From 763d7f5f2718f085bab5a9e63308349728f3ad12 Mon Sep 17 00:00:00 2001 From: Masanari Iida Date: Sun, 20 Apr 2014 00:16:41 +0900 Subject: [PATCH 01/12] perf tests x86: Fix memory leak in sample_ustack() The buf is not freed, when kernel failed to get stack map and return. Signed-off-by: Masanari Iida Link: http://lkml.kernel.org/r/1398091024-7901-1-git-send-email-standby24x7@gmail.com Signed-off-by: Jiri Olsa --- tools/perf/arch/x86/tests/dwarf-unwind.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/perf/arch/x86/tests/dwarf-unwind.c b/tools/perf/arch/x86/tests/dwarf-unwind.c index b602ad93ce63..b8c0102c70c8 100644 --- a/tools/perf/arch/x86/tests/dwarf-unwind.c +++ b/tools/perf/arch/x86/tests/dwarf-unwind.c @@ -26,6 +26,7 @@ static int sample_ustack(struct perf_sample *sample, map = map_groups__find(&thread->mg, MAP__FUNCTION, (u64) sp); if (!map) { pr_debug("failed to get stack map\n"); + free(buf); return -1; } From 4dc549e58b6ebf63554cd466d5ceb0e9c70ab859 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 20 Apr 2014 10:50:00 +0200 Subject: [PATCH 02/12] perf tools: Disable libdw unwind for all but x86 arch So far there's only x86 libdw unwind support merged in perf. Disable it on all other architectures in case libdw unwind support is detected in system. Cc: Arnaldo Carvalho de Melo Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jean Pihet Cc: Josh Boyer Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1397988006-14158-1-git-send-email-jolsa@redhat.com Signed-off-by: Jiri Olsa --- tools/perf/config/Makefile | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile index ee21fa95ebcf..a71fb395e38f 100644 --- a/tools/perf/config/Makefile +++ b/tools/perf/config/Makefile @@ -34,6 +34,14 @@ ifeq ($(ARCH),arm) LIBUNWIND_LIBS = -lunwind -lunwind-arm endif +# So far there's only x86 libdw unwind support merged in perf. +# Disable it on all other architectures in case libdw unwind +# support is detected in system. Add supported architectures +# to the check. +ifneq ($(ARCH),x86) + NO_LIBDW_DWARF_UNWIND := 1 +endif + ifeq ($(LIBUNWIND_LIBS),) NO_LIBUNWIND := 1 else From 4508793541140723ab862d83095f6b5141d24e49 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Mon, 21 Apr 2014 22:23:46 -0400 Subject: [PATCH 03/12] tools lib traceevent: Fix backward compatibility macros for pevent filter enums The return value for pevent_filter_match() is suppose to return FILTER_NONE if the event doesn't have a filter, and FILTER_NOEXIST if there is no filter at all. But the change 41e12e580a7 "tools lib traceevent: Refactor pevent_filter_match() to get rid of die()" replaced the return value with PEVENT_ERRNO__* values and added "backward compatibility" macros that used the old names. Unfortunately, the NOEXIST and NONE macros were swapped, and this broke users that use the old return names. Signed-off-by: Steven Rostedt Link: http://lkml.kernel.org/r/20140421222346.0351ced4@gandalf.local.home Signed-off-by: Jiri Olsa --- tools/lib/traceevent/event-parse.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/lib/traceevent/event-parse.h b/tools/lib/traceevent/event-parse.h index 791c539374c7..feab94281634 100644 --- a/tools/lib/traceevent/event-parse.h +++ b/tools/lib/traceevent/event-parse.h @@ -876,8 +876,8 @@ struct event_filter { struct event_filter *pevent_filter_alloc(struct pevent *pevent); /* for backward compatibility */ -#define FILTER_NONE PEVENT_ERRNO__FILTER_NOT_FOUND -#define FILTER_NOEXIST PEVENT_ERRNO__NO_FILTER +#define FILTER_NONE PEVENT_ERRNO__NO_FILTER +#define FILTER_NOEXIST PEVENT_ERRNO__FILTER_NOT_FOUND #define FILTER_MISS PEVENT_ERRNO__FILTER_MISS #define FILTER_MATCH PEVENT_ERRNO__FILTER_MATCH From de04f8657de9d3351a2d5880f1f7080b23b798cf Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 22 Apr 2014 19:23:30 -0400 Subject: [PATCH 04/12] tools lib traceevent: Fix memory leak in pretty_print() Commit 12e55569a244 "tools lib traceevent: Use helper trace-seq in print functions like kernel does" added a extra trace_seq helper to process string arguments like the kernel does it. But the difference between the kernel and the userspace library is that the kernel's trace_seq structure has a static allocated buffer. The userspace one has a dynamically allocated one. It requires a trace_seq_destroy(), otherwise it produces a nasty memory leak. Cc: stable@vger.kernel.org # 3.14+ Signed-off-by: Steven Rostedt Link: http://lkml.kernel.org/r/20140422192330.6bb09bf8@gandalf.local.home Signed-off-by: Jiri Olsa --- tools/lib/traceevent/event-parse.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/lib/traceevent/event-parse.c b/tools/lib/traceevent/event-parse.c index baec7d887da4..b83184f2d484 100644 --- a/tools/lib/traceevent/event-parse.c +++ b/tools/lib/traceevent/event-parse.c @@ -4344,6 +4344,7 @@ static void pretty_print(struct trace_seq *s, void *data, int size, struct event format, len_arg, arg); trace_seq_terminate(&p); trace_seq_puts(s, p.buffer); + trace_seq_destroy(&p); arg = arg->next; break; default: From 9f7ff8931e3c5ddc8535476971ec9501e9555c05 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Wed, 23 Apr 2014 19:04:19 +0200 Subject: [PATCH 05/12] perf/x86: Fix RAPL rdmsrl_safe() usage This patch fixes a bug introduced by: 24223657806a ("perf/x86/intel: Use rdmsrl_safe() when initializing RAPL PMU") The rdmsrl_safe() function returns 0 on success. The current code was failing to detect the RAPL PMU on real hardware (missing /sys/devices/power) because the return value of rdmsrl_safe() was misinterpreted. Signed-off-by: Stephane Eranian Acked-by: Borislav Petkov Acked-by: Venkatesh Srinivas Cc: peterz@infradead.org Cc: zheng.z.yan@intel.com Link: http://lkml.kernel.org/r/20140423170418.GA12767@quad Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel_rapl.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/x86/kernel/cpu/perf_event_intel_rapl.c b/arch/x86/kernel/cpu/perf_event_intel_rapl.c index 7c87424d4140..619f7699487a 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_rapl.c +++ b/arch/x86/kernel/cpu/perf_event_intel_rapl.c @@ -543,7 +543,8 @@ static int rapl_cpu_prepare(int cpu) if (phys_id < 0) return -1; - if (!rdmsrl_safe(MSR_RAPL_POWER_UNIT, &msr_rapl_power_unit_bits)) + /* protect rdmsrl() to handle virtualization */ + if (rdmsrl_safe(MSR_RAPL_POWER_UNIT, &msr_rapl_power_unit_bits)) return -1; pmu = kzalloc_node(sizeof(*pmu), GFP_KERNEL, cpu_to_node(cpu)); From e27a08f53d081b33c988db954008d879b2ab8b39 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 27 Apr 2014 11:12:21 +0200 Subject: [PATCH 06/12] perf tools: Use LDFLAGS instead of ALL_LDFLAGS We no longer use ALL_LDFLAGS, Replacing with LDFLAGS. Acked-by: Namhyung Kim Cc: Arnaldo Carvalho de Melo Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1398675770-3109-1-git-send-email-jolsa@kernel.org Signed-off-by: Jiri Olsa --- tools/perf/Makefile.perf | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index e96923310d57..895edd32930c 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -589,7 +589,7 @@ $(GTK_OBJS): $(OUTPUT)%.o: %.c $(LIB_H) $(QUIET_CC)$(CC) -o $@ -c -fPIC $(CFLAGS) $(GTK_CFLAGS) $< $(OUTPUT)libperf-gtk.so: $(GTK_OBJS) $(PERFLIBS) - $(QUIET_LINK)$(CC) -o $@ -shared $(ALL_LDFLAGS) $(filter %.o,$^) $(GTK_LIBS) + $(QUIET_LINK)$(CC) -o $@ -shared $(LDFLAGS) $(filter %.o,$^) $(GTK_LIBS) $(OUTPUT)builtin-help.o: builtin-help.c $(OUTPUT)common-cmds.h $(OUTPUT)PERF-CFLAGS $(QUIET_CC)$(CC) -o $@ -c $(CFLAGS) \ From 2cf9040714f3e63f94935c9f1ed4e5bc54265ddb Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Wed, 23 Apr 2014 16:53:25 +0200 Subject: [PATCH 07/12] perf tools: Fix bfd dependency libraries detection There's false assumption in the library detection code assuming -liberty and -lz are always present once bfd is detected. The fails on Ubuntu (14.04) as reported by Ingo. Forcing the bdf dependency libraries detection any time bfd library is detected. Reported-by: Ingo Molnar Tested-by: Ingo Molnar Cc: Arnaldo Carvalho de Melo Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1398676935-6615-1-git-send-email-jolsa@kernel.org Signed-off-by: Jiri Olsa --- tools/perf/config/Makefile | 34 +++++++++++++++++++++++----------- 1 file changed, 23 insertions(+), 11 deletions(-) diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile index a71fb395e38f..a57d59e7fe4b 100644 --- a/tools/perf/config/Makefile +++ b/tools/perf/config/Makefile @@ -194,7 +194,10 @@ VF_FEATURE_TESTS = \ stackprotector-all \ timerfd \ libunwind-debug-frame \ - bionic + bionic \ + liberty \ + liberty-z \ + cplus-demangle # Set FEATURE_CHECK_(C|LD)FLAGS-all for all CORE_FEATURE_TESTS features. # If in the future we need per-feature checks/flags for features not @@ -512,7 +515,21 @@ else endif ifeq ($(feature-libbfd), 1) - EXTLIBS += -lbfd -lz -liberty + EXTLIBS += -lbfd + + # call all detections now so we get correct + # status in VF output + $(call feature_check,liberty) + $(call feature_check,liberty-z) + $(call feature_check,cplus-demangle) + + ifeq ($(feature-liberty), 1) + EXTLIBS += -liberty + else + ifeq ($(feature-liberty-z), 1) + EXTLIBS += -liberty -lz + endif + endif endif ifdef NO_DEMANGLE @@ -523,15 +540,10 @@ else CFLAGS += -DHAVE_CPLUS_DEMANGLE_SUPPORT else ifneq ($(feature-libbfd), 1) - $(call feature_check,liberty) - ifeq ($(feature-liberty), 1) - EXTLIBS += -lbfd -liberty - else - $(call feature_check,liberty-z) - ifeq ($(feature-liberty-z), 1) - EXTLIBS += -lbfd -liberty -lz - else - $(call feature_check,cplus-demangle) + ifneq ($(feature-liberty), 1) + ifneq ($(feature-liberty-z), 1) + # we dont have neither HAVE_CPLUS_DEMANGLE_SUPPORT + # or any of 'bfd iberty z' trinity ifeq ($(feature-cplus-demangle), 1) EXTLIBS += -liberty CFLAGS += -DHAVE_CPLUS_DEMANGLE_SUPPORT From 611ec127165b572f02823e152d6774e2145305a3 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Tue, 29 Apr 2014 09:53:40 +0200 Subject: [PATCH 08/12] perf tests: Add static build make test Adding test for building static perf build into the automated suite. Also available via following commands: $ make -f tests/make make_static - make_static: cd . && make -f Makefile DESTDIR=/tmp/tmp.7u5MlB4njo LDFLAGS=-static $ make -f tests/make make_static_O - make_static_O: cd . && make -f Makefile O=/tmp/tmp.Ay6r3wEmtX DESTDIR=/tmp/tmp.vK0KQwO0Vi LDFLAGS=-static Acked-by: David Ahern Cc: Arnaldo Carvalho de Melo Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1398760413-7574-1-git-send-email-jolsa@kernel.org Signed-off-by: Jiri Olsa --- tools/perf/tests/make | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/perf/tests/make b/tools/perf/tests/make index 5daeae1cb4c0..2f92d6e7ee00 100644 --- a/tools/perf/tests/make +++ b/tools/perf/tests/make @@ -46,6 +46,7 @@ make_install_man := install-man make_install_html := install-html make_install_info := install-info make_install_pdf := install-pdf +make_static := LDFLAGS=-static # all the NO_* variable combined make_minimal := NO_LIBPERL=1 NO_LIBPYTHON=1 NO_NEWT=1 NO_GTK2=1 @@ -87,6 +88,7 @@ run += make_install_bin # run += make_install_info # run += make_install_pdf run += make_minimal +run += make_static ifneq ($(call has,ctags),) run += make_tags From 61d4290cc1f10588147b76b385875f06827d47ff Mon Sep 17 00:00:00 2001 From: Richard Yao Date: Sat, 26 Apr 2014 13:17:55 -0400 Subject: [PATCH 09/12] perf machine: Search for modules in %s/lib/modules/%s Modules installed outside of the kernel's build system should go into "%s/lib/modules/%s/extra", but at present, perf will only look at them when they are in "%s/lib/modules/%s/kernel". Lets encourage good citizenship by relaxing this requirement to "%s/lib/modules/%s". This way open source modules that are out-of-tree have no incentive to start populating a directory reserved for in-kernel modules and I can stop hex-editing my system's perf binary when profiling OSS out-of-tree modules. Feedback from Namhyung Kim correctly revealed that the hex-edits that I had been doing meant that perf was also traversing the build and source symlinks in %s/lib/modules/%s. That is undesireable, so we explicitly exclude them from traversal with a minor tweak to the traversal routine. Signed-off-by: Richard Yao Acked-by: Namhyung kim Link: http://lkml.kernel.org/r/1398532675-13684-1-git-send-email-ryao@gentoo.org Signed-off-by: Jiri Olsa --- tools/perf/util/machine.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c index a53cd0b8c151..27c2a5efe450 100644 --- a/tools/perf/util/machine.c +++ b/tools/perf/util/machine.c @@ -717,7 +717,7 @@ static char *get_kernel_version(const char *root_dir) } static int map_groups__set_modules_path_dir(struct map_groups *mg, - const char *dir_name) + const char *dir_name, int depth) { struct dirent *dent; DIR *dir = opendir(dir_name); @@ -742,7 +742,15 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, !strcmp(dent->d_name, "..")) continue; - ret = map_groups__set_modules_path_dir(mg, path); + /* Do not follow top-level source and build symlinks */ + if (depth == 0) { + if (!strcmp(dent->d_name, "source") || + !strcmp(dent->d_name, "build")) + continue; + } + + ret = map_groups__set_modules_path_dir(mg, path, + depth + 1); if (ret < 0) goto out; } else { @@ -786,11 +794,11 @@ static int machine__set_modules_path(struct machine *machine) if (!version) return -1; - snprintf(modules_path, sizeof(modules_path), "%s/lib/modules/%s/kernel", + snprintf(modules_path, sizeof(modules_path), "%s/lib/modules/%s", machine->root_dir, version); free(version); - return map_groups__set_modules_path_dir(&machine->kmaps, modules_path); + return map_groups__set_modules_path_dir(&machine->kmaps, modules_path, 0); } static int machine__create_module(void *arg, const char *name, u64 start) From 603940b6b8b1347ec13a628165c00194ebc17ed2 Mon Sep 17 00:00:00 2001 From: Xia Kaixu Date: Sat, 26 Apr 2014 15:55:12 +0800 Subject: [PATCH 10/12] perf tools: Remove extra '/' character in events file path The array debugfs_known_mountpoints[] will cause extra '/' character output. Remove it. pre: $ perf probe -l /sys/kernel/debug//tracing/uprobe_events file does not exist - please rebuild kernel with CONFIG_UPROBE_EVENTS. post: $ perf probe -l /sys/kernel/debug/tracing/uprobe_events file does not exist - please rebuild kernel with CONFIG_UPROBE_EVENTS. Signed-off-by: Xia Kaixu Acked-by: Namhyung Kim Link: http://lkml.kernel.org/r/535B6660.2060001@huawei.com Signed-off-by: Jiri Olsa --- tools/lib/api/fs/debugfs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/lib/api/fs/debugfs.c b/tools/lib/api/fs/debugfs.c index 7c4347962353..a74fba6d7743 100644 --- a/tools/lib/api/fs/debugfs.c +++ b/tools/lib/api/fs/debugfs.c @@ -12,8 +12,8 @@ char debugfs_mountpoint[PATH_MAX + 1] = "/sys/kernel/debug"; static const char * const debugfs_known_mountpoints[] = { - "/sys/kernel/debug/", - "/debug/", + "/sys/kernel/debug", + "/debug", 0, }; From 6392b4ebdc1acfbed605d68c437ff5597d0a8d95 Mon Sep 17 00:00:00 2001 From: Mathias Krause Date: Sun, 27 Apr 2014 18:51:05 +0200 Subject: [PATCH 11/12] perf x86: Fix perf to use non-executable stack, again arch/x86/tests/regs_load.S is missing the linker note about the stack requirements, therefore making the linker fall back to an executable stack. As this object gets linked against the final perf binary, it'll needlessly end up with an executable stack. Fix this by adding the appropriate linker note. Also add a global linker flag to prevent future regressions, as suggested by Jiri. This way perf won't get an executable stack even if we fail to add the .GNU-stack linker note to future assembler files. Though, doing so might create regressions the other way around, when (statically) linking against libraries needing an executable stack. But, apparently, regressing in that direction is wanted as it is an indicator of poor code quality -- or just missing linker notes. Fixes: 3c8b06f981 ("perf tests x86: Introduce perf_regs_load function") Signed-off-by: Mathias Krause Acked-by: Ingo Molnar Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1398617466-22749-1-git-send-email-minipli@googlemail.com Signed-off-by: Jiri Olsa --- tools/perf/arch/x86/tests/regs_load.S | 8 +++++++- tools/perf/config/Makefile | 4 ++++ 2 files changed, 11 insertions(+), 1 deletion(-) diff --git a/tools/perf/arch/x86/tests/regs_load.S b/tools/perf/arch/x86/tests/regs_load.S index 99167bf644ea..60875d5c556c 100644 --- a/tools/perf/arch/x86/tests/regs_load.S +++ b/tools/perf/arch/x86/tests/regs_load.S @@ -1,4 +1,3 @@ - #include #define AX 0 @@ -90,3 +89,10 @@ ENTRY(perf_regs_load) ret ENDPROC(perf_regs_load) #endif + +/* + * We need to provide note.GNU-stack section, saying that we want + * NOT executable stack. Otherwise the final linking will assume that + * the ELF stack should not be restricted at all and set it RWX. + */ +.section .note.GNU-stack,"",@progbits diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile index a57d59e7fe4b..802cf544202b 100644 --- a/tools/perf/config/Makefile +++ b/tools/perf/config/Makefile @@ -117,6 +117,10 @@ CFLAGS += -Wall CFLAGS += -Wextra CFLAGS += -std=gnu99 +# Enforce a non-executable stack, as we may regress (again) in the future by +# adding assembler files missing the .GNU-stack linker note. +LDFLAGS += -Wl,-z,noexecstack + EXTLIBS = -lelf -lpthread -lrt -lm -ldl ifneq ($(OUTPUT),) From 9a126728165eef9b1e9349fe57b9747ac6c6f7e3 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Wed, 30 Apr 2014 16:39:44 +0200 Subject: [PATCH 12/12] perf tests x86: Fix stack map lookup in dwarf unwind test Previous commit 'perf x86: Fix perf to use non-executable stack, again' moved stack map into MAP__VARIABLE map type again. Fixing the dwarf unwind test stack map lookup appropriately. Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Jean Pihet Link: http://lkml.kernel.org/n/tip-ttzyhbe4zls24z7ednkmhvxl@git.kernel.org Signed-off-by: Jiri Olsa --- tools/perf/arch/x86/tests/dwarf-unwind.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/arch/x86/tests/dwarf-unwind.c b/tools/perf/arch/x86/tests/dwarf-unwind.c index b8c0102c70c8..83bc2385e6d3 100644 --- a/tools/perf/arch/x86/tests/dwarf-unwind.c +++ b/tools/perf/arch/x86/tests/dwarf-unwind.c @@ -23,7 +23,7 @@ static int sample_ustack(struct perf_sample *sample, sp = (unsigned long) regs[PERF_REG_X86_SP]; - map = map_groups__find(&thread->mg, MAP__FUNCTION, (u64) sp); + map = map_groups__find(&thread->mg, MAP__VARIABLE, (u64) sp); if (!map) { pr_debug("failed to get stack map\n"); free(buf);