@@ -617,7 +617,7 @@ config ARCH_SUPPORTS_CLANG_LTO
config CLANG_LTO
bool "Use clang Link Time Optimization (LTO)"
depends on ARCH_SUPPORTS_CLANG_LTO
- depends on !FTRACE_MCOUNT_RECORD
+ depends on !FTRACE_MCOUNT_RECORD || HAVE_C_RECORDMCOUNT
select LD_DEAD_CODE_DATA_ELIMINATION
help
This option enables clang's Link Time Optimization (LTO), which allows
@@ -458,7 +458,7 @@
*/
#define TEXT_TEXT \
ALIGN_FUNCTION(); \
- *(.text.hot TEXT_MAIN .text.fixup .text.unlikely) \
+ *(.text.hot TEXT_MAIN .text.fixup .text.unlikely .text..ftrace) \
*(.ref.text) \
MEM_KEEP(init.text) \
MEM_KEEP(exit.text) \
@@ -16,3 +16,10 @@
* with any version that can compile the kernel
*/
#define __UNIQUE_ID(prefix) __PASTE(__PASTE(__UNIQUE_ID_, prefix), __COUNTER__)
+
+#ifdef CONFIG_CLANG_LTO
+#ifdef CONFIG_FTRACE_MCOUNT_RECORD
+#define __norecordmcount \
+ __attribute__((__section__(".text..ftrace")))
+#endif
+#endif
@@ -502,6 +502,10 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
#define __visible
#endif
+#ifndef __norecordmcount
+#define __norecordmcount
+#endif
+
#ifndef __nostackprotector
# define __nostackprotector
#endif
@@ -5573,9 +5573,9 @@ static int ftrace_cmp_ips(const void *a, const void *b)
return 0;
}
-static int ftrace_process_locs(struct module *mod,
- unsigned long *start,
- unsigned long *end)
+static int __norecordmcount ftrace_process_locs(struct module *mod,
+ unsigned long *start,
+ unsigned long *end)
{
struct ftrace_page *start_pg;
struct ftrace_page *pg;
@@ -246,6 +246,12 @@ ifdef BUILD_C_RECORDMCOUNT
ifeq ("$(origin RECORDMCOUNT_WARN)", "command line")
RECORDMCOUNT_FLAGS = -w
endif
+
+ifdef CONFIG_CLANG_LTO
+# With LTO, we postpone running recordmcount until after the LTO link step, so
+# let's export the parameters for the link script.
+export RECORDMCOUNT_FLAGS
+else
# Due to recursion, we must skip empty.o.
# The empty.o file is created in the make process in order to determine
# the target endianness and word size. It is made before all other C
@@ -254,17 +260,22 @@ sub_cmd_record_mcount = \
if [ $(@) != "scripts/mod/empty.o" ]; then \
$(objtree)/scripts/recordmcount $(RECORDMCOUNT_FLAGS) "$(@)"; \
fi;
+endif
+
recordmcount_source := $(srctree)/scripts/recordmcount.c \
$(srctree)/scripts/recordmcount.h
-else
+else # !BUILD_C_RECORDMCOUNT
sub_cmd_record_mcount = set -e ; perl $(srctree)/scripts/recordmcount.pl "$(ARCH)" \
"$(if $(CONFIG_CPU_BIG_ENDIAN),big,little)" \
"$(if $(CONFIG_64BIT),64,32)" \
"$(OBJDUMP)" "$(OBJCOPY)" "$(CC) $(KBUILD_CFLAGS)" \
"$(LD)" "$(NM)" "$(RM)" "$(MV)" \
"$(if $(part-of-module),1,0)" "$(@)";
+
recordmcount_source := $(srctree)/scripts/recordmcount.pl
endif # BUILD_C_RECORDMCOUNT
+
+ifndef CONFIG_CLANG_LTO
cmd_record_mcount = \
if [ "$(findstring $(CC_FLAGS_FTRACE),$(_c_flags))" = \
"$(CC_FLAGS_FTRACE)" ]; then \
@@ -287,6 +298,7 @@ objtool_args += --no-unreachable
else
objtool_args += $(call cc-ifversion, -lt, 0405, --no-unreachable)
endif
+endif # CONFIG_FTRACE_MCOUNT_RECORD
# 'OBJECT_FILES_NON_STANDARD := y': skip objtool checking for a directory
# 'OBJECT_FILES_NON_STANDARD_foo.o := 'y': skip objtool checking for a file
@@ -164,6 +164,10 @@ lto_ko_objects = $(foreach o,$(1:$(modpost-ext).o=.o), \
$(shell [ -s $(@:.ko=.modversions) ] && \
echo -T $(@:.ko=.modversions)) \
-o $@ $(call lto_ko_objects, $(filter-out FORCE,$^))
+
+ifdef CONFIG_FTRACE_MCOUNT_RECORD
+cmd_ld_ko_o += ; $(objtree)/scripts/recordmcount $(RECORDMCOUNT_FLAGS) $@
+endif
else
cmd_ld_ko_o = \
$(LD) -r $(LDFLAGS) \
@@ -143,6 +143,19 @@ modpost_link()
${LD} ${LDFLAGS} -r -o ${1} $(modversions) ${objects}
}
+# If CONFIG_CLANG_LTO is selected, we postpone running recordmcount until
+# we have compiled LLVM IR to an object file.
+recordmcount()
+{
+ if [ -z "${CONFIG_CLANG_LTO}" ]; then
+ return
+ fi
+
+ if [ -n "${CONFIG_FTRACE_MCOUNT_RECORD}" ]; then
+ scripts/recordmcount ${RECORDMCOUNT_FLAGS} $*
+ fi
+}
+
# Link of vmlinux
# ${1} - optional extra .o files
# ${2} - output file
@@ -327,6 +340,9 @@ if [ -n "${CONFIG_CLANG_LTO}" ]; then
KBUILD_VMLINUX_INIT=
KBUILD_VMLINUX_MAIN=vmlinux.o
KBUILD_VMLINUX_LIBS=
+
+ # Call recordmcount if needed
+ recordmcount vmlinux.o
fi
kallsymso=""
@@ -420,7 +420,8 @@ is_mcounted_section_name(char const *const txtname)
strcmp(".softirqentry.text", txtname) == 0 ||
strcmp(".kprobes.text", txtname) == 0 ||
strcmp(".cpuidle.text", txtname) == 0 ||
- strcmp(".text.unlikely", txtname) == 0;
+ (strncmp(".text.", txtname, 6) == 0 &&
+ strcmp(".text..ftrace", txtname) != 0);
}
/* 32 bit and 64 bit are very similar */
With CONFIG_CLANG_LTO enabled, LLVM IR won't be compiled into object files until modpost_link. This change postpones calls to recordmcount until after this step. In order to exclude ftrace_process_locs from inspection, we add a new code section .text..ftrace, which we tell recordmcount to ignore, and a __norecordmcount attribute for moving functions to this section. Signed-off-by: Sami Tolvanen <samitolvanen@google.com> --- arch/Kconfig | 2 +- include/asm-generic/vmlinux.lds.h | 2 +- include/linux/compiler-clang.h | 7 +++++++ include/linux/compiler.h | 4 ++++ kernel/trace/ftrace.c | 6 +++--- scripts/Makefile.build | 14 +++++++++++++- scripts/Makefile.modpost | 4 ++++ scripts/link-vmlinux.sh | 16 ++++++++++++++++ scripts/recordmcount.c | 3 ++- 9 files changed, 51 insertions(+), 7 deletions(-)