package/perf: migrate perf to use the new linux-tools infrastructure

Remove the perf package and add legacy handling.

[Thomas:
  - improve the Config.in.legacy help text
  - improve the comment explaining why we pass O= when building perf]

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux/Config.tools.in

@@ -8,4 +8,22 @@ config BR2_LINUX_KERNEL_TOOL_CPUPOWER
 	  cpupower is a collection of tools to examine and tune power
 	  saving related features of your processor.
 
+config BR2_LINUX_KERNEL_TOOL_PERF
+	bool "perf"
+	help
+	  perf (sometimes "Perf Events" or perf tools, originally
+	  "Performance Counters for Linux") - is a performance
+	  analyzing tool in Linux, available from kernel version
+	  2.6.31. User-space controlling utility, called 'perf' has
+	  git-like interface with subcommands. It is capable of
+	  statistical profiling of entire system (both kernel and user
+	  code), single CPU or severals threads.
+
+	  This will build and install the userspace 'perf'
+	  command. It is up to the user to ensure that the kernel
+	  configuration has all the suitable options enabled to allow a
+	  proper operation of 'perf'.
+
+	  https://perf.wiki.kernel.org/
+
 endmenu

linux/linux-tool-perf.mk

@@ -0,0 +1,87 @@
+################################################################################
+#
+# perf
+#
+################################################################################
+
+LINUX_TOOLS += perf
+
+PERF_DEPENDENCIES = host-flex host-bison
+
+ifeq ($(KERNEL_ARCH),x86_64)
+PERF_ARCH=x86
+else
+PERF_ARCH=$(KERNEL_ARCH)
+endif
+
+PERF_MAKE_FLAGS = \
+	$(LINUX_MAKE_FLAGS) \
+	ARCH=$(PERF_ARCH) \
+	NO_LIBAUDIT=1 \
+	NO_NEWT=1 \
+	NO_GTK2=1 \
+	NO_LIBPERL=1 \
+	NO_LIBPYTHON=1 \
+	DESTDIR=$(TARGET_DIR) \
+	prefix=/usr \
+	WERROR=0 \
+	ASCIIDOC=
+
+# The call to backtrace() function fails for ARC, because for some
+# reason the unwinder from libgcc returns early. Thus the usage of
+# backtrace() should be disabled in perf explicitly: at build time
+# backtrace() appears to be available, but it fails at runtime: the
+# backtrace will contain only several functions from the top of stack,
+# instead of the complete backtrace.
+ifeq ($(BR2_arc),y)
+PERF_MAKE_FLAGS += NO_BACKTRACE=1
+endif
+
+ifeq ($(BR2_PACKAGE_SLANG),y)
+PERF_DEPENDENCIES += slang
+endif
+
+ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
+PERF_DEPENDENCIES += libunwind
+endif
+
+ifeq ($(BR2_PACKAGE_NUMACTL),y)
+PERF_DEPENDENCIES += numactl
+endif
+
+ifeq ($(BR2_PACKAGE_ELFUTILS),y)
+PERF_DEPENDENCIES += elfutils
+else
+PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
+endif
+
+# O must be redefined here to overwrite the one used by Buildroot for
+# out of tree build. We build perf in $(@D)/tools/perf/ and not just
+# $(@D) so that it isn't built in the root directory of the kernel
+# sources.
+define PERF_BUILD_CMDS
+	$(Q)if test ! -f $(@D)/tools/perf/Makefile ; then \
+		echo "Your kernel version is too old and does not have the perf tool." ; \
+		echo "At least kernel 2.6.31 must be used." ; \
+		exit 1 ; \
+	fi
+	$(Q)if test "$(BR2_PACKAGE_ELFUTILS)" = "" ; then \
+		if ! grep -q NO_LIBELF $(@D)/tools/perf/Makefile* ; then \
+			if ! test -r $(@D)/tools/perf/config/Makefile ; then \
+				echo "The perf tool in your kernel cannot be built without libelf." ; \
+				echo "Either upgrade your kernel to >= 3.7, or enable the elfutils package." ; \
+				exit 1 ; \
+			fi \
+		fi \
+	fi
+	$(TARGET_MAKE_ENV) $(MAKE1) $(PERF_MAKE_FLAGS) \
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/
+endef
+
+# After installation, we remove the Perl and Python scripts from the
+# target.
+define PERF_INSTALL_TARGET_CMDS
+	$(TARGET_MAKE_ENV) $(MAKE1) $(PERF_MAKE_FLAGS) \
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/ install
+	$(RM) -rf $(TARGET_DIR)/usr/libexec/perf-core/scripts/
+endef