perf build: Use libtraceevent from the system
authorIan Rogers <irogers@google.com>
Mon, 5 Dec 2022 22:59:39 +0000 (14:59 -0800)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 14 Dec 2022 14:16:12 +0000 (11:16 -0300)
Remove the LIBTRACEEVENT_DYNAMIC and LIBTRACEFS_DYNAMIC make command
line variables.

If libtraceevent isn't installed or NO_LIBTRACEEVENT=1 is passed to the
build, don't compile in libtraceevent and libtracefs support.

This also disables CONFIG_TRACE that controls "perf trace".

CONFIG_LIBTRACEEVENT is used to control enablement in Build/Makefiles,
HAVE_LIBTRACEEVENT is used in C code.

Without HAVE_LIBTRACEEVENT tracepoints are disabled and as such the
commands kmem, kwork, lock, sched and timechart are removed.  The
majority of commands continue to work including "perf test".

Committer notes:

Fixed up a tools/perf/util/Build reject and added:

  #include <traceevent/event-parse.h>

to tools/perf/util/scripting-engines/trace-event-perl.c.

Committer testing:

  $ rpm -qi libtraceevent-devel
  Name        : libtraceevent-devel
  Version     : 1.5.3
  Release     : 2.fc36
  Architecture: x86_64
  Install Date: Mon 25 Jul 2022 03:20:19 PM -03
  Group       : Unspecified
  Size        : 27728
  License     : LGPLv2+ and GPLv2+
  Signature   : RSA/SHA256, Fri 15 Apr 2022 02:11:58 PM -03, Key ID 999f7cbf38ab71f4
  Source RPM  : libtraceevent-1.5.3-2.fc36.src.rpm
  Build Date  : Fri 15 Apr 2022 10:57:01 AM -03
  Build Host  : buildvm-x86-05.iad2.fedoraproject.org
  Packager    : Fedora Project
  Vendor      : Fedora Project
  URL         : https://git.kernel.org/pub/scm/libs/libtrace/libtraceevent.git/
  Bug URL     : https://bugz.fedoraproject.org/libtraceevent
  Summary     : Development headers of libtraceevent
  Description :
  Development headers of libtraceevent-libs
  $

Default build:

  $ ldd ~/bin/perf | grep tracee
   libtraceevent.so.1 => /lib64/libtraceevent.so.1 (0x00007f1dcaf8f000)
  $

  # perf trace -e sched:* --max-events 10
       0.000 migration/0/17 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, dest_cpu: 1)
       0.005 migration/0/17 sched:sched_wake_idle_without_ipi(cpu: 1)
       0.011 migration/0/17 sched:sched_switch(prev_comm: "", prev_pid: 17 (migration/0), prev_state: 1, next_comm: "", next_prio: 120)
       1.173 :0/0 sched:sched_wakeup(comm: "", pid: 3138 (gnome-terminal-), prio: 120)
       1.180 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 3138 (gnome-terminal-), next_prio: 120)
       0.156 migration/1/21 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, orig_cpu: 1, dest_cpu: 2)
       0.160 migration/1/21 sched:sched_wake_idle_without_ipi(cpu: 2)
       0.166 migration/1/21 sched:sched_switch(prev_comm: "", prev_pid: 21 (migration/1), prev_state: 1, next_comm: "", next_prio: 120)
       1.183 :0/0 sched:sched_wakeup(comm: "", pid: 1602985 (kworker/u16:0-f), prio: 120, target_cpu: 1)
       1.186 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 1602985 (kworker/u16:0-f), next_prio: 120)
  #

Had to tweak tools/perf/util/setup.py to make sure the python binding
shared object links with libtraceevent if -DHAVE_LIBTRACEEVENT is
present in CFLAGS.

Building with NO_LIBTRACEEVENT=1 uncovered some more build failures:

- Make building of data-convert-bt.c to CONFIG_LIBTRACEEVENT=y

- perf-$(CONFIG_LIBTRACEEVENT) += scripts/

- bpf_kwork.o needs also to be dependent on CONFIG_LIBTRACEEVENT=y

- The python binding needed some fixups and util/trace-event.c can't be
  built and linked with the python binding shared object, so remove it
  in tools/perf/util/setup.py and exclude it from the list of
  dependencies in the python/perf.so Makefile.perf target.

Building without libtraceevent-devel installed uncovered more build
failures:

- The python binding tools/perf/util/python.c was assuming that
  traceevent/parse-events.h was always available, which was the case
  when we defaulted to using the in-kernel tools/lib/traceevent/ files,
  now we need to enclose it under ifdef HAVE_LIBTRACEEVENT, just like
  the other parts of it that deal with tracepoints.

- We have to ifdef the rules in the Build files with
  CONFIG_LIBTRACEEVENT=y to build builtin-trace.c and
  tools/perf/trace/beauty/ as we only ifdef setting CONFIG_TRACE=y when
  setting NO_LIBTRACEEVENT=1 in the make command line, not when we don't
  detect libtraceevent-devel installed in the system. Simplification here
  to avoid these two ways of disabling builtin-trace.c and not having
  CONFIG_TRACE=y when libtraceevent-devel isn't installed is the clean
  way.

From Athira:

<quote>
tools/perf/arch/powerpc/util/Build
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
</quote>

Then, ditto for arm64 and s390, detected by container cross build tests.

- s/390 uses test__checkevent_tracepoint() that is now only available if
  HAVE_LIBTRACEEVENT is defined, enclose the callsite with ifder HAVE_LIBTRACEEVENT.

Also from Athira:

<quote>
With this change, I could successfully compile in these environment:
- Without libtraceevent-devel installed
- With libtraceevent-devel installed
- With “make NO_LIBTRACEEVENT=1”
</quote>

Then, finally rename CONFIG_TRACEEVENT to CONFIG_LIBTRACEEVENT for
consistency with other libraries detected in tools/perf/.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: bpf@vger.kernel.org
Link: http://lore.kernel.org/lkml/20221205225940.3079667-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
52 files changed:
tools/perf/Build
tools/perf/Makefile.config
tools/perf/Makefile.perf
tools/perf/arch/arm64/util/Build
tools/perf/arch/powerpc/util/Build
tools/perf/arch/s390/util/Build
tools/perf/arch/x86/util/Build
tools/perf/arch/x86/util/intel-pt.c
tools/perf/builtin-annotate.c
tools/perf/builtin-data.c
tools/perf/builtin-inject.c
tools/perf/builtin-kmem.c
tools/perf/builtin-kvm.c
tools/perf/builtin-kwork.c
tools/perf/builtin-record.c
tools/perf/builtin-report.c
tools/perf/builtin-script.c
tools/perf/builtin-timechart.c
tools/perf/builtin-trace.c
tools/perf/builtin-version.c
tools/perf/perf.c
tools/perf/scripts/python/Perf-Trace-Util/Build
tools/perf/tests/Build
tools/perf/tests/builtin-test.c
tools/perf/tests/parse-events.c
tools/perf/util/Build
tools/perf/util/data-convert-bt.c
tools/perf/util/data-convert-json.c
tools/perf/util/evlist.c
tools/perf/util/evlist.h
tools/perf/util/evsel.c
tools/perf/util/evsel.h
tools/perf/util/evsel_fprintf.c
tools/perf/util/header.c
tools/perf/util/header.h
tools/perf/util/intel-pt.c
tools/perf/util/parse-events.c
tools/perf/util/parse-events.h
tools/perf/util/python.c
tools/perf/util/scripting-engines/Build
tools/perf/util/scripting-engines/trace-event-perl.c
tools/perf/util/scripting-engines/trace-event-python.c
tools/perf/util/session.c
tools/perf/util/session.h
tools/perf/util/setup.py
tools/perf/util/sort.c
tools/perf/util/synthetic-events.c
tools/perf/util/trace-event-parse.c
tools/perf/util/trace-event-read.c
tools/perf/util/trace-event-scripting.c
tools/perf/util/trace-event.c
tools/perf/util/trace-event.h

index 496b096..6dd67e5 100644 (file)
@@ -5,7 +5,6 @@ perf-y += builtin-diff.o
 perf-y += builtin-evlist.o
 perf-y += builtin-ftrace.o
 perf-y += builtin-help.o
-perf-y += builtin-sched.o
 perf-y += builtin-buildid-list.o
 perf-y += builtin-buildid-cache.o
 perf-y += builtin-kallsyms.o
@@ -13,11 +12,8 @@ perf-y += builtin-list.o
 perf-y += builtin-record.o
 perf-y += builtin-report.o
 perf-y += builtin-stat.o
-perf-y += builtin-timechart.o
 perf-y += builtin-top.o
 perf-y += builtin-script.o
-perf-y += builtin-kmem.o
-perf-y += builtin-lock.o
 perf-y += builtin-kvm.o
 perf-y += builtin-inject.o
 perf-y += builtin-mem.o
@@ -25,9 +21,18 @@ perf-y += builtin-data.o
 perf-y += builtin-version.o
 perf-y += builtin-c2c.o
 perf-y += builtin-daemon.o
-perf-y += builtin-kwork.o
 
-perf-$(CONFIG_TRACE) += builtin-trace.o
+perf-$(CONFIG_LIBTRACEEVENT) += builtin-kmem.o
+perf-$(CONFIG_LIBTRACEEVENT) += builtin-kwork.o
+perf-$(CONFIG_LIBTRACEEVENT) += builtin-lock.o
+perf-$(CONFIG_LIBTRACEEVENT) += builtin-sched.o
+perf-$(CONFIG_LIBTRACEEVENT) += builtin-timechart.o
+
+ifeq ($(CONFIG_LIBTRACEEVENT),y)
+  perf-$(CONFIG_TRACE) += builtin-trace.o
+  perf-$(CONFIG_TRACE) += trace/beauty/
+endif
+
 perf-$(CONFIG_LIBELF) += builtin-probe.o
 
 perf-y += bench/
@@ -51,7 +56,6 @@ CFLAGS_builtin-report.o          += -DDOCDIR="BUILD_STR($(srcdir_SQ)/Documentation)"
 perf-y += util/
 perf-y += arch/
 perf-y += ui/
-perf-y += scripts/
-perf-$(CONFIG_TRACE) += trace/beauty/
+perf-$(CONFIG_LIBTRACEEVENT) += scripts/
 
 gtk-y += ui/gtk/
index 9cc3c48..680228e 100644 (file)
@@ -763,18 +763,20 @@ ifndef NO_LIBUNWIND
   EXTLIBS += $(EXTLIBS_LIBUNWIND)
 endif
 
-ifeq ($(NO_SYSCALL_TABLE),0)
-  $(call detected,CONFIG_TRACE)
-else
-  ifndef NO_LIBAUDIT
-    $(call feature_check,libaudit)
-    ifneq ($(feature-libaudit), 1)
-      msg := $(warning No libaudit.h found, disables 'trace' tool, please install audit-libs-devel or libaudit-dev);
-      NO_LIBAUDIT := 1
-    else
-      CFLAGS += -DHAVE_LIBAUDIT_SUPPORT
-      EXTLIBS += -laudit
-      $(call detected,CONFIG_TRACE)
+ifneq ($(NO_LIBTRACEEVENT),1)
+  ifeq ($(NO_SYSCALL_TABLE),0)
+    $(call detected,CONFIG_TRACE)
+  else
+    ifndef NO_LIBAUDIT
+      $(call feature_check,libaudit)
+      ifneq ($(feature-libaudit), 1)
+        msg := $(warning No libaudit.h found, disables 'trace' tool, please install audit-libs-devel or libaudit-dev);
+        NO_LIBAUDIT := 1
+      else
+        CFLAGS += -DHAVE_LIBAUDIT_SUPPORT
+        EXTLIBS += -laudit
+        $(call detected,CONFIG_TRACE)
+      endif
     endif
   endif
 endif
@@ -1182,9 +1184,11 @@ ifdef LIBPFM4
   endif
 endif
 
-ifdef LIBTRACEEVENT_DYNAMIC
+# libtraceevent is a recommended dependency picked up from the system.
+ifneq ($(NO_LIBTRACEEVENT),1)
   $(call feature_check,libtraceevent)
   ifeq ($(feature-libtraceevent), 1)
+    CFLAGS += -DHAVE_LIBTRACEEVENT
     EXTLIBS += -ltraceevent
     LIBTRACEEVENT_VERSION := $(shell $(PKG_CONFIG) --modversion libtraceevent)
     LIBTRACEEVENT_VERSION_1 := $(word 1, $(subst ., ,$(LIBTRACEEVENT_VERSION)))
@@ -1192,12 +1196,11 @@ ifdef LIBTRACEEVENT_DYNAMIC
     LIBTRACEEVENT_VERSION_3 := $(word 3, $(subst ., ,$(LIBTRACEEVENT_VERSION)))
     LIBTRACEEVENT_VERSION_CPP := $(shell expr $(LIBTRACEEVENT_VERSION_1) \* 255 \* 255 + $(LIBTRACEEVENT_VERSION_2) \* 255 + $(LIBTRACEEVENT_VERSION_3))
     CFLAGS += -DLIBTRACEEVENT_VERSION=$(LIBTRACEEVENT_VERSION_CPP)
+    $(call detected,CONFIG_LIBTRACEEVENT)
   else
-    dummy := $(error Error: No libtraceevent devel library found, please install libtraceevent-devel);
+    dummy := $(warning Warning: libtraceevent is missing limiting functionality, please install libtraceevent-dev/libtraceevent-devel)
   endif
-endif
 
-ifdef LIBTRACEFS_DYNAMIC
   $(call feature_check,libtracefs)
   ifeq ($(feature-libtracefs), 1)
     EXTLIBS += -ltracefs
@@ -1207,8 +1210,6 @@ ifdef LIBTRACEFS_DYNAMIC
     LIBTRACEFS_VERSION_3 := $(word 3, $(subst ., ,$(LIBTRACEFS_VERSION)))
     LIBTRACEFS_VERSION_CPP := $(shell expr $(LIBTRACEFS_VERSION_1) \* 255 \* 255 + $(LIBTRACEFS_VERSION_2) \* 255 + $(LIBTRACEFS_VERSION_3))
     CFLAGS += -DLIBTRACEFS_VERSION=$(LIBTRACEFS_VERSION_CPP)
-  else
-    dummy := $(error Error: No libtracefs devel library found, please install libtracefs-dev);
   endif
 endif
 
index 6689f64..98f629b 100644 (file)
@@ -128,10 +128,6 @@ include ../scripts/utilities.mak
 #
 # Define BUILD_BPF_SKEL to enable BPF skeletons
 #
-# Define LIBTRACEEVENT_DYNAMIC to enable libtraceevent dynamic linking
-#
-# Define LIBTRACEFS_DYNAMIC to enable libtracefs dynamic linking
-#
 
 # As per kernel Makefile, avoid funny character set dependencies
 unexport LC_ALL
@@ -242,10 +238,6 @@ sub-make: fixdep
 else # force_fixdep
 
 LIBAPI_DIR      = $(srctree)/tools/lib/api/
-ifndef LIBTRACEEVENT_DYNAMIC
-LIBTRACEEVENT_DIR = $(srctree)/tools/lib/traceevent/
-LIBTRACEEVENT_PLUGINS_DIR = $(LIBTRACEEVENT_DIR)/plugins
-endif
 LIBBPF_DIR      = $(srctree)/tools/lib/bpf/
 LIBSUBCMD_DIR   = $(srctree)/tools/lib/subcmd/
 LIBSYMBOL_DIR   = $(srctree)/tools/lib/symbol/
@@ -295,31 +287,6 @@ SCRIPT_SH += perf-iostat.sh
 grep-libs = $(filter -l%,$(1))
 strip-libs = $(filter-out -l%,$(1))
 
-ifndef LIBTRACEEVENT_DYNAMIC
-ifneq ($(OUTPUT),)
-  LIBTRACEEVENT_OUTPUT = $(abspath $(OUTPUT))/libtraceevent
-else
-  LIBTRACEEVENT_OUTPUT = $(CURDIR)/libtraceevent
-endif
-LIBTRACEEVENT_PLUGINS_OUTPUT = $(LIBTRACEEVENT_OUTPUT)_plugins
-LIBTRACEEVENT_DESTDIR = $(LIBTRACEEVENT_OUTPUT)
-LIBTRACEEVENT_PLUGINS_DESTDIR = $(LIBTRACEEVENT_PLUGINS_OUTPUT)
-LIBTRACEEVENT_INCLUDE = $(LIBTRACEEVENT_DESTDIR)/include
-LIBTRACEEVENT = $(LIBTRACEEVENT_OUTPUT)/libtraceevent.a
-export LIBTRACEEVENT
-LIBTRACEEVENT_DYNAMIC_LIST = $(LIBTRACEEVENT_PLUGINS_OUTPUT)/libtraceevent-dynamic-list
-CFLAGS += -I$(LIBTRACEEVENT_OUTPUT)/include
-#
-# The static build has no dynsym table, so this does not work for
-# static build. Looks like linker starts to scream about that now
-# (in Fedora 26) so we need to switch it off for static build.
-DYNAMIC_LIST_LDFLAGS               = -Xlinker --dynamic-list=$(LIBTRACEEVENT_DYNAMIC_LIST)
-LIBTRACEEVENT_DYNAMIC_LIST_LDFLAGS = $(if $(findstring -static,$(LDFLAGS)),,$(DYNAMIC_LIST_LDFLAGS))
-else
-LIBTRACEEVENT_DYNAMIC_LIST =
-LIBTRACEEVENT_DYNAMIC_LIST_LDFLAGS =
-endif
-
 ifneq ($(OUTPUT),)
   LIBAPI_OUTPUT = $(abspath $(OUTPUT))/libapi
 else
@@ -380,13 +347,14 @@ export PYTHON_EXTBUILD_LIB PYTHON_EXTBUILD_TMP
 
 python-clean := $(call QUIET_CLEAN, python) $(RM) -r $(PYTHON_EXTBUILD) $(OUTPUT)python/perf*.so
 
-PYTHON_EXT_SRCS := $(shell grep -v ^\# util/python-ext-sources)
-ifndef LIBTRACEEVENT_DYNAMIC
-PYTHON_EXT_DEPS := util/python-ext-sources util/setup.py $(LIBTRACEEVENT) $(LIBAPI)
+ifeq ($(CONFIG_LIBTRACEEVENT),y)
+  PYTHON_EXT_SRCS := $(shell grep -v ^\# util/python-ext-sources)
 else
-PYTHON_EXT_DEPS := util/python-ext-sources util/setup.py $(LIBAPI)
+  PYTHON_EXT_SRCS := $(shell grep -v '^\#\|util/trace-event.c' util/python-ext-sources)
 endif
 
+PYTHON_EXT_DEPS := util/python-ext-sources util/setup.py $(LIBAPI)
+
 SCRIPTS = $(patsubst %.sh,%,$(SCRIPT_SH))
 
 PROGRAMS += $(OUTPUT)perf
@@ -430,9 +398,6 @@ ifndef NO_LIBBPF
     PERFLIBS += $(LIBBPF)
   endif
 endif
-ifndef LIBTRACEEVENT_DYNAMIC
-  PERFLIBS += $(LIBTRACEEVENT)
-endif
 
 # We choose to avoid "if .. else if .. else .. endif endif"
 # because maintaining the nesting to match is a pain.  If
@@ -682,9 +647,9 @@ all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
 # Create python binding output directory if not already present
 _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
 
-$(OUTPUT)python/perf.so: $(PYTHON_EXT_SRCS) $(PYTHON_EXT_DEPS) $(LIBTRACEEVENT_DYNAMIC_LIST) $(LIBPERF)
+$(OUTPUT)python/perf.so: $(PYTHON_EXT_SRCS) $(PYTHON_EXT_DEPS) $(LIBPERF)
        $(QUIET_GEN)LDSHARED="$(CC) -pthread -shared" \
-        CFLAGS='$(CFLAGS)' LDFLAGS='$(LDFLAGS) $(LIBTRACEEVENT_DYNAMIC_LIST_LDFLAGS)' \
+        CFLAGS='$(CFLAGS)' LDFLAGS='$(LDFLAGS)' \
          $(PYTHON_WORD) util/setup.py \
          --quiet build_ext; \
        cp $(PYTHON_EXTBUILD_LIB)perf*.so $(OUTPUT)python/
@@ -710,8 +675,8 @@ $(PERF_IN): prepare FORCE
 $(PMU_EVENTS_IN): FORCE prepare
        $(Q)$(MAKE) -f $(srctree)/tools/build/Makefile.build dir=pmu-events obj=pmu-events
 
-$(OUTPUT)perf: $(PERFLIBS) $(PERF_IN) $(PMU_EVENTS_IN) $(LIBTRACEEVENT_DYNAMIC_LIST)
-       $(QUIET_LINK)$(CC) $(CFLAGS) $(LDFLAGS) $(LIBTRACEEVENT_DYNAMIC_LIST_LDFLAGS) \
+$(OUTPUT)perf: $(PERFLIBS) $(PERF_IN) $(PMU_EVENTS_IN)
+       $(QUIET_LINK)$(CC) $(CFLAGS) $(LDFLAGS) \
                $(PERF_IN) $(PMU_EVENTS_IN) $(LIBS) -o $@
 
 $(GTK_IN): FORCE prepare
@@ -797,10 +762,6 @@ prepare: $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)common-cmds.h archheaders $(drm_ioc
        $(LIBSYMBOL) \
        bpf-skel
 
-ifndef LIBTRACEEVENT_DYNAMIC
-prepare: $(LIBTRACEEVENT)
-endif
-
 $(OUTPUT)%.o: %.c prepare FORCE
        $(Q)$(MAKE) -f $(srctree)/tools/build/Makefile.build dir=$(build-dir) $@
 
@@ -856,38 +817,6 @@ endif
 
 $(patsubst perf-%,%.o,$(PROGRAMS)): $(wildcard */*.h)
 
-ifndef LIBTRACEEVENT_DYNAMIC
-LIBTRACEEVENT_FLAGS += plugin_dir=$(plugindir_SQ) 'EXTRA_CFLAGS=$(EXTRA_CFLAGS)' 'LDFLAGS=$(filter-out -static,$(LDFLAGS))'
-
-$(LIBTRACEEVENT): FORCE | $(LIBTRACEEVENT_OUTPUT)
-       $(Q)$(MAKE) -C $(LIBTRACEEVENT_DIR) O=$(LIBTRACEEVENT_OUTPUT) \
-               DESTDIR=$(LIBTRACEEVENT_DESTDIR) prefix= \
-               $@ install_headers
-
-$(LIBTRACEEVENT)-clean:
-       $(call QUIET_CLEAN, libtraceevent)
-       $(Q)$(RM) -r -- $(LIBTRACEEVENT_OUTPUT)
-
-libtraceevent_plugins: FORCE | $(LIBTRACEEVENT_PLUGINS_OUTPUT)
-       $(Q)$(MAKE) -C $(LIBTRACEEVENT_PLUGINS_DIR) O=$(LIBTRACEEVENT_PLUGINS_OUTPUT) \
-               DESTDIR=$(LIBTRACEEVENT_PLUGINS_DESTDIR) prefix= \
-               plugins
-
-libtraceevent_plugins-clean:
-       $(call QUIET_CLEAN, libtraceevent_plugins)
-       $(Q)$(RM) -r -- $(LIBTRACEEVENT_PLUGINS_OUTPUT)
-
-$(LIBTRACEEVENT_DYNAMIC_LIST): libtraceevent_plugins
-       $(Q)$(MAKE) -C $(LIBTRACEEVENT_PLUGINS_DIR) O=$(LIBTRACEEVENT_PLUGINS_OUTPUT) \
-               DESTDIR=$(LIBTRACEEVENT_PLUGINS_DESTDIR) prefix= \
-               $(LIBTRACEEVENT_FLAGS) $@
-
-install-traceevent-plugins: libtraceevent_plugins
-       $(Q)$(MAKE) -C $(LIBTRACEEVENT_PLUGINS_DIR) O=$(LIBTRACEEVENT_PLUGINS_OUTPUT) \
-               DESTDIR=$(DESTDIR_SQ)$(prefix) prefix= \
-               $(LIBTRACEEVENT_FLAGS) install
-endif
-
 $(LIBAPI): FORCE | $(LIBAPI_OUTPUT)
        $(Q)$(MAKE) -C $(LIBAPI_DIR) O=$(LIBAPI_OUTPUT) \
                DESTDIR=$(LIBAPI_DESTDIR) prefix= \
@@ -1095,10 +1024,6 @@ install-tests: all install-gtk
 
 install-bin: install-tools install-tests
 
-ifndef LIBTRACEEVENT_DYNAMIC
-install-bin: install-traceevent-plugins
-endif
-
 install: install-bin try-install-man
 
 install-python_ext:
@@ -1124,11 +1049,6 @@ SKELETONS += $(SKEL_OUT)/kwork_trace.skel.h
 $(SKEL_TMP_OUT) $(LIBAPI_OUTPUT) $(LIBBPF_OUTPUT) $(LIBPERF_OUTPUT) $(LIBSUBCMD_OUTPUT) $(LIBSYMBOL_OUTPUT):
        $(Q)$(MKDIR) -p $@
 
-ifndef LIBTRACEEVENT_DYNAMIC
-$(LIBTRACEEVENT_OUTPUT) $(LIBTRACEEVENT_PLUGINS_OUTPUT):
-       $(Q)$(MKDIR) -p $@
-endif
-
 ifdef BUILD_BPF_SKEL
 BPFTOOL := $(SKEL_TMP_OUT)/bootstrap/bpftool
 BPF_INCLUDE := -I$(SKEL_TMP_OUT)/.. -I$(LIBBPF_INCLUDE)
@@ -1211,10 +1131,6 @@ clean:: $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBSYMBOL)-clean $(
        $(call QUIET_CLEAN, Documentation) \
        $(MAKE) -C $(DOC_DIR) O=$(OUTPUT) clean >/dev/null
 
-ifndef LIBTRACEEVENT_DYNAMIC
-clean:: $(LIBTRACEEVENT)-clean libtraceevent_plugins-clean
-endif
-
 #
 # To provide FEATURE-DUMP into $(FEATURE_DUMP_COPY)
 # file if defined, with no further action.
@@ -1232,6 +1148,6 @@ FORCE:
 .PHONY: all install clean config-clean strip install-gtk
 .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
 .PHONY: .FORCE-PERF-VERSION-FILE TAGS tags cscope FORCE prepare
-.PHONY: libtraceevent_plugins archheaders
+.PHONY: archheaders
 
 endif # force_fixdep
index 337aa9b..78ef711 100644 (file)
@@ -3,7 +3,7 @@ perf-y += machine.o
 perf-y += perf_regs.o
 perf-y += tsc.o
 perf-y += pmu.o
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
 perf-$(CONFIG_DWARF)     += dwarf-regs.o
 perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o
 perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
index 0115f31..9889245 100644 (file)
@@ -1,5 +1,5 @@
 perf-y += header.o
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
 perf-y += perf_regs.o
 perf-y += mem-events.o
 perf-y += sym-handling.o
index 3d9d0f4..db68840 100644 (file)
@@ -1,5 +1,5 @@
 perf-y += header.o
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
 perf-y += perf_regs.o
 
 perf-$(CONFIG_DWARF) += dwarf-regs.o
index dbeb04c..195ccfd 100644 (file)
@@ -1,7 +1,7 @@
 perf-y += header.o
 perf-y += tsc.o
 perf-y += pmu.o
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
 perf-y += perf_regs.o
 perf-y += topdown.o
 perf-y += machine.o
index af102f4..1e39a03 100644 (file)
@@ -418,6 +418,7 @@ static int intel_pt_info_fill(struct auxtrace_record *itr,
        return 0;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int intel_pt_track_switches(struct evlist *evlist)
 {
        const char *sched_switch = "sched:sched_switch";
@@ -439,6 +440,7 @@ static int intel_pt_track_switches(struct evlist *evlist)
 
        return 0;
 }
+#endif
 
 static void intel_pt_valid_str(char *str, size_t len, u64 valid)
 {
@@ -829,6 +831,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
                                        ptr->have_sched_switch = 2;
                        }
                } else {
+#ifdef HAVE_LIBTRACEEVENT
                        err = intel_pt_track_switches(evlist);
                        if (err == -EPERM)
                                pr_debug2("Unable to select sched:sched_switch\n");
@@ -836,6 +839,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
                                return err;
                        else
                                ptr->have_sched_switch = 1;
+#endif
                }
        }
 
index 517d928..90458ca 100644 (file)
@@ -499,7 +499,9 @@ int cmd_annotate(int argc, const char **argv)
                        .namespaces = perf_event__process_namespaces,
                        .attr   = perf_event__process_attr,
                        .build_id = perf_event__process_build_id,
+#ifdef HAVE_LIBTRACEEVENT
                        .tracing_data   = perf_event__process_tracing_data,
+#endif
                        .id_index       = perf_event__process_id_index,
                        .auxtrace_info  = perf_event__process_auxtrace_info,
                        .auxtrace       = perf_event__process_auxtrace,
index c22d82d..b2a9a3b 100644 (file)
@@ -78,12 +78,13 @@ static int cmd_data_convert(int argc, const char **argv)
                return bt_convert__perf2json(input_name, to_json, &opts);
 
        if (to_ctf) {
-#ifdef HAVE_LIBBABELTRACE_SUPPORT
+#if defined(HAVE_LIBBABELTRACE_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
                return bt_convert__perf2ctf(input_name, to_ctf, &opts);
 #else
                pr_err("The libbabeltrace support is not compiled in. perf should be "
                       "compiled with environment variables LIBBABELTRACE=1 and "
-                      "LIBBABELTRACE_DIR=/path/to/libbabeltrace/\n");
+                      "LIBBABELTRACE_DIR=/path/to/libbabeltrace/.\n"
+                      "Check also if libbtraceevent devel files are available.\n");
                return -1;
 #endif
        }
index e254f18..3f4e4dd 100644 (file)
@@ -607,6 +607,7 @@ static int perf_event__repipe_exit(struct perf_tool *tool,
        return err;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int perf_event__repipe_tracing_data(struct perf_session *session,
                                           union perf_event *event)
 {
@@ -614,6 +615,7 @@ static int perf_event__repipe_tracing_data(struct perf_session *session,
 
        return perf_event__process_tracing_data(session, event);
 }
+#endif
 
 static int dso__read_build_id(struct dso *dso)
 {
@@ -807,6 +809,7 @@ static int perf_inject__sched_switch(struct perf_tool *tool,
        return 0;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int perf_inject__sched_stat(struct perf_tool *tool,
                                   union perf_event *event __maybe_unused,
                                   struct perf_sample *sample,
@@ -836,6 +839,7 @@ found:
        build_id__mark_dso_hit(tool, event_sw, &sample_sw, evsel, machine);
        return perf_event__repipe(tool, event_sw, &sample_sw, machine);
 }
+#endif
 
 static struct guest_vcpu *guest_session__vcpu(struct guest_session *gs, u32 vcpu)
 {
@@ -1961,7 +1965,9 @@ static int __cmd_inject(struct perf_inject *inject)
                inject->tool.mmap         = perf_event__repipe_mmap;
                inject->tool.mmap2        = perf_event__repipe_mmap2;
                inject->tool.fork         = perf_event__repipe_fork;
+#ifdef HAVE_LIBTRACEEVENT
                inject->tool.tracing_data = perf_event__repipe_tracing_data;
+#endif
        }
 
        output_data_offset = perf_session__data_offset(session->evlist);
@@ -1984,8 +1990,10 @@ static int __cmd_inject(struct perf_inject *inject)
                                evsel->handler = perf_inject__sched_switch;
                        } else if (!strcmp(name, "sched:sched_process_exit"))
                                evsel->handler = perf_inject__sched_process_exit;
+#ifdef HAVE_LIBTRACEEVENT
                        else if (!strncmp(name, "sched:sched_stat_", 17))
                                evsel->handler = perf_inject__sched_stat;
+#endif
                }
        } else if (inject->itrace_synth_opts.vm_time_correlation) {
                session->itrace_synth_opts = &inject->itrace_synth_opts;
index ebfab2c..e20656c 100644 (file)
@@ -35,6 +35,7 @@
 #include <regex.h>
 
 #include <linux/ctype.h>
+#include <traceevent/event-parse.h>
 
 static int     kmem_slab;
 static int     kmem_page;
index 7d9ec1b..641e739 100644 (file)
@@ -63,7 +63,7 @@ static const char *get_filename_for_perf_kvm(void)
        return filename;
 }
 
-#ifdef HAVE_KVM_STAT_SUPPORT
+#if defined(HAVE_KVM_STAT_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
 
 void exit_event_get_key(struct evsel *evsel,
                        struct perf_sample *sample,
@@ -654,7 +654,7 @@ static void print_result(struct perf_kvm_stat *kvm)
                pr_info("\nLost events: %" PRIu64 "\n\n", kvm->lost_events);
 }
 
-#ifdef HAVE_TIMERFD_SUPPORT
+#if defined(HAVE_TIMERFD_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
 static int process_lost_event(struct perf_tool *tool,
                              union perf_event *event __maybe_unused,
                              struct perf_sample *sample __maybe_unused,
@@ -742,7 +742,7 @@ static bool verify_vcpu(int vcpu)
        return true;
 }
 
-#ifdef HAVE_TIMERFD_SUPPORT
+#if defined(HAVE_TIMERFD_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
 /* keeping the max events to a modest level to keep
  * the processing of samples per mmap smooth.
  */
@@ -1290,7 +1290,7 @@ kvm_events_report(struct perf_kvm_stat *kvm, int argc, const char **argv)
        return kvm_events_report_vcpu(kvm);
 }
 
-#ifdef HAVE_TIMERFD_SUPPORT
+#if defined(HAVE_TIMERFD_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
 static struct evlist *kvm_live_event_list(void)
 {
        struct evlist *evlist;
@@ -1507,7 +1507,7 @@ static int kvm_cmd_stat(const char *file_name, int argc, const char **argv)
        if (strlen(argv[1]) > 2 && strstarts("report", argv[1]))
                return kvm_events_report(&kvm, argc - 1 , argv + 1);
 
-#ifdef HAVE_TIMERFD_SUPPORT
+#if defined(HAVE_TIMERFD_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
        if (!strncmp(argv[1], "live", 4))
                return kvm_events_live(&kvm, argc - 1 , argv + 1);
 #endif
@@ -1644,7 +1644,7 @@ int cmd_kvm(int argc, const char **argv)
                return cmd_top(argc, argv);
        else if (strlen(argv[0]) > 2 && strstarts("buildid-list", argv[0]))
                return __cmd_buildid_list(file_name, argc, argv);
-#ifdef HAVE_KVM_STAT_SUPPORT
+#if defined(HAVE_KVM_STAT_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
        else if (strlen(argv[0]) > 2 && strstarts("stat", argv[0]))
                return kvm_cmd_stat(file_name, argc, argv);
 #endif
index 0e02b80..dc59d75 100644 (file)
@@ -23,6 +23,7 @@
 
 #include <subcmd/pager.h>
 #include <subcmd/parse-options.h>
+#include <traceevent/event-parse.h>
 
 #include <errno.h>
 #include <inttypes.h>
index b7fd7ec..7e17374 100644 (file)
@@ -1701,8 +1701,10 @@ static void record__init_features(struct record *rec)
        if (rec->no_buildid)
                perf_header__clear_feat(&session->header, HEADER_BUILD_ID);
 
+#ifdef HAVE_LIBTRACEEVENT
        if (!have_tracepoints(&rec->evlist->core.entries))
                perf_header__clear_feat(&session->header, HEADER_TRACING_DATA);
+#endif
 
        if (!rec->opts.branch_stack)
                perf_header__clear_feat(&session->header, HEADER_BRANCH_STACK);
index b6d77d3..2ee2ecc 100644 (file)
 #include <unistd.h>
 #include <linux/mman.h>
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 struct report {
        struct perf_tool        tool;
        struct perf_session     *session;
@@ -1199,7 +1203,9 @@ int cmd_report(int argc, const char **argv)
                        .lost            = perf_event__process_lost,
                        .read            = process_read_event,
                        .attr            = process_attr,
+#ifdef HAVE_LIBTRACEEVENT
                        .tracing_data    = perf_event__process_tracing_data,
+#endif
                        .build_id        = perf_event__process_build_id,
                        .id_index        = perf_event__process_id_index,
                        .auxtrace_info   = perf_event__process_auxtrace_info,
@@ -1660,6 +1666,7 @@ repeat:
                                                  report.range_num);
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        if (session->tevent.pevent &&
            tep_set_function_resolver(session->tevent.pevent,
                                      machine__resolve_kernel_addr,
@@ -1668,7 +1675,7 @@ repeat:
                       __func__);
                return -1;
        }
-
+#endif
        sort__setup_elide(stdout);
 
        ret = __cmd_report(&report);
index d7ec8c1..88888fb 100644 (file)
@@ -62,6 +62,9 @@
 #include "perf.h"
 
 #include <linux/ctype.h>
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
 
 static char const              *script_name;
 static char const              *generate_script_lang;
@@ -2154,12 +2157,12 @@ static void process_event(struct perf_script *script,
                perf_sample__fprintf_bts(sample, evsel, thread, al, addr_al, machine, fp);
                return;
        }
-
+#ifdef HAVE_LIBTRACEEVENT
        if (PRINT_FIELD(TRACE) && sample->raw_data) {
                event_format__fprintf(evsel->tp_format, sample->cpu,
                                      sample->raw_data, sample->raw_size, fp);
        }
-
+#endif
        if (attr->type == PERF_TYPE_SYNTH && PRINT_FIELD(SYNTH))
                perf_sample__fprintf_synth(sample, evsel, fp);
 
@@ -2283,8 +2286,10 @@ static void process_stat_interval(u64 tstamp)
 
 static void setup_scripting(void)
 {
+#ifdef HAVE_LIBTRACEEVENT
        setup_perl_scripting();
        setup_python_scripting();
+#endif
 }
 
 static int flush_scripting(void)
@@ -3784,7 +3789,9 @@ int cmd_script(int argc, const char **argv)
                        .fork            = perf_event__process_fork,
                        .attr            = process_attr,
                        .event_update   = perf_event__process_event_update,
+#ifdef HAVE_LIBTRACEEVENT
                        .tracing_data    = perf_event__process_tracing_data,
+#endif
                        .feature         = process_feature_event,
                        .build_id        = perf_event__process_build_id,
                        .id_index        = perf_event__process_id_index,
@@ -4215,6 +4222,7 @@ script_found:
        else
                symbol_conf.use_callchain = false;
 
+#ifdef HAVE_LIBTRACEEVENT
        if (session->tevent.pevent &&
            tep_set_function_resolver(session->tevent.pevent,
                                      machine__resolve_kernel_addr,
@@ -4223,7 +4231,7 @@ script_found:
                err = -1;
                goto out_delete;
        }
-
+#endif
        if (generate_script_lang) {
                struct stat perf_stat;
                int input;
@@ -4259,9 +4267,12 @@ script_found:
                        err = -ENOENT;
                        goto out_delete;
                }
-
+#ifdef HAVE_LIBTRACEEVENT
                err = scripting_ops->generate_script(session->tevent.pevent,
                                                     "perf-script");
+#else
+               err = scripting_ops->generate_script(NULL, "perf-script");
+#endif
                goto out_delete;
        }
 
index c36296b..6c629e7 100644 (file)
@@ -38,6 +38,7 @@
 #include "util/string2.h"
 #include "util/tracepoint.h"
 #include <linux/err.h>
+#include <traceevent/event-parse.h>
 
 #ifdef LACKS_OPEN_MEMSTREAM_PROTOTYPE
 FILE *open_memstream(char **ptr, size_t *sizeloc);
index 543c379..6909cd9 100644 (file)
@@ -15,7 +15,6 @@
  */
 
 #include "util/record.h"
-#include <traceevent/event-parse.h>
 #include <api/fs/tracing_path.h>
 #include <bpf/bpf.h>
 #include "util/bpf_map.h"
 #include <linux/ctype.h>
 #include <perf/mmap.h>
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 #ifndef O_CLOEXEC
 # define O_CLOEXEC             02000000
 #endif
index a71f491..a886929 100644 (file)
@@ -82,6 +82,7 @@ static void library_status(void)
        STATUS(HAVE_AIO_SUPPORT, aio);
        STATUS(HAVE_ZSTD_SUPPORT, zstd);
        STATUS(HAVE_LIBPFM, libpfm4);
+       STATUS(HAVE_LIBTRACEEVENT, libtraceevent);
 }
 
 int cmd_version(int argc, const char **argv)
index 7af135d..82bbe0c 100644 (file)
@@ -70,20 +70,26 @@ static struct cmd_struct commands[] = {
        { "report",     cmd_report,     0 },
        { "bench",      cmd_bench,      0 },
        { "stat",       cmd_stat,       0 },
+#ifdef HAVE_LIBTRACEEVENT
        { "timechart",  cmd_timechart,  0 },
+#endif
        { "top",        cmd_top,        0 },
        { "annotate",   cmd_annotate,   0 },
        { "version",    cmd_version,    0 },
        { "script",     cmd_script,     0 },
+#ifdef HAVE_LIBTRACEEVENT
        { "sched",      cmd_sched,      0 },
+#endif
 #ifdef HAVE_LIBELF_SUPPORT
        { "probe",      cmd_probe,      0 },
 #endif
+#ifdef HAVE_LIBTRACEEVENT
        { "kmem",       cmd_kmem,       0 },
        { "lock",       cmd_lock,       0 },
+#endif
        { "kvm",        cmd_kvm,        0 },
        { "test",       cmd_test,       0 },
-#if defined(HAVE_LIBAUDIT_SUPPORT) || defined(HAVE_SYSCALL_TABLE_SUPPORT)
+#if defined(HAVE_LIBTRACEEVENT) && (defined(HAVE_LIBAUDIT_SUPPORT) || defined(HAVE_SYSCALL_TABLE_SUPPORT))
        { "trace",      cmd_trace,      0 },
 #endif
        { "inject",     cmd_inject,     0 },
@@ -91,7 +97,9 @@ static struct cmd_struct commands[] = {
        { "data",       cmd_data,       0 },
        { "ftrace",     cmd_ftrace,     0 },
        { "daemon",     cmd_daemon,     0 },
+#ifdef HAVE_LIBTRACEEVENT
        { "kwork",      cmd_kwork,      0 },
+#endif
 };
 
 struct pager_config {
@@ -500,14 +508,18 @@ int main(int argc, const char **argv)
                argv[0] = cmd;
        }
        if (strstarts(cmd, "trace")) {
-#if defined(HAVE_LIBAUDIT_SUPPORT) || defined(HAVE_SYSCALL_TABLE_SUPPORT)
-               setup_path();
-               argv[0] = "trace";
-               return cmd_trace(argc, argv);
-#else
+#ifndef HAVE_LIBTRACEEVENT
+               fprintf(stderr,
+                       "trace command not available: missing libtraceevent devel package at build time.\n");
+               goto out;
+#elif !defined(HAVE_LIBAUDIT_SUPPORT) && !defined(HAVE_SYSCALL_TABLE_SUPPORT)
                fprintf(stderr,
                        "trace command not available: missing audit-libs devel package at build time.\n");
                goto out;
+#else
+               setup_path();
+               argv[0] = "trace";
+               return cmd_trace(argc, argv);
 #endif
        }
        /* Look for flags.. */
index 7d0e33c..d5fed4e 100644 (file)
@@ -1,3 +1,3 @@
-perf-y += Context.o
+perf-$(CONFIG_LIBTRACEEVENT) += Context.o
 
 CFLAGS_Context.o += $(PYTHON_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs
index 658b505..90fd1eb 100644 (file)
@@ -6,13 +6,13 @@ perf-y += parse-events.o
 perf-y += dso-data.o
 perf-y += attr.o
 perf-y += vmlinux-kallsyms.o
-perf-y += openat-syscall.o
-perf-y += openat-syscall-all-cpus.o
-perf-y += openat-syscall-tp-fields.o
-perf-y += mmap-basic.o
+perf-$(CONFIG_LIBTRACEEVENT) += openat-syscall.o
+perf-$(CONFIG_LIBTRACEEVENT) += openat-syscall-all-cpus.o
+perf-$(CONFIG_LIBTRACEEVENT) += openat-syscall-tp-fields.o
+perf-$(CONFIG_LIBTRACEEVENT) += mmap-basic.o
 perf-y += perf-record.o
 perf-y += evsel-roundtrip-name.o
-perf-y += evsel-tp-sched.o
+perf-$(CONFIG_LIBTRACEEVENT) += evsel-tp-sched.o
 perf-y += fdarray.o
 perf-y += pmu.o
 perf-y += pmu-events.o
@@ -30,7 +30,7 @@ perf-y += task-exit.o
 perf-y += sw-clock.o
 perf-y += mmap-thread-lookup.o
 perf-y += thread-maps-share.o
-perf-y += switch-tracking.o
+perf-$(CONFIG_LIBTRACEEVENT) += switch-tracking.o
 perf-y += keep-tracking.o
 perf-y += code-reading.o
 perf-y += sample-parsing.o
index ddd8262..f6c16ad 100644 (file)
@@ -38,9 +38,11 @@ struct test_suite *__weak arch_tests[] = {
 
 static struct test_suite *generic_tests[] = {
        &suite__vmlinux_matches_kallsyms,
+#ifdef HAVE_LIBTRACEEVENT
        &suite__openat_syscall_event,
        &suite__openat_syscall_event_on_all_cpus,
        &suite__basic_mmap,
+#endif
        &suite__mem,
        &suite__parse_events,
        &suite__expr,
@@ -51,8 +53,10 @@ static struct test_suite *generic_tests[] = {
        &suite__dso_data_cache,
        &suite__dso_data_reopen,
        &suite__perf_evsel__roundtrip_name_test,
+#ifdef HAVE_LIBTRACEEVENT
        &suite__perf_evsel__tp_sched_test,
        &suite__syscall_openat_tp_fields,
+#endif
        &suite__attr,
        &suite__hists_link,
        &suite__python_use,
@@ -71,7 +75,9 @@ static struct test_suite *generic_tests[] = {
        &suite__thread_maps_share,
        &suite__hists_output,
        &suite__hists_cumulate,
+#ifdef HAVE_LIBTRACEEVENT
        &suite__switch_tracking,
+#endif
        &suite__fdarray__filter,
        &suite__fdarray__add,
        &suite__kmod_path__parse,
index 3440dd2..71a5cb3 100644 (file)
@@ -20,6 +20,8 @@
 #define PERF_TP_SAMPLE_TYPE (PERF_SAMPLE_RAW | PERF_SAMPLE_TIME | \
                             PERF_SAMPLE_CPU | PERF_SAMPLE_PERIOD)
 
+#ifdef HAVE_LIBTRACEEVENT
+
 #if defined(__s390x__)
 /* Return true if kvm module is available and loaded. Test this
  * and return success when trace point kvm_s390_create_vm
@@ -76,6 +78,7 @@ static int test__checkevent_tracepoint_multi(struct evlist *evlist)
        }
        return TEST_OK;
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 static int test__checkevent_raw(struct evlist *evlist)
 {
@@ -222,6 +225,7 @@ static int test__checkevent_breakpoint_rw(struct evlist *evlist)
        return TEST_OK;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int test__checkevent_tracepoint_modifier(struct evlist *evlist)
 {
        struct evsel *evsel = evlist__first(evlist);
@@ -252,6 +256,7 @@ test__checkevent_tracepoint_multi_modifier(struct evlist *evlist)
 
        return test__checkevent_tracepoint_multi(evlist);
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 static int test__checkevent_raw_modifier(struct evlist *evlist)
 {
@@ -453,6 +458,7 @@ static int test__checkevent_pmu(struct evlist *evlist)
        return TEST_OK;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int test__checkevent_list(struct evlist *evlist)
 {
        struct evsel *evsel = evlist__first(evlist);
@@ -491,6 +497,7 @@ static int test__checkevent_list(struct evlist *evlist)
 
        return TEST_OK;
 }
+#endif
 
 static int test__checkevent_pmu_name(struct evlist *evlist)
 {
@@ -762,6 +769,7 @@ static int test__group2(struct evlist *evlist)
        return TEST_OK;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int test__group3(struct evlist *evlist __maybe_unused)
 {
        struct evsel *evsel, *leader;
@@ -853,6 +861,7 @@ static int test__group3(struct evlist *evlist __maybe_unused)
 
        return TEST_OK;
 }
+#endif
 
 static int test__group4(struct evlist *evlist __maybe_unused)
 {
@@ -1460,6 +1469,7 @@ static int test__sym_event_dc(struct evlist *evlist)
        return TEST_OK;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int count_tracepoints(void)
 {
        struct dirent *events_ent;
@@ -1513,6 +1523,7 @@ static int test__all_tracepoints(struct evlist *evlist)
 
        return test__checkevent_tracepoint_multi(evlist);
 }
+#endif /* HAVE_LIBTRACEVENT */
 
 static int test__hybrid_hw_event_with_pmu(struct evlist *evlist)
 {
@@ -1642,6 +1653,7 @@ struct evlist_test {
 };
 
 static const struct evlist_test test__events[] = {
+#ifdef HAVE_LIBTRACEEVENT
        {
                .name  = "syscalls:sys_enter_openat",
                .check = test__checkevent_tracepoint,
@@ -1652,6 +1664,7 @@ static const struct evlist_test test__events[] = {
                .check = test__checkevent_tracepoint_multi,
                /* 1 */
        },
+#endif
        {
                .name  = "r1a",
                .check = test__checkevent_raw,
@@ -1702,6 +1715,7 @@ static const struct evlist_test test__events[] = {
                .check = test__checkevent_breakpoint_w,
                /* 1 */
        },
+#ifdef HAVE_LIBTRACEEVENT
        {
                .name  = "syscalls:sys_enter_openat:k",
                .check = test__checkevent_tracepoint_modifier,
@@ -1712,6 +1726,7 @@ static const struct evlist_test test__events[] = {
                .check = test__checkevent_tracepoint_multi_modifier,
                /* 3 */
        },
+#endif
        {
                .name  = "r1a:kp",
                .check = test__checkevent_raw_modifier,
@@ -1757,11 +1772,13 @@ static const struct evlist_test test__events[] = {
                .check = test__checkevent_breakpoint_w_modifier,
                /* 2 */
        },
+#ifdef HAVE_LIBTRACEEVENT
        {
                .name  = "r1,syscalls:sys_enter_openat:k,1:1:hp",
                .check = test__checkevent_list,
                /* 3 */
        },
+#endif
        {
                .name  = "instructions:G",
                .check = test__checkevent_exclude_host_modifier,
@@ -1792,11 +1809,13 @@ static const struct evlist_test test__events[] = {
                .check = test__group2,
                /* 9 */
        },
+#ifdef HAVE_LIBTRACEEVENT
        {
                .name  = "group1{syscalls:sys_enter_openat:H,cycles:kppp},group2{cycles,1:3}:G,instructions:u",
                .check = test__group3,
                /* 0 */
        },
+#endif
        {
                .name  = "{cycles:u,instructions:kp}:p",
                .check = test__group4,
@@ -1807,11 +1826,13 @@ static const struct evlist_test test__events[] = {
                .check = test__group5,
                /* 2 */
        },
+#ifdef HAVE_LIBTRACEEVENT
        {
                .name  = "*:*",
                .check = test__all_tracepoints,
                /* 3 */
        },
+#endif
        {
                .name  = "{cycles,cache-misses:G}:H",
                .check = test__group_gh1,
@@ -1867,7 +1888,7 @@ static const struct evlist_test test__events[] = {
                .check = test__checkevent_breakpoint_len_rw_modifier,
                /* 4 */
        },
-#if defined(__s390x__)
+#if defined(__s390x__) && defined(HAVE_LIBTRACEEVENT)
        {
                .name  = "kvm-s390:kvm_s390_create_vm",
                .check = test__checkevent_tracepoint,
index d04802b..8345464 100644 (file)
@@ -69,7 +69,6 @@ perf-y += namespaces.o
 perf-y += comm.o
 perf-y += thread.o
 perf-y += thread_map.o
-perf-y += trace-event-parse.o
 perf-y += parse-events-flex.o
 perf-y += parse-events-bison.o
 perf-y += pmu.o
@@ -77,11 +76,12 @@ perf-y += pmus.o
 perf-y += pmu-flex.o
 perf-y += pmu-bison.o
 perf-y += pmu-hybrid.o
-perf-y += trace-event-read.o
-perf-y += trace-event-info.o
-perf-y += trace-event-scripting.o
-perf-y += trace-event.o
 perf-y += svghelper.o
+perf-$(CONFIG_LIBTRACEEVENT) += trace-event-info.o
+perf-$(CONFIG_LIBTRACEEVENT) += trace-event-scripting.o
+perf-$(CONFIG_LIBTRACEEVENT) += trace-event.o
+perf-$(CONFIG_LIBTRACEEVENT) += trace-event-parse.o
+perf-$(CONFIG_LIBTRACEEVENT) += trace-event-read.o
 perf-y += sort.o
 perf-y += hist.o
 perf-y += util.o
@@ -153,8 +153,12 @@ perf-$(CONFIG_PERF_BPF_SKEL) += bpf_counter.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_counter_cgroup.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_ftrace.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_off_cpu.o
-perf-$(CONFIG_PERF_BPF_SKEL) += bpf_kwork.o
 perf-$(CONFIG_PERF_BPF_SKEL) += bpf_lock_contention.o
+
+ifeq ($(CONFIG_LIBTRACEEVENT),y)
+  perf-$(CONFIG_PERF_BPF_SKEL) += bpf_kwork.o
+endif
+
 perf-$(CONFIG_BPF_PROLOGUE) += bpf-prologue.o
 perf-$(CONFIG_LIBELF) += symbol-elf.o
 perf-$(CONFIG_LIBELF) += probe-file.o
@@ -189,7 +193,10 @@ perf-$(CONFIG_LIBUNWIND)          += unwind-libunwind.o
 perf-$(CONFIG_LIBUNWIND_X86)      += libunwind/x86_32.o
 perf-$(CONFIG_LIBUNWIND_AARCH64)  += libunwind/arm64.o
 
-perf-$(CONFIG_LIBBABELTRACE) += data-convert-bt.o
+ifeq ($(CONFIG_LIBTRACEEVENT),y)
+  perf-$(CONFIG_LIBBABELTRACE) += data-convert-bt.o
+endif
+
 perf-y += data-convert-json.o
 
 perf-y += scripting-engines/
index c65cdaf..8031b58 100644 (file)
@@ -19,7 +19,6 @@
 #include <babeltrace/ctf-writer/event-fields.h>
 #include <babeltrace/ctf-ir/utils.h>
 #include <babeltrace/ctf/events.h>
-#include <traceevent/event-parse.h>
 #include "asm/bug.h"
 #include "data-convert.h"
 #include "session.h"
 #include "clockid.h"
 #include "util/sample.h"
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 #define pr_N(n, fmt, ...) \
        eprintf(n, debug_data_convert, fmt, ##__VA_ARGS__)
 
index 57db590..ba9d93c 100644 (file)
 #include "util/thread.h"
 #include "util/tool.h"
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 struct convert_json {
        struct perf_tool tool;
        FILE *out;
@@ -217,6 +221,7 @@ static int process_sample_event(struct perf_tool *tool,
        }
        output_json_format(out, false, 3, "]");
 
+#ifdef HAVE_LIBTRACEEVENT
        if (sample->raw_data) {
                int i;
                struct tep_format_field **fields;
@@ -236,7 +241,7 @@ static int process_sample_event(struct perf_tool *tool,
                        free(fields);
                }
        }
-
+#endif
        output_json_format(out, false, 2, "}");
        return 0;
 }
@@ -313,7 +318,9 @@ int bt_convert__perf2json(const char *input_name, const char *output_name,
                        .exit           = perf_event__process_exit,
                        .fork           = perf_event__process_fork,
                        .lost           = perf_event__process_lost,
+#ifdef HAVE_LIBTRACEEVENT
                        .tracing_data   = perf_event__process_tracing_data,
+#endif
                        .build_id       = perf_event__process_build_id,
                        .id_index       = perf_event__process_id_index,
                        .auxtrace_info  = perf_event__process_auxtrace_info,
index fbf3192..590d4e7 100644 (file)
@@ -290,6 +290,7 @@ struct evsel *evlist__add_aux_dummy(struct evlist *evlist, bool system_wide)
        return evsel;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide)
 {
        struct evsel *evsel = evsel__newtp_idx("sched", "sched_switch", 0);
@@ -305,7 +306,8 @@ struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide)
 
        evlist__add(evlist, evsel);
        return evsel;
-};
+}
+#endif
 
 int evlist__add_attrs(struct evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs)
 {
@@ -376,6 +378,7 @@ struct evsel *evlist__find_tracepoint_by_name(struct evlist *evlist, const char
        return NULL;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 int evlist__add_newtp(struct evlist *evlist, const char *sys, const char *name, void *handler)
 {
        struct evsel *evsel = evsel__newtp(sys, name);
@@ -387,6 +390,7 @@ int evlist__add_newtp(struct evlist *evlist, const char *sys, const char *name,
        evlist__add(evlist, evsel);
        return 0;
 }
+#endif
 
 struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affinity *affinity)
 {
index 16734c6..e5b84ea 100644 (file)
@@ -127,7 +127,9 @@ static inline struct evsel *evlist__add_dummy_on_all_cpus(struct evlist *evlist)
 {
        return evlist__add_aux_dummy(evlist, true);
 }
+#ifdef HAVE_LIBTRACEEVENT
 struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide);
+#endif
 
 int evlist__add_sb_event(struct evlist *evlist, struct perf_event_attr *attr,
                         evsel__sb_cb_t cb, void *data);
@@ -135,7 +137,9 @@ void evlist__set_cb(struct evlist *evlist, evsel__sb_cb_t cb, void *data);
 int evlist__start_sb_thread(struct evlist *evlist, struct target *target);
 void evlist__stop_sb_thread(struct evlist *evlist);
 
+#ifdef HAVE_LIBTRACEEVENT
 int evlist__add_newtp(struct evlist *evlist, const char *sys, const char *name, void *handler);
+#endif
 
 int __evlist__set_tracepoints_handlers(struct evlist *evlist,
                                       const struct evsel_str_handler *assocs,
index 0f61735..ca91185 100644 (file)
@@ -12,7 +12,6 @@
 #include <linux/bitops.h>
 #include <api/fs/fs.h>
 #include <api/fs/tracing_path.h>
-#include <traceevent/event-parse.h>
 #include <linux/hw_breakpoint.h>
 #include <linux/perf_event.h>
 #include <linux/compiler.h>
 
 #include <linux/ctype.h>
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 struct perf_missing_features perf_missing_features;
 
 static clockid_t clockid;
@@ -439,7 +442,9 @@ struct evsel *evsel__clone(struct evsel *orig)
                        goto out_err;
        }
        evsel->cgrp = cgroup__get(orig->cgrp);
+#ifdef HAVE_LIBTRACEEVENT
        evsel->tp_format = orig->tp_format;
+#endif
        evsel->handler = orig->handler;
        evsel->core.leader = orig->core.leader;
 
@@ -479,6 +484,7 @@ out_err:
 /*
  * Returns pointer with encoded error via <linux/err.h> interface.
  */
+#ifdef HAVE_LIBTRACEEVENT
 struct evsel *evsel__newtp_idx(const char *sys, const char *name, int idx)
 {
        struct evsel *evsel = zalloc(perf_evsel__object.size);
@@ -516,6 +522,7 @@ out_free:
 out_err:
        return ERR_PTR(err);
 }
+#endif
 
 const char *const evsel__hw_names[PERF_COUNT_HW_MAX] = {
        "cycles",
@@ -2758,6 +2765,7 @@ u16 evsel__id_hdr_size(struct evsel *evsel)
        return size;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 struct tep_format_field *evsel__field(struct evsel *evsel, const char *name)
 {
        return tep_find_field(evsel->tp_format, name);
@@ -2831,6 +2839,7 @@ u64 evsel__intval(struct evsel *evsel, struct perf_sample *sample, const char *n
 
        return field ? format_field__intval(field, sample, evsel->needs_swap) : 0;
 }
+#endif
 
 bool evsel__fallback(struct evsel *evsel, int err, char *msg, size_t msgsize)
 {
index f348579..d572be4 100644 (file)
@@ -72,7 +72,9 @@ struct evsel {
                char                    *name;
                char                    *group_name;
                const char              *pmu_name;
+#ifdef HAVE_LIBTRACEEVENT
                struct tep_event        *tp_format;
+#endif
                char                    *filter;
                unsigned long           max_events;
                double                  scale;
@@ -223,11 +225,14 @@ static inline struct evsel *evsel__new(struct perf_event_attr *attr)
 }
 
 struct evsel *evsel__clone(struct evsel *orig);
-struct evsel *evsel__newtp_idx(const char *sys, const char *name, int idx);
 
 int copy_config_terms(struct list_head *dst, struct list_head *src);
 void free_config_terms(struct list_head *config_terms);
 
+
+#ifdef HAVE_LIBTRACEEVENT
+struct evsel *evsel__newtp_idx(const char *sys, const char *name, int idx);
+
 /*
  * Returns pointer with encoded error via <linux/err.h> interface.
  */
@@ -235,10 +240,13 @@ static inline struct evsel *evsel__newtp(const char *sys, const char *name)
 {
        return evsel__newtp_idx(sys, name, 0);
 }
+#endif
 
 struct evsel *evsel__new_cycles(bool precise, __u32 type, __u64 config);
 
+#ifdef HAVE_LIBTRACEEVENT
 struct tep_event *event_format__new(const char *sys, const char *name);
+#endif
 
 void evsel__init(struct evsel *evsel, struct perf_event_attr *attr, int idx);
 void evsel__exit(struct evsel *evsel);
@@ -323,6 +331,7 @@ bool evsel__precise_ip_fallback(struct evsel *evsel);
 
 struct perf_sample;
 
+#ifdef HAVE_LIBTRACEEVENT
 void *evsel__rawptr(struct evsel *evsel, struct perf_sample *sample, const char *name);
 u64 evsel__intval(struct evsel *evsel, struct perf_sample *sample, const char *name);
 
@@ -330,6 +339,7 @@ static inline char *evsel__strval(struct evsel *evsel, struct perf_sample *sampl
 {
        return evsel__rawptr(evsel, sample, name);
 }
+#endif
 
 struct tep_format_field;
 
index 8c2ea80..bd22c49 100644 (file)
@@ -2,7 +2,6 @@
 #include <inttypes.h>
 #include <stdio.h>
 #include <stdbool.h>
-#include <traceevent/event-parse.h>
 #include "evsel.h"
 #include "util/evsel_fprintf.h"
 #include "util/event.h"
 #include "srcline.h"
 #include "dso.h"
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 static int comma_fprintf(FILE *fp, bool *first, const char *fmt, ...)
 {
        va_list args;
@@ -74,6 +77,7 @@ int evsel__fprintf(struct evsel *evsel, struct perf_attr_details *details, FILE
                                         term, (u64)evsel->core.attr.sample_freq);
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        if (details->trace_fields) {
                struct tep_format_field *field;
 
@@ -96,6 +100,7 @@ int evsel__fprintf(struct evsel *evsel, struct perf_attr_details *details, FILE
                        field = field->next;
                }
        }
+#endif
 out:
        fputc('\n', fp);
        return ++printed;
index dc2ae39..404d816 100644 (file)
@@ -6,6 +6,7 @@
 #include <sys/types.h>
 #include <byteswap.h>
 #include <unistd.h>
+#include <regex.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <linux/compiler.h>
 #include <linux/ctype.h>
 #include <internal/lib.h>
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 /*
  * magic2 = "PERFILE2"
  * must be a numerical value to let the endianness
@@ -298,6 +303,7 @@ static int do_read_bitmap(struct feat_fd *ff, unsigned long **pset, u64 *psize)
        return 0;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int write_tracing_data(struct feat_fd *ff,
                              struct evlist *evlist)
 {
@@ -306,6 +312,7 @@ static int write_tracing_data(struct feat_fd *ff,
 
        return read_tracing_data(ff->fd, &evlist->core.entries);
 }
+#endif
 
 static int write_build_id(struct feat_fd *ff,
                          struct evlist *evlist __maybe_unused)
@@ -2394,12 +2401,14 @@ FEAT_PROCESS_STR_FUN(arch, arch);
 FEAT_PROCESS_STR_FUN(cpudesc, cpu_desc);
 FEAT_PROCESS_STR_FUN(cpuid, cpuid);
 
+#ifdef HAVE_LIBTRACEEVENT
 static int process_tracing_data(struct feat_fd *ff, void *data)
 {
        ssize_t ret = trace_report(ff->fd, data, false);
 
        return ret < 0 ? -1 : 0;
 }
+#endif
 
 static int process_build_id(struct feat_fd *ff, void *data __maybe_unused)
 {
@@ -3366,7 +3375,9 @@ err:
 const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE];
 
 const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE] = {
+#ifdef HAVE_LIBTRACEEVENT
        FEAT_OPN(TRACING_DATA,  tracing_data,   false),
+#endif
        FEAT_OPN(BUILD_ID,      build_id,       false),
        FEAT_OPR(HOSTNAME,      hostname,       false),
        FEAT_OPR(OSRELEASE,     osrelease,      false),
@@ -4082,6 +4093,7 @@ static int read_attr(int fd, struct perf_header *ph,
        return ret <= 0 ? -1 : 0;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int evsel__prepare_tracepoint_event(struct evsel *evsel, struct tep_handle *pevent)
 {
        struct tep_event *event;
@@ -4125,6 +4137,7 @@ static int evlist__prepare_tracepoint_events(struct evlist *evlist, struct tep_h
 
        return 0;
 }
+#endif
 
 int perf_session__read_header(struct perf_session *session, int repipe_fd)
 {
@@ -4230,11 +4243,15 @@ int perf_session__read_header(struct perf_session *session, int repipe_fd)
                lseek(fd, tmp, SEEK_SET);
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        perf_header__process_sections(header, fd, &session->tevent,
                                      perf_file_section__process);
 
        if (evlist__prepare_tracepoint_events(session->evlist, session->tevent.pevent))
                goto out_delete_evlist;
+#else
+       perf_header__process_sections(header, fd, NULL, perf_file_section__process);
+#endif
 
        return 0;
 out_errno:
@@ -4412,6 +4429,7 @@ int perf_event__process_event_update(struct perf_tool *tool __maybe_unused,
        return 0;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 int perf_event__process_tracing_data(struct perf_session *session,
                                     union perf_event *event)
 {
@@ -4459,6 +4477,7 @@ int perf_event__process_tracing_data(struct perf_session *session,
 
        return size_read + padding;
 }
+#endif
 
 int perf_event__process_build_id(struct perf_session *session,
                                 union perf_event *event)
index 2d5e601..e3861ae 100644 (file)
@@ -160,8 +160,10 @@ int perf_event__process_event_update(struct perf_tool *tool,
                                     union perf_event *event,
                                     struct evlist **pevlist);
 size_t perf_event__fprintf_event_update(union perf_event *event, FILE *fp);
+#ifdef HAVE_LIBTRACEEVENT
 int perf_event__process_tracing_data(struct perf_session *session,
                                     union perf_event *event);
+#endif
 int perf_event__process_build_id(struct perf_session *session,
                                 union perf_event *event);
 bool is_perf_magic(u64 magic);
index e3548dd..6d39216 100644 (file)
@@ -3142,6 +3142,7 @@ static int intel_pt_sync_switch(struct intel_pt *pt, int cpu, pid_t tid,
        return 1;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int intel_pt_process_switch(struct intel_pt *pt,
                                   struct perf_sample *sample)
 {
@@ -3165,6 +3166,7 @@ static int intel_pt_process_switch(struct intel_pt *pt,
 
        return machine__set_current_tid(pt->machine, cpu, -1, tid);
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 static int intel_pt_context_switch_in(struct intel_pt *pt,
                                      struct perf_sample *sample)
@@ -3433,9 +3435,12 @@ static int intel_pt_process_event(struct perf_session *session,
                        return err;
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        if (pt->switch_evsel && event->header.type == PERF_RECORD_SAMPLE)
                err = intel_pt_process_switch(pt, sample);
-       else if (event->header.type == PERF_RECORD_ITRACE_START)
+       else
+#endif
+       if (event->header.type == PERF_RECORD_ITRACE_START)
                err = intel_pt_process_itrace_start(pt, event, sample);
        else if (event->header.type == PERF_RECORD_AUX_OUTPUT_HW_ID)
                err = intel_pt_process_aux_output_hw_id(pt, event, sample);
index 6502cd6..21cce83 100644 (file)
@@ -445,6 +445,7 @@ out_free_terms:
        return ret;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static void tracepoint_error(struct parse_events_error *e, int err,
                             const char *sys, const char *name)
 {
@@ -593,6 +594,7 @@ static int add_tracepoint_multi_sys(struct list_head *list, int *idx,
        closedir(events_dir);
        return ret;
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 #ifdef HAVE_LIBBPF_SUPPORT
 struct __add_bpf_event_param {
@@ -1143,6 +1145,7 @@ static int config_term_pmu(struct perf_event_attr *attr,
                return config_term_common(attr, term, err);
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int config_term_tracepoint(struct perf_event_attr *attr,
                                  struct parse_events_term *term,
                                  struct parse_events_error *err)
@@ -1170,6 +1173,7 @@ static int config_term_tracepoint(struct perf_event_attr *attr,
 
        return 0;
 }
+#endif
 
 static int config_attr(struct perf_event_attr *attr,
                       struct list_head *head,
@@ -1325,6 +1329,7 @@ int parse_events_add_tracepoint(struct list_head *list, int *idx,
                                struct parse_events_error *err,
                                struct list_head *head_config)
 {
+#ifdef HAVE_LIBTRACEEVENT
        if (head_config) {
                struct perf_event_attr attr;
 
@@ -1339,6 +1344,16 @@ int parse_events_add_tracepoint(struct list_head *list, int *idx,
        else
                return add_tracepoint_event(list, idx, sys, event,
                                            err, head_config);
+#else
+       (void)list;
+       (void)idx;
+       (void)sys;
+       (void)event;
+       (void)head_config;
+       parse_events_error__handle(err, 0, strdup("unsupported tracepoint"),
+                               strdup("libtraceevent is necessary for tracepoint support"));
+       return -1;
+#endif
 }
 
 int parse_events_add_numeric(struct parse_events_state *parse_state,
index 07df7bb..428e72e 100644 (file)
@@ -18,7 +18,6 @@ struct parse_events_error;
 struct option;
 struct perf_pmu;
 
-bool have_tracepoints(struct list_head *evlist);
 bool is_event_supported(u8 type, u64 config);
 
 const char *event_type(int type);
index b5941c7..6fb84b7 100644 (file)
@@ -5,7 +5,9 @@
 #include <poll.h>
 #include <linux/err.h>
 #include <perf/cpumap.h>
+#ifdef HAVE_LIBTRACEEVENT
 #include <traceevent/event-parse.h>
+#endif
 #include <perf/mmap.h>
 #include "evlist.h"
 #include "callchain.h"
@@ -417,6 +419,7 @@ static PyObject *pyrf_sample_event__repr(struct pyrf_event *pevent)
        return ret;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static bool is_tracepoint(struct pyrf_event *pevent)
 {
        return pevent->evsel->core.attr.type == PERF_TYPE_TRACEPOINT;
@@ -486,14 +489,17 @@ get_tracepoint_field(struct pyrf_event *pevent, PyObject *attr_name)
 
        return tracepoint_field(pevent, field);
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 static PyObject*
 pyrf_sample_event__getattro(struct pyrf_event *pevent, PyObject *attr_name)
 {
        PyObject *obj = NULL;
 
+#ifdef HAVE_LIBTRACEEVENT
        if (is_tracepoint(pevent))
                obj = get_tracepoint_field(pevent, attr_name);
+#endif
 
        return obj ?: PyObject_GenericGetAttr((PyObject *) pevent, attr_name);
 }
@@ -1326,6 +1332,9 @@ static struct {
 static PyObject *pyrf__tracepoint(struct pyrf_evsel *pevsel,
                                  PyObject *args, PyObject *kwargs)
 {
+#ifndef HAVE_LIBTRACEEVENT
+       return NULL;
+#else
        struct tep_event *tp_format;
        static char *kwlist[] = { "sys", "name", NULL };
        char *sys  = NULL;
@@ -1340,6 +1349,7 @@ static PyObject *pyrf__tracepoint(struct pyrf_evsel *pevsel,
                return _PyLong_FromLong(-1);
 
        return _PyLong_FromLong(tp_format->id);
+#endif // HAVE_LIBTRACEEVENT
 }
 
 static PyMethodDef perf__methods[] = {
index 0f5ba28..d47820c 100644 (file)
@@ -1,5 +1,7 @@
-perf-$(CONFIG_LIBPERL)   += trace-event-perl.o
-perf-$(CONFIG_LIBPYTHON) += trace-event-python.o
+ifeq ($(CONFIG_LIBTRACEEVENT),y)
+  perf-$(CONFIG_LIBPERL)   += trace-event-perl.o
+  perf-$(CONFIG_LIBPYTHON) += trace-event-python.o
+endif
 
 CFLAGS_trace-event-perl.o += $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow -Wno-nested-externs -Wno-undef -Wno-switch-default -Wno-bad-function-cast -Wno-declaration-after-statement -Wno-switch-enum
 
index 5b602b6..0bacb49 100644 (file)
@@ -27,6 +27,7 @@
 #include <errno.h>
 #include <linux/bitmap.h>
 #include <linux/time64.h>
+#include <traceevent/event-parse.h>
 
 #include <stdbool.h>
 /* perl needs the following define, right after including stdbool.h */
index d685a73..fabba21 100644 (file)
@@ -30,6 +30,7 @@
 #include <linux/bitmap.h>
 #include <linux/compiler.h>
 #include <linux/time64.h>
+#include <traceevent/event-parse.h>
 
 #include "../build-id.h"
 #include "../counts.h"
index 1facd46..7c021c6 100644 (file)
@@ -314,7 +314,9 @@ void perf_session__delete(struct perf_session *session)
                        evlist__delete(session->evlist);
                perf_data__close(session->data);
        }
+#ifdef HAVE_LIBTRACEEVENT
        trace_event__cleanup(&session->tevent);
+#endif
        free(session);
 }
 
index be5871e..ee3715e 100644 (file)
@@ -33,7 +33,9 @@ struct perf_session {
        struct auxtrace         *auxtrace;
        struct itrace_synth_opts *itrace_synth_opts;
        struct list_head        auxtrace_index;
+#ifdef HAVE_LIBTRACEEVENT
        struct trace_event      tevent;
+#endif
        struct perf_record_time_conv    time_conv;
        bool                    repipe;
        bool                    one_mmap;
index 43e7ca4..e80ffbb 100644 (file)
@@ -63,12 +63,18 @@ libperf = getenv('LIBPERF')
 ext_sources = [f.strip() for f in open('util/python-ext-sources')
                                if len(f.strip()) > 0 and f[0] != '#']
 
+extra_libraries = []
+
+if '-DHAVE_LIBTRACEEVENT' in cflags:
+    extra_libraries += [ 'traceevent' ]
+else:
+    ext_sources.remove('util/trace-event.c')
+
 # use full paths with source files
 ext_sources = list(map(lambda x: '%s/%s' % (src_perf, x) , ext_sources))
 
-extra_libraries = []
 if '-DHAVE_LIBNUMA_SUPPORT' in cflags:
-    extra_libraries = [ 'numa' ]
+    extra_libraries += [ 'numa' ]
 if '-DHAVE_LIBCAP_SUPPORT' in cflags:
     extra_libraries += [ 'cap' ]
 
index 2e73308..c7a97b3 100644 (file)
@@ -22,7 +22,6 @@
 #include "srcline.h"
 #include "strlist.h"
 #include "strbuf.h"
-#include <traceevent/event-parse.h>
 #include "mem-events.h"
 #include "annotate.h"
 #include "event.h"
 #include <linux/kernel.h>
 #include <linux/string.h>
 
+#ifdef HAVE_LIBTRACEEVENT
+#include <traceevent/event-parse.h>
+#endif
+
 regex_t                parent_regex;
 const char     default_parent_pattern[] = "^sys_|^do_page_fault";
 const char     *parent_pattern = default_parent_pattern;
@@ -743,6 +746,7 @@ struct sort_entry sort_time = {
 
 /* --sort trace */
 
+#ifdef HAVE_LIBTRACEEVENT
 static char *get_trace_output(struct hist_entry *he)
 {
        struct trace_seq seq;
@@ -806,6 +810,7 @@ struct sort_entry sort_trace = {
        .se_snprintf    = hist_entry__trace_snprintf,
        .se_width_idx   = HISTC_TRACE,
 };
+#endif /* HAVE_LIBTRACEEVENT */
 
 /* sort keys for branch stacks */
 
@@ -2022,7 +2027,9 @@ static struct sort_dimension common_sort_dimensions[] = {
        DIM(SORT_LOCAL_WEIGHT, "local_weight", sort_local_weight),
        DIM(SORT_GLOBAL_WEIGHT, "weight", sort_global_weight),
        DIM(SORT_TRANSACTION, "transaction", sort_transaction),
+#ifdef HAVE_LIBTRACEEVENT
        DIM(SORT_TRACE, "trace", sort_trace),
+#endif
        DIM(SORT_SYM_SIZE, "symbol_size", sort_sym_size),
        DIM(SORT_DSO_SIZE, "dso_size", sort_dso_size),
        DIM(SORT_CGROUP, "cgroup", sort_cgroup),
@@ -2206,7 +2213,14 @@ bool perf_hpp__is_ ## key ## _entry(struct perf_hpp_fmt *fmt)    \
        return hse->se == &sort_ ## key ;                       \
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 MK_SORT_ENTRY_CHK(trace)
+#else
+bool perf_hpp__is_trace_entry(struct perf_hpp_fmt *fmt __maybe_unused)
+{
+       return false;
+}
+#endif
 MK_SORT_ENTRY_CHK(srcline)
 MK_SORT_ENTRY_CHK(srcfile)
 MK_SORT_ENTRY_CHK(thread)
@@ -2347,6 +2361,17 @@ static int __sort_dimension__add_hpp_output(struct sort_dimension *sd,
        return 0;
 }
 
+#ifndef HAVE_LIBTRACEEVENT
+bool perf_hpp__is_dynamic_entry(struct perf_hpp_fmt *fmt __maybe_unused)
+{
+       return false;
+}
+bool perf_hpp__defined_dynamic_entry(struct perf_hpp_fmt *fmt __maybe_unused,
+                                    struct hists *hists __maybe_unused)
+{
+       return false;
+}
+#else
 struct hpp_dynamic_entry {
        struct perf_hpp_fmt hpp;
        struct evsel *evsel;
@@ -2621,6 +2646,7 @@ __alloc_dynamic_entry(struct evsel *evsel, struct tep_format_field *field,
 
        return hde;
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 struct perf_hpp_fmt *perf_hpp_fmt__dup(struct perf_hpp_fmt *fmt)
 {
@@ -2633,6 +2659,7 @@ struct perf_hpp_fmt *perf_hpp_fmt__dup(struct perf_hpp_fmt *fmt)
                new_hse = memdup(hse, sizeof(*hse));
                if (new_hse)
                        new_fmt = &new_hse->hpp;
+#ifdef HAVE_LIBTRACEEVENT
        } else if (perf_hpp__is_dynamic_entry(fmt)) {
                struct hpp_dynamic_entry *hde, *new_hde;
 
@@ -2640,6 +2667,7 @@ struct perf_hpp_fmt *perf_hpp_fmt__dup(struct perf_hpp_fmt *fmt)
                new_hde = memdup(hde, sizeof(*hde));
                if (new_hde)
                        new_fmt = &new_hde->hpp;
+#endif
        } else {
                new_fmt = memdup(fmt, sizeof(*fmt));
        }
@@ -2719,6 +2747,7 @@ static struct evsel *find_evsel(struct evlist *evlist, char *event_name)
        return evsel;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 static int __dynamic_dimension__add(struct evsel *evsel,
                                    struct tep_format_field *field,
                                    bool raw_trace, int level)
@@ -2789,13 +2818,13 @@ static int add_all_matching_fields(struct evlist *evlist,
        }
        return ret;
 }
+#endif /* HAVE_LIBTRACEEVENT */
 
 static int add_dynamic_entry(struct evlist *evlist, const char *tok,
                             int level)
 {
        char *str, *event_name, *field_name, *opt_name;
        struct evsel *evsel;
-       struct tep_format_field *field;
        bool raw_trace = symbol_conf.raw_trace;
        int ret = 0;
 
@@ -2820,6 +2849,7 @@ static int add_dynamic_entry(struct evlist *evlist, const char *tok,
                raw_trace = true;
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        if (!strcmp(field_name, "trace_fields")) {
                ret = add_all_dynamic_fields(evlist, raw_trace, level);
                goto out;
@@ -2829,6 +2859,7 @@ static int add_dynamic_entry(struct evlist *evlist, const char *tok,
                ret = add_all_matching_fields(evlist, field_name, raw_trace, level);
                goto out;
        }
+#endif
 
        evsel = find_evsel(evlist, event_name);
        if (evsel == NULL) {
@@ -2843,10 +2874,12 @@ static int add_dynamic_entry(struct evlist *evlist, const char *tok,
                goto out;
        }
 
+#ifdef HAVE_LIBTRACEEVENT
        if (!strcmp(field_name, "*")) {
                ret = add_evsel_fields(evsel, raw_trace, level);
        } else {
-               field = tep_find_any_field(evsel->tp_format, field_name);
+               struct tep_format_field *field = tep_find_any_field(evsel->tp_format, field_name);
+
                if (field == NULL) {
                        pr_debug("Cannot find event field for %s.%s\n",
                                 event_name, field_name);
@@ -2855,6 +2888,10 @@ static int add_dynamic_entry(struct evlist *evlist, const char *tok,
 
                ret = __dynamic_dimension__add(evsel, field, raw_trace, level);
        }
+#else
+       (void)level;
+       (void)raw_trace;
+#endif /* HAVE_LIBTRACEEVENT */
 
 out:
        free(str);
@@ -2955,11 +2992,11 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
        for (i = 0; i < ARRAY_SIZE(common_sort_dimensions); i++) {
                struct sort_dimension *sd = &common_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                for (j = 0; j < ARRAY_SIZE(dynamic_headers); j++) {
-                       if (!strcmp(dynamic_headers[j], sd->name))
+                       if (sd->name && !strcmp(dynamic_headers[j], sd->name))
                                sort_dimension_add_dynamic_header(sd);
                }
 
@@ -3009,7 +3046,7 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
        for (i = 0; i < ARRAY_SIZE(bstack_sort_dimensions); i++) {
                struct sort_dimension *sd = &bstack_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                if (sort__mode != SORT_MODE__BRANCH)
@@ -3025,7 +3062,7 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
        for (i = 0; i < ARRAY_SIZE(memory_sort_dimensions); i++) {
                struct sort_dimension *sd = &memory_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                if (sort__mode != SORT_MODE__MEMORY)
@@ -3339,7 +3376,7 @@ int output_field_add(struct perf_hpp_list *list, char *tok)
        for (i = 0; i < ARRAY_SIZE(common_sort_dimensions); i++) {
                struct sort_dimension *sd = &common_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                return __sort_dimension__add_output(list, sd);
@@ -3357,7 +3394,7 @@ int output_field_add(struct perf_hpp_list *list, char *tok)
        for (i = 0; i < ARRAY_SIZE(bstack_sort_dimensions); i++) {
                struct sort_dimension *sd = &bstack_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                if (sort__mode != SORT_MODE__BRANCH)
@@ -3369,7 +3406,7 @@ int output_field_add(struct perf_hpp_list *list, char *tok)
        for (i = 0; i < ARRAY_SIZE(memory_sort_dimensions); i++) {
                struct sort_dimension *sd = &memory_sort_dimensions[i];
 
-               if (strncasecmp(tok, sd->name, strlen(tok)))
+               if (!sd->name || strncasecmp(tok, sd->name, strlen(tok)))
                        continue;
 
                if (sort__mode != SORT_MODE__MEMORY)
@@ -3508,6 +3545,9 @@ void reset_output_field(void)
 
 static void add_key(struct strbuf *sb, const char *str, int *llen)
 {
+       if (!str)
+               return;
+
        if (*llen >= 75) {
                strbuf_addstr(sb, "\n\t\t\t ");
                *llen = INDENT;
index 0645795..3ab6a92 100644 (file)
@@ -2157,6 +2157,7 @@ int perf_event__synthesize_attr(struct perf_tool *tool, struct perf_event_attr *
        return err;
 }
 
+#ifdef HAVE_LIBTRACEEVENT
 int perf_event__synthesize_tracing_data(struct perf_tool *tool, int fd, struct evlist *evlist,
                                        perf_event__handler_t process)
 {
@@ -2203,6 +2204,7 @@ int perf_event__synthesize_tracing_data(struct perf_tool *tool, int fd, struct e
 
        return aligned_size;
 }
+#endif
 
 int perf_event__synthesize_build_id(struct perf_tool *tool, struct dso *pos, u16 misc,
                                    perf_event__handler_t process, struct machine *machine)
@@ -2355,6 +2357,7 @@ int perf_event__synthesize_for_pipe(struct perf_tool *tool,
        }
        ret += err;
 
+#ifdef HAVE_LIBTRACEEVENT
        if (have_tracepoints(&evlist->core.entries)) {
                int fd = perf_data__fd(data);
 
@@ -2374,6 +2377,9 @@ int perf_event__synthesize_for_pipe(struct perf_tool *tool,
                }
                ret += err;
        }
+#else
+       (void)data;
+#endif
 
        return ret;
 }
index c9c83a4..2d3c257 100644 (file)
@@ -11,6 +11,8 @@
 #include "trace-event.h"
 
 #include <linux/ctype.h>
+#include <linux/kernel.h>
+#include <traceevent/event-parse.h>
 
 static int get_common_field(struct scripting_context *context,
                            int *offset, int *size, const char *type)
index 43146a4..1162c49 100644 (file)
@@ -11,6 +11,7 @@
 #include <sys/stat.h>
 #include <sys/wait.h>
 #include <sys/mman.h>
+#include <traceevent/event-parse.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <errno.h>
index 636a010..56175c5 100644 (file)
@@ -9,6 +9,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <errno.h>
+#include <traceevent/event-parse.h>
 
 #include "debug.h"
 #include "trace-event.h"
index b3ee651..8ad75b3 100644 (file)
@@ -1,5 +1,4 @@
 // SPDX-License-Identifier: GPL-2.0
-
 #include <stdio.h>
 #include <unistd.h>
 #include <stdlib.h>
index 8f39f5b..add6c5d 100644 (file)
@@ -2,9 +2,11 @@
 #ifndef _PERF_UTIL_TRACE_EVENT_H
 #define _PERF_UTIL_TRACE_EVENT_H
 
-#include <traceevent/event-parse.h>
-#include "parse-events.h"
+#include <stdbool.h>
+#include <stdio.h>
+#include <linux/types.h>
 
+struct evlist;
 struct machine;
 struct perf_sample;
 union perf_event;
@@ -18,6 +20,11 @@ struct trace_event {
        struct tep_plugin_list  *plugin_list;
 };
 
+typedef char *(tep_func_resolver_t)(void *priv,
+                                   unsigned long long *addrp, char **modp);
+
+bool have_tracepoints(struct list_head *evlist);
+
 int trace_event__init(struct trace_event *t);
 void trace_event__cleanup(struct trace_event *t);
 int trace_event__register_resolver(struct machine *machine,