[PATCH 04/11] objtool: Fix false positive warnings related to sibling calls

From: Josh Poimboeuf
Date: Wed Mar 09 2016 - 01:08:28 EST


With some configs [1], objtool prints a bunch of false positive warnings
like:

arch/x86/events/core.o: warning: objtool: x86_del_exclusive()+0x0: frame pointer state mismatch

For some reason this config has a bunch of sibling calls. When objtool
follows a sibling call jump, it attempts to compare the frame pointer
state. But it also accidentally compares the FENTRY state, resulting in
a false positive warning.

[1] https://lkml.kernel.org/r/20160308154909.GA20956@xxxxxxxxx

Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
---
tools/objtool/builtin-check.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/tools/objtool/builtin-check.c b/tools/objtool/builtin-check.c
index 51da270..fe24804 100644
--- a/tools/objtool/builtin-check.c
+++ b/tools/objtool/builtin-check.c
@@ -729,6 +729,11 @@ static bool has_valid_stack_frame(struct instruction *insn)
(insn->state & STATE_FP_SETUP);
}

+static unsigned int frame_state(unsigned long state)
+{
+ return (state & (STATE_FP_SAVED | STATE_FP_SETUP));
+}
+
/*
* Follow the branch starting at the given instruction, and recursively follow
* any other branches (jumps). Meanwhile, track the frame pointer state at
@@ -756,7 +761,7 @@ static int validate_branch(struct objtool_file *file,

while (1) {
if (insn->visited) {
- if (insn->state != state) {
+ if (frame_state(insn->state) != frame_state(state)) {
WARN_FUNC("frame pointer state mismatch",
sec, insn->offset);
warnings++;
--
2.4.3