Re: [RFC PATCH v2 13/15] arm64: mm: Guard page table writes with kpkeys

[Qi Zheng <zhengqi.arch@bytedance.com>]

Hi Kevin,

On 2025/1/8 18:32, Kevin Brodsky wrote:
> When CONFIG_KPKEYS_HARDENED_PGTABLES is enabled, page tables (both
> user and kernel) are mapped with a privileged pkey in the linear
> mapping. As a result, they can only be written under the
> kpkeys_hardened_pgtables guard, which sets POR_EL1 appropriately to
> allow such writes.
> 
> Use this guard wherever page tables genuinely need to be written,
> keeping its scope as small as possible (so that POR_EL1 is reset as
> fast as possible). Where atomics are involved, the guard's scope
> encompasses the whole loop to avoid switching POR_EL1 unnecessarily.
> 
> This patch is a no-op if CONFIG_KPKEYS_HARDENED_PGTABLES is disabled
> (default).
> 
> Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
> ---
>   arch/arm64/include/asm/pgtable.h | 19 +++++++++++++++++--
>   arch/arm64/mm/fault.c            |  2 ++
>   2 files changed, 19 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> index f8dac6673887..0d60a49dc234 100644
> --- a/arch/arm64/include/asm/pgtable.h
> +++ b/arch/arm64/include/asm/pgtable.h
> @@ -39,6 +39,7 @@
>   #include <linux/mm_types.h>
>   #include <linux/sched.h>
>   #include <linux/page_table_check.h>
> +#include <linux/kpkeys.h>
>   
>   #ifdef CONFIG_TRANSPARENT_HUGEPAGE
>   #define __HAVE_ARCH_FLUSH_PMD_TLB_RANGE
> @@ -314,6 +315,7 @@ static inline pte_t pte_clear_uffd_wp(pte_t pte)
>   
>   static inline void __set_pte_nosync(pte_t *ptep, pte_t pte)
>   {
> +	guard(kpkeys_hardened_pgtables)();
>   	WRITE_ONCE(*ptep, pte);
>   }
>   
> @@ -758,6 +760,7 @@ static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
>   	}
>   #endif /* __PAGETABLE_PMD_FOLDED */
>   
> +	guard(kpkeys_hardened_pgtables)();
>   	WRITE_ONCE(*pmdp, pmd);
>   
>   	if (pmd_valid(pmd)) {

I noticed a long time ago that set_pte/set_pmd/... was implemented
separately by each architecture without a unified entry point. This
makes it difficult to add some hooks for them.

Taking set_pte() as an example, is it possible to do the following:

1) add a generic set_pte() in include/asm-generic/tlb.h (Or other more
    appropriate files)

static inline void set_pte(pte_t *ptep, pte_t pte)
{
	arch_set_pte(ptep, pte);
}

2) let each architecture include this file and rename the original
    set_pte() to arch_set_pte().

3) then we can add hooks for generic set_pte():

static inline void set_pte(pte_t *ptep, pte_t pte)
{
	guard(kpkeys_hardened_pgtables)();
	arch_set_pte(ptep, pte);
}

4) in this way, the architecture that supports
    ARCH_HAS_KPKEYS_HARDENED_PGTABLES only needs to implement
    the kpkeys_hardened_pgtables(), otherwise it is no-op.

Just some immature ideas, and the related set/clear interfaces
are currently quite messy. ;)

Of course, this does not affect the feature to be implemented
in this patch series.

Thanks!