[PATCH] powerpc/64: Use -mprofile-kernel for big endian ELFv2 kernels

-mprofile-kernel is an optimised calling convention for mcount that
Linux  has only implemented with the ELFv2 ABI, so it was disabled for
big endian kernels. However it does work with ELFv2 big endian, so let's
allow that if the compiler supports it.

Cc: Naveen N. Rao <redacted>
Suggested-by: Christophe Leroy <redacted>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
Christophe had the good idea that we could use -mprofile-kernel for
ELFv2 BE. Unfortunately can't remove -pg due to lack of -mprofile-kernel
in clang, but this gives BE the nicer ftrace code with GCC at least.
Function tracer works for me with a BE kernel.

Thanks,
Nick

 arch/powerpc/Kconfig                            | 6 ++++--
 arch/powerpc/tools/gcc-check-mprofile-kernel.sh | 4 ++--
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index a64bfd9b8a1d..bd2ee7af1342 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig

@@ -547,8 +547,10 @@ config LD_HEAD_STUB_CATCH
 	  If unsure, say "N".
 
 config MPROFILE_KERNEL
-	depends on PPC64 && CPU_LITTLE_ENDIAN && FUNCTION_TRACER
-	def_bool $(success,$(srctree)/arch/powerpc/tools/gcc-check-mprofile-kernel.sh $(CC) -I$(srctree)/include -D__KERNEL__)
+	depends on PPC64 && FUNCTION_TRACER
+	depends on CPU_LITTLE_ENDIAN || PPC64_BIG_ENDIAN_ELF_ABI_V2
+	def_bool $(success,$(srctree)/arch/powerpc/tools/gcc-check-mprofile-kernel.sh $(CC) -mlittle-endian) if CPU_LITTLE_ENDIAN
+	def_bool $(success,$(srctree)/arch/powerpc/tools/gcc-check-mprofile-kernel.sh $(CC) -mbig-endian) if CPU_BIG_ENDIAN
 
 config HOTPLUG_CPU
 	bool "Support for enabling/disabling CPUs"

diff --git a/arch/powerpc/tools/gcc-check-mprofile-kernel.sh b/arch/powerpc/tools/gcc-check-mprofile-kernel.sh
index 137f3376ac2b..e78c599251ff 100755
--- a/arch/powerpc/tools/gcc-check-mprofile-kernel.sh
+++ b/arch/powerpc/tools/gcc-check-mprofile-kernel.sh

@@ -14,13 +14,13 @@ set -o pipefail
 # Test whether the compile option -mprofile-kernel exists and generates
 # profiling code (ie. a call to _mcount()).
 echo "int func() { return 0; }" | \
-    $* -m64 -mlittle-endian -S -x c -O2 -p -mprofile-kernel - -o - \
+    $* -m64 -S -x c -O2 -p -mprofile-kernel - -o - \
     2> /dev/null | grep -q "_mcount"
 
 # Test whether the notrace attribute correctly suppresses calls to _mcount().
 
 echo -e "#include <linux/compiler.h>\nnotrace int func() { return 0; }" | \
-    $* -m64 -mlittle-endian -S -x c -O2 -p -mprofile-kernel - -o - \
+    $* -m64 -S -x c -O2 -p -mprofile-kernel - -o - \
     2> /dev/null | grep -q "_mcount" && \
     exit 1
 

-- 
2.40.1

Nicholas Piggin wrote:

-mprofile-kernel is an optimised calling convention for mcount that
Linux  has only implemented with the ELFv2 ABI, so it was disabled for
big endian kernels. However it does work with ELFv2 big endian, so let's
allow that if the compiler supports it.

Cc: Naveen N. Rao <redacted>
Suggested-by: Christophe Leroy <redacted>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
Christophe had the good idea that we could use -mprofile-kernel for
ELFv2 BE. Unfortunately can't remove -pg due to lack of -mprofile-kernel
in clang, but this gives BE the nicer ftrace code with GCC at least.
Function tracer works for me with a BE kernel.

LGTM. With a few minor nits below:
Acked-by: Naveen N. Rao <redacted>

Can't we just check for PPC64_ELF_ABI_V2?

We have this comment before the below code, which should also be updated/removed:
  # -mprofile-kernel is only supported on 64le, so this should not be invoked
  # for other targets. Therefore we can pass in -m64 and -mlittle-endian
  # explicitly, to take care of toolchains defaulting to other targets.

 # Test whether the compile option -mprofile-kernel exists and generates
 # profiling code (ie. a call to _mcount()).
 echo "int func() { return 0; }" | \
-    $* -m64 -mlittle-endian -S -x c -O2 -p -mprofile-kernel - -o - \
+    $* -m64 -S -x c -O2 -p -mprofile-kernel - -o - \
     2> /dev/null | grep -q "_mcount"

 # Test whether the notrace attribute correctly suppresses calls to _mcount().

 echo -e "#include <linux/compiler.h>\nnotrace int func() { return 0; }" | \
-    $* -m64 -mlittle-endian -S -x c -O2 -p -mprofile-kernel - -o - \
+    $* -m64 -S -x c -O2 -p -mprofile-kernel - -o - \
     2> /dev/null | grep -q "_mcount" && \
     exit 1

- Naveen