skip_annotation:

1.9.3

From: Arnaldo Carvalho de Melo <[email protected]>

Since hist_entry__delete() nowadays doesn't actually frees anything that may be in use by the annotation code.

Eventually we will solve this for good by reference counting struct symbol.

---
 tools/perf/ui/browsers/hists.c | 5 -----
 tools/perf/util/hist.c         | 8 +-------
 tools/perf/util/sort.h         | 1 -
 3 files changed, 1 insertion(+), 13 deletions(-)

index ebd9d3afe2d6..c37e70ae0b20 100644
@@ -1731,12 +1731,7 @@ do_annotate:
 			if (!notes->src)
 				continue;
 
-			/*
-			 * Don't let this be freed, say, by hists__decay_entry.
-			 */
-			he->used = true;
 			err = hist_entry__tui_annotate(he, evsel, hbt);
-			he->used = false;
 			/*
 			 * offer option to annotate the other branch source or target
 			 * (if they exists) when returning from annotate

index d9a6d35eda17..cc22b9158b93 100644
@@ -263,15 +263,9 @@ void hists__decay_entries(struct hists *hists, bool zap_user, bool zap_kernel)
 	while (next) {
 		n = rb_entry(next, struct hist_entry, rb_node);
 		next = rb_next(&n->rb_node);
-		/*
-		 * We may be annotating this, for instance, so keep it here in
-		 * case some it gets new samples, we'll eventually free it when
-		 * the user stops browsing and it agains gets fully decayed.
-		 */
 		if (((zap_user && n->level == '.') ||
 		     (zap_kernel && n->level != '.') ||
-		     hists__decay_entry(hists, n)) &&
-		    !n->used) {
+		     hists__decay_entry(hists, n))) {
 			hists__delete_entry(hists, n);
 		}
 	}

index c03e4ff8beff..9c01b83eebca 100644
@@ -102,7 +102,6 @@ struct hist_entry {
 
 	bool			init_have_children;
 	char			level;
-	bool			used;
 	u8			filtered;
 	char			*srcline;
 	struct symbol		*parent;

1.9.3

From: Arnaldo Carvalho de Melo <[email protected]>

Instead of annotating just the top level hist_entry, allow instead annotating a map_symbol, i.e. the top level hist_entry or one of the callchains for which there were samples.

---
 tools/perf/ui/browsers/annotate.c |  8 +++++++-
 tools/perf/ui/browsers/hists.c    | 23 ++++++++++-------------
 tools/perf/util/hist.h            |  9 +++++++++
 3 files changed, 26 insertions(+), 14 deletions(-)

index 9d32e3c0cfee..e5250eb2dd57 100644
@@ -829,10 +829,16 @@ out:
 	return key;
 }
 
+int map_symbol__tui_annotate(struct map_symbol *ms, struct perf_evsel *evsel,
+			     struct hist_browser_timer *hbt)
+{
+	return symbol__tui_annotate(ms->sym, ms->map, evsel, hbt);
+}
+
 int hist_entry__tui_annotate(struct hist_entry *he, struct perf_evsel *evsel,
 			     struct hist_browser_timer *hbt)
 {
-	return symbol__tui_annotate(he->ms.sym, he->ms.map, evsel, hbt);
+	return map_symbol__tui_annotate(&he->ms, evsel, hbt);
 }
 
 static void annotate_browser__mark_jump_targets(struct annotate_browser *browser,

index c37e70ae0b20..cd7350aeb8e7 100644
@@ -1704,6 +1704,7 @@ retry_popup_menu:
 		if (choice == annotate || choice == annotate_t || choice == annotate_f) {
 			struct hist_entry *he;
 			struct annotation *notes;
+			struct map_symbol ms;
 			int err;
 do_annotate:
 			if (!objdump_path && perf_session_env__lookup_objdump(env))

@@ -1713,25 +1714,21 @@ do_annotate:
 			if (he == NULL)
 				continue;
 
-			/*
-			 * we stash the branch_info symbol + map into the
-			 * the ms so we don't have to rewrite all the annotation
-			 * code to use branch_info.
-			 * in branch mode, the ms struct is not used
-			 */
 			if (choice == annotate_f) {
-				he->ms.sym = he->branch_info->from.sym;
-				he->ms.map = he->branch_info->from.map;
-			}  else if (choice == annotate_t) {
-				he->ms.sym = he->branch_info->to.sym;
-				he->ms.map = he->branch_info->to.map;
+				ms.map = he->branch_info->from.map;
+				ms.sym = he->branch_info->from.sym;
+			} else if (choice == annotate_t) {
+				ms.map = he->branch_info->to.map;
+				ms.sym = he->branch_info->to.sym;
+			} else {
+				ms = *browser->selection;
 			}
 
-			notes = symbol__annotation(he->ms.sym);
+			notes = symbol__annotation(ms.sym);
 			if (!notes->src)
 				continue;
 
-			err = hist_entry__tui_annotate(he, evsel, hbt);
+			err = map_symbol__tui_annotate(&ms, evsel, hbt);
 			/*
 			 * offer option to annotate the other branch source or target
 			 * (if they exists) when returning from annotate

index e988c9fcd1bc..9f31b89a527a 100644
@@ -303,6 +303,9 @@ struct hist_browser_timer {
 
 #ifdef HAVE_SLANG_SUPPORT
 #include "../ui/keysyms.h"
+int map_symbol__tui_annotate(struct map_symbol *ms, struct perf_evsel *evsel,
+			     struct hist_browser_timer *hbt);
+
 int hist_entry__tui_annotate(struct hist_entry *he, struct perf_evsel *evsel,
 			     struct hist_browser_timer *hbt);
 

@@ -321,6 +324,12 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
 {
 	return 0;
 }
+static inline int map_symbol__tui_annotate(struct map_symbol *ms __maybe_unused,
+					   struct perf_evsel *evsel __maybe_unused,
+					   struct hist_browser_timer *hbt __maybe_unused)
+{
+	return 0;
+}
 
 static inline int hist_entry__tui_annotate(struct hist_entry *he __maybe_unused,
 					   struct perf_evsel *evsel __maybe_unused,

1.9.3

From: Wang Nan <[email protected]>

Commit b11db6581beaccef8ae9a388ae96074aa5cc144f ("perf tools: Fix build error on ARCH=i386/x86_64/sparc64") uses sed on ARCH, which triggers a bug in sequence of sed expression, where 's/arm.*/arm/' will replace 'arm64' to 'arm', causes arm64 building failure.

This patch prevent 'arm64' to be mached for 'arm.*' case.

---
 tools/perf/config/Makefile.arch | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index e9720571341d..e11fbd6fae78 100644
@@ -4,7 +4,7 @@ endif
 
 ARCH := $(shell echo $(ARCH) | sed -e s/i.86/x86/ -e s/x86_64/x86/ \
                                   -e s/sun4u/sparc/ -e s/sparc64/sparc/ \
-                                  -e s/arm.*/arm/ -e s/sa110/arm/ \
+                                  -e /arm64/!s/arm.*/arm/ -e s/sa110/arm/ \
                                   -e s/s390x/s390/ -e s/parisc64/parisc/ \
                                   -e s/ppc.*/powerpc/ -e s/mips.*/mips/ \
                                   -e s/sh[234].*/sh/ -e s/aarch64.*/arm64/ \

1.9.3

From: Jiri Olsa <[email protected]>

Remove libbabeltrace check from default features set, because the requested version is not released yet in most distributions. We'll enable later.

Calling libbabeltrace check manually via feature_check before $(feature-libbabeltrace) is used.

---
 tools/perf/config/Makefile                  | 11 +++++------
 tools/perf/config/feature-checks/Makefile   |  2 +-
 tools/perf/config/feature-checks/test-all.c |  9 ++++++++-
 3 files changed, 14 insertions(+), 8 deletions(-)

index bd097187724b..b799649ab0e7 100644
@@ -227,7 +227,6 @@ CORE_FEATURE_TESTS =			\
 	stackprotector-all		\
 	timerfd				\
 	libdw-dwarf-unwind		\
-	libbabeltrace			\
 	zlib
 
 LIB_FEATURE_TESTS =			\

@@ -243,7 +242,6 @@ LIB_FEATURE_TESTS =			\
 	libslang			\
 	libunwind			\
 	libdw-dwarf-unwind		\
-	libbabeltrace			\
 	zlib
 
 VF_FEATURE_TESTS =			\

@@ -705,14 +703,15 @@ else
 endif
 
 ifndef NO_LIBBABELTRACE
-  ifeq ($(feature-libbabeltrace), 0)
-    msg := $(warning No libbabeltrace found, disables 'perf data' CTF format support, please install libbabeltrace-dev[el]/libbabeltrace-ctf-dev);
-    NO_LIBBABELTRACE := 1
-  else
+  $(call feature_check,libbabeltrace)
+  ifeq ($(feature-libbabeltrace), 1)
     CFLAGS += -DHAVE_LIBBABELTRACE_SUPPORT $(LIBBABELTRACE_CFLAGS)
     LDFLAGS += $(LIBBABELTRACE_LDFLAGS)
     EXTLIBS += -lbabeltrace-ctf
     $(call detected,CONFIG_LIBBABELTRACE)
+  else
+    msg := $(warning No libbabeltrace found, disables 'perf data' CTF format support, please install libbabeltrace-dev[el]/libbabeltrace-ctf-dev);
+    NO_LIBBABELTRACE := 1
   endif
 endif
 

index 8fe067864957..b6e2e312ba54 100644
@@ -44,7 +44,7 @@ BUILD = $(CC) $(CFLAGS) -Wall -Werror -o $(OUTPUT)$@ $(patsubst %.bin,%.c,$@) $(
 ###############################
 
 test-all.bin:
-	$(BUILD) -fstack-protector-all -O2 -D_FORTIFY_SOURCE=2 -ldw -lelf -lnuma -lelf -laudit -I/usr/include/slang -lslang $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null) $(FLAGS_PERL_EMBED) $(FLAGS_PYTHON_EMBED) -DPACKAGE='"perf"' -lbfd -ldl -lz -lbabeltrace
+	$(BUILD) -fstack-protector-all -O2 -D_FORTIFY_SOURCE=2 -ldw -lelf -lnuma -lelf -laudit -I/usr/include/slang -lslang $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null) $(FLAGS_PERL_EMBED) $(FLAGS_PYTHON_EMBED) -DPACKAGE='"perf"' -lbfd -ldl -lz
 
 test-hello.bin:
 	$(BUILD)

index 1ffc3da5ca10..06f5c8a67cb2 100644
@@ -101,9 +101,17 @@
 # include "test-pthread_attr_setaffinity_np.c"
 #undef main
 
+# if 0
+/*
+ * Disable libbabeltrace check for test-all, because the requested
+ * library version is not released yet in most distributions. Will
+ * reenable later.
+ */
+
 #define main main_test_libbabeltrace
 # include "test-libbabeltrace.c"
 #undef main
+#endif
 
 int main(int argc, char *argv[])
 {

@@ -130,7 +138,6 @@ int main(int argc, char *argv[])
 	main_test_sync_compare_and_swap(argc, argv);
 	main_test_zlib();
 	main_test_pthread_attr_setaffinity_np();
-	main_test_libbabeltrace();
 
 	return 0;
 }

1.9.3

From: Jiri Olsa <[email protected]>

Following commit introduced features build dump:
443a70541c56 perf tools: Output feature detection's gcc output to a file

Moving them into to have code more compact and renaming build dump files. For each feature 'test-X' new file 'test-X.make.output' is created and contains the build out. It's created in the same directory as the feature itself.

---
 tools/perf/Makefile.perf                  | 1 -
 tools/perf/config/Makefile                | 2 +-
 tools/perf/config/feature-checks/Makefile | 5 +++--
 3 files changed, 4 insertions(+), 4 deletions(-)

index 933d70345f87..ec4c063ed9f3 100644
@@ -529,7 +529,6 @@ clean: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean config-clean
 	$(Q)$(RM) .config-detected
 	$(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf perf-read-vdso32 perf-read-vdsox32
 	$(call QUIET_CLEAN, core-gen)   $(RM)  *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope* $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)PERF-FEATURES $(OUTPUT)util/*-bison* $(OUTPUT)util/*-flex*
-	$(call QUIET_CLEAN, feature-detect) $(RM) $(OUTPUT)config/feature-checks/.make-*.output
 	$(QUIET_SUBDIR0)Documentation $(QUIET_SUBDIR1) clean
 	$(python-clean)
 

index b799649ab0e7..45f61253ccce 100644
@@ -184,7 +184,7 @@ endif
 
 feature_check = $(eval $(feature_check_code))
 define feature_check_code
-  feature-$(1) := $(shell $(MAKE) OUTPUT=$(OUTPUT_FEATURES) CFLAGS="$(EXTRA_CFLAGS) $(FEATURE_CHECK_CFLAGS-$(1))" LDFLAGS="$(LDFLAGS) $(FEATURE_CHECK_LDFLAGS-$(1))" -C config/feature-checks test-$1.bin > $(OUTPUT)config/feature-checks/.make-$(1).output 2>&1 && echo 1 || echo 0)
+  feature-$(1) := $(shell $(MAKE) OUTPUT=$(OUTPUT_FEATURES) CFLAGS="$(EXTRA_CFLAGS) $(FEATURE_CHECK_CFLAGS-$(1))" LDFLAGS="$(LDFLAGS) $(FEATURE_CHECK_LDFLAGS-$(1))" -C config/feature-checks test-$1.bin >/dev/null 2>/dev/null && echo 1 || echo 0)
 endef
 
 feature_set = $(eval $(feature_set_code))

index b6e2e312ba54..0d694a94cda2 100644
@@ -39,7 +39,8 @@ PKG_CONFIG := $(CROSS_COMPILE)pkg-config
 
 all: $(FILES)
 
-BUILD = $(CC) $(CFLAGS) -Wall -Werror -o $(OUTPUT)$@ $(patsubst %.bin,%.c,$@) $(LDFLAGS)
+__BUILD = $(CC) $(CFLAGS) -Wall -Werror -o $(OUTPUT)$@ $(patsubst %.bin,%.c,$@) $(LDFLAGS)
+  BUILD = $(__BUILD) > $(OUTPUT)$(@:.bin=.make.output) 2>&1
 
 ###############################
 

@@ -156,4 +157,4 @@ test-zlib.bin:
 ###############################
 
 clean:
-	rm -f $(FILES) *.d
+	rm -f $(FILES) *.d $(FILES:.bin=.make.output)

1.9.3

From: Jiri Olsa <[email protected]>

The test-all fails to build due to type in pthread-attr-setaffinity-np include.

---
 tools/perf/config/feature-checks/test-all.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index 06f5c8a67cb2..a61c20456705 100644
@@ -98,7 +98,7 @@
 #undef main
 
 #define main main_test_pthread_attr_setaffinity_np
-# include "test-pthread_attr_setaffinity_np.c"
+# include "test-pthread-attr-setaffinity-np.c"
 #undef main
 
 # if 0

1.9.3

From: Jiri Olsa <[email protected]>

It has no use, so we can directly use the value for CFLAGS.

---
 tools/perf/config/Makefile | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

index 45f61253ccce..467106bc5fab 100644
@@ -15,7 +15,6 @@ $(shell echo -n > .config-detected)
 detected     = $(shell echo "$(1)=y"       >> .config-detected)
 detected_var = $(shell echo "$(1)=$($(1))" >> .config-detected)
 
-LIB_INCLUDE := $(srctree)/tools/lib/
 CFLAGS := $(EXTRA_CFLAGS) $(EXTRA_WARNINGS)
 
 include $(src-perf)/config/Makefile.arch

@@ -320,7 +319,7 @@ endif
 
 CFLAGS += -I$(src-perf)/util
 CFLAGS += -I$(src-perf)
-CFLAGS += -I$(LIB_INCLUDE)
+CFLAGS += -I$(srctree)/tools/lib/
 
 CFLAGS += -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE
 

1.9.3

From: Jiri Olsa <[email protected]>

Preparing for feature checks separation, moving related stuff under 'FEATURE*' namespace.

---
 tools/perf/config/Makefile | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

index 467106bc5fab..507350566e7f 100644
@@ -203,7 +203,7 @@ endef
 #   the rule that uses them - an example for that is the 'bionic'
 #   feature check. ]
 #
-CORE_FEATURE_TESTS =			\
+FEATURE_TESTS =			\
 	backtrace			\
 	dwarf				\
 	fortify-source			\

@@ -262,7 +262,7 @@ VF_FEATURE_TESTS =			\
 	compile-32			\
 	compile-x32
 
-# Set FEATURE_CHECK_(C|LD)FLAGS-all for all CORE_FEATURE_TESTS features.
+# Set FEATURE_CHECK_(C|LD)FLAGS-all for all FEATURE_TESTS features.
 # If in the future we need per-feature checks/flags for features not
 # mentioned in this list we need to refactor this ;-).
 set_test_all_flags = $(eval $(set_test_all_flags_code))

@@ -271,7 +271,7 @@ define set_test_all_flags_code
   FEATURE_CHECK_LDFLAGS-all += $(FEATURE_CHECK_LDFLAGS-$(1))
 endef
 
-$(foreach feat,$(CORE_FEATURE_TESTS),$(call set_test_all_flags,$(feat)))
+$(foreach feat,$(FEATURE_TESTS),$(call set_test_all_flags,$(feat)))
 
 #
 # Special fast-path for the 'all features are available' case:

@@ -286,10 +286,10 @@ ifeq ($(feature-all), 1)
   #
   # test-all.c passed - just set all the core feature flags to 1:
   #
-  $(foreach feat,$(CORE_FEATURE_TESTS),$(call feature_set,$(feat)))
+  $(foreach feat,$(FEATURE_TESTS),$(call feature_set,$(feat)))
 else
-  $(shell $(MAKE) OUTPUT=$(OUTPUT_FEATURES) CFLAGS="$(EXTRA_CFLAGS)" LDFLAGS=$(LDFLAGS) -i -j -C config/feature-checks $(addsuffix .bin,$(CORE_FEATURE_TESTS)) >/dev/null 2>&1)
-  $(foreach feat,$(CORE_FEATURE_TESTS),$(call feature_check,$(feat)))
+  $(shell $(MAKE) OUTPUT=$(OUTPUT_FEATURES) CFLAGS="$(EXTRA_CFLAGS)" LDFLAGS=$(LDFLAGS) -i -j -C config/feature-checks $(addsuffix .bin,$(FEATURE_TESTS)) >/dev/null 2>&1)
+  $(foreach feat,$(FEATURE_TESTS),$(call feature_check,$(feat)))
 endif
 
 ifeq ($(feature-stackprotector-all), 1)

1.9.3

From: Jiri Olsa <[email protected]>

It only contains (FEATURE_TESTS - FEATURE_DISPLAY) tests to display the rest of the checks on 'make VF=1'. But we can actually compute this list, which is less confusing.

Also renaming LIB_FEATURE_TESTS into FEATURE_DISPLAY, so it reflects what this variable actually does - display its tests status to user.

---
 tools/perf/config/Makefile | 32 +++++++-------------------------
 1 file changed, 7 insertions(+), 25 deletions(-)

index 507350566e7f..198609e1d696 100644
@@ -228,7 +228,7 @@ FEATURE_TESTS =			\
 	libdw-dwarf-unwind		\
 	zlib
 
-LIB_FEATURE_TESTS =			\
+FEATURE_DISPLAY =			\
 	dwarf				\
 	glibc				\
 	gtk2				\

@@ -243,25 +243,6 @@ LIB_FEATURE_TESTS =			\
 	libdw-dwarf-unwind		\
 	zlib
 
-VF_FEATURE_TESTS =			\
-	backtrace			\
-	fortify-source			\
-	sync-compare-and-swap		\
-	gtk2-infobar			\
-	libelf-getphdrnum		\
-	libelf-mmap			\
-	libpython-version		\
-	pthread-attr-setaffinity-np	\
-	stackprotector-all		\
-	timerfd				\
-	libunwind-debug-frame		\
-	bionic				\
-	liberty				\
-	liberty-z			\
-	cplus-demangle			\
-	compile-32			\
-	compile-x32
-
 # Set FEATURE_CHECK_(C|LD)FLAGS-all for all FEATURE_TESTS features.
 # If in the future we need per-feature checks/flags for features not
 # mentioned in this list we need to refactor this ;-).

@@ -805,7 +786,7 @@ define feature_print_text_code
     MSG = $(shell printf '...%30s: %s' $(1) $(2))
 endef
 
-PERF_FEATURES := $(foreach feat,$(LIB_FEATURE_TESTS),feature-$(feat)($(feature-$(feat))))
+PERF_FEATURES := $(foreach feat,$(FEATURE_DISPLAY),feature-$(feat)($(feature-$(feat))))
 PERF_FEATURES_FILE := $(shell touch $(OUTPUT)PERF-FEATURES; cat $(OUTPUT)PERF-FEATURES)
 
 ifeq ($(dwarf-post-unwind),1)

@@ -816,7 +797,7 @@ endif
 # output. It's set if:
 # - detected features differes from stored features from
 #   last build (in PERF-FEATURES file)
-# - one of the $(LIB_FEATURE_TESTS) is not detected
+# - one of the $(FEATURE_DISPLAY) is not detected
 # - VF is enabled
 
 ifneq ("$(PERF_FEATURES)","$(PERF_FEATURES_FILE)")

@@ -831,7 +812,7 @@ define feature_check_code
   endif
 endef
 
-$(foreach feat,$(LIB_FEATURE_TESTS),$(call feature_check,$(feat)))
+$(foreach feat,$(FEATURE_DISPLAY),$(call feature_check,$(feat)))
 
 ifeq ($(VF),1)
   display_lib := 1

@@ -841,7 +822,7 @@ endif
 ifeq ($(display_lib),1)
   $(info )
   $(info Auto-detecting system features:)
-  $(foreach feat,$(LIB_FEATURE_TESTS),$(call feature_print_status,$(feat),))
+  $(foreach feat,$(FEATURE_DISPLAY),$(call feature_print_status,$(feat),))
 
   ifeq ($(dwarf-post-unwind),1)
     $(call feature_print_text,"DWARF post unwind library", $(dwarf-post-unwind-text))

@@ -849,7 +830,8 @@ ifeq ($(display_lib),1)
 endif
 
 ifeq ($(display_vf),1)
-  $(foreach feat,$(VF_FEATURE_TESTS),$(call feature_print_status,$(feat),))
+  TMP := $(filter-out $(FEATURE_DISPLAY),$(FEATURE_TESTS))
+  $(foreach feat,$(TMP),$(call feature_print_status,$(feat),))
   $(info )
   $(call feature_print_var,prefix)
   $(call feature_print_var,bindir)

1.9.3

From: Jiri Olsa <[email protected]>

Preparing for feature checks separation, moving related stuff under 'feature*' namespace.

---
 tools/perf/config/Makefile | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

index 198609e1d696..f80fc7ec8036 100644
@@ -793,7 +793,7 @@ ifeq ($(dwarf-post-unwind),1)
   PERF_FEATURES += dwarf-post-unwind($(dwarf-post-unwind-text))
 endif
 
-# The $(display_lib) controls the default detection message
+# The $(feature_display) controls the default detection message
 # output. It's set if:
 # - detected features differes from stored features from
 #   last build (in PERF-FEATURES file)

@@ -802,24 +802,24 @@ endif
 
 ifneq ("$(PERF_FEATURES)","$(PERF_FEATURES_FILE)")
   $(shell echo "$(PERF_FEATURES)" > $(OUTPUT)PERF-FEATURES)
-  display_lib := 1
+  feature_display := 1
 endif
 
 feature_check = $(eval $(feature_check_code))
 define feature_check_code
   ifneq ($(feature-$(1)), 1)
-    display_lib := 1
+    feature_display := 1
   endif
 endef
 
 $(foreach feat,$(FEATURE_DISPLAY),$(call feature_check,$(feat)))
 
 ifeq ($(VF),1)
-  display_lib := 1
+  feature_display := 1
   display_vf := 1
 endif
 
-ifeq ($(display_lib),1)
+ifeq ($(feature_display),1)
   $(info )
   $(info Auto-detecting system features:)
   $(foreach feat,$(FEATURE_DISPLAY),$(call feature_print_status,$(feat),))

@@ -841,7 +841,7 @@ ifeq ($(display_vf),1)
   $(call feature_print_var,LIBDW_DIR)
 endif
 
-ifeq ($(display_lib),1)
+ifeq ($(feature_display),1)
   $(info )
 endif
 

1.9.3

From: Jiri Olsa <[email protected]>

Preparing for feature checks separation, moving related stuff under 'feature*' namespace.

---
 tools/perf/config/Makefile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

index f80fc7ec8036..370ad509422b 100644
@@ -816,7 +816,7 @@ $(foreach feat,$(FEATURE_DISPLAY),$(call feature_check,$(feat)))
 
 ifeq ($(VF),1)
   feature_display := 1
-  display_vf := 1
+  feature_verbose := 1
 endif
 
 ifeq ($(feature_display),1)

@@ -829,7 +829,7 @@ ifeq ($(feature_display),1)
   endif
 endif
 
-ifeq ($(display_vf),1)
+ifeq ($(feature_verbose),1)
   TMP := $(filter-out $(FEATURE_DISPLAY),$(FEATURE_TESTS))
   $(foreach feat,$(TMP),$(call feature_print_status,$(feat),))
   $(info )

1.9.3

From: Jiri Olsa <[email protected]>

Preparing for feature checks separation, moving related stuff under 'FEATURE*' namespace.

---
 tools/perf/Makefile.perf   |  2 +-
 tools/perf/config/Makefile | 12 ++++++------
 2 files changed, 7 insertions(+), 7 deletions(-)

index ec4c063ed9f3..e323eab10694 100644
@@ -528,7 +528,7 @@ clean: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean config-clean
 	$(Q)find . -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
 	$(Q)$(RM) .config-detected
 	$(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf perf-read-vdso32 perf-read-vdsox32
-	$(call QUIET_CLEAN, core-gen)   $(RM)  *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope* $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)PERF-FEATURES $(OUTPUT)util/*-bison* $(OUTPUT)util/*-flex*
+	$(call QUIET_CLEAN, core-gen)   $(RM)  *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope* $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)FEATURE-DUMP $(OUTPUT)util/*-bison* $(OUTPUT)util/*-flex*
 	$(QUIET_SUBDIR0)Documentation $(QUIET_SUBDIR1) clean
 	$(python-clean)
 

index 370ad509422b..b98ab77ca179 100644
@@ -786,22 +786,22 @@ define feature_print_text_code
     MSG = $(shell printf '...%30s: %s' $(1) $(2))
 endef
 
-PERF_FEATURES := $(foreach feat,$(FEATURE_DISPLAY),feature-$(feat)($(feature-$(feat))))
-PERF_FEATURES_FILE := $(shell touch $(OUTPUT)PERF-FEATURES; cat $(OUTPUT)PERF-FEATURES)
+FEATURE_DUMP := $(foreach feat,$(FEATURE_DISPLAY),feature-$(feat)($(feature-$(feat))))
+FEATURE_DUMP_FILE := $(shell touch $(OUTPUT)FEATURE-DUMP; cat $(OUTPUT)FEATURE-DUMP)
 
 ifeq ($(dwarf-post-unwind),1)
-  PERF_FEATURES += dwarf-post-unwind($(dwarf-post-unwind-text))
+  FEATURE_DUMP += dwarf-post-unwind($(dwarf-post-unwind-text))
 endif
 
 # The $(feature_display) controls the default detection message
 # output. It's set if:
 # - detected features differes from stored features from
-#   last build (in PERF-FEATURES file)
+#   last build (in FEATURE-DUMP file)
 # - one of the $(FEATURE_DISPLAY) is not detected
 # - VF is enabled
 
-ifneq ("$(PERF_FEATURES)","$(PERF_FEATURES_FILE)")
-  $(shell echo "$(PERF_FEATURES)" > $(OUTPUT)PERF-FEATURES)
+ifneq ("$(FEATURE_DUMP)","$(FEATURE_DUMP_FILE)")
+  $(shell echo "$(FEATURE_DUMP)" > $(OUTPUT)FEATURE-DUMP)
   feature_display := 1
 endif
 

1.9.3

So I think the various capital letter files in $(cwd) are highly suboptimal:

FEATURE-DUMP
PERF-CFLAGS
PERF-VERSION-FILE

I think we should standardize on something less pushy (no shouting), plus more standardized, more logical, more coherent names, i.e. something like:

.build.features.cache
.build.cflags
.build.version

?

.build.features.* would be reserved to the generic feature detection machinery.

.build.cflags and .build.version could be used by the perf build - hopefully exported to tools/build/ as well in the future! :-)

That way the .gitignore could just ignore .build.* and we'd not have to update it much.

Btw., another small nit: could we please do a (git-)rename of tools/build/feature/ to tools/build/features/?

Plural is what is used typically for generic frameworks that deal with a set of items (features in this case) - for example we have tools/tests/, and tools/perf/scripts/, not tools/test/ and tools/perf/script/.

Thanks,

Ingo

we no longer have the PERF-CFLAGS file:
db8486626246 perf build: Remove PERF-CFLAGS file

forgot to update the .gitignore file, will do

ok, noted ;-)

agreed

ok, I'll try to address all this in the next build update

jirka

Em Sun, Mar 22, 2015 at 11:19:33AM +0100, Ingo Molnar escreveu:

Agreed, Jiri?

This one is easy enough, will do in my next pull req.

- Arnaldo

From: Jiri Olsa <[email protected]>

As it has nothing to do with features and won't be moved into tools/build.

---
 tools/perf/config/Makefile | 23 ++++++++++++-----------
 1 file changed, 12 insertions(+), 11 deletions(-)

index b98ab77ca179..489d33305b16 100644
@@ -776,8 +776,8 @@ define feature_print_status_code
   endif
 endef
 
-feature_print_var = $(eval $(feature_print_var_code)) $(info $(MSG))
-define feature_print_var_code
+print_var = $(eval $(print_var_code)) $(info $(MSG))
+define print_var_code
     MSG = $(shell printf '...%30s: %s' $(1) $($(1)))
 endef
 

@@ -827,21 +827,22 @@ ifeq ($(feature_display),1)
   ifeq ($(dwarf-post-unwind),1)
     $(call feature_print_text,"DWARF post unwind library", $(dwarf-post-unwind-text))
   endif
+
+  ifneq ($(feature_verbose),1)
+    $(info )
+  endif
 endif
 
 ifeq ($(feature_verbose),1)
   TMP := $(filter-out $(FEATURE_DISPLAY),$(FEATURE_TESTS))
   $(foreach feat,$(TMP),$(call feature_print_status,$(feat),))
   $(info )
-  $(call feature_print_var,prefix)
-  $(call feature_print_var,bindir)
-  $(call feature_print_var,libdir)
-  $(call feature_print_var,sysconfdir)
-  $(call feature_print_var,LIBUNWIND_DIR)
-  $(call feature_print_var,LIBDW_DIR)
-endif
-
-ifeq ($(feature_display),1)
+  $(call print_var,prefix)
+  $(call print_var,bindir)
+  $(call print_var,libdir)
+  $(call print_var,sysconfdir)
+  $(call print_var,LIBUNWIND_DIR)
+  $(call print_var,LIBDW_DIR)
   $(info )
 endif
 

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing options for 'perf kvm|kmem|mem|lock|sched --<TAB>', where 'kvm|kmem|mem|lock|sched' are all subcommands of perf.

Example:

Before this patch:

 $ perf kvm --<TAB>
 $

As shown above, the options of perf kvm does not come out.

After this patch:

 $ perf kvm --<TAB>
 --alloc    --caller   --input    --line     --raw-ip   --sort
 --verbose

As shown above, the options of perf kvm can come out now.

---
 tools/perf/perf-completion.sh | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

index c2595e9bd69b..4822ed38e23e 100644
@@ -119,15 +119,18 @@ __perf_main ()
 	elif [[ $prev == "-e" && "${words[1]}" == @(record|stat|top) ]]; then
 		evts=$($cmd list --raw-dump)
 		__perfcomp_colon "$evts" "$cur"
-	# List subcommands for perf commands
-	elif [[ $prev == @(kvm|kmem|mem|lock|sched) ]]; then
-		subcmds=$($cmd $prev --list-cmds)
-		__perfcomp_colon "$subcmds" "$cur"
-	# List long option names
-	elif [[ $cur == --* ]];  then
-		subcmd=${words[1]}
-		opts=$($cmd $subcmd --list-opts)
-		__perfcomp "$opts" "$cur"
+	else
+		# List subcommands for perf commands
+		if [[ $prev == @(kvm|kmem|mem|lock|sched) ]]; then
+			subcmds=$($cmd $prev --list-cmds)
+			__perfcomp_colon "$subcmds" "$cur"
+		fi
+		# List long option names
+		if [[ $cur == --* ]];  then
+			subcmd=${words[1]}
+			opts=$($cmd $subcmd --list-opts)
+			__perfcomp "$opts" "$cur"
+		fi
 	fi
 }
 

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf kvm|kmem|mem|lock|sched --<long option> <TAB>', where 'kvm|kmem|mem| lock|sched' are all subcommands of perf.

Example:

Before this patch:

 $ perf kvm --verbose <TAB>
 $

As shown above, the subsubcommands of perf kvm does not come out.

After this patch:

 $ perf kvm --verbose <TAB>
 buildid-list  diff          record        report        stat
 top

As shown above, the subsubcommands of perf kvm can come out now.

---
 tools/perf/perf-completion.sh | 23 +++++++++++++++++++++--
 1 file changed, 21 insertions(+), 2 deletions(-)

index 4822ed38e23e..bbb61d0b79cd 100644
@@ -100,6 +100,23 @@ __perfcomp_colon ()
 	__ltrim_colon_completions $cur
 }
 
+__perf_prev_skip_opts ()
+{
+	local i cmd_ cmds_
+
+	let i=cword-1
+	cmds_=$($cmd --list-cmds)
+	prev_skip_opts=()
+	while [ $i -ge 0 ]; do
+		for cmd_ in $cmds_; do
+			if [[ ${words[i]} == $cmd_ ]]; then
+				prev_skip_opts=${words[i]}
+				return
+			fi
+		done
+		((i--))
+	done
+}
 __perf_main ()
 {
 	local cmd

@@ -107,6 +124,8 @@ __perf_main ()
 	cmd=${words[0]}
 	COMPREPLY=()
 
+	# Skip options backward and find the last perf command
+	__perf_prev_skip_opts
 	# List perf subcommands or long options
 	if [ $cword -eq 1 ]; then
 		if [[ $cur == --* ]]; then

@@ -121,8 +140,8 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev == @(kvm|kmem|mem|lock|sched) ]]; then
-			subcmds=$($cmd $prev --list-cmds)
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched) ]]; then
+			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi
 		# List long option names

1.9.3

From: Yunlong Song <[email protected]>

The bash completion gives wrong options for 'perf kvm|kmem|mem|lock| sched subsubcommand --<TAB>', where 'kvm|kmem|mem|lock|sched' are all subcommands of perf and 'subsubcommand' is a subcommand of 'kvm|kmem|mem |lock|sched'. In fact, the result incorrectly lists the bash completion of 'perf subcommand' rather than 'perf subcommand subsubcommand'.

Example:

Before this patch:

 $ perf kvm record --<TAB>
 --guest          --guestkallsyms  --guestmodules   --guestmount
 --guestvmlinux   --host           --input          --output
 --verbose

As shown above, the result is the options of kvm rather than record.

After this patch:

 $ perf kvm record --<TAB>
 --all-cpus          --cgroup            --delay             --group
 --no-buildid        --output            --quiet             --stat
 --uid
 --branch-any        --count             --event             --intr-regs
 --no-buildid-cache  --period            --raw-samples       --tid
 --verbose
 --branch-filter     --cpu               --filter            --mmap-pages
 --no-inherit        --per-thread        --realtime          --timestamp
 --weight
 --call-graph        --data              --freq
 --no-buffering      --no-samples        --pid
 --running-time      --transaction

As shown above, the result is exactly the options of record as we wished.

---
 tools/perf/perf-completion.sh | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

index bbb61d0b79cd..01ce841f155b 100644
@@ -105,9 +105,12 @@ __perf_prev_skip_opts ()
 	local i cmd_ cmds_
 
 	let i=cword-1
-	cmds_=$($cmd --list-cmds)
+	cmds_=$($cmd $1 --list-cmds)
 	prev_skip_opts=()
 	while [ $i -ge 0 ]; do
+		if [[ ${words[i]} == $1 ]]; then
+			return
+		fi
 		for cmd_ in $cmds_; do
 			if [[ ${words[i]} == $cmd_ ]]; then
 				prev_skip_opts=${words[i]}

@@ -146,7 +149,9 @@ __perf_main ()
 		fi
 		# List long option names
 		if [[ $cur == --* ]];  then
-			subcmd=${words[1]}
+			subcmd=$prev_skip_opts
+			__perf_prev_skip_opts $subcmd
+			subcmd=$subcmd" "$prev_skip_opts
 			opts=$($cmd $subcmd --list-opts)
 			__perfcomp "$opts" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing events for 'perf kvm|kmem| mem|lock|sched record|stat|top -e <TAB>', where 'kvm|kmem|mem|lock|sched' are all subcommands of perf.

Example:

Before this patch:

 $ perf kvm record -e <TAB>
 $

As shown above, the events of record does not come out.

After this patch:

 $ perf kvm record -e <TAB>
 alignment-faults                   cpu/instructions/
 L1-dcache-prefetch-misses          node-prefetches
 uncore_rbox_0/qpi0_idle_filt/
 branch-instructions                cpu/mem-loads/
 L1-dcache-store-misses             node-prefetch-misses
 uncore_rbox_0/qpi1_date_response/
 branch-load-misses                 cpu-migrations
 L1-dcache-stores                   node-store-misses
 uncore_rbox_0/qpi1_filt_send/
 branch-loads                       dTLB-load-misses
 L1-icache-load-misses              node-stores
 uncore_rbox_0/qpi1_idle_filt/
 ...

As shown above, the events of record can come out now.

---
 tools/perf/perf-completion.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index 01ce841f155b..4b58ac2ae578 100644
@@ -138,7 +138,7 @@ __perf_main ()
 		fi
 		__perfcomp "$cmds" "$cur"
 	# List possible events for -e option
-	elif [[ $prev == "-e" && "${words[1]}" == @(record|stat|top) ]]; then
+	elif [[ $prev == "-e" && $prev_skip_opts == @(record|stat|top) ]]; then
 		evts=$($cmd list --raw-dump)
 		__perfcomp_colon "$evts" "$cur"
 	else

1.9.3

From: Yunlong Song <[email protected]>

The bash completion only supports -e rather than --event, so fix it.

Example:

Before this patch:

 $ perf record --event <TAB>
 $

As shown above, the events of record does not come out.

After this patch:

 $ perf record --event <TAB>
 lignment-faults                   cpu/instructions/
 L1-dcache-prefetch-misses          node-prefetches
 uncore_rbox_0/qpi0_idle_filt/
 branch-instructions                cpu/mem-loads/
 L1-dcache-store-misses             node-prefetch-misses
 uncore_rbox_0/qpi1_date_response/
 branch-load-misses                 cpu-migrations
 L1-dcache-stores                   node-store-misses
 uncore_rbox_0/qpi1_filt_send/
 branch-loads                       dTLB-load-misses
 L1-icache-load-misses              node-stores
 uncore_rbox_0/qpi1_idle_filt/
 ...

As shown above, the events of record can come out now.

---
 tools/perf/perf-completion.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

index 4b58ac2ae578..3e25d3ea17f9 100644
@@ -137,8 +137,8 @@ __perf_main ()
 			cmds=$($cmd --list-cmds)
 		fi
 		__perfcomp "$cmds" "$cur"
-	# List possible events for -e option
-	elif [[ $prev == "-e" && $prev_skip_opts == @(record|stat|top) ]]; then
+	# List possible events for -e and --event option
+	elif [[ $prev == @("-e"|"--event") && $prev_skip_opts == @(record|stat|top) ]]; then
 		evts=$($cmd list --raw-dump)
 		__perfcomp_colon "$evts" "$cur"
 	else

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subcommands for 'perf --<long option> <TAB>'.

Example:

Before this patch:

 $ perf --debug <TAB>
 $

As shown above, the subcommands of perf does not come out.

After this patch:

 $ perf --debug <TAB>
 annotate       buildid-cache  data           evlist         inject
 kvm            lock           probe          report         script
 test           top            version
 bench          buildid-list   diff           help           kmem
 list           mem            record         sched          stat
 timechart      trace

As shown above, the subcommands of perf can come out now.

---
 tools/perf/perf-completion.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index 3e25d3ea17f9..7b98ae48f0bf 100644
@@ -130,7 +130,7 @@ __perf_main ()
 	# Skip options backward and find the last perf command
 	__perf_prev_skip_opts
 	# List perf subcommands or long options
-	if [ $cword -eq 1 ]; then
+	if [ -z $prev_skip_opts ]; then
 		if [[ $cur == --* ]]; then
 			cmds=$($cmd --list-opts)
 		else

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf data <TAB>', so fix it.

Example:

Before this patch:

 $ perf data <TAB>
 $

As shown above, the subsubcommands of perf data does not come out.

After this patch:

 $ perf data <TAB>
 convert

As shown above, the subsubcommands of perf data can come out now.

---
 tools/perf/builtin-data.c     | 6 ++++--
 tools/perf/perf-completion.sh | 2 +-
 2 files changed, 5 insertions(+), 3 deletions(-)

index 155cf75b8199..709152a7b408 100644
@@ -22,7 +22,9 @@ static const struct option data_options[] = {
 	OPT_END()
 };
 
-static const char * const data_usage[] = {
+static const char * const data_subcommands[] = { "convert", NULL };
+
+static const char *data_usage[] = {
 	"perf data [<common options>] <command> [<options>]",
 	NULL
 };

@@ -98,7 +100,7 @@ int cmd_data(int argc, const char **argv, const char *prefix)
 	if (argc < 2)
 		goto usage;
 
-	argc = parse_options(argc, argv, data_options, data_usage,
+	argc = parse_options_subcommand(argc, argv, data_options, data_subcommands, data_usage,
 			     PARSE_OPT_STOP_AT_NON_OPTION);
 	if (argc < 1)
 		goto usage;

index 7b98ae48f0bf..535ff72babc7 100644
@@ -143,7 +143,7 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf help <TAB>', so fix it.

Example:

Before this patch:

 $ perf help <TAB>
 $

As shown above, the subsubcommands of perf help does not come out.

After this patch:

 $ perf help <TAB>
 annotate       buildid-cache  data           evlist         inject
 kvm            lock           probe          report         script
 test           top
 bench          buildid-list   diff           help           kmem
 list           mem            record         sched          stat
 timechart      trace

As shown above, the subsubcommands of perf help can come out now.

---
 tools/perf/builtin-help.c     | 17 ++++++++++++++---
 tools/perf/perf-completion.sh |  2 +-
 2 files changed, 15 insertions(+), 4 deletions(-)

index 25d20628212e..36486eade1ef 100644
@@ -437,7 +437,18 @@ int cmd_help(int argc, const char **argv, const char *prefix __maybe_unused)
 			HELP_FORMAT_INFO),
 	OPT_END(),
 	};
-	const char * const builtin_help_usage[] = {
+	const char * const builtin_help_subcommands[] = {
+		"buildid-cache", "buildid-list", "diff", "evlist", "help", "list",
+		"record", "report", "bench", "stat", "timechart", "top", "annotate",
+		"script", "sched", "kmem", "lock", "kvm", "test", "inject", "mem", "data",
+#ifdef HAVE_LIBELF_SUPPORT
+		"probe",
+#endif
+#ifdef HAVE_LIBAUDIT_SUPPORT
+		"trace",
+#endif
+	NULL };
+	const char *builtin_help_usage[] = {
 		"perf help [--all] [--man|--web|--info] [command]",
 		NULL
 	};

@@ -448,8 +459,8 @@ int cmd_help(int argc, const char **argv, const char *prefix __maybe_unused)
 
 	perf_config(perf_help_config, &help_format);
 
-	argc = parse_options(argc, argv, builtin_help_options,
-			builtin_help_usage, 0);
+	argc = parse_options_subcommand(argc, argv, builtin_help_options,
+			builtin_help_subcommands, builtin_help_usage, 0);
 
 	if (show_all) {
 		printf("\n usage: %s\n\n", perf_usage_string);

index 535ff72babc7..dbca6a66f401 100644
@@ -143,7 +143,7 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf script <TAB>', so fix it.

Example:

Before this patch:

 $ perf script <TAB>
 $

As shown above, the subsubcommands of perf script does not come out.

After this patch:

 $ perf script <TAB>
 record  report

As shown above, the subsubcommands of perf script can come out now.

---
 tools/perf/builtin-script.c   | 5 +++--
 tools/perf/perf-completion.sh | 2 +-
 2 files changed, 4 insertions(+), 3 deletions(-)

index c7e6750923ef..f2a348b57b8f 100644
@@ -1572,7 +1572,8 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 		    "Show the mmap events"),
 	OPT_END()
 	};
-	const char * const script_usage[] = {
+	const char * const script_subcommands[] = { "record", "report", NULL };
+	const char *script_usage[] = {
 		"perf script [<options>]",
 		"perf script [<options>] record <script> [<record-options>] <command>",
 		"perf script [<options>] report <script> [script-args]",

@@ -1586,7 +1587,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 
 	setup_scripting();
 
-	argc = parse_options(argc, argv, options, script_usage,
+	argc = parse_options_subcommand(argc, argv, options, script_subcommands, script_usage,
 			     PARSE_OPT_STOP_AT_NON_OPTION);
 
 	file.path = input_name;

index dbca6a66f401..8739a1345017 100644
@@ -143,7 +143,7 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf test <TAB>', so fix it.

Example:

Before this patch:

 $ perf test <TAB>
 $

As shown above, the subsubcommands of perf test does not come out.

After this patch:

 $ perf test <TAB>
 list

As shown above, the subsubcommands of perf test can come out now.

---
 tools/perf/perf-completion.sh   | 2 +-
 tools/perf/tests/builtin-test.c | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)

index 8739a1345017..35f43fd7a38b 100644
@@ -143,7 +143,7 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script|test) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

index 4b7d9ab0f049..ed8e05c6839a 100644
@@ -291,7 +291,7 @@ static int perf_test__list(int argc, const char **argv)
 
 int cmd_test(int argc, const char **argv, const char *prefix __maybe_unused)
 {
-	const char * const test_usage[] = {
+	const char *test_usage[] = {
 	"perf test [<options>] [{list <test-name-fragment>|[<test-name-fragments>|<test-numbers>]}]",
 	NULL,
 	};

@@ -302,13 +302,14 @@ int cmd_test(int argc, const char **argv, const char *prefix __maybe_unused)
 		    "be more verbose (show symbol address, etc)"),
 	OPT_END()
 	};
+	const char * const test_subcommands[] = { "list", NULL };
 	struct intlist *skiplist = NULL;
         int ret = hists__init();
 
         if (ret < 0)
                 return ret;
 
-	argc = parse_options(argc, argv, test_options, test_usage, 0);
+	argc = parse_options_subcommand(argc, argv, test_options, test_subcommands, test_usage, 0);
 	if (argc >= 1 && !strcmp(argv[0], "list"))
 		return perf_test__list(argc, argv);
 

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf timechart <TAB>', so fix it.

Example:

Before this patch:

 $ perf timechart <TAB>
 $

As shown above, the subsubcommands of perf timechart does not come out.

After this patch:

 $ perf timechart <TAB>
 record

As shown above, the subsubcommands of perf timechart can come out now.

---
 tools/perf/builtin-timechart.c | 7 ++++---
 tools/perf/perf-completion.sh  | 2 +-
 2 files changed, 5 insertions(+), 4 deletions(-)

index 51440d1fc722..494b3bbe5ea4 100644
@@ -1958,7 +1958,8 @@ int cmd_timechart(int argc, const char **argv,
 		     parse_time),
 	OPT_END()
 	};
-	const char * const timechart_usage[] = {
+	const char * const timechart_subcommands[] = { "record", NULL };
+	const char *timechart_usage[] = {
 		"perf timechart [<options>] {record}",
 		NULL
 	};

@@ -1976,8 +1977,8 @@ int cmd_timechart(int argc, const char **argv,
 		"perf timechart record [<options>]",
 		NULL
 	};
-	argc = parse_options(argc, argv, timechart_options, timechart_usage,
-			PARSE_OPT_STOP_AT_NON_OPTION);
+	argc = parse_options_subcommand(argc, argv, timechart_options, timechart_subcommands,
+			timechart_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 
 	if (tchart.power_only && tchart.tasks_only) {
 		pr_err("-P and -T options cannot be used at the same time.\n");

index 35f43fd7a38b..a33d2ef53ae4 100644
@@ -143,7 +143,7 @@ __perf_main ()
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script|test) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script|test|timechart) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

The bash completion does not support listing subsubcommands for 'perf trace <TAB>', so fix it.

Example:

Before this patch:

 $ perf trace <TAB>
 $

As shown above, the subsubcommands of perf trace does not come out.

After this patch:

 $ perf trace <TAB>
 record

As shown above, the subsubcommands of perf trace can come out now.

---
 tools/perf/builtin-trace.c    | 7 ++++---
 tools/perf/perf-completion.sh | 9 ++++++---
 2 files changed, 10 insertions(+), 6 deletions(-)

index 6969ba98ff2f..0b3b4e4984b9 100644
@@ -2609,7 +2609,7 @@ static void evlist__set_evsel_handler(struct perf_evlist *evlist, void *handler)
 
 int cmd_trace(int argc, const char **argv, const char *prefix __maybe_unused)
 {
-	const char * const trace_usage[] = {
+	const char *trace_usage[] = {
 		"perf trace [<options>] [<command>]",
 		"perf trace [<options>] -- <command> [<options>]",
 		"perf trace record [<options>] [<command>]",

@@ -2684,6 +2684,7 @@ int cmd_trace(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_BOOLEAN(0, "syscalls", &trace.trace_syscalls, "Trace syscalls"),
 	OPT_END()
 	};
+	const char * const trace_subcommands[] = { "record", NULL };
 	int err;
 	char bf[BUFSIZ];
 

@@ -2699,8 +2700,8 @@ int cmd_trace(int argc, const char **argv, const char *prefix __maybe_unused)
 		goto out;
 	}
 
-	argc = parse_options(argc, argv, trace_options, trace_usage,
-			     PARSE_OPT_STOP_AT_NON_OPTION);
+	argc = parse_options_subcommand(argc, argv, trace_options, trace_subcommands,
+				 trace_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 
 	if (trace.trace_pgfaults) {
 		trace.opts.sample_address = true;

index a33d2ef53ae4..bdd40350870a 100644
@@ -120,6 +120,7 @@ __perf_prev_skip_opts ()
 		((i--))
 	done
 }
+
 __perf_main ()
 {
 	local cmd

@@ -137,13 +138,15 @@ __perf_main ()
 			cmds=$($cmd --list-cmds)
 		fi
 		__perfcomp "$cmds" "$cur"
-	# List possible events for -e and --event option
-	elif [[ $prev == @("-e"|"--event") && $prev_skip_opts == @(record|stat|top) ]]; then
+	# List possible events for -e option
+	elif [[ $prev == @("-e"|"--event") &&
+		$prev_skip_opts == @(record|stat|top) ]]; then
 		evts=$($cmd list --raw-dump)
 		__perfcomp_colon "$evts" "$cur"
 	else
 		# List subcommands for perf commands
-		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|data|help|script|test|timechart) ]]; then
+		if [[ $prev_skip_opts == @(kvm|kmem|mem|lock|sched|
+			|data|help|script|test|timechart|trace) ]]; then
 			subcmds=$($cmd $prev_skip_opts --list-cmds)
 			__perfcomp_colon "$subcmds" "$cur"
 		fi

1.9.3

From: Yunlong Song <[email protected]>

Since some functions (e.g. '_get_comp_words_by_ref()') in perf bash completion script are originally taken from git bash completion script, these functions may be preloaded before perf bash completion script runs.

In order to avoid repeating loading the same function twice, some test constraints are used before these function definitions in the perf bash completion script (e.g. 'type _get_comp_words_by_ref &>/dev/null ||').

The problem is that, if these functions in perf bash completion script are changed for some reason, perf will still use the preloaded bash functions rather than the customized functions of its own.

As a result, the perf bash completion will behave incorrectly. To get rid of this problem, a flag can be defined to determine the proper situation.

And to avoid overwriting the preloaded functions, the names of these functions in perf bash completion script should be renamed to the perf-customized ones.

Example:

Before this patch:

 $ type _get_comp_words_by_ref
 _get_comp_words_by_ref is a function
 _get_comp_words_by_ref ()
 {
     local exclude flag i OPTIND=1;
     local cur cword words=();
     local upargs=() upvars=() vcur vcword vprev vwords;
     while getopts "c:i:n:p:w:" flag "$@"; do
         case $flag in
             c)
                 vcur=$OPTARG
             ;;
             i)
                 vcword=$OPTARG
             ;;
             n)
                 exclude=$OPTARG
             ;;
             p)
                 vprev=$OPTARG
             ;;
             w)
                 vwords=$OPTARG
             ;;
         esac;
     done;
     while [[ $# -ge $OPTIND ]]; do
         case ${!OPTIND} in
             cur)
                 vcur=cur
             ;;
             prev)
                 vprev=prev
             ;;
             cword)
                 vcword=cword
             ;;
             words)
                 vwords=words
             ;;
             *)
                 echo "bash: $FUNCNAME(): \`${!OPTIND}': unknown argument" 1>&2;
                 return 1
             ;;
         esac;
         let "OPTIND += 1";
     done;
     __get_cword_at_cursor_by_ref "$exclude" words cword cur;
     [[ -n $vcur ]] && {
         upvars+=("$vcur");
         upargs+=(-v $vcur "$cur")
     };
     [[ -n $vcword ]] && {
         upvars+=("$vcword");
         upargs+=(-v $vcword "$cword")
     };
     [[ -n $vprev && $cword -ge 1 ]] && {
         upvars+=("$vprev");
         upargs+=(-v $vprev "${words[cword - 1]}")
     };
     [[ -n $vwords ]] && {
         upvars+=("$vwords");
         upargs+=(-a${#words[@]} $vwords "${words[@]}")
     };
     (( ${#upvars[@]} )) && local "${upvars[@]}" && _upvars "${upargs[@]}"
 }

As shown above, the _get_comp_words_by_ref is the preloaded function in fact, rather than the function defined in perf-completion.sh. So if we happen to change the function for some reason, the result will behave in a wrong state.

After this patch:

We can set preload_get_comp_words_by_ref="false" to not use the preloaded function. Instead, it will use the function defined in perf-completion.sh, which is renamed as __perf_get_comp_words_by_ref to avoid overwriting the preloaded function _get_comp_words_by_ref.

 $ type __perf_get_comp_words_by_ref
 __perf_get_comp_words_by_ref is a function
 __perf_get_comp_words_by_ref ()
 {
     local exclude cur_ words_ cword_;
     if [ "$1" = "-n" ]; then
         exclude=$2;
         shift 2;
     fi;
     __my_reassemble_comp_words_by_ref "$exclude";
     cur_=${words_[cword_]};
     while [ $# -gt 0 ]; do
         case "$1" in
             cur)
                 cur=$cur_
             ;;
             prev)
                 prev=${words_[$cword_-1]}
             ;;
             words)
                 words=("${words_[@]}")
             ;;
             cword)
                 cword=$cword_
             ;;
         esac;
         shift;
     done
 }

As shown above, the function __perf_get_comp_words_by_ref is loaded and can work this time.

Note that we do not change the original behavior when those functions are not preloaded before perf bash completion script runs. In this case, although the flag is set to "true", the code will still change it to "false" to use the function defined in perf-completion.sh.

---
 tools/perf/perf-completion.sh | 36 ++++++++++++++++++++++++++++++------
 1 file changed, 30 insertions(+), 6 deletions(-)

index bdd40350870a..3ba80b2359cc 100644
@@ -47,8 +47,16 @@ __my_reassemble_comp_words_by_ref()
 	done
 }
 
-type _get_comp_words_by_ref &>/dev/null ||
-_get_comp_words_by_ref()
+# Define preload_get_comp_words_by_ref="false", if the function
+# __perf_get_comp_words_by_ref() is required instead.
+preload_get_comp_words_by_ref="true"
+
+if [ $preload_get_comp_words_by_ref = "true" ]; then
+	type _get_comp_words_by_ref &>/dev/null ||
+	preload_get_comp_words_by_ref="false"
+fi
+[ $preload_get_comp_words_by_ref = "true" ] ||
+__perf_get_comp_words_by_ref()
 {
 	local exclude cur_ words_ cword_
 	if [ "$1" = "-n" ]; then

@@ -76,8 +84,16 @@ _get_comp_words_by_ref()
 	done
 }
 
-type __ltrim_colon_completions &>/dev/null ||
-__ltrim_colon_completions()
+# Define preload__ltrim_colon_completions="false", if the function
+# __perf__ltrim_colon_completions() is required instead.
+preload__ltrim_colon_completions="true"
+
+if [ $preload__ltrim_colon_completions = "true" ]; then
+	type __ltrim_colon_completions &>/dev/null ||
+	preload__ltrim_colon_completions="false"
+fi
+[ $preload__ltrim_colon_completions = "true" ] ||
+__perf__ltrim_colon_completions()
 {
 	if [[ "$1" == *:* && "$COMP_WORDBREAKS" == *:* ]]; then
 		# Remove colon-word prefix from COMPREPLY items

@@ -97,7 +113,11 @@ __perfcomp ()
 __perfcomp_colon ()
 {
 	__perfcomp "$1" "$2"
-	__ltrim_colon_completions $cur
+	if [ $preload__ltrim_colon_completions = "true" ]; then
+		__ltrim_colon_completions $cur
+	else
+		__perf__ltrim_colon_completions $cur
+	fi
 }
 
 __perf_prev_skip_opts ()

@@ -226,7 +246,11 @@ type perf &>/dev/null &&
 _perf()
 {
 	local cur words cword prev
-	_get_comp_words_by_ref -n =: cur words cword prev
+	if [ $preload_get_comp_words_by_ref = "true" ]; then
+		_get_comp_words_by_ref -n =: cur words cword prev
+	else
+		__perf_get_comp_words_by_ref -n =: cur words cword prev
+	fi
 	__perf_main
 } &&
 

1.9.3

From: Wang Nan <[email protected]>

Commit f1f13af99a90 ("perf callchain: Cache eh/debug frame offset for dwarf unwind") introduces a cache for .debug_frame and .eh_frame_hdr. Unfortunately, it makes them share a same cache (dso->frame_offset). Which causes unwind failure on ARM:

   $ perf test unwind
  Test dwarf unwind: FAILED!

The reason is that, if a dso has '.debug_frame' but doesn't have '.eh_frame_hdr' (like ARM), dso->frame_offset will be filled by offset of '.debug_frame' during the first time calling of find_proc_info() -> read_unwind_spec_debug_frame(), and be regarded to '.eh_frame_hdr' when the second time calling of find_proc_info() -> read_unwind_spec_eh_frame(), since '.eh_frame_hdr' is checked prior to '.debug_frame'.

This patch solves the problem by creating two cache fields for '.eh_frame_hdr' and '.debug_frame'.

---
 tools/perf/util/dso.h              | 3 ++-
 tools/perf/util/unwind-libunwind.c | 8 ++++----
 2 files changed, 6 insertions(+), 5 deletions(-)

index ced92841ff97..408c65f1a757 100644
@@ -139,7 +139,8 @@ struct dso {
 		u32		 status_seen;
 		size_t		 file_size;
 		struct list_head open_entry;
-		u64		 frame_offset;
+		u64		 debug_frame_offset;
+		u64		 eh_frame_hdr_offset;
 	} data;
 
 	union { /* Tool specific area */

index e3c40a520a25..7b09a443a280 100644
@@ -266,7 +266,7 @@ static int read_unwind_spec_eh_frame(struct dso *dso, struct machine *machine,
 				     u64 *fde_count)
 {
 	int ret = -EINVAL, fd;
-	u64 offset = dso->data.frame_offset;
+	u64 offset = dso->data.eh_frame_hdr_offset;
 
 	if (offset == 0) {
 		fd = dso__data_fd(dso, machine);

@@ -275,7 +275,7 @@ static int read_unwind_spec_eh_frame(struct dso *dso, struct machine *machine,
 
 		/* Check the .eh_frame section for unwinding info */
 		offset = elf_section_offset(fd, ".eh_frame_hdr");
-		dso->data.frame_offset = offset;
+		dso->data.eh_frame_hdr_offset = offset;
 	}
 
 	if (offset)

@@ -291,7 +291,7 @@ static int read_unwind_spec_debug_frame(struct dso *dso,
 					struct machine *machine, u64 *offset)
 {
 	int fd;
-	u64 ofs = dso->data.frame_offset;
+	u64 ofs = dso->data.debug_frame_offset;
 
 	if (ofs == 0) {
 		fd = dso__data_fd(dso, machine);

@@ -300,7 +300,7 @@ static int read_unwind_spec_debug_frame(struct dso *dso,
 
 		/* Check the .debug_frame section for unwinding info */
 		ofs = elf_section_offset(fd, ".debug_frame");
-		dso->data.frame_offset = ofs;
+		dso->data.debug_frame_offset = ofs;
 	}
 
 	*offset = ofs;

1.9.3

From: Wang Nan <[email protected]>

Without this patch, perf report cause segfault if pass "" as '-t':

  $ perf report -t ""

# To display the perf.data header info, please use --header/--header-only options. #
# Samples: 37 of event 'syscalls:sys_enter_write'
# Event count (approx.): 37
#
# Children SelfCommand Shared Object Symbol
Segmentation fault

Since -t is used to add field-separator for generate table, -t "" is actually meanless. This patch defines a new OPT_STRING_NOEMPTY() option generator to ensure user never pass empty string to that option.

---
 tools/perf/builtin-report.c     |  2 +-
 tools/perf/util/parse-options.c | 21 +++++++++++++++++++--
 tools/perf/util/parse-options.h |  2 ++
 3 files changed, 22 insertions(+), 3 deletions(-)

index 52f74e1367e9..0ae482689e3c 100644
@@ -676,7 +676,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
 		   "width[,width...]",
 		   "don't try to adjust column width, use these fixed values"),
-	OPT_STRING('t', "field-separator", &symbol_conf.field_sep, "separator",
+	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
 		   "separator for columns, no spaces will be added between "
 		   "columns '.' is reserved."),
 	OPT_BOOLEAN('U', "hide-unresolved", &report.hide_unresolved,

index 1457d6639b60..01626be2a8eb 100644
@@ -37,6 +37,7 @@ static int get_value(struct parse_opt_ctx_t *p,
 {
 	const char *s, *arg = NULL;
 	const int unset = flags & OPT_UNSET;
+	int err;
 
 	if (unset && p->opt)
 		return opterror(opt, "takes no value", flags);

@@ -114,13 +115,29 @@ static int get_value(struct parse_opt_ctx_t *p,
 		return 0;
 
 	case OPTION_STRING:
+		err = 0;
 		if (unset)
 			*(const char **)opt->value = NULL;
 		else if (opt->flags & PARSE_OPT_OPTARG && !p->opt)
 			*(const char **)opt->value = (const char *)opt->defval;
 		else
-			return get_arg(p, opt, flags, (const char **)opt->value);
-		return 0;
+			err = get_arg(p, opt, flags, (const char **)opt->value);
+
+		/* PARSE_OPT_NOEMPTY: Allow NULL but disallow empty string. */
+		if (opt->flags & PARSE_OPT_NOEMPTY) {
+			const char *val = *(const char **)opt->value;
+
+			if (!val)
+				return err;
+
+			/* Similar to unset if we are given an empty string. */
+			if (val[0] == '\0') {
+				*(const char **)opt->value = NULL;
+				return 0;
+			}
+		}
+
+		return err;
 
 	case OPTION_CALLBACK:
 		if (unset)

index 97b153fb4999..59561fd86278 100644
@@ -40,6 +40,7 @@ enum parse_opt_option_flags {
 	PARSE_OPT_LASTARG_DEFAULT = 16,
 	PARSE_OPT_DISABLED = 32,
 	PARSE_OPT_EXCLUSIVE = 64,
+	PARSE_OPT_NOEMPTY  = 128,
 };
 
 struct option;

@@ -122,6 +123,7 @@ struct option {
 #define OPT_LONG(s, l, v, h)        { .type = OPTION_LONG, .short_name = (s), .long_name = (l), .value = check_vtype(v, long *), .help = (h) }
 #define OPT_U64(s, l, v, h)         { .type = OPTION_U64, .short_name = (s), .long_name = (l), .value = check_vtype(v, u64 *), .help = (h) }
 #define OPT_STRING(s, l, v, a, h)   { .type = OPTION_STRING,  .short_name = (s), .long_name = (l), .value = check_vtype(v, const char **), (a), .help = (h) }
+#define OPT_STRING_NOEMPTY(s, l, v, a, h)   { .type = OPTION_STRING,  .short_name = (s), .long_name = (l), .value = check_vtype(v, const char **), (a), .help = (h), .flags = PARSE_OPT_NOEMPTY}
 #define OPT_DATE(s, l, v, h) \
 	{ .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), .argh = "time", .help = (h), .callback = parse_opt_approxidate_cb }
 #define OPT_CALLBACK(s, l, v, a, h, f) \

1.9.3

Pulled, thanks a lot Arnaldo!

Ingo