[PATCH v3] mm: huge-vmap: fail gracefully on unexpected huge vmap mappings

[mark.rutland@arm.com (Mark Rutland)]

On Thu, Jun 08, 2017 at 11:35:48AM +0000, Ard Biesheuvel wrote:
> Existing code that uses vmalloc_to_page() may assume that any
> address for which is_vmalloc_addr() returns true may be passed
> into vmalloc_to_page() to retrieve the associated struct page.
> 
> This is not un unreasonable assumption to make, but on architectures
> that have CONFIG_HAVE_ARCH_HUGE_VMAP=y, it no longer holds, and we
> need to ensure that vmalloc_to_page() does not go off into the weeds
> trying to dereference huge PUDs or PMDs as table entries.
> 
> Given that vmalloc() and vmap() themselves never create huge
> mappings or deal with compound pages at all, there is no correct
> value to return in this case, so return NULL instead, and issue a
> warning.
> 
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> ---
> This is a followup to '[PATCH v2] mm: vmalloc: make vmalloc_to_page()
> deal with PMD/PUD mappings', hence the v3. The root issue with /proc/kcore
> on arm64 is now handled by '[PATCH] mm: vmalloc: simplify vread/vwrite to
> use existing mappings' [1], and this patch now only complements it by
> taking care of other vmalloc_to_page() users.
> 
> [0] http://marc.info/?l=linux-mm&m=149641886821855&w=2
> [1] http://marc.info/?l=linux-mm&m=149685966530180&w=2
> 
>  include/linux/hugetlb.h | 13 +++++++++----
>  mm/vmalloc.c            |  5 +++--
>  2 files changed, 12 insertions(+), 6 deletions(-)
> 
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index b857fc8cc2ec..b7166e5426b6 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -121,8 +121,6 @@ struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
>  				pmd_t *pmd, int flags);
>  struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
>  				pud_t *pud, int flags);
> -int pmd_huge(pmd_t pmd);
> -int pud_huge(pud_t pud);
>  unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
>  		unsigned long address, unsigned long end, pgprot_t newprot);
>  
> @@ -150,8 +148,6 @@ static inline void hugetlb_show_meminfo(void)
>  #define follow_huge_pmd(mm, addr, pmd, flags)	NULL
>  #define follow_huge_pud(mm, addr, pud, flags)	NULL
>  #define prepare_hugepage_range(file, addr, len)	(-EINVAL)
> -#define pmd_huge(x)	0
> -#define pud_huge(x)	0
>  #define is_hugepage_only_range(mm, addr, len)	0
>  #define hugetlb_free_pgd_range(tlb, addr, end, floor, ceiling) ({BUG(); 0; })
>  #define hugetlb_fault(mm, vma, addr, flags)	({ BUG(); 0; })
> @@ -190,6 +186,15 @@ static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
>  }
>  
>  #endif /* !CONFIG_HUGETLB_PAGE */
> +
> +#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_HAVE_ARCH_HUGE_VMAP)
> +int pmd_huge(pmd_t pmd);
> +int pud_huge(pud_t pud);
> +#else
> +#define pmd_huge(x)	0
> +#define pud_huge(x)	0
> +#endif
> +
>  /*
>   * hugepages at page global directory. If arch support
>   * hugepages at pgd level, they need to define this.
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 982d29511f92..67e1a304c467 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -12,6 +12,7 @@
>  #include <linux/mm.h>
>  #include <linux/module.h>
>  #include <linux/highmem.h>
> +#include <linux/hugetlb.h>
>  #include <linux/sched/signal.h>
>  #include <linux/slab.h>
>  #include <linux/spinlock.h>
> @@ -287,10 +288,10 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  	if (p4d_none(*p4d))
>  		return NULL;
>  	pud = pud_offset(p4d, addr);
> -	if (pud_none(*pud))
> +	if (pud_none(*pud) || WARN_ON_ONCE(pud_huge(*pud)))
>  		return NULL;
>  	pmd = pmd_offset(pud, addr);
> -	if (pmd_none(*pmd))
> +	if (pmd_none(*pmd) || WARN_ON_ONCE(pmd_huge(*pmd)))
>  		return NULL;

I think it might be better to use p*d_bad() here, since that doesn't
depend on CONFIG_HUGETLB_PAGE.

While the cross-arch semantics are a little fuzzy, my understanding is
those should return true if an entry is not a pointer to a next level of
table (so pXd_huge(p) implies pXd_bad(p)).

Otherwise, this looks sensible to me.

Thanks,
Mark.