diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index 690364577611..9f5a47e5c07e 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -1,3 +1,7 @@
+ifeq ("$(origin O)", "command line")
+	OUTPUT := $(O)/
+endif
+
 # The default target of this Makefile is...
 all::
 
@@ -153,9 +157,13 @@ all::
 #
 # Define NO_DWARF if you do not want debug-info analysis feature at all.
 
-PERF-VERSION-FILE: .FORCE-PERF-VERSION-FILE
-	@$(SHELL_PATH) util/PERF-VERSION-GEN
--include PERF-VERSION-FILE
+$(shell sh -c 'mkdir -p $(OUTPUT)scripts/python/Perf-Trace-Util/' 2> /dev/null)
+$(shell sh -c 'mkdir -p $(OUTPUT)util/scripting-engines/' 2> /dev/null)
+$(shell sh -c 'mkdir $(OUTPUT)bench' 2> /dev/null)
+
+$(OUTPUT)PERF-VERSION-FILE: .FORCE-PERF-VERSION-FILE
+	@$(SHELL_PATH) util/PERF-VERSION-GEN $(OUTPUT)
+-include $(OUTPUT)PERF-VERSION-FILE
 
 uname_S := $(shell sh -c 'uname -s 2>/dev/null || echo not')
 uname_M := $(shell sh -c 'uname -m 2>/dev/null || echo not')
@@ -310,7 +318,7 @@ PROGRAMS += $(EXTRA_PROGRAMS)
 #
 # Single 'perf' binary right now:
 #
-PROGRAMS += perf
+PROGRAMS += $(OUTPUT)perf
 
 # List built-in command $C whose implementation cmd_$C() is not in
 # builtin-$C.o but is linked in as part of some other command.
@@ -320,7 +328,7 @@ PROGRAMS += perf
 ALL_PROGRAMS = $(PROGRAMS) $(SCRIPTS)
 
 # what 'all' will build but not install in perfexecdir
-OTHER_PROGRAMS = perf$X
+OTHER_PROGRAMS = $(OUTPUT)perf$X
 
 # Set paths to tools early so that they can be used for version tests.
 ifndef SHELL_PATH
@@ -332,7 +340,7 @@ endif
 
 export PERL_PATH
 
-LIB_FILE=libperf.a
+LIB_FILE=$(OUTPUT)libperf.a
 
 LIB_H += ../../include/linux/perf_event.h
 LIB_H += ../../include/linux/rbtree.h
@@ -377,7 +385,6 @@ LIB_H += util/header.h
 LIB_H += util/help.h
 LIB_H += util/session.h
 LIB_H += util/strbuf.h
-LIB_H += util/string.h
 LIB_H += util/strlist.h
 LIB_H += util/svghelper.h
 LIB_H += util/run-command.h
@@ -393,77 +400,77 @@ LIB_H += util/probe-finder.h
 LIB_H += util/probe-event.h
 LIB_H += util/cpumap.h
 
-LIB_OBJS += util/abspath.o
-LIB_OBJS += util/alias.o
-LIB_OBJS += util/build-id.o
-LIB_OBJS += util/config.o
-LIB_OBJS += util/ctype.o
-LIB_OBJS += util/debugfs.o
-LIB_OBJS += util/environment.o
-LIB_OBJS += util/event.o
-LIB_OBJS += util/exec_cmd.o
-LIB_OBJS += util/help.o
-LIB_OBJS += util/levenshtein.o
-LIB_OBJS += util/parse-options.o
-LIB_OBJS += util/parse-events.o
-LIB_OBJS += util/path.o
-LIB_OBJS += util/rbtree.o
-LIB_OBJS += util/bitmap.o
-LIB_OBJS += util/hweight.o
-LIB_OBJS += util/find_next_bit.o
-LIB_OBJS += util/run-command.o
-LIB_OBJS += util/quote.o
-LIB_OBJS += util/strbuf.o
-LIB_OBJS += util/string.o
-LIB_OBJS += util/strlist.o
-LIB_OBJS += util/usage.o
-LIB_OBJS += util/wrapper.o
-LIB_OBJS += util/sigchain.o
-LIB_OBJS += util/symbol.o
-LIB_OBJS += util/color.o
-LIB_OBJS += util/pager.o
-LIB_OBJS += util/header.o
-LIB_OBJS += util/callchain.o
-LIB_OBJS += util/values.o
-LIB_OBJS += util/debug.o
-LIB_OBJS += util/map.o
-LIB_OBJS += util/session.o
-LIB_OBJS += util/thread.o
-LIB_OBJS += util/trace-event-parse.o
-LIB_OBJS += util/trace-event-read.o
-LIB_OBJS += util/trace-event-info.o
-LIB_OBJS += util/trace-event-scripting.o
-LIB_OBJS += util/svghelper.o
-LIB_OBJS += util/sort.o
-LIB_OBJS += util/hist.o
-LIB_OBJS += util/probe-event.o
-LIB_OBJS += util/util.o
-LIB_OBJS += util/cpumap.o
+LIB_OBJS += $(OUTPUT)util/abspath.o
+LIB_OBJS += $(OUTPUT)util/alias.o
+LIB_OBJS += $(OUTPUT)util/build-id.o
+LIB_OBJS += $(OUTPUT)util/config.o
+LIB_OBJS += $(OUTPUT)util/ctype.o
+LIB_OBJS += $(OUTPUT)util/debugfs.o
+LIB_OBJS += $(OUTPUT)util/environment.o
+LIB_OBJS += $(OUTPUT)util/event.o
+LIB_OBJS += $(OUTPUT)util/exec_cmd.o
+LIB_OBJS += $(OUTPUT)util/help.o
+LIB_OBJS += $(OUTPUT)util/levenshtein.o
+LIB_OBJS += $(OUTPUT)util/parse-options.o
+LIB_OBJS += $(OUTPUT)util/parse-events.o
+LIB_OBJS += $(OUTPUT)util/path.o
+LIB_OBJS += $(OUTPUT)util/rbtree.o
+LIB_OBJS += $(OUTPUT)util/bitmap.o
+LIB_OBJS += $(OUTPUT)util/hweight.o
+LIB_OBJS += $(OUTPUT)util/find_next_bit.o
+LIB_OBJS += $(OUTPUT)util/run-command.o
+LIB_OBJS += $(OUTPUT)util/quote.o
+LIB_OBJS += $(OUTPUT)util/strbuf.o
+LIB_OBJS += $(OUTPUT)util/string.o
+LIB_OBJS += $(OUTPUT)util/strlist.o
+LIB_OBJS += $(OUTPUT)util/usage.o
+LIB_OBJS += $(OUTPUT)util/wrapper.o
+LIB_OBJS += $(OUTPUT)util/sigchain.o
+LIB_OBJS += $(OUTPUT)util/symbol.o
+LIB_OBJS += $(OUTPUT)util/color.o
+LIB_OBJS += $(OUTPUT)util/pager.o
+LIB_OBJS += $(OUTPUT)util/header.o
+LIB_OBJS += $(OUTPUT)util/callchain.o
+LIB_OBJS += $(OUTPUT)util/values.o
+LIB_OBJS += $(OUTPUT)util/debug.o
+LIB_OBJS += $(OUTPUT)util/map.o
+LIB_OBJS += $(OUTPUT)util/session.o
+LIB_OBJS += $(OUTPUT)util/thread.o
+LIB_OBJS += $(OUTPUT)util/trace-event-parse.o
+LIB_OBJS += $(OUTPUT)util/trace-event-read.o
+LIB_OBJS += $(OUTPUT)util/trace-event-info.o
+LIB_OBJS += $(OUTPUT)util/trace-event-scripting.o
+LIB_OBJS += $(OUTPUT)util/svghelper.o
+LIB_OBJS += $(OUTPUT)util/sort.o
+LIB_OBJS += $(OUTPUT)util/hist.o
+LIB_OBJS += $(OUTPUT)util/probe-event.o
+LIB_OBJS += $(OUTPUT)util/util.o
+LIB_OBJS += $(OUTPUT)util/cpumap.o
 
-BUILTIN_OBJS += builtin-annotate.o
+BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
 
-BUILTIN_OBJS += builtin-bench.o
+BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
 
 # Benchmark modules
-BUILTIN_OBJS += bench/sched-messaging.o
-BUILTIN_OBJS += bench/sched-pipe.o
-BUILTIN_OBJS += bench/mem-memcpy.o
+BUILTIN_OBJS += $(OUTPUT)bench/sched-messaging.o
+BUILTIN_OBJS += $(OUTPUT)bench/sched-pipe.o
+BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy.o
 
-BUILTIN_OBJS += builtin-diff.o
-BUILTIN_OBJS += builtin-help.o
-BUILTIN_OBJS += builtin-sched.o
-BUILTIN_OBJS += builtin-buildid-list.o
-BUILTIN_OBJS += builtin-buildid-cache.o
-BUILTIN_OBJS += builtin-list.o
-BUILTIN_OBJS += builtin-record.o
-BUILTIN_OBJS += builtin-report.o
-BUILTIN_OBJS += builtin-stat.o
-BUILTIN_OBJS += builtin-timechart.o
-BUILTIN_OBJS += builtin-top.o
-BUILTIN_OBJS += builtin-trace.o
-BUILTIN_OBJS += builtin-probe.o
-BUILTIN_OBJS += builtin-kmem.o
-BUILTIN_OBJS += builtin-lock.o
+BUILTIN_OBJS += $(OUTPUT)builtin-diff.o
+BUILTIN_OBJS += $(OUTPUT)builtin-help.o
+BUILTIN_OBJS += $(OUTPUT)builtin-sched.o
+BUILTIN_OBJS += $(OUTPUT)builtin-buildid-list.o
+BUILTIN_OBJS += $(OUTPUT)builtin-buildid-cache.o
+BUILTIN_OBJS += $(OUTPUT)builtin-list.o
+BUILTIN_OBJS += $(OUTPUT)builtin-record.o
+BUILTIN_OBJS += $(OUTPUT)builtin-report.o
+BUILTIN_OBJS += $(OUTPUT)builtin-stat.o
+BUILTIN_OBJS += $(OUTPUT)builtin-timechart.o
+BUILTIN_OBJS += $(OUTPUT)builtin-top.o
+BUILTIN_OBJS += $(OUTPUT)builtin-trace.o
+BUILTIN_OBJS += $(OUTPUT)builtin-probe.o
+BUILTIN_OBJS += $(OUTPUT)builtin-kmem.o
+BUILTIN_OBJS += $(OUTPUT)builtin-lock.o
 
 PERFLIBS = $(LIB_FILE)
 
@@ -494,6 +501,10 @@ ifeq ($(uname_S),Darwin)
 	PTHREAD_LIBS =
 endif
 
+ifneq ($(OUTPUT),)
+	BASIC_CFLAGS += -I$(OUTPUT)
+endif
+
 ifeq ($(shell sh -c "(echo '\#include <libelf.h>'; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o $(BITBUCKET) $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y)
 ifneq ($(shell sh -c "(echo '\#include <gnu/libc-version.h>'; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o $(BITBUCKET) $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y)
 	msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]/glibc-static);
@@ -512,7 +523,7 @@ else
 ifndef NO_DWARF
 	BASIC_CFLAGS += -I/usr/include/elfutils -DDWARF_SUPPORT
 	EXTLIBS += -lelf -ldw
-	LIB_OBJS += util/probe-finder.o
+	LIB_OBJS += $(OUTPUT)util/probe-finder.o
 endif
 endif
 
@@ -521,7 +532,7 @@ ifneq ($(shell sh -c "(echo '\#include <newt.h>'; echo 'int main(void) { newtIni
 	BASIC_CFLAGS += -DNO_NEWT_SUPPORT
 else
 	EXTLIBS += -lnewt
-	LIB_OBJS += util/newt.o
+	LIB_OBJS += $(OUTPUT)util/newt.o
 endif
 
 ifndef NO_LIBPERL
@@ -533,8 +544,8 @@ ifneq ($(shell sh -c "(echo '\#include <EXTERN.h>'; echo '\#include <perl.h>'; e
 	BASIC_CFLAGS += -DNO_LIBPERL
 else
 	ALL_LDFLAGS += $(PERL_EMBED_LDOPTS)
-	LIB_OBJS += util/scripting-engines/trace-event-perl.o
-	LIB_OBJS += scripts/perl/Perf-Trace-Util/Context.o
+	LIB_OBJS += $(OUTPUT)util/scripting-engines/trace-event-perl.o
+	LIB_OBJS += $(OUTPUT)scripts/perl/Perf-Trace-Util/Context.o
 endif
 
 ifndef NO_LIBPYTHON
@@ -542,12 +553,12 @@ PYTHON_EMBED_LDOPTS = `python-config --ldflags 2>/dev/null`
 PYTHON_EMBED_CCOPTS = `python-config --cflags 2>/dev/null`
 endif
 
-ifneq ($(shell sh -c "(echo '\#include <Python.h>'; echo 'int main(void) { Py_Initialize(); return 0; }') | $(CC) -x c - $(PYTHON_EMBED_CCOPTS) -o /dev/null $(PYTHON_EMBED_LDOPTS) > /dev/null 2>&1 && echo y"), y)
+ifneq ($(shell sh -c "(echo '\#include <Python.h>'; echo 'int main(void) { Py_Initialize(); return 0; }') | $(CC) -x c - $(PYTHON_EMBED_CCOPTS) -o $(BITBUCKET) $(PYTHON_EMBED_LDOPTS) > /dev/null 2>&1 && echo y"), y)
 	BASIC_CFLAGS += -DNO_LIBPYTHON
 else
 	ALL_LDFLAGS += $(PYTHON_EMBED_LDOPTS)
-	LIB_OBJS += util/scripting-engines/trace-event-python.o
-	LIB_OBJS += scripts/python/Perf-Trace-Util/Context.o
+	LIB_OBJS += $(OUTPUT)util/scripting-engines/trace-event-python.o
+	LIB_OBJS += $(OUTPUT)scripts/python/Perf-Trace-Util/Context.o
 endif
 
 ifdef NO_DEMANGLE
@@ -618,53 +629,53 @@ ifdef NO_C99_FORMAT
 endif
 ifdef SNPRINTF_RETURNS_BOGUS
 	COMPAT_CFLAGS += -DSNPRINTF_RETURNS_BOGUS
-	COMPAT_OBJS += compat/snprintf.o
+	COMPAT_OBJS += $(OUTPUT)compat/snprintf.o
 endif
 ifdef FREAD_READS_DIRECTORIES
 	COMPAT_CFLAGS += -DFREAD_READS_DIRECTORIES
-	COMPAT_OBJS += compat/fopen.o
+	COMPAT_OBJS += $(OUTPUT)compat/fopen.o
 endif
 ifdef NO_SYMLINK_HEAD
 	BASIC_CFLAGS += -DNO_SYMLINK_HEAD
 endif
 ifdef NO_STRCASESTR
 	COMPAT_CFLAGS += -DNO_STRCASESTR
-	COMPAT_OBJS += compat/strcasestr.o
+	COMPAT_OBJS += $(OUTPUT)compat/strcasestr.o
 endif
 ifdef NO_STRTOUMAX
 	COMPAT_CFLAGS += -DNO_STRTOUMAX
-	COMPAT_OBJS += compat/strtoumax.o
+	COMPAT_OBJS += $(OUTPUT)compat/strtoumax.o
 endif
 ifdef NO_STRTOULL
 	COMPAT_CFLAGS += -DNO_STRTOULL
 endif
 ifdef NO_SETENV
 	COMPAT_CFLAGS += -DNO_SETENV
-	COMPAT_OBJS += compat/setenv.o
+	COMPAT_OBJS += $(OUTPUT)compat/setenv.o
 endif
 ifdef NO_MKDTEMP
 	COMPAT_CFLAGS += -DNO_MKDTEMP
-	COMPAT_OBJS += compat/mkdtemp.o
+	COMPAT_OBJS += $(OUTPUT)compat/mkdtemp.o
 endif
 ifdef NO_UNSETENV
 	COMPAT_CFLAGS += -DNO_UNSETENV
-	COMPAT_OBJS += compat/unsetenv.o
+	COMPAT_OBJS += $(OUTPUT)compat/unsetenv.o
 endif
 ifdef NO_SYS_SELECT_H
 	BASIC_CFLAGS += -DNO_SYS_SELECT_H
 endif
 ifdef NO_MMAP
 	COMPAT_CFLAGS += -DNO_MMAP
-	COMPAT_OBJS += compat/mmap.o
+	COMPAT_OBJS += $(OUTPUT)compat/mmap.o
 else
 	ifdef USE_WIN32_MMAP
 		COMPAT_CFLAGS += -DUSE_WIN32_MMAP
-		COMPAT_OBJS += compat/win32mmap.o
+		COMPAT_OBJS += $(OUTPUT)compat/win32mmap.o
 	endif
 endif
 ifdef NO_PREAD
 	COMPAT_CFLAGS += -DNO_PREAD
-	COMPAT_OBJS += compat/pread.o
+	COMPAT_OBJS += $(OUTPUT)compat/pread.o
 endif
 ifdef NO_FAST_WORKING_DIRECTORY
 	BASIC_CFLAGS += -DNO_FAST_WORKING_DIRECTORY
@@ -686,10 +697,10 @@ else
 endif
 endif
 ifdef NO_INET_NTOP
-	LIB_OBJS += compat/inet_ntop.o
+	LIB_OBJS += $(OUTPUT)compat/inet_ntop.o
 endif
 ifdef NO_INET_PTON
-	LIB_OBJS += compat/inet_pton.o
+	LIB_OBJS += $(OUTPUT)compat/inet_pton.o
 endif
 
 ifdef NO_ICONV
@@ -706,15 +717,15 @@ endif
 
 ifdef PPC_SHA1
 	SHA1_HEADER = "ppc/sha1.h"
-	LIB_OBJS += ppc/sha1.o ppc/sha1ppc.o
+	LIB_OBJS += $(OUTPUT)ppc/sha1.o ppc/sha1ppc.o
 else
 ifdef ARM_SHA1
 	SHA1_HEADER = "arm/sha1.h"
-	LIB_OBJS += arm/sha1.o arm/sha1_arm.o
+	LIB_OBJS += $(OUTPUT)arm/sha1.o $(OUTPUT)arm/sha1_arm.o
 else
 ifdef MOZILLA_SHA1
 	SHA1_HEADER = "mozilla-sha1/sha1.h"
-	LIB_OBJS += mozilla-sha1/sha1.o
+	LIB_OBJS += $(OUTPUT)mozilla-sha1/sha1.o
 else
 	SHA1_HEADER = <openssl/sha.h>
 	EXTLIBS += $(LIB_4_CRYPTO)
@@ -726,15 +737,15 @@ ifdef NO_PERL_MAKEMAKER
 endif
 ifdef NO_HSTRERROR
 	COMPAT_CFLAGS += -DNO_HSTRERROR
-	COMPAT_OBJS += compat/hstrerror.o
+	COMPAT_OBJS += $(OUTPUT)compat/hstrerror.o
 endif
 ifdef NO_MEMMEM
 	COMPAT_CFLAGS += -DNO_MEMMEM
-	COMPAT_OBJS += compat/memmem.o
+	COMPAT_OBJS += $(OUTPUT)compat/memmem.o
 endif
 ifdef INTERNAL_QSORT
 	COMPAT_CFLAGS += -DINTERNAL_QSORT
-	COMPAT_OBJS += compat/qsort.o
+	COMPAT_OBJS += $(OUTPUT)compat/qsort.o
 endif
 ifdef RUNTIME_PREFIX
 	COMPAT_CFLAGS += -DRUNTIME_PREFIX
@@ -814,7 +825,7 @@ export TAR INSTALL DESTDIR SHELL_PATH
 
 SHELL = $(SHELL_PATH)
 
-all:: .perf.dev.null shell_compatibility_test $(ALL_PROGRAMS) $(BUILT_INS) $(OTHER_PROGRAMS) PERF-BUILD-OPTIONS
+all:: .perf.dev.null shell_compatibility_test $(ALL_PROGRAMS) $(BUILT_INS) $(OTHER_PROGRAMS) $(OUTPUT)PERF-BUILD-OPTIONS
 ifneq (,$X)
 	$(foreach p,$(patsubst %$X,%,$(filter %$X,$(ALL_PROGRAMS) $(BUILT_INS) perf$X)), test '$p' -ef '$p$X' || $(RM) '$p';)
 endif
@@ -826,39 +837,39 @@ please_set_SHELL_PATH_to_a_more_modern_shell:
 
 shell_compatibility_test: please_set_SHELL_PATH_to_a_more_modern_shell
 
-strip: $(PROGRAMS) perf$X
-	$(STRIP) $(STRIP_OPTS) $(PROGRAMS) perf$X
+strip: $(PROGRAMS) $(OUTPUT)perf$X
+	$(STRIP) $(STRIP_OPTS) $(PROGRAMS) $(OUTPUT)perf$X
 
-perf.o: perf.c common-cmds.h PERF-CFLAGS
+$(OUTPUT)perf.o: perf.c $(OUTPUT)common-cmds.h $(OUTPUT)PERF-CFLAGS
 	$(QUIET_CC)$(CC) -DPERF_VERSION='"$(PERF_VERSION)"' \
 		'-DPERF_HTML_PATH="$(htmldir_SQ)"' \
-		$(ALL_CFLAGS) -c $(filter %.c,$^)
+		$(ALL_CFLAGS) -c $(filter %.c,$^) -o $@
 
-perf$X: perf.o $(BUILTIN_OBJS) $(PERFLIBS)
-	$(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ perf.o \
+$(OUTPUT)perf$X: $(OUTPUT)perf.o $(BUILTIN_OBJS) $(PERFLIBS)
+	$(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(OUTPUT)perf.o \
 		$(BUILTIN_OBJS) $(ALL_LDFLAGS) $(LIBS)
 
-builtin-help.o: builtin-help.c common-cmds.h PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) \
+$(OUTPUT)builtin-help.o: builtin-help.c $(OUTPUT)common-cmds.h $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) \
 		'-DPERF_HTML_PATH="$(htmldir_SQ)"' \
 		'-DPERF_MAN_PATH="$(mandir_SQ)"' \
 		'-DPERF_INFO_PATH="$(infodir_SQ)"' $<
 
-builtin-timechart.o: builtin-timechart.c common-cmds.h PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) \
+$(OUTPUT)builtin-timechart.o: builtin-timechart.c $(OUTPUT)common-cmds.h $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) \
 		'-DPERF_HTML_PATH="$(htmldir_SQ)"' \
 		'-DPERF_MAN_PATH="$(mandir_SQ)"' \
 		'-DPERF_INFO_PATH="$(infodir_SQ)"' $<
 
-$(BUILT_INS): perf$X
+$(BUILT_INS): $(OUTPUT)perf$X
 	$(QUIET_BUILT_IN)$(RM) $@ && \
 	ln perf$X $@ 2>/dev/null || \
 	ln -s perf$X $@ 2>/dev/null || \
 	cp perf$X $@
 
-common-cmds.h: util/generate-cmdlist.sh command-list.txt
+$(OUTPUT)common-cmds.h: util/generate-cmdlist.sh command-list.txt
 
-common-cmds.h: $(wildcard Documentation/perf-*.txt)
+$(OUTPUT)common-cmds.h: $(wildcard Documentation/perf-*.txt)
 	$(QUIET_GEN). util/generate-cmdlist.sh > $@+ && mv $@+ $@
 
 $(patsubst %.sh,%,$(SCRIPT_SH)) : % : %.sh
@@ -870,7 +881,7 @@ $(patsubst %.sh,%,$(SCRIPT_SH)) : % : %.sh
 	    -e 's/@@NO_CURL@@/$(NO_CURL)/g' \
 	    $@.sh >$@+ && \
 	chmod +x $@+ && \
-	mv $@+ $@
+	mv $@+ $(OUTPUT)$@
 
 configure: configure.ac
 	$(QUIET_GEN)$(RM) $@ $<+ && \
@@ -880,60 +891,60 @@ configure: configure.ac
 	$(RM) $<+
 
 # These can record PERF_VERSION
-perf.o perf.spec \
+$(OUTPUT)perf.o perf.spec \
 	$(patsubst %.sh,%,$(SCRIPT_SH)) \
 	$(patsubst %.perl,%,$(SCRIPT_PERL)) \
-	: PERF-VERSION-FILE
+	: $(OUTPUT)PERF-VERSION-FILE
 
-%.o: %.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) $<
-%.s: %.c PERF-CFLAGS
+$(OUTPUT)%.o: %.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $<
+$(OUTPUT)%.s: %.c $(OUTPUT)PERF-CFLAGS
 	$(QUIET_CC)$(CC) -S $(ALL_CFLAGS) $<
-%.o: %.S
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) $<
+$(OUTPUT)%.o: %.S
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $<
 
-util/exec_cmd.o: util/exec_cmd.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) \
+$(OUTPUT)util/exec_cmd.o: util/exec_cmd.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) \
 		'-DPERF_EXEC_PATH="$(perfexecdir_SQ)"' \
 		'-DBINDIR="$(bindir_relative_SQ)"' \
 		'-DPREFIX="$(prefix_SQ)"' \
 		$<
 
-builtin-init-db.o: builtin-init-db.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) -DDEFAULT_PERF_TEMPLATE_DIR='"$(template_dir_SQ)"' $<
+$(OUTPUT)builtin-init-db.o: builtin-init-db.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DDEFAULT_PERF_TEMPLATE_DIR='"$(template_dir_SQ)"' $<
 
-util/config.o: util/config.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
+$(OUTPUT)util/config.o: util/config.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
 
-util/rbtree.o: ../../lib/rbtree.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/rbtree.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
+$(OUTPUT)util/rbtree.o: ../../lib/rbtree.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
 
 # some perf warning policies can't fit to lib/bitmap.c, eg: it warns about variable shadowing
 # from <string.h> that comes from kernel headers wrapping.
 KBITMAP_FLAGS=`echo $(ALL_CFLAGS) | sed s/-Wshadow// | sed s/-Wswitch-default// | sed s/-Wextra//`
 
-util/bitmap.o: ../../lib/bitmap.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/bitmap.o -c $(KBITMAP_FLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
+$(OUTPUT)util/bitmap.o: ../../lib/bitmap.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(KBITMAP_FLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
 
-util/hweight.o: ../../lib/hweight.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/hweight.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
+$(OUTPUT)util/hweight.o: ../../lib/hweight.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
 
-util/find_next_bit.o: ../../lib/find_next_bit.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/find_next_bit.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
+$(OUTPUT)util/find_next_bit.o: ../../lib/find_next_bit.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
 
-util/scripting-engines/trace-event-perl.o: util/scripting-engines/trace-event-perl.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/scripting-engines/trace-event-perl.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $<
+$(OUTPUT)util/scripting-engines/trace-event-perl.o: util/scripting-engines/trace-event-perl.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $<
 
-scripts/perl/Perf-Trace-Util/Context.o: scripts/perl/Perf-Trace-Util/Context.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o scripts/perl/Perf-Trace-Util/Context.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $<
+$(OUTPUT)scripts/perl/Perf-Trace-Util/Context.o: scripts/perl/Perf-Trace-Util/Context.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $<
 
-util/scripting-engines/trace-event-python.o: util/scripting-engines/trace-event-python.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o util/scripting-engines/trace-event-python.o -c $(ALL_CFLAGS) $(PYTHON_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $<
+$(OUTPUT)util/scripting-engines/trace-event-python.o: util/scripting-engines/trace-event-python.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PYTHON_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $<
 
-scripts/python/Perf-Trace-Util/Context.o: scripts/python/Perf-Trace-Util/Context.c PERF-CFLAGS
-	$(QUIET_CC)$(CC) -o scripts/python/Perf-Trace-Util/Context.o -c $(ALL_CFLAGS) $(PYTHON_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $<
+$(OUTPUT)scripts/python/Perf-Trace-Util/Context.o: scripts/python/Perf-Trace-Util/Context.c $(OUTPUT)PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PYTHON_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $<
 
-perf-%$X: %.o $(PERFLIBS)
+$(OUTPUT)perf-%$X: %.o $(PERFLIBS)
 	$(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS)
 
 $(LIB_OBJS) $(BUILTIN_OBJS): $(LIB_H)
@@ -974,17 +985,17 @@ cscope:
 TRACK_CFLAGS = $(subst ','\'',$(ALL_CFLAGS)):\
              $(bindir_SQ):$(perfexecdir_SQ):$(template_dir_SQ):$(prefix_SQ)
 
-PERF-CFLAGS: .FORCE-PERF-CFLAGS
+$(OUTPUT)PERF-CFLAGS: .FORCE-PERF-CFLAGS
 	@FLAGS='$(TRACK_CFLAGS)'; \
-	    if test x"$$FLAGS" != x"`cat PERF-CFLAGS 2>/dev/null`" ; then \
+	    if test x"$$FLAGS" != x"`cat $(OUTPUT)PERF-CFLAGS 2>/dev/null`" ; then \
 		echo 1>&2 "    * new build flags or prefix"; \
-		echo "$$FLAGS" >PERF-CFLAGS; \
+		echo "$$FLAGS" >$(OUTPUT)PERF-CFLAGS; \
             fi
 
 # We need to apply sq twice, once to protect from the shell
-# that runs PERF-BUILD-OPTIONS, and then again to protect it
+# that runs $(OUTPUT)PERF-BUILD-OPTIONS, and then again to protect it
 # and the first level quoting from the shell that runs "echo".
-PERF-BUILD-OPTIONS: .FORCE-PERF-BUILD-OPTIONS
+$(OUTPUT)PERF-BUILD-OPTIONS: .FORCE-PERF-BUILD-OPTIONS
 	@echo SHELL_PATH=\''$(subst ','\'',$(SHELL_PATH_SQ))'\' >$@
 	@echo TAR=\''$(subst ','\'',$(subst ','\'',$(TAR)))'\' >>$@
 	@echo NO_CURL=\''$(subst ','\'',$(subst ','\'',$(NO_CURL)))'\' >>$@
@@ -1005,7 +1016,7 @@ all:: $(TEST_PROGRAMS)
 
 export NO_SVN_TESTS
 
-check: common-cmds.h
+check: $(OUTPUT)common-cmds.h
 	if sparse; \
 	then \
 		for i in *.c */*.c; \
@@ -1039,10 +1050,10 @@ export perfexec_instdir
 
 install: all
 	$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(bindir_SQ)'
-	$(INSTALL) perf$X '$(DESTDIR_SQ)$(bindir_SQ)'
+	$(INSTALL) $(OUTPUT)perf$X '$(DESTDIR_SQ)$(bindir_SQ)'
 	$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace'
 	$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/bin'
-	$(INSTALL) perf-archive -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
+	$(INSTALL) $(OUTPUT)perf-archive -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
 	$(INSTALL) scripts/perl/Perf-Trace-Util/lib/Perf/Trace/* -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace'
 	$(INSTALL) scripts/perl/*.pl -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl'
 	$(INSTALL) scripts/perl/bin/* -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/bin'
@@ -1056,7 +1067,7 @@ ifdef BUILT_INS
 	$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
 	$(INSTALL) $(BUILT_INS) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
 ifneq (,$X)
-	$(foreach p,$(patsubst %$X,%,$(filter %$X,$(ALL_PROGRAMS) $(BUILT_INS) perf$X)), $(RM) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$p';)
+	$(foreach p,$(patsubst %$X,%,$(filter %$X,$(ALL_PROGRAMS) $(BUILT_INS) $(OUTPUT)perf$X)), $(RM) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$p';)
 endif
 endif
 
@@ -1140,14 +1151,14 @@ clean:
 	$(RM) *.o */*.o */*/*.o */*/*/*.o $(LIB_FILE)
 	$(RM) $(ALL_PROGRAMS) $(BUILT_INS) perf$X
 	$(RM) $(TEST_PROGRAMS)
-	$(RM) *.spec *.pyc *.pyo */*.pyc */*.pyo common-cmds.h TAGS tags cscope*
+	$(RM) *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope*
 	$(RM) -r autom4te.cache
 	$(RM) config.log config.mak.autogen config.mak.append config.status config.cache
 	$(RM) -r $(PERF_TARNAME) .doc-tmp-dir
 	$(RM) $(PERF_TARNAME).tar.gz perf-core_$(PERF_VERSION)-*.tar.gz
 	$(RM) $(htmldocs).tar.gz $(manpages).tar.gz
 	$(MAKE) -C Documentation/ clean
-	$(RM) PERF-VERSION-FILE PERF-CFLAGS PERF-BUILD-OPTIONS
+	$(RM) $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)PERF-CFLAGS $(OUTPUT)PERF-BUILD-OPTIONS
 
 .PHONY: all install clean strip
 .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c
index 89773178e894..52e646e3e873 100644
--- a/tools/perf/bench/mem-memcpy.c
+++ b/tools/perf/bench/mem-memcpy.c
@@ -10,7 +10,6 @@
 #include "../perf.h"
 #include "../util/util.h"
 #include "../util/parse-options.h"
-#include "../util/string.h"
 #include "../util/header.h"
 #include "bench.h"
 
diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c
index 887e8e04a6f9..ee0d91726991 100644
--- a/tools/perf/builtin-annotate.c
+++ b/tools/perf/builtin-annotate.c
@@ -14,7 +14,6 @@
 #include "util/cache.h"
 #include <linux/rbtree.h>
 #include "util/symbol.h"
-#include "util/string.h"
 
 #include "perf.h"
 #include "util/debug.h"
diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c
index 924a9518931a..513aa8a55db6 100644
--- a/tools/perf/builtin-kmem.c
+++ b/tools/perf/builtin-kmem.c
@@ -363,19 +363,21 @@ static void __print_result(struct rb_root *root, struct perf_session *session,
 		struct alloc_stat *data = rb_entry(next, struct alloc_stat,
 						   node);
 		struct symbol *sym = NULL;
+		struct map *map;
 		char buf[BUFSIZ];
 		u64 addr;
 
 		if (is_caller) {
 			addr = data->call_site;
 			if (!raw_ip)
-				sym = map_groups__find_function(&session->kmaps, addr, NULL);
+				sym = map_groups__find_function(&session->kmaps,
+								addr, &map, NULL);
 		} else
 			addr = data->ptr;
 
 		if (sym != NULL)
 			snprintf(buf, sizeof(buf), "%s+%Lx", sym->name,
-				 addr - sym->start);
+				 addr - map->unmap_ip(map, sym->start));
 		else
 			snprintf(buf, sizeof(buf), "%#Lx", addr);
 		printf(" %-34s |", buf);
@@ -488,6 +490,9 @@ static int __cmd_kmem(void)
 	if (session == NULL)
 		return -ENOMEM;
 
+	if (perf_session__create_kernel_maps(session) < 0)
+		goto out_delete;
+
 	if (!perf_session__has_traces(session, "kmem record"))
 		goto out_delete;
 
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 60ecdd3dd26d..dc61f1b68b40 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -15,7 +15,6 @@
 #include "util/util.h"
 #include "util/parse-options.h"
 #include "util/parse-events.h"
-#include "util/string.h"
 
 #include "util/header.h"
 #include "util/event.h"
@@ -575,6 +574,9 @@ static int __cmd_record(int argc, const char **argv)
 
 	err = event__synthesize_kernel_mmap(process_synthesized_event,
 					    session, "_text");
+	if (err < 0)
+		err = event__synthesize_kernel_mmap(process_synthesized_event,
+						    session, "_stext");
 	if (err < 0) {
 		pr_err("Couldn't record kernel reference relocation symbol.\n");
 		return err;
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 6ab16980dd66..e93c69a8e720 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -14,7 +14,6 @@
 #include "util/cache.h"
 #include <linux/rbtree.h>
 #include "util/symbol.h"
-#include "util/string.h"
 #include "util/callchain.h"
 #include "util/strlist.h"
 #include "util/values.h"
@@ -89,9 +88,12 @@ static int perf_session__add_hist_entry(struct perf_session *self,
 	struct event_stat_id *stats;
 	struct perf_event_attr *attr;
 
-	if ((sort__has_parent || symbol_conf.use_callchain) && data->callchain)
+	if ((sort__has_parent || symbol_conf.use_callchain) && data->callchain) {
 		syms = perf_session__resolve_callchain(self, al->thread,
 						       data->callchain, &parent);
+		if (syms == NULL)
+			return -ENOMEM;
+	}
 
 	attr = perf_header__find_attr(data->id, &self->header);
 	if (attr)
@@ -110,8 +112,8 @@ static int perf_session__add_hist_entry(struct perf_session *self,
 
 	if (symbol_conf.use_callchain) {
 		if (!hit)
-			callchain_init(&he->callchain);
-		err = append_chain(&he->callchain, data->callchain, syms);
+			callchain_init(he->callchain);
+		err = append_chain(he->callchain, data->callchain, syms);
 		free(syms);
 
 		if (err)
@@ -303,14 +305,16 @@ static int __cmd_report(void)
 	next = rb_first(&session->stats_by_id);
 	while (next) {
 		struct event_stat_id *stats;
+		u64 nr_hists;
 
 		stats = rb_entry(next, struct event_stat_id, rb_node);
 		perf_session__collapse_resort(&stats->hists);
-		perf_session__output_resort(&stats->hists, stats->stats.total);
-
+		nr_hists = perf_session__output_resort(&stats->hists,
+						       stats->stats.total);
 		if (use_browser)
-			perf_session__browse_hists(&stats->hists,
-						   stats->stats.total, help);
+			perf_session__browse_hists(&stats->hists, nr_hists,
+						   stats->stats.total, help,
+						   input_name);
 		else {
 			if (rb_first(&session->stats_by_id) ==
 			    rb_last(&session->stats_by_id))
@@ -469,7 +473,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __used)
 	setup_sorting(report_usage, options);
 
 	if (parent_pattern != default_parent_pattern) {
-		sort_dimension__add("parent");
+		if (sort_dimension__add("parent") < 0)
+			return -1;
 		sort_parent.elide = 1;
 	} else
 		symbol_conf.exclude_other = false;
diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c
index 0d4d8ff7914b..266e7aa996d3 100644
--- a/tools/perf/builtin-timechart.c
+++ b/tools/perf/builtin-timechart.c
@@ -21,7 +21,6 @@
 #include "util/cache.h"
 #include <linux/rbtree.h>
 #include "util/symbol.h"
-#include "util/string.h"
 #include "util/callchain.h"
 #include "util/strlist.h"
 
diff --git a/tools/perf/perf.c b/tools/perf/perf.c
index 2826e702986e..d4be55b6cd34 100644
--- a/tools/perf/perf.c
+++ b/tools/perf/perf.c
@@ -13,7 +13,6 @@
 #include "util/quote.h"
 #include "util/run-command.h"
 #include "util/parse-events.h"
-#include "util/string.h"
 #include "util/debugfs.h"
 
 bool use_browser;
diff --git a/tools/perf/util/PERF-VERSION-GEN b/tools/perf/util/PERF-VERSION-GEN
index 54552a00a117..49ece7921914 100755
--- a/tools/perf/util/PERF-VERSION-GEN
+++ b/tools/perf/util/PERF-VERSION-GEN
@@ -1,6 +1,10 @@
 #!/bin/sh
 
-GVF=PERF-VERSION-FILE
+if [ $# -eq 1 ]  ; then
+	OUTPUT=$1
+fi
+
+GVF=${OUTPUT}PERF-VERSION-FILE
 DEF_VER=v0.0.2.PERF
 
 LF='
diff --git a/tools/perf/util/color.c b/tools/perf/util/color.c
index 9da01914e0af..e191eb9a667f 100644
--- a/tools/perf/util/color.c
+++ b/tools/perf/util/color.c
@@ -166,6 +166,31 @@ int perf_color_default_config(const char *var, const char *value, void *cb)
 	return perf_default_config(var, value, cb);
 }
 
+static int __color_vsnprintf(char *bf, size_t size, const char *color,
+			     const char *fmt, va_list args, const char *trail)
+{
+	int r = 0;
+
+	/*
+	 * Auto-detect:
+	 */
+	if (perf_use_color_default < 0) {
+		if (isatty(1) || pager_in_use())
+			perf_use_color_default = 1;
+		else
+			perf_use_color_default = 0;
+	}
+
+	if (perf_use_color_default && *color)
+		r += snprintf(bf, size, "%s", color);
+	r += vsnprintf(bf + r, size - r, fmt, args);
+	if (perf_use_color_default && *color)
+		r += snprintf(bf + r, size - r, "%s", PERF_COLOR_RESET);
+	if (trail)
+		r += snprintf(bf + r, size - r, "%s", trail);
+	return r;
+}
+
 static int __color_vfprintf(FILE *fp, const char *color, const char *fmt,
 		va_list args, const char *trail)
 {
@@ -191,11 +216,28 @@ static int __color_vfprintf(FILE *fp, const char *color, const char *fmt,
 	return r;
 }
 
+int color_vsnprintf(char *bf, size_t size, const char *color,
+		    const char *fmt, va_list args)
+{
+	return __color_vsnprintf(bf, size, color, fmt, args, NULL);
+}
+
 int color_vfprintf(FILE *fp, const char *color, const char *fmt, va_list args)
 {
 	return __color_vfprintf(fp, color, fmt, args, NULL);
 }
 
+int color_snprintf(char *bf, size_t size, const char *color,
+		   const char *fmt, ...)
+{
+	va_list args;
+	int r;
+
+	va_start(args, fmt);
+	r = color_vsnprintf(bf, size, color, fmt, args);
+	va_end(args);
+	return r;
+}
 
 int color_fprintf(FILE *fp, const char *color, const char *fmt, ...)
 {
@@ -203,10 +245,7 @@ int color_fprintf(FILE *fp, const char *color, const char *fmt, ...)
 	int r;
 
 	va_start(args, fmt);
-	if (use_browser)
-		r = vfprintf(fp, fmt, args);
-	else
-		r = color_vfprintf(fp, color, fmt, args);
+	r = color_vfprintf(fp, color, fmt, args);
 	va_end(args);
 	return r;
 }
@@ -277,3 +316,9 @@ int percent_color_fprintf(FILE *fp, const char *fmt, double percent)
 
 	return r;
 }
+
+int percent_color_snprintf(char *bf, size_t size, const char *fmt, double percent)
+{
+	const char *color = get_percent_color(percent);
+	return color_snprintf(bf, size, color, fmt, percent);
+}
diff --git a/tools/perf/util/color.h b/tools/perf/util/color.h
index 24e8809210bb..dea082b79602 100644
--- a/tools/perf/util/color.h
+++ b/tools/perf/util/color.h
@@ -32,10 +32,14 @@ int perf_color_default_config(const char *var, const char *value, void *cb);
 int perf_config_colorbool(const char *var, const char *value, int stdout_is_tty);
 void color_parse(const char *value, const char *var, char *dst);
 void color_parse_mem(const char *value, int len, const char *var, char *dst);
+int color_vsnprintf(char *bf, size_t size, const char *color,
+		    const char *fmt, va_list args);
 int color_vfprintf(FILE *fp, const char *color, const char *fmt, va_list args);
 int color_fprintf(FILE *fp, const char *color, const char *fmt, ...);
+int color_snprintf(char *bf, size_t size, const char *color, const char *fmt, ...);
 int color_fprintf_ln(FILE *fp, const char *color, const char *fmt, ...);
 int color_fwrite_lines(FILE *fp, const char *color, size_t count, const char *buf);
+int percent_color_snprintf(char *bf, size_t size, const char *fmt, double percent);
 int percent_color_fprintf(FILE *fp, const char *fmt, double percent);
 const char *get_percent_color(double percent);
 
diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h
index 0172edf3f153..5cb0a1b1401a 100644
--- a/tools/perf/util/debug.h
+++ b/tools/perf/util/debug.h
@@ -10,13 +10,29 @@ extern int dump_trace;
 int dump_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2)));
 void trace_event(event_t *event);
 
+struct ui_progress;
+
 #ifdef NO_NEWT_SUPPORT
 static inline int browser__show_help(const char *format __used, va_list ap __used)
 {
 	return 0;
 }
+
+static inline struct ui_progress *ui_progress__new(const char *title __used,
+						   u64 total __used)
+{
+	return (struct ui_progress *)1;
+}
+
+static inline void ui_progress__update(struct ui_progress *self __used,
+				       u64 curr __used) {}
+
+static inline void ui_progress__delete(struct ui_progress *self __used) {}
 #else
 int browser__show_help(const char *format, va_list ap);
+struct ui_progress *ui_progress__new(const char *title, u64 total);
+void ui_progress__update(struct ui_progress *self, u64 curr);
+void ui_progress__delete(struct ui_progress *self);
 #endif
 
 #endif	/* __PERF_DEBUG_H */
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index 052eaeccc202..571fb25f7eb9 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -130,6 +130,7 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
 			continue;
 		pbf += n + 3;
 		if (*pbf == 'x') { /* vm_exec */
+			u64 vm_pgoff;
 			char *execname = strchr(bf, '/');
 
 			/* Catch VDSO */
@@ -139,6 +140,14 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
 			if (execname == NULL)
 				continue;
 
+			pbf += 3;
+			n = hex2u64(pbf, &vm_pgoff);
+			/* pgoff is in bytes, not pages */
+			if (n >= 0)
+				ev.mmap.pgoff = vm_pgoff << getpagesize();
+			else
+				ev.mmap.pgoff = 0;
+
 			size = strlen(execname);
 			execname[size - 1] = '\0'; /* Remove \n */
 			memcpy(ev.mmap.filename, execname, size);
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 09e09e78cb62..18cf8b321608 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -50,7 +50,8 @@ struct hist_entry *__perf_session__add_hist_entry(struct rb_root *hists,
 			p = &(*p)->rb_right;
 	}
 
-	he = malloc(sizeof(*he));
+	he = malloc(sizeof(*he) + (symbol_conf.use_callchain ?
+				    sizeof(struct callchain_node) : 0));
 	if (!he)
 		return NULL;
 	*he = entry;
@@ -168,7 +169,7 @@ static void perf_session__insert_output_hist_entry(struct rb_root *root,
 	struct hist_entry *iter;
 
 	if (symbol_conf.use_callchain)
-		callchain_param.sort(&he->sorted_chain, &he->callchain,
+		callchain_param.sort(&he->sorted_chain, he->callchain,
 				      min_callchain_hits, &callchain_param);
 
 	while (*p != NULL) {
@@ -185,12 +186,13 @@ static void perf_session__insert_output_hist_entry(struct rb_root *root,
 	rb_insert_color(&he->rb_node, root);
 }
 
-void perf_session__output_resort(struct rb_root *hists, u64 total_samples)
+u64 perf_session__output_resort(struct rb_root *hists, u64 total_samples)
 {
 	struct rb_root tmp;
 	struct rb_node *next;
 	struct hist_entry *n;
 	u64 min_callchain_hits;
+	u64 nr_hists = 0;
 
 	min_callchain_hits =
 		total_samples * (callchain_param.min_percent / 100);
@@ -205,9 +207,11 @@ void perf_session__output_resort(struct rb_root *hists, u64 total_samples)
 		rb_erase(&n->rb_node, hists);
 		perf_session__insert_output_hist_entry(&tmp, n,
 						       min_callchain_hits);
+		++nr_hists;
 	}
 
 	*hists = tmp;
+	return nr_hists;
 }
 
 static size_t callchain__fprintf_left_margin(FILE *fp, int left_margin)
@@ -452,16 +456,17 @@ static size_t hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self,
 	return ret;
 }
 
-size_t hist_entry__fprintf(struct hist_entry *self,
+int hist_entry__snprintf(struct hist_entry *self,
+			   char *s, size_t size,
 			   struct perf_session *pair_session,
 			   bool show_displacement,
-			   long displacement, FILE *fp,
+			   long displacement, bool color,
 			   u64 session_total)
 {
 	struct sort_entry *se;
 	u64 count, total;
 	const char *sep = symbol_conf.field_sep;
-	size_t ret;
+	int ret;
 
 	if (symbol_conf.exclude_other && !self->parent)
 		return 0;
@@ -474,17 +479,22 @@ size_t hist_entry__fprintf(struct hist_entry *self,
 		total = session_total;
 	}
 
-	if (total)
-		ret = percent_color_fprintf(fp, sep ? "%.2f" : "   %6.2f%%",
-					    (count * 100.0) / total);
-	else
-		ret = fprintf(fp, sep ? "%lld" : "%12lld ", count);
+	if (total) {
+		if (color)
+			ret = percent_color_snprintf(s, size,
+						     sep ? "%.2f" : "   %6.2f%%",
+						     (count * 100.0) / total);
+		else
+			ret = snprintf(s, size, sep ? "%.2f" : "   %6.2f%%",
+				       (count * 100.0) / total);
+	} else
+		ret = snprintf(s, size, sep ? "%lld" : "%12lld ", count);
 
 	if (symbol_conf.show_nr_samples) {
 		if (sep)
-			ret += fprintf(fp, "%c%lld", *sep, count);
+			ret += snprintf(s + ret, size - ret, "%c%lld", *sep, count);
 		else
-			ret += fprintf(fp, "%11lld", count);
+			ret += snprintf(s + ret, size - ret, "%11lld", count);
 	}
 
 	if (pair_session) {
@@ -504,9 +514,9 @@ size_t hist_entry__fprintf(struct hist_entry *self,
 			snprintf(bf, sizeof(bf), " ");
 
 		if (sep)
-			ret += fprintf(fp, "%c%s", *sep, bf);
+			ret += snprintf(s + ret, size - ret, "%c%s", *sep, bf);
 		else
-			ret += fprintf(fp, "%11.11s", bf);
+			ret += snprintf(s + ret, size - ret, "%11.11s", bf);
 
 		if (show_displacement) {
 			if (displacement)
@@ -515,9 +525,9 @@ size_t hist_entry__fprintf(struct hist_entry *self,
 				snprintf(bf, sizeof(bf), " ");
 
 			if (sep)
-				ret += fprintf(fp, "%c%s", *sep, bf);
+				ret += snprintf(s + ret, size - ret, "%c%s", *sep, bf);
 			else
-				ret += fprintf(fp, "%6.6s", bf);
+				ret += snprintf(s + ret, size - ret, "%6.6s", bf);
 		}
 	}
 
@@ -525,11 +535,25 @@ size_t hist_entry__fprintf(struct hist_entry *self,
 		if (se->elide)
 			continue;
 
-		ret += fprintf(fp, "%s", sep ?: "  ");
-		ret += se->print(fp, self, se->width ? *se->width : 0);
+		ret += snprintf(s + ret, size - ret, "%s", sep ?: "  ");
+		ret += se->snprintf(self, s + ret, size - ret,
+				    se->width ? *se->width : 0);
 	}
 
-	return ret + fprintf(fp, "\n");
+	return ret;
+}
+
+int hist_entry__fprintf(struct hist_entry *self,
+			struct perf_session *pair_session,
+			bool show_displacement,
+			long displacement, FILE *fp,
+			u64 session_total)
+{
+	char bf[512];
+	hist_entry__snprintf(self, bf, sizeof(bf), pair_session,
+			     show_displacement, displacement,
+			     true, session_total);
+	return fprintf(fp, "%s\n", bf);
 }
 
 static size_t hist_entry__fprintf_callchain(struct hist_entry *self, FILE *fp,
@@ -658,7 +682,7 @@ size_t perf_session__fprintf_hists(struct rb_root *hists,
 
 		if (h->ms.map == NULL && verbose > 1) {
 			__map_groups__fprintf_maps(&h->thread->mg,
-						   MAP__FUNCTION, fp);
+						   MAP__FUNCTION, verbose, fp);
 			fprintf(fp, "%.10s end\n", graph_dotted_line);
 		}
 	}
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index fe366ce5db45..ad17f0ad798b 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -18,14 +18,19 @@ struct hist_entry *__perf_session__add_hist_entry(struct rb_root *hists,
 						  u64 count, bool *hit);
 extern int64_t hist_entry__cmp(struct hist_entry *, struct hist_entry *);
 extern int64_t hist_entry__collapse(struct hist_entry *, struct hist_entry *);
-size_t hist_entry__fprintf(struct hist_entry *self,
+int hist_entry__fprintf(struct hist_entry *self,
 			   struct perf_session *pair_session,
 			   bool show_displacement,
 			   long displacement, FILE *fp,
 			   u64 session_total);
+int hist_entry__snprintf(struct hist_entry *self,
+			 char *bf, size_t size,
+			 struct perf_session *pair_session,
+			 bool show_displacement, long displacement,
+			 bool color, u64 session_total);
 void hist_entry__free(struct hist_entry *);
 
-void perf_session__output_resort(struct rb_root *hists, u64 total_samples);
+u64 perf_session__output_resort(struct rb_root *hists, u64 total_samples);
 void perf_session__collapse_resort(struct rb_root *hists);
 size_t perf_session__fprintf_hists(struct rb_root *hists,
 				   struct perf_session *pair,
diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
index 9f2963f9ee9a..37913b241bdf 100644
--- a/tools/perf/util/map.c
+++ b/tools/perf/util/map.c
@@ -1,4 +1,5 @@
 #include "symbol.h"
+#include <errno.h>
 #include <limits.h>
 #include <stdlib.h>
 #include <string.h>
@@ -234,18 +235,211 @@ u64 map__objdump_2ip(struct map *map, u64 addr)
 	return ip;
 }
 
+void map_groups__init(struct map_groups *self)
+{
+	int i;
+	for (i = 0; i < MAP__NR_TYPES; ++i) {
+		self->maps[i] = RB_ROOT;
+		INIT_LIST_HEAD(&self->removed_maps[i]);
+	}
+}
+
+void map_groups__flush(struct map_groups *self)
+{
+	int type;
+
+	for (type = 0; type < MAP__NR_TYPES; type++) {
+		struct rb_root *root = &self->maps[type];
+		struct rb_node *next = rb_first(root);
+
+		while (next) {
+			struct map *pos = rb_entry(next, struct map, rb_node);
+			next = rb_next(&pos->rb_node);
+			rb_erase(&pos->rb_node, root);
+			/*
+			 * We may have references to this map, for
+			 * instance in some hist_entry instances, so
+			 * just move them to a separate list.
+			 */
+			list_add_tail(&pos->node, &self->removed_maps[pos->type]);
+		}
+	}
+}
+
 struct symbol *map_groups__find_symbol(struct map_groups *self,
 				       enum map_type type, u64 addr,
+				       struct map **mapp,
 				       symbol_filter_t filter)
 {
 	struct map *map = map_groups__find(self, type, addr);
 
-	if (map != NULL)
+	if (map != NULL) {
+		if (mapp != NULL)
+			*mapp = map;
 		return map__find_symbol(map, map->map_ip(map, addr), filter);
+	}
 
 	return NULL;
 }
 
+struct symbol *map_groups__find_symbol_by_name(struct map_groups *self,
+					       enum map_type type,
+					       const char *name,
+					       struct map **mapp,
+					       symbol_filter_t filter)
+{
+	struct rb_node *nd;
+
+	for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) {
+		struct map *pos = rb_entry(nd, struct map, rb_node);
+		struct symbol *sym = map__find_symbol_by_name(pos, name, filter);
+
+		if (sym == NULL)
+			continue;
+		if (mapp != NULL)
+			*mapp = pos;
+		return sym;
+	}
+
+	return NULL;
+}
+
+size_t __map_groups__fprintf_maps(struct map_groups *self,
+				  enum map_type type, int verbose, FILE *fp)
+{
+	size_t printed = fprintf(fp, "%s:\n", map_type__name[type]);
+	struct rb_node *nd;
+
+	for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) {
+		struct map *pos = rb_entry(nd, struct map, rb_node);
+		printed += fprintf(fp, "Map:");
+		printed += map__fprintf(pos, fp);
+		if (verbose > 2) {
+			printed += dso__fprintf(pos->dso, type, fp);
+			printed += fprintf(fp, "--\n");
+		}
+	}
+
+	return printed;
+}
+
+size_t map_groups__fprintf_maps(struct map_groups *self, int verbose, FILE *fp)
+{
+	size_t printed = 0, i;
+	for (i = 0; i < MAP__NR_TYPES; ++i)
+		printed += __map_groups__fprintf_maps(self, i, verbose, fp);
+	return printed;
+}
+
+static size_t __map_groups__fprintf_removed_maps(struct map_groups *self,
+						 enum map_type type,
+						 int verbose, FILE *fp)
+{
+	struct map *pos;
+	size_t printed = 0;
+
+	list_for_each_entry(pos, &self->removed_maps[type], node) {
+		printed += fprintf(fp, "Map:");
+		printed += map__fprintf(pos, fp);
+		if (verbose > 1) {
+			printed += dso__fprintf(pos->dso, type, fp);
+			printed += fprintf(fp, "--\n");
+		}
+	}
+	return printed;
+}
+
+static size_t map_groups__fprintf_removed_maps(struct map_groups *self,
+					       int verbose, FILE *fp)
+{
+	size_t printed = 0, i;
+	for (i = 0; i < MAP__NR_TYPES; ++i)
+		printed += __map_groups__fprintf_removed_maps(self, i, verbose, fp);
+	return printed;
+}
+
+size_t map_groups__fprintf(struct map_groups *self, int verbose, FILE *fp)
+{
+	size_t printed = map_groups__fprintf_maps(self, verbose, fp);
+	printed += fprintf(fp, "Removed maps:\n");
+	return printed + map_groups__fprintf_removed_maps(self, verbose, fp);
+}
+
+int map_groups__fixup_overlappings(struct map_groups *self, struct map *map,
+				   int verbose, FILE *fp)
+{
+	struct rb_root *root = &self->maps[map->type];
+	struct rb_node *next = rb_first(root);
+
+	while (next) {
+		struct map *pos = rb_entry(next, struct map, rb_node);
+		next = rb_next(&pos->rb_node);
+
+		if (!map__overlap(pos, map))
+			continue;
+
+		if (verbose >= 2) {
+			fputs("overlapping maps:\n", fp);
+			map__fprintf(map, fp);
+			map__fprintf(pos, fp);
+		}
+
+		rb_erase(&pos->rb_node, root);
+		/*
+		 * We may have references to this map, for instance in some
+		 * hist_entry instances, so just move them to a separate
+		 * list.
+		 */
+		list_add_tail(&pos->node, &self->removed_maps[map->type]);
+		/*
+		 * Now check if we need to create new maps for areas not
+		 * overlapped by the new map:
+		 */
+		if (map->start > pos->start) {
+			struct map *before = map__clone(pos);
+
+			if (before == NULL)
+				return -ENOMEM;
+
+			before->end = map->start - 1;
+			map_groups__insert(self, before);
+			if (verbose >= 2)
+				map__fprintf(before, fp);
+		}
+
+		if (map->end < pos->end) {
+			struct map *after = map__clone(pos);
+
+			if (after == NULL)
+				return -ENOMEM;
+
+			after->start = map->end + 1;
+			map_groups__insert(self, after);
+			if (verbose >= 2)
+				map__fprintf(after, fp);
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * XXX This should not really _copy_ te maps, but refcount them.
+ */
+int map_groups__clone(struct map_groups *self,
+		      struct map_groups *parent, enum map_type type)
+{
+	struct rb_node *nd;
+	for (nd = rb_first(&parent->maps[type]); nd; nd = rb_next(nd)) {
+		struct map *map = rb_entry(nd, struct map, rb_node);
+		struct map *new = map__clone(map);
+		if (new == NULL)
+			return -ENOMEM;
+		map_groups__insert(self, new);
+	}
+	return 0;
+}
+
 static u64 map__reloc_map_ip(struct map *map, u64 ip)
 {
 	return ip + (s64)map->pgoff;
diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h
index 6a703fa74707..2031278cc06a 100644
--- a/tools/perf/util/map.h
+++ b/tools/perf/util/map.h
@@ -97,11 +97,14 @@ struct map_groups {
 };
 
 size_t __map_groups__fprintf_maps(struct map_groups *self,
-				  enum map_type type, FILE *fp);
+				  enum map_type type, int verbose, FILE *fp);
 void maps__insert(struct rb_root *maps, struct map *map);
 struct map *maps__find(struct rb_root *maps, u64 addr);
 void map_groups__init(struct map_groups *self);
-size_t map_groups__fprintf_maps(struct map_groups *self, FILE *fp);
+int map_groups__clone(struct map_groups *self,
+		      struct map_groups *parent, enum map_type type);
+size_t map_groups__fprintf(struct map_groups *self, int verbose, FILE *fp);
+size_t map_groups__fprintf_maps(struct map_groups *self, int verbose, FILE *fp);
 
 static inline void map_groups__insert(struct map_groups *self, struct map *map)
 {
@@ -116,15 +119,33 @@ static inline struct map *map_groups__find(struct map_groups *self,
 
 struct symbol *map_groups__find_symbol(struct map_groups *self,
 				       enum map_type type, u64 addr,
+				       struct map **mapp,
 				       symbol_filter_t filter);
 
-static inline struct symbol *map_groups__find_function(struct map_groups *self,
-						       u64 addr,
-						       symbol_filter_t filter)
+struct symbol *map_groups__find_symbol_by_name(struct map_groups *self,
+					       enum map_type type,
+					       const char *name,
+					       struct map **mapp,
+					       symbol_filter_t filter);
+
+static inline
+struct symbol *map_groups__find_function(struct map_groups *self, u64 addr,
+					 struct map **mapp, symbol_filter_t filter)
 {
-	return map_groups__find_symbol(self, MAP__FUNCTION, addr, filter);
+	return map_groups__find_symbol(self, MAP__FUNCTION, addr, mapp, filter);
 }
 
+static inline
+struct symbol *map_groups__find_function_by_name(struct map_groups *self,
+						 const char *name, struct map **mapp,
+						 symbol_filter_t filter)
+{
+	return map_groups__find_symbol_by_name(self, MAP__FUNCTION, name, mapp, filter);
+}
+
+int map_groups__fixup_overlappings(struct map_groups *self, struct map *map,
+				   int verbose, FILE *fp);
+
 struct map *map_groups__find_by_name(struct map_groups *self,
 				     enum map_type type, const char *name);
 int __map_groups__create_kernel_maps(struct map_groups *self,
@@ -134,5 +155,6 @@ int map_groups__create_kernel_maps(struct map_groups *self,
 				   struct map *vmlinux_maps[MAP__NR_TYPES]);
 struct map *map_groups__new_module(struct map_groups *self, u64 start,
 				   const char *filename);
+void map_groups__flush(struct map_groups *self);
 
 #endif /* __PERF_MAP_H */
diff --git a/tools/perf/util/newt.c b/tools/perf/util/newt.c
index e99bcc8d1939..c93bc2a2d137 100644
--- a/tools/perf/util/newt.c
+++ b/tools/perf/util/newt.c
@@ -12,6 +12,72 @@
 #include "sort.h"
 #include "symbol.h"
 
+struct ui_progress {
+	newtComponent form, scale;
+};
+
+struct ui_progress *ui_progress__new(const char *title, u64 total)
+{
+	struct ui_progress *self = malloc(sizeof(*self));
+
+	if (self != NULL) {
+		int cols;
+		newtGetScreenSize(&cols, NULL);
+		cols -= 4;
+		newtCenteredWindow(cols, 1, title);
+		self->form  = newtForm(NULL, NULL, 0);
+		if (self->form == NULL)
+			goto out_free_self;
+		self->scale = newtScale(0, 0, cols, total);
+		if (self->scale == NULL)
+			goto out_free_form;
+		newtFormAddComponents(self->form, self->scale, NULL);
+		newtRefresh();
+	}
+
+	return self;
+
+out_free_form:
+	newtFormDestroy(self->form);
+out_free_self:
+	free(self);
+	return NULL;
+}
+
+void ui_progress__update(struct ui_progress *self, u64 curr)
+{
+	newtScaleSet(self->scale, curr);
+	newtRefresh();
+}
+
+void ui_progress__delete(struct ui_progress *self)
+{
+	newtFormDestroy(self->form);
+	newtPopWindow();
+	free(self);
+}
+
+static char browser__last_msg[1024];
+
+int browser__show_help(const char *format, va_list ap)
+{
+	int ret;
+	static int backlog;
+
+        ret = vsnprintf(browser__last_msg + backlog,
+			sizeof(browser__last_msg) - backlog, format, ap);
+	backlog += ret;
+
+	if (browser__last_msg[backlog - 1] == '\n') {
+		newtPopHelpLine();
+		newtPushHelpLine(browser__last_msg);
+		newtRefresh();
+		backlog = 0;
+	}
+
+	return ret;
+}
+
 static void newt_form__set_exit_keys(newtComponent self)
 {
 	newtFormAddHotKey(self, NEWT_KEY_ESCAPE);
@@ -228,60 +294,17 @@ static void hist_entry__append_callchain_browser(struct hist_entry *self,
 	}
 }
 
-/*
- * FIXME: get lib/string.c linked with perf somehow
- */
-static char *skip_spaces(const char *str)
-{
-	while (isspace(*str))
-		++str;
-	return (char *)str;
-}
-
-static char *strim(char *s)
-{
-	size_t size;
-	char *end;
-
-	s = skip_spaces(s);
-	size = strlen(s);
-	if (!size)
-		return s;
-
-	end = s + size - 1;
-	while (end >= s && isspace(*end))
-		end--;
-	*(end + 1) = '\0';
-
-	return s;
-}
-
 static size_t hist_entry__append_browser(struct hist_entry *self,
 					 newtComponent tree, u64 total)
 {
-	char bf[1024], *s;
-	FILE *fp;
+	char s[256];
+	size_t ret;
 
 	if (symbol_conf.exclude_other && !self->parent)
 		return 0;
 
-	fp = fmemopen(bf, sizeof(bf), "w");
-	if (fp == NULL)
-		return 0;
-
-	hist_entry__fprintf(self, NULL, false, 0, fp, total);
-	fclose(fp);
-
-	/*
-	 * FIXME: We shouldn't need to trim, as the printing routines shouldn't
-	 * add spaces it in the first place, the stdio output routines should
-	 * call a __snprintf method instead of the current __print (that
-	 * actually is a __fprintf) one, but get the raw string and _then_ add
-	 * the newline, as this is a detail of stdio printing, not needed in
-	 * other UIs, e.g. newt.
-	 */
-	s = strim(bf);
-
+	ret = hist_entry__snprintf(self, s, sizeof(s), NULL,
+				   false, 0, false, total);
 	if (symbol_conf.use_callchain) {
 		int indexes[2];
 
@@ -291,10 +314,11 @@ static size_t hist_entry__append_browser(struct hist_entry *self,
 	} else
 		newtListboxAppendEntry(tree, s, &self->ms);
 
-	return strlen(s);
+	return ret;
 }
 
-static void map_symbol__annotate_browser(const struct map_symbol *self)
+static void map_symbol__annotate_browser(const struct map_symbol *self,
+					 const char *input_name)
 {
 	FILE *fp;
 	int cols, rows;
@@ -308,8 +332,8 @@ static void map_symbol__annotate_browser(const struct map_symbol *self)
 	if (self->sym == NULL)
 		return;
 
-	if (asprintf(&str, "perf annotate -d \"%s\" %s 2>&1 | expand",
-		     self->map->dso->name, self->sym->name) < 0)
+	if (asprintf(&str, "perf annotate -i \"%s\" -d \"%s\" %s 2>&1 | expand",
+		     input_name, self->map->dso->name, self->sym->name) < 0)
 		return;
 
 	fp = popen(str, "r");
@@ -358,90 +382,121 @@ static const void *newt__symbol_tree_get_current(newtComponent self)
 	return newtListboxGetCurrent(self);
 }
 
-static void perf_session__selection(newtComponent self, void *data)
+static void hist_browser__selection(newtComponent self, void *data)
 {
 	const struct map_symbol **symbol_ptr = data;
 	*symbol_ptr = newt__symbol_tree_get_current(self);
 }
 
-void perf_session__browse_hists(struct rb_root *hists, u64 session_total,
-				const char *helpline)
-{
-	struct sort_entry *se;
-	struct rb_node *nd;
-	char seq[] = ".";
-	unsigned int width;
-	char *col_width = symbol_conf.col_width_list_str;
-	int rows, cols, idx;
-	int max_len = 0;
-	char str[1024];
-	newtComponent form, tree;
-	struct newtExitStruct es;
+struct hist_browser {
+	newtComponent		form, tree;
 	const struct map_symbol *selection;
+};
 
-	snprintf(str, sizeof(str), "Samples: %Ld", session_total);
-	newtDrawRootText(0, 0, str);
-	newtPushHelpLine(helpline);
+static struct hist_browser *hist_browser__new(void)
+{
+	struct hist_browser *self = malloc(sizeof(*self));
 
-	newtGetScreenSize(&cols, &rows);
+	if (self != NULL) {
+		char seq[] = ".";
+		int rows;
 
-	if (symbol_conf.use_callchain)
-		tree = newtCheckboxTreeMulti(0, 0, rows - 5, seq,
-						NEWT_FLAG_SCROLL);
-	else
-		tree = newtListbox(0, 0, rows - 5, (NEWT_FLAG_SCROLL |
-						       NEWT_FLAG_RETURNEXIT));
+		newtGetScreenSize(NULL, &rows);
 
-	newtComponentAddCallback(tree, perf_session__selection, &selection);
-
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (se->elide)
-			continue;
-		width = strlen(se->header);
-		if (se->width) {
-			if (symbol_conf.col_width_list_str) {
-				if (col_width) {
-					*se->width = atoi(col_width);
-					col_width = strchr(col_width, ',');
-					if (col_width)
-						++col_width;
-				}
-			}
-			*se->width = max(*se->width, width);
-		}
+		if (symbol_conf.use_callchain)
+			self->tree = newtCheckboxTreeMulti(0, 0, rows - 5, seq,
+							   NEWT_FLAG_SCROLL);
+		else
+			self->tree = newtListbox(0, 0, rows - 5,
+						(NEWT_FLAG_SCROLL |
+						 NEWT_FLAG_RETURNEXIT));
+		newtComponentAddCallback(self->tree, hist_browser__selection,
+					 &self->selection);
 	}
 
+	return self;
+}
+
+static void hist_browser__delete(struct hist_browser *self)
+{
+	newtFormDestroy(self->form);
+	newtPopWindow();
+	free(self);
+}
+
+static int hist_browser__populate(struct hist_browser *self, struct rb_root *hists,
+				  u64 nr_hists, u64 session_total)
+{
+	int max_len = 0, idx, cols, rows;
+	struct ui_progress *progress;
+	struct rb_node *nd;
+	u64 curr_hist = 0;
+
+	progress = ui_progress__new("Adding entries to the browser...", nr_hists);
+	if (progress == NULL)
+		return -1;
+
 	idx = 0;
 	for (nd = rb_first(hists); nd; nd = rb_next(nd)) {
 		struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
-		int len = hist_entry__append_browser(h, tree, session_total);
+		int len = hist_entry__append_browser(h, self->tree, session_total);
 		if (len > max_len)
 			max_len = len;
 		if (symbol_conf.use_callchain)
-			hist_entry__append_callchain_browser(h, tree, session_total, idx++);
+			hist_entry__append_callchain_browser(h, self->tree,
+							     session_total, idx++);
+		++curr_hist;
+		if (curr_hist % 5)
+			ui_progress__update(progress, curr_hist);
 	}
 
+	ui_progress__delete(progress);
+
+	newtGetScreenSize(&cols, &rows);
+
 	if (max_len > cols)
 		max_len = cols - 3;
 
 	if (!symbol_conf.use_callchain)
-		newtListboxSetWidth(tree, max_len);
+		newtListboxSetWidth(self->tree, max_len);
 
 	newtCenteredWindow(max_len + (symbol_conf.use_callchain ? 5 : 0),
 			   rows - 5, "Report");
-	form = newt_form__new();
-	newtFormAddHotKey(form, 'A');
-	newtFormAddHotKey(form, 'a');
-	newtFormAddHotKey(form, NEWT_KEY_RIGHT);
-	newtFormAddComponents(form, tree, NULL);
-	selection = newt__symbol_tree_get_current(tree);
+	self->form = newt_form__new();
+	newtFormAddHotKey(self->form, 'A');
+	newtFormAddHotKey(self->form, 'a');
+	newtFormAddHotKey(self->form, NEWT_KEY_RIGHT);
+	newtFormAddComponents(self->form, self->tree, NULL);
+	self->selection = newt__symbol_tree_get_current(self->tree);
+
+	return 0;
+}
+
+int perf_session__browse_hists(struct rb_root *hists, u64 nr_hists,
+			       u64 session_total, const char *helpline,
+			       const char *input_name)
+{
+	struct newtExitStruct es;
+	char str[1024];
+	int err = -1;
+	struct hist_browser *browser = hist_browser__new();
+
+	if (browser == NULL)
+		return -1;
+
+	snprintf(str, sizeof(str), "Samples: %Ld", session_total);
+	newtDrawRootText(0, 0, str);
+	newtPushHelpLine(helpline);
+
+	if (hist_browser__populate(browser, hists, nr_hists, session_total) < 0)
+		goto out;
 
 	while (1) {
 		char annotate[512];
 		const char *options[2];
 		int nr_options = 0, choice = 0;
 
-		newtFormRun(form, &es);
+		newtFormRun(browser->form, &es);
 		if (es.reason == NEWT_EXIT_HOTKEY) {
 			if (toupper(es.u.key) == 'A')
 				goto do_annotate;
@@ -455,9 +510,9 @@ void perf_session__browse_hists(struct rb_root *hists, u64 session_total,
 			}
 		}
 
-		if (selection->sym != NULL) {
+		if (browser->selection->sym != NULL) {
 			snprintf(annotate, sizeof(annotate),
-				 "Annotate %s", selection->sym->name);
+				 "Annotate %s", browser->selection->sym->name);
 			options[nr_options++] = annotate;
 		}
 
@@ -466,41 +521,22 @@ void perf_session__browse_hists(struct rb_root *hists, u64 session_total,
 		if (choice == nr_options - 1)
 			break;
 do_annotate:
-		if (selection->sym != NULL && choice >= 0) {
-			if (selection->map->dso->origin == DSO__ORIG_KERNEL) {
+		if (browser->selection->sym != NULL && choice >= 0) {
+			if (browser->selection->map->dso->origin == DSO__ORIG_KERNEL) {
 				newtPopHelpLine();
 				newtPushHelpLine("No vmlinux file found, can't "
 						 "annotate with just a "
 						 "kallsyms file");
 				continue;
 			}
-			map_symbol__annotate_browser(selection);
+			map_symbol__annotate_browser(browser->selection,
+						     input_name);
 		}
 	}
-
-	newtFormDestroy(form);
-	newtPopWindow();
-}
-
-static char browser__last_msg[1024];
-
-int browser__show_help(const char *format, va_list ap)
-{
-	int ret;
-	static int backlog;
-
-        ret = vsnprintf(browser__last_msg + backlog,
-			sizeof(browser__last_msg) - backlog, format, ap);
-	backlog += ret;
-
-	if (browser__last_msg[backlog - 1] == '\n') {
-		newtPopHelpLine();
-		newtPushHelpLine(browser__last_msg);
-		newtRefresh();
-		backlog = 0;
-	}
-
-	return ret;
+	err = 0;
+out:
+	hist_browser__delete(browser);
+	return err;
 }
 
 void setup_browser(void)
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 76b4ac689df9..ddf288fca3eb 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -52,11 +52,6 @@ static int perf_session__open(struct perf_session *self, bool force)
 	return -1;
 }
 
-static inline int perf_session__create_kernel_maps(struct perf_session *self)
-{
-	return map_groups__create_kernel_maps(&self->kmaps, self->vmlinux_maps);
-}
-
 struct perf_session *perf_session__new(const char *filename, int mode, bool force)
 {
 	size_t len = filename ? strlen(filename) + 1 : 0;
@@ -123,16 +118,11 @@ struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
 						   struct symbol **parent)
 {
 	u8 cpumode = PERF_RECORD_MISC_USER;
-	struct map_symbol *syms = NULL;
 	unsigned int i;
+	struct map_symbol *syms = calloc(chain->nr, sizeof(*syms));
 
-	if (symbol_conf.use_callchain) {
-		syms = calloc(chain->nr, sizeof(*syms));
-		if (!syms) {
-			fprintf(stderr, "Can't allocate memory for symbols\n");
-			exit(-1);
-		}
-	}
+	if (!syms)
+		return NULL;
 
 	for (i = 0; i < chain->nr; i++) {
 		u64 ip = chain->ips[i];
@@ -397,6 +387,10 @@ int __perf_session__process_events(struct perf_session *self,
 	event_t *event;
 	uint32_t size;
 	char *buf;
+	struct ui_progress *progress = ui_progress__new("Processing events...",
+							self->size);
+	if (progress == NULL)
+		return -1;
 
 	perf_event_ops__fill_defaults(ops);
 
@@ -425,6 +419,7 @@ int __perf_session__process_events(struct perf_session *self,
 
 more:
 	event = (event_t *)(buf + head);
+	ui_progress__update(progress, offset);
 
 	if (self->header.needs_swap)
 		perf_event_header__bswap(&event->header);
@@ -475,6 +470,7 @@ int __perf_session__process_events(struct perf_session *self,
 done:
 	err = 0;
 out_err:
+	ui_progress__delete(progress);
 	return err;
 }
 
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index 631f8157fc17..27f4c2dc715b 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -80,6 +80,11 @@ static inline int __perf_session__create_kernel_maps(struct perf_session *self,
 						self->vmlinux_maps, kernel);
 }
 
+static inline int perf_session__create_kernel_maps(struct perf_session *self)
+{
+	return map_groups__create_kernel_maps(&self->kmaps, self->vmlinux_maps);
+}
+
 static inline struct map *
 	perf_session__new_module_map(struct perf_session *self,
 				     u64 start, const char *filename)
@@ -88,11 +93,17 @@ static inline struct map *
 }
 
 #ifdef NO_NEWT_SUPPORT
-static inline void perf_session__browse_hists(struct rb_root *hists __used,
+static inline int perf_session__browse_hists(struct rb_root *hists __used,
+					      u64 nr_hists __used,
 					      u64 session_total __used,
-					      const char *helpline __used) {}
+					     const char *helpline __used,
+					     const char *input_name __used)
+{
+	return 0;
+}
 #else
-void perf_session__browse_hists(struct rb_root *hists, u64 session_total,
-				const char *helpline);
+int perf_session__browse_hists(struct rb_root *hists, u64 nr_hists,
+			       u64 session_total, const char *helpline,
+			       const char *input_name);
 #endif
 #endif /* __PERF_SESSION_H */
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 9b80c13cae46..9d24d4b2c8fb 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -18,10 +18,21 @@ char * field_sep;
 
 LIST_HEAD(hist_entry__sort_list);
 
+static int hist_entry__thread_snprintf(struct hist_entry *self, char *bf,
+				       size_t size, unsigned int width);
+static int hist_entry__comm_snprintf(struct hist_entry *self, char *bf,
+				     size_t size, unsigned int width);
+static int hist_entry__dso_snprintf(struct hist_entry *self, char *bf,
+				    size_t size, unsigned int width);
+static int hist_entry__sym_snprintf(struct hist_entry *self, char *bf,
+				    size_t size, unsigned int width);
+static int hist_entry__parent_snprintf(struct hist_entry *self, char *bf,
+				       size_t size, unsigned int width);
+
 struct sort_entry sort_thread = {
 	.header = "Command:  Pid",
 	.cmp	= sort__thread_cmp,
-	.print	= sort__thread_print,
+	.snprintf = hist_entry__thread_snprintf,
 	.width	= &threads__col_width,
 };
 
@@ -29,27 +40,27 @@ struct sort_entry sort_comm = {
 	.header		= "Command",
 	.cmp		= sort__comm_cmp,
 	.collapse	= sort__comm_collapse,
-	.print		= sort__comm_print,
+	.snprintf	= hist_entry__comm_snprintf,
 	.width		= &comms__col_width,
 };
 
 struct sort_entry sort_dso = {
 	.header = "Shared Object",
 	.cmp	= sort__dso_cmp,
-	.print	= sort__dso_print,
+	.snprintf = hist_entry__dso_snprintf,
 	.width	= &dsos__col_width,
 };
 
 struct sort_entry sort_sym = {
 	.header = "Symbol",
 	.cmp	= sort__sym_cmp,
-	.print	= sort__sym_print,
+	.snprintf = hist_entry__sym_snprintf,
 };
 
 struct sort_entry sort_parent = {
 	.header = "Parent symbol",
 	.cmp	= sort__parent_cmp,
-	.print	= sort__parent_print,
+	.snprintf = hist_entry__parent_snprintf,
 	.width	= &parent_symbol__col_width,
 };
 
@@ -85,45 +96,38 @@ sort__thread_cmp(struct hist_entry *left, struct hist_entry *right)
 	return right->thread->pid - left->thread->pid;
 }
 
-int repsep_fprintf(FILE *fp, const char *fmt, ...)
+static int repsep_snprintf(char *bf, size_t size, const char *fmt, ...)
 {
 	int n;
 	va_list ap;
 
 	va_start(ap, fmt);
-	if (!field_sep)
-		n = vfprintf(fp, fmt, ap);
-	else {
-		char *bf = NULL;
-		n = vasprintf(&bf, fmt, ap);
-		if (n > 0) {
-			char *sep = bf;
+	n = vsnprintf(bf, size, fmt, ap);
+	if (field_sep && n > 0) {
+		char *sep = bf;
 
-			while (1) {
-				sep = strchr(sep, *field_sep);
-				if (sep == NULL)
-					break;
-				*sep = '.';
-			}
+		while (1) {
+			sep = strchr(sep, *field_sep);
+			if (sep == NULL)
+				break;
+			*sep = '.';
 		}
-		fputs(bf, fp);
-		free(bf);
 	}
 	va_end(ap);
 	return n;
 }
 
-size_t
-sort__thread_print(FILE *fp, struct hist_entry *self, unsigned int width)
+static int hist_entry__thread_snprintf(struct hist_entry *self, char *bf,
+				       size_t size, unsigned int width)
 {
-	return repsep_fprintf(fp, "%*s:%5d", width - 6,
+	return repsep_snprintf(bf, size, "%*s:%5d", width,
 			      self->thread->comm ?: "", self->thread->pid);
 }
 
-size_t
-sort__comm_print(FILE *fp, struct hist_entry *self, unsigned int width)
+static int hist_entry__comm_snprintf(struct hist_entry *self, char *bf,
+				     size_t size, unsigned int width)
 {
-	return repsep_fprintf(fp, "%*s", width, self->thread->comm);
+	return repsep_snprintf(bf, size, "%*s", width, self->thread->comm);
 }
 
 /* --sort dso */
@@ -149,16 +153,16 @@ sort__dso_cmp(struct hist_entry *left, struct hist_entry *right)
 	return strcmp(dso_name_l, dso_name_r);
 }
 
-size_t
-sort__dso_print(FILE *fp, struct hist_entry *self, unsigned int width)
+static int hist_entry__dso_snprintf(struct hist_entry *self, char *bf,
+				    size_t size, unsigned int width)
 {
 	if (self->ms.map && self->ms.map->dso) {
 		const char *dso_name = !verbose ? self->ms.map->dso->short_name :
 						  self->ms.map->dso->long_name;
-		return repsep_fprintf(fp, "%-*s", width, dso_name);
+		return repsep_snprintf(bf, size, "%-*s", width, dso_name);
 	}
 
-	return repsep_fprintf(fp, "%*llx", width, (u64)self->ip);
+	return repsep_snprintf(bf, size, "%*Lx", width, self->ip);
 }
 
 /* --sort symbol */
@@ -177,22 +181,22 @@ sort__sym_cmp(struct hist_entry *left, struct hist_entry *right)
 	return (int64_t)(ip_r - ip_l);
 }
 
-
-size_t
-sort__sym_print(FILE *fp, struct hist_entry *self, unsigned int width __used)
+static int hist_entry__sym_snprintf(struct hist_entry *self, char *bf,
+				    size_t size, unsigned int width __used)
 {
 	size_t ret = 0;
 
 	if (verbose) {
 		char o = self->ms.map ? dso__symtab_origin(self->ms.map->dso) : '!';
-		ret += repsep_fprintf(fp, "%#018llx %c ", (u64)self->ip, o);
+		ret += repsep_snprintf(bf, size, "%#018llx %c ", self->ip, o);
 	}
 
-	ret += repsep_fprintf(fp, "[%c] ", self->level);
+	ret += repsep_snprintf(bf + ret, size - ret, "[%c] ", self->level);
 	if (self->ms.sym)
-		ret += repsep_fprintf(fp, "%s", self->ms.sym->name);
+		ret += repsep_snprintf(bf + ret, size - ret, "%s",
+				       self->ms.sym->name);
 	else
-		ret += repsep_fprintf(fp, "%#016llx", (u64)self->ip);
+		ret += repsep_snprintf(bf + ret, size - ret, "%#016llx", self->ip);
 
 	return ret;
 }
@@ -231,10 +235,10 @@ sort__parent_cmp(struct hist_entry *left, struct hist_entry *right)
 	return strcmp(sym_l->name, sym_r->name);
 }
 
-size_t
-sort__parent_print(FILE *fp, struct hist_entry *self, unsigned int width)
+static int hist_entry__parent_snprintf(struct hist_entry *self, char *bf,
+				       size_t size, unsigned int width)
 {
-	return repsep_fprintf(fp, "%-*s", width,
+	return repsep_snprintf(bf, size, "%-*s", width,
 			      self->parent ? self->parent->name : "[other]");
 }
 
@@ -260,9 +264,8 @@ int sort_dimension__add(const char *tok)
 				char err[BUFSIZ];
 
 				regerror(ret, &parent_regex, err, sizeof(err));
-				fprintf(stderr, "Invalid regex: %s\n%s",
-					parent_pattern, err);
-				exit(-1);
+				pr_err("Invalid regex: %s\n%s", parent_pattern, err);
+				return -EINVAL;
 			}
 			sort__has_parent = 1;
 		}
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 598568696f97..5bf2b744e7b2 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -49,12 +49,12 @@ struct hist_entry {
 	u64			ip;
 	char			level;
 	struct symbol	  *parent;
-	struct callchain_node	callchain;
 	union {
 		unsigned long	  position;
 		struct hist_entry *pair;
 		struct rb_root	  sorted_chain;
 	};
+	struct callchain_node	callchain[0];
 };
 
 enum sort_type {
@@ -76,7 +76,8 @@ struct sort_entry {
 
 	int64_t (*cmp)(struct hist_entry *, struct hist_entry *);
 	int64_t (*collapse)(struct hist_entry *, struct hist_entry *);
-	size_t	(*print)(FILE *fp, struct hist_entry *, unsigned int width);
+	int	(*snprintf)(struct hist_entry *self, char *bf, size_t size,
+			    unsigned int width);
 	unsigned int *width;
 	bool	elide;
 };
@@ -86,7 +87,6 @@ extern struct list_head hist_entry__sort_list;
 
 void setup_sorting(const char * const usagestr[], const struct option *opts);
 
-extern int repsep_fprintf(FILE *fp, const char *fmt, ...);
 extern size_t sort__thread_print(FILE *, struct hist_entry *, unsigned int);
 extern size_t sort__comm_print(FILE *, struct hist_entry *, unsigned int);
 extern size_t sort__dso_print(FILE *, struct hist_entry *, unsigned int);
diff --git a/tools/perf/util/string.h b/tools/perf/util/string.h
deleted file mode 100644
index 700582416664..000000000000
--- a/tools/perf/util/string.h
+++ /dev/null
@@ -1,16 +0,0 @@
-#ifndef __PERF_STRING_H_
-#define __PERF_STRING_H_
-
-#include <stdbool.h>
-#include "types.h"
-
-s64 perf_atoll(const char *str);
-char **argv_split(const char *str, int *argcp);
-void argv_free(char **argv);
-bool strglobmatch(const char *str, const char *pat);
-bool strlazymatch(const char *str, const char *pat);
-
-#define _STR(x) #x
-#define STR(x) _STR(x)
-
-#endif /* __PERF_STRING_H */
diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
index 9bbe27d75306..1f7ecd47f499 100644
--- a/tools/perf/util/thread.c
+++ b/tools/perf/util/thread.c
@@ -38,15 +38,6 @@ int find_all_tid(int pid, pid_t ** all_tid)
 	return ret;
 }
 
-void map_groups__init(struct map_groups *self)
-{
-	int i;
-	for (i = 0; i < MAP__NR_TYPES; ++i) {
-		self->maps[i] = RB_ROOT;
-		INIT_LIST_HEAD(&self->removed_maps[i]);
-	}
-}
-
 static struct thread *thread__new(pid_t pid)
 {
 	struct thread *self = zalloc(sizeof(*self));
@@ -62,28 +53,6 @@ static struct thread *thread__new(pid_t pid)
 	return self;
 }
 
-static void map_groups__flush(struct map_groups *self)
-{
-	int type;
-
-	for (type = 0; type < MAP__NR_TYPES; type++) {
-		struct rb_root *root = &self->maps[type];
-		struct rb_node *next = rb_first(root);
-
-		while (next) {
-			struct map *pos = rb_entry(next, struct map, rb_node);
-			next = rb_next(&pos->rb_node);
-			rb_erase(&pos->rb_node, root);
-			/*
-			 * We may have references to this map, for
-			 * instance in some hist_entry instances, so
-			 * just move them to a separate list.
-			 */
-			list_add_tail(&pos->node, &self->removed_maps[pos->type]);
-		}
-	}
-}
-
 int thread__set_comm(struct thread *self, const char *comm)
 {
 	int err;
@@ -110,69 +79,10 @@ int thread__comm_len(struct thread *self)
 	return self->comm_len;
 }
 
-size_t __map_groups__fprintf_maps(struct map_groups *self,
-				  enum map_type type, FILE *fp)
-{
-	size_t printed = fprintf(fp, "%s:\n", map_type__name[type]);
-	struct rb_node *nd;
-
-	for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) {
-		struct map *pos = rb_entry(nd, struct map, rb_node);
-		printed += fprintf(fp, "Map:");
-		printed += map__fprintf(pos, fp);
-		if (verbose > 2) {
-			printed += dso__fprintf(pos->dso, type, fp);
-			printed += fprintf(fp, "--\n");
-		}
-	}
-
-	return printed;
-}
-
-size_t map_groups__fprintf_maps(struct map_groups *self, FILE *fp)
-{
-	size_t printed = 0, i;
-	for (i = 0; i < MAP__NR_TYPES; ++i)
-		printed += __map_groups__fprintf_maps(self, i, fp);
-	return printed;
-}
-
-static size_t __map_groups__fprintf_removed_maps(struct map_groups *self,
-						 enum map_type type, FILE *fp)
-{
-	struct map *pos;
-	size_t printed = 0;
-
-	list_for_each_entry(pos, &self->removed_maps[type], node) {
-		printed += fprintf(fp, "Map:");
-		printed += map__fprintf(pos, fp);
-		if (verbose > 1) {
-			printed += dso__fprintf(pos->dso, type, fp);
-			printed += fprintf(fp, "--\n");
-		}
-	}
-	return printed;
-}
-
-static size_t map_groups__fprintf_removed_maps(struct map_groups *self, FILE *fp)
-{
-	size_t printed = 0, i;
-	for (i = 0; i < MAP__NR_TYPES; ++i)
-		printed += __map_groups__fprintf_removed_maps(self, i, fp);
-	return printed;
-}
-
-static size_t map_groups__fprintf(struct map_groups *self, FILE *fp)
-{
-	size_t printed = map_groups__fprintf_maps(self, fp);
-	printed += fprintf(fp, "Removed maps:\n");
-	return printed + map_groups__fprintf_removed_maps(self, fp);
-}
-
 static size_t thread__fprintf(struct thread *self, FILE *fp)
 {
 	return fprintf(fp, "Thread %d %s\n", self->pid, self->comm) +
-	       map_groups__fprintf(&self->mg, fp);
+	       map_groups__fprintf(&self->mg, verbose, fp);
 }
 
 struct thread *perf_session__findnew(struct perf_session *self, pid_t pid)
@@ -214,87 +124,12 @@ struct thread *perf_session__findnew(struct perf_session *self, pid_t pid)
 	return th;
 }
 
-static int map_groups__fixup_overlappings(struct map_groups *self,
-					  struct map *map)
-{
-	struct rb_root *root = &self->maps[map->type];
-	struct rb_node *next = rb_first(root);
-
-	while (next) {
-		struct map *pos = rb_entry(next, struct map, rb_node);
-		next = rb_next(&pos->rb_node);
-
-		if (!map__overlap(pos, map))
-			continue;
-
-		if (verbose >= 2) {
-			fputs("overlapping maps:\n", stderr);
-			map__fprintf(map, stderr);
-			map__fprintf(pos, stderr);
-		}
-
-		rb_erase(&pos->rb_node, root);
-		/*
-		 * We may have references to this map, for instance in some
-		 * hist_entry instances, so just move them to a separate
-		 * list.
-		 */
-		list_add_tail(&pos->node, &self->removed_maps[map->type]);
-		/*
-		 * Now check if we need to create new maps for areas not
-		 * overlapped by the new map:
-		 */
-		if (map->start > pos->start) {
-			struct map *before = map__clone(pos);
-
-			if (before == NULL)
-				return -ENOMEM;
-
-			before->end = map->start - 1;
-			map_groups__insert(self, before);
-			if (verbose >= 2)
-				map__fprintf(before, stderr);
-		}
-
-		if (map->end < pos->end) {
-			struct map *after = map__clone(pos);
-
-			if (after == NULL)
-				return -ENOMEM;
-
-			after->start = map->end + 1;
-			map_groups__insert(self, after);
-			if (verbose >= 2)
-				map__fprintf(after, stderr);
-		}
-	}
-
-	return 0;
-}
-
 void thread__insert_map(struct thread *self, struct map *map)
 {
-	map_groups__fixup_overlappings(&self->mg, map);
+	map_groups__fixup_overlappings(&self->mg, map, verbose, stderr);
 	map_groups__insert(&self->mg, map);
 }
 
-/*
- * XXX This should not really _copy_ te maps, but refcount them.
- */
-static int map_groups__clone(struct map_groups *self,
-			     struct map_groups *parent, enum map_type type)
-{
-	struct rb_node *nd;
-	for (nd = rb_first(&parent->maps[type]); nd; nd = rb_next(nd)) {
-		struct map *map = rb_entry(nd, struct map, rb_node);
-		struct map *new = map__clone(map);
-		if (new == NULL)
-			return -ENOMEM;
-		map_groups__insert(self, new);
-	}
-	return 0;
-}
-
 int thread__fork(struct thread *self, struct thread *parent)
 {
 	int i;
diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h
index 52701087ce04..fbf45d1b26f7 100644
--- a/tools/perf/util/util.h
+++ b/tools/perf/util/util.h
@@ -42,12 +42,14 @@
 #define _ALL_SOURCE 1
 #define _GNU_SOURCE 1
 #define _BSD_SOURCE 1
+#define HAS_BOOL
 
 #include <unistd.h>
 #include <stdio.h>
 #include <sys/stat.h>
 #include <sys/statfs.h>
 #include <fcntl.h>
+#include <stdbool.h>
 #include <stddef.h>
 #include <stdlib.h>
 #include <stdarg.h>
@@ -78,6 +80,7 @@
 #include <pwd.h>
 #include <inttypes.h>
 #include "../../../include/linux/magic.h"
+#include "types.h"
 
 
 #ifndef NO_ICONV
@@ -415,4 +418,13 @@ void git_qsort(void *base, size_t nmemb, size_t size,
 int mkdir_p(char *path, mode_t mode);
 int copyfile(const char *from, const char *to);
 
+s64 perf_atoll(const char *str);
+char **argv_split(const char *str, int *argcp);
+void argv_free(char **argv);
+bool strglobmatch(const char *str, const char *pat);
+bool strlazymatch(const char *str, const char *pat);
+
+#define _STR(x) #x
+#define STR(x) _STR(x)
+
 #endif