arm64: mm: detect bad __create_mapping uses

If a caller of __create_mapping provides a PA and VA which have
different sub-page offsets, it is not clear which offset they expect to
apply to the mapping, and is indicative of a bad caller.

In some cases, the region we wish to map may validly have a sub-page
offset in the physical and virtual addresses. For example, EFI runtime
regions have 4K granularity, yet may be mapped by a 64K page kernel. So
long as the physical and virtual offsets are the same, the region will
be mapped at the expected VAs.

Disallow calls with differing sub-page offsets, and WARN when they are
encountered, so that we can detect and fix such cases.

Cc: Laura Abbott <[email protected]>
Acked-by: Ard Biesheuvel <[email protected]>
Acked-by: Catalin Marinas <[email protected]>
Reviewed-by: Steve Capper <[email protected]>
Signed-off-by: Mark Rutland <[email protected]>
Signed-off-by: Will Deacon <[email protected]>

diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 873e363048c6c661eda620bbd4f826ba568c5ba7..6e4b4351ce9e69919d3b653c72eb8b0b87501a01 100644 (file)
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -251,6 +251,13 @@ static void  __create_mapping(struct mm_struct *mm, pgd_t *pgd,
 {
        unsigned long addr, length, end, next;
 
+       /*
+        * If the virtual and physical address don't have the same offset
+        * within a page, we cannot map the region as the caller expects.
+        */
+       if (WARN_ON((phys ^ virt) & ~PAGE_MASK))
+               return;
+
        addr = virt & PAGE_MASK;
        length = PAGE_ALIGN(size + (virt & ~PAGE_MASK));