[PATCH] arm64: Enforce BBM for huge IO/VMAP mappings

[Ard Biesheuvel]

On 21 February 2018 at 12:59, Will Deacon <will.deacon at arm.com> wrote:
> ioremap_page_range doesn't honour break-before-make and attempts to put
> down huge mappings (using p*d_set_huge) over the top of pre-existing
> table entries. This leads to us leaking page table memory and also gives
> rise to TLB conflicts and spurious aborts, which have been seen in
> practice on Cortex-A75.
>
> Until this has been resolved, refuse to put block mappings when the
> existing entry is found to be present.
>
> Cc: Ard Biesheuvel <ard.biesheuvel at linaro.org>
> Reported-by: Hanjun Guo <hanjun.guo at linaro.org>
> Reported-by: Lei Li <lious.lilei at hisilicon.com>
> Signed-off-by: Will Deacon <will.deacon at arm.com>
> ---
>  arch/arm64/mm/mmu.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index 3161b853f29e..84a019f55022 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -933,6 +933,11 @@ int pud_set_huge(pud_t *pudp, phys_addr_t phys, pgprot_t prot)
>  {
>         pgprot_t sect_prot = __pgprot(PUD_TYPE_SECT |
>                                         pgprot_val(mk_sect_prot(prot)));
> +
> +       /* ioremap_page_range doesn't honour BBM */
> +       if (pud_present(READ_ONCE(*pudp)))
> +               return 0;
> +
>         BUG_ON(phys & ~PUD_MASK);
>         set_pud(pudp, pfn_pud(__phys_to_pfn(phys), sect_prot));
>         return 1;
> @@ -942,6 +947,11 @@ int pmd_set_huge(pmd_t *pmdp, phys_addr_t phys, pgprot_t prot)
>  {
>         pgprot_t sect_prot = __pgprot(PMD_TYPE_SECT |
>                                         pgprot_val(mk_sect_prot(prot)));
> +
> +       /* ioremap_page_range doesn't honour BBM */
> +       if (pmd_present(READ_ONCE(*pmdp)))
> +               return 0;
> +
>         BUG_ON(phys & ~PMD_MASK);
>         set_pmd(pmdp, pfn_pmd(__phys_to_pfn(phys), sect_prot));
>         return 1;

This is a nice way of dealing with this IMO, as large mappings that
are only created at boot (e.g., for the ECAM space or gfx BARs) will
still use huge pages.

Acked-by: Ard Biesheuvel <ard.biesheuvel at linaro.org>