Re: [PATCH] arm64: remove unneeded BUILD_BUG_ON assertion

[Mark Rutland]

On Fri, Feb 02, 2024 at 12:02:11PM +0800, Dawei Li wrote:
> Since commit c02433dd6de3 ("arm64: split thread_info from task stack"),
> CONFIG_THREAD_INFO_IN_TASK is enabled unconditionally for arm64. So
> remove this always-true assertion from arch_dup_task_struct.
> 
> Signed-off-by: Dawei Li <dawei.li@xxxxxxxxxxxx>
> ---
>  arch/arm64/kernel/process.c | 3 ---
>  1 file changed, 3 deletions(-)
> 
> diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
> index 7387b68c745b..4ae31b7af6c3 100644
> --- a/arch/arm64/kernel/process.c
> +++ b/arch/arm64/kernel/process.c
> @@ -290,9 +290,6 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
>  		fpsimd_preserve_current_state();
>  	*dst = *src;
>  
> -	/* We rely on the above assignment to initialize dst's thread_flags: */
> -	BUILD_BUG_ON(!IS_ENABLED(CONFIG_THREAD_INFO_IN_TASK));
> -

Does the above cause any problem today, or is this patch just a cleanup?

For the benefit of other reviewers, the assertion and comment were added in
commit:

  4585fc59c0e8 ("arm64/sve: Fix wrong free for task->thread.sve_state")

.. back in 2019, 3 years after commit:

  c02433dd6de3 ("arm64: split thread_info from task stack")

The comment and assertion were a safety-net for backports, since commit
4585fc59c0e8 was a fix which dependend upon the thread_info being contained
within task_struct, and couldn't be backported to kernels without
CONFIG_THREAD_INFO_IN_TASK.

I'm sure that we currently have plenty of other code with a similar (but
undocumented) dependency. Given we've unconditionally selected
CONFIG_THREAD_INFO_IN_TASK since v4.10, and the oldest longterm stable kernel
is v4.19 (with v4.14 having EOL'd last month), I think it makes sense to delete
the assertion and comment.

So FWIW:

Acked-by: Mark Rutland <mark.rutland@xxxxxxx>

Mark.