[PATCH 3.18 54/85] branch-check: fix long->int truncation when profiling branches

From: Greg Kroah-Hartman
Date: Sun Jul 01 2018 - 12:15:13 EST

3.18-stable review patch. If anyone has any objections, please let me know.


From: Mikulas Patocka <mpatocka@xxxxxxxxxx>

commit 2026d35741f2c3ece73c11eb7e4a15d7c2df9ebe upstream.

The function __builtin_expect returns long type (see the gcc
documentation), and so do macros likely and unlikely. Unfortunatelly, when
CONFIG_PROFILE_ANNOTATED_BRANCHES is selected, the macros likely and
unlikely expand to __branch_check__ and __branch_check__ truncates the
long type to int. This unintended truncation may cause bugs in various
kernel code (we found a bug in dm-writecache because of it), so it's
better to fix __branch_check__ to return long.

Link: http://lkml.kernel.org/r/alpine.LRH.2.02.1805300818140.24812@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx
Fixes: 1f0d69a9fc815 ("tracing: profile likely and unlikely annotations")
Signed-off-by: Mikulas Patocka <mpatocka@xxxxxxxxxx>
Signed-off-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

include/linux/compiler.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

--- a/include/linux/compiler.h
+++ b/include/linux/compiler.h
@@ -105,7 +105,7 @@ void ftrace_likely_update(struct ftrace_
#define unlikely_notrace(x) __builtin_expect(!!(x), 0)

#define __branch_check__(x, expect) ({ \
- int ______r; \
+ long ______r; \
static struct ftrace_branch_data \
__attribute__((__aligned__(4))) \
__attribute__((section("_ftrace_annotated_branch"))) \