Re: [PATCH 1/1] hung_task: fix objtool sibling call warning in watchdog()

[Nathan Chancellor <nathan@kernel.org>]

On Thu, Jul 31, 2025 at 10:39:34AM +0800, Lance Yang wrote:
> From: Lance Yang <lance.yang@linux.dev>
> 
> The kernel test robot reported an objtool warning on the loongarch
> architecture with clang:
> 
>         vmlinux.o: warning: objtool: watchdog+0x418: sibling call from
> callable instruction with modified stack frame
> 
> To resolve this, explicitly prevent the compiler from inlining or
> performing a tail call on check_hung_uninterruptible_tasks() by marking
> it with the noinline attribute. This ensures a standard function call
> with a proper stack frame, satisfying objtool's validation requirements.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202507301256.cZlxQ10s-lkp@intel.com/
> Signed-off-by: Lance Yang <lance.yang@linux.dev>

While this may solve that particular instance of the warning, the fact
there are many, many more with that configuration indicates that there
will need to be a more generalized fix. I have started a separate thread
with some LoongArch folks around this issue:

https://lore.kernel.org/20250731175655.GA1455142@ax162/

Cheers,
Nathan

> ---
>  kernel/hung_task.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
> index 8708a1205f82..a5b5a0a2c262 100644
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -283,7 +283,7 @@ static bool rcu_lock_break(struct task_struct *g, struct task_struct *t)
>   * a really long time (120 seconds). If that happens, print out
>   * a warning.
>   */
> -static void check_hung_uninterruptible_tasks(unsigned long timeout)
> +static noinline void check_hung_uninterruptible_tasks(unsigned long timeout)
>  {
>  	int max_count = sysctl_hung_task_check_count;
>  	unsigned long last_break = jiffies;
> -- 
> 2.49.0
> 
>