Re: [PATCH 02/31] arm64: fix the flush_icache_range arguments in machine_kexec

[Will Deacon]

[+James and Catalin]

On Sun, May 10, 2020 at 09:54:41AM +0200, Christoph Hellwig wrote:
> The second argument is the end "pointer", not the length.
> 
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
>  arch/arm64/kernel/machine_kexec.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm64/kernel/machine_kexec.c b/arch/arm64/kernel/machine_kexec.c
> index 8e9c924423b4e..a0b144cfaea71 100644
> --- a/arch/arm64/kernel/machine_kexec.c
> +++ b/arch/arm64/kernel/machine_kexec.c
> @@ -177,6 +177,7 @@ void machine_kexec(struct kimage *kimage)
>  	 * the offline CPUs. Therefore, we must use the __* variant here.
>  	 */
>  	__flush_icache_range((uintptr_t)reboot_code_buffer,
> +			     (uintptr_t)reboot_code_buffer +
>  			     arm64_relocate_new_kernel_size);

Urgh, well spotted. It's annoyingly different from __flush_dcache_area().

But now I'm wondering what this code actually does... the loop condition
in invalidate_icache_by_line works with 64-bit arithmetic, so we could
spend a /very/ long time here afaict. It's also a bit annoying that we
do a bunch of redundant D-cache maintenance too.

Should we use invalidate_icache_range() here instead? (and why does that
thing need to toggle uaccess)? Argh, too many questions!

Will