On Tue, May 26, 2026 at 07:59:00PM +0200, Ard Biesheuvel wrote:
> From: Ard Biesheuvel <[email protected]>
> 
> SH performs cache maintenance on the zero page during boot, presumably
> to ensure that any clearing of BSS that has occurred at startup is
> visible to other CPUs and DMA devices.

It flushes the cache there because before 6215d9f4470f ("arch, mm:
consolidate empty_zero_page") empty_zero_page used was for boot parameters
at the very early boot.

I removed memset(empty_zero_page, 0, PAGE_SIZE) that was a line above but
missed the cache flush. 

This whole "clear zero-page" can be removed IMO.
 
> The __flush_wback_region() function takes a void* argument, which is
> conceptually sound, but given that empty_zero_page[] must never be
> modified, it is being repainted as const, making it incompatible with a
> void* formal parameter.
> 
> Given the above, and the fact that __flush_wback_region() is in fact a
> function pointer variable with multiple implementations, take the easy
> way out, and cast away the constness in this particular invocation.
> 
> Cc: Yoshinori Sato <[email protected]>
> Cc: Rich Felker <[email protected]>
> Cc: John Paul Adrian Glaubitz <[email protected]>
> Signed-off-by: Ard Biesheuvel <[email protected]>
> ---
>  arch/sh/mm/init.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/sh/mm/init.c b/arch/sh/mm/init.c
> index 4e40d5e96be9..acbb481cdbfe 100644
> --- a/arch/sh/mm/init.c
> +++ b/arch/sh/mm/init.c
> @@ -332,7 +332,7 @@ void __init mem_init(void)
>       cpu_cache_init();
>  
>       /* clear the zero-page */
> -     __flush_wback_region(empty_zero_page, PAGE_SIZE);
> +     __flush_wback_region((void *)empty_zero_page, PAGE_SIZE);
>  
>       vsyscall_init();
>  
> -- 
> 2.54.0.794.g4f17f83d09-goog
> 

-- 
Sincerely yours,
Mike.

Reply via email to