Re: [PATCH] [RFC]arm64:Mark __stack_chk_guard as __ro_after_init

From: Mark Rutland
Date: Tue Sep 14 2021 - 06:17:19 EST


On Tue, Sep 14, 2021 at 05:44:02PM +0800, Dan Li wrote:
> __stack_chk_guard is setup once while init stage and never changed
> after that.
>
> Although the modification of this variable at runtime will usually
> cause the kernel to crash (so dose the attacker), it should be marked
> as _ro_after_init, and it should not affect performance if it is
> placed in the ro_after_init section.
>
> This should also be the case on the ARM platform, or am I missing
> something?
>
> Signed-off-by: Dan Li <ashimida@xxxxxxxxxxxxxxxxx>

FWIW, this makes sense to me:

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

Looking at the history, this was added to arm64 in commit:

c0c264ae5112d1cd ("arm64: Add CONFIG_CC_STACKPROTECTOR")

... whereas __ro_after_init was introduced around 2 years later in
commit:

c74ba8b3480da6dd ("arch: Introduce post-init read-only memory")

... so we weren't deliberately avoiding __ro_after_init, and there are
probably a significant number of other variables we could apply it to.

Mark.

> ---
> arch/arm64/kernel/process.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
> index c8989b9..c858b85 100644
> --- a/arch/arm64/kernel/process.c
> +++ b/arch/arm64/kernel/process.c
> @@ -60,7 +60,7 @@
>
> #if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_STACKPROTECTOR_PER_TASK)
> #include <linux/stackprotector.h>
> -unsigned long __stack_chk_guard __read_mostly;
> +unsigned long __stack_chk_guard __ro_after_init;
> EXPORT_SYMBOL(__stack_chk_guard);
> #endif
>
> --
> 2.7.4
>