[tip: core/urgent] smpboot: Mark idle_init() as __always_inlined to work around aggressive compiler un-inlining

From: tip-bot2 for Ingo Molnar
Date: Tue Jul 13 2021 - 00:36:09 EST


The following commit has been merged into the core/urgent branch of tip:

Commit-ID: e9ba16e68cce2f85e9f5d2eba5c0453f1a741fd2
Gitweb: https://git.kernel.org/tip/e9ba16e68cce2f85e9f5d2eba5c0453f1a741fd2
Author: Ingo Molnar <mingo@xxxxxxxxxx>
AuthorDate: Sun, 11 Jul 2021 08:26:45 +02:00
Committer: Ingo Molnar <mingo@xxxxxxxxxx>
CommitterDate: Tue, 13 Jul 2021 06:32:59 +02:00

smpboot: Mark idle_init() as __always_inlined to work around aggressive compiler un-inlining

While this function is a static inline, and is only used once in
local scope, certain Kconfig variations may cause it to be compiled
as a standalone function:

89231bf0 <idle_init>:
89231bf0: 83 05 60 d9 45 89 01 addl $0x1,0x8945d960
89231bf7: 55 push %ebp

Resulting in this build failure:

WARNING: modpost: vmlinux.o(.text.unlikely+0x7fd5): Section mismatch in reference from the function idle_init() to the function .init.text:fork_idle()
The function idle_init() references
the function __init fork_idle().
This is often because idle_init lacks a __init
annotation or the annotation of fork_idle is wrong.
ERROR: modpost: Section mismatches detected.

Certain USBSAN options x86-32 builds with CONFIG_CC_OPTIMIZE_FOR_SIZE=y
seem to be causing this.

So mark idle_init() as __always_inline to work around this compiler
bug/feature.

Signed-off-by: Ingo Molnar <mingo@xxxxxxxxxx>
---
kernel/smpboot.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/smpboot.c b/kernel/smpboot.c
index e416304..21b7953 100644
--- a/kernel/smpboot.c
+++ b/kernel/smpboot.c
@@ -47,7 +47,7 @@ void __init idle_thread_set_boot_cpu(void)
*
* Creates the thread if it does not exist.
*/
-static inline void idle_init(unsigned int cpu)
+static inline void __always_inline idle_init(unsigned int cpu)
{
struct task_struct *tsk = per_cpu(idle_threads, cpu);