[PATCH 2/2] arm64: Allow changing of attributes outside of modules

[Ard Biesheuvel]

On 3 November 2015 at 22:48, Laura Abbott <labbott at fedoraproject.org> wrote:
>
> Currently, the set_memory_* functions that are implemented for arm64
> are restricted to module addresses only. This was mostly done
> because arm64 maps normal zone memory with larger page sizes to
> improve TLB performance. This has the side effect though of making it
> difficult to adjust attributes at the PAGE_SIZE granularity. There are
> an increasing number of use cases related to security where it is
> necessary to change the attributes of kernel memory. Add functionality
> to the page attribute changing code under a Kconfig to let systems
> designers decide if they want to make the trade off of security for TLB
> pressure.
>
> Signed-off-by: Laura Abbott <labbott at fedoraproject.org>
> ---
>  arch/arm64/Kconfig.debug | 11 +++++++
>  arch/arm64/mm/mm.h       |  3 ++
>  arch/arm64/mm/mmu.c      |  2 +-
>  arch/arm64/mm/pageattr.c | 74 ++++++++++++++++++++++++++++++++++++++++++++----
>  4 files changed, 84 insertions(+), 6 deletions(-)
>
> diff --git a/arch/arm64/Kconfig.debug b/arch/arm64/Kconfig.debug
> index d6285ef..abc6922 100644
> --- a/arch/arm64/Kconfig.debug
> +++ b/arch/arm64/Kconfig.debug
> @@ -89,6 +89,17 @@ config DEBUG_ALIGN_RODATA
>
>           If in doubt, say N
>
> +config DEBUG_CHANGE_PAGEATTR

I don't think this belongs in Kconfig.debug, and I don't think this
should default to off.

We know we currently have no users of set_memory_xx() in arch/arm64
that operate on linear mapping addresses, so we will not introduce any
performance regressions by adding this functionality now. By putting
this feature behind a debug option, every newly introduced call
set_memory_xx() that operates on linear or vmalloc() addresses needs
to deal with -EINVAL (or depend on DEBUG_CHANGE_PAGEATTR), or it may
error out at runtime if the feature is not enabled.

> +       bool "Allow all kernel memory to have attributes changed"
> +       help
> +         If this option is selected, APIs that change page attributes
> +         (RW <-> RO, X <-> NX) will be valid for all memory mapped in
> +         the kernel space. The trade off is that there may be increased
> +         TLB pressure from finer grained page mapping. Turn on this option
> +         if performance is more important than security
> +

This is backwards

> +         If in doubt, say N
> +
>  source "drivers/hwtracing/coresight/Kconfig"
>
>  endmenu

[...]

> diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c
> index e47ed1c..48a4ce9 100644
> --- a/arch/arm64/mm/pageattr.c
> +++ b/arch/arm64/mm/pageattr.c

[...]

> @@ -45,17 +108,18 @@ static int change_memory_common(unsigned long addr, int numpages,
>         int ret;
>         struct page_change_data data;
>
> +       if (addr < PAGE_OFFSET && !is_vmalloc_addr((void *)addr))
> +               return -EINVAL;
> +

Doesn't this exclude the module area?

>         if (!IS_ALIGNED(addr, PAGE_SIZE)) {
>                 start &= PAGE_MASK;
>                 end = start + size;
>                 WARN_ON_ONCE(1);
>         }
>
> -       if (start < MODULES_VADDR || start >= MODULES_END)
> -               return -EINVAL;
> -
> -       if (end < MODULES_VADDR || end >= MODULES_END)
> -               return -EINVAL;
> +       ret = check_address(addr);
> +       if (ret)
> +               return ret;
>
>         data.set_mask = set_mask;
>         data.clear_mask = clear_mask;
> --
> 2.4.3
>