arm64: Invalidate the TLB when replacing pmd entries during boot
authorCatalin Marinas <catalin.marinas@arm.com>
Tue, 4 Feb 2014 16:01:31 +0000 (16:01 +0000)
committerCatalin Marinas <catalin.marinas@arm.com>
Wed, 5 Feb 2014 10:30:51 +0000 (10:30 +0000)
With the 64K page size configuration, __create_page_tables in head.S
maps enough memory to get started but using 64K pages rather than 512M
sections with a single pgd/pud/pmd entry pointing to a pte table.
create_mapping() may override the pgd/pud/pmd table entry with a block
(section) one if the RAM size is more than 512MB and aligned correctly.
For the end of this block to be accessible, the old TLB entry must be
invalidated.

Cc: <stable@vger.kernel.org>
Reported-by: Mark Salter <msalter@redhat.com>
Tested-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
arch/arm64/mm/mmu.c

index f557ebbe7013edb71599570f0e4904a318b9dafc..f8dc7e8fce6fea147a9842824a8046822b70b180 100644 (file)
@@ -203,10 +203,18 @@ static void __init alloc_init_pmd(pud_t *pud, unsigned long addr,
        do {
                next = pmd_addr_end(addr, end);
                /* try section mapping first */
-               if (((addr | next | phys) & ~SECTION_MASK) == 0)
+               if (((addr | next | phys) & ~SECTION_MASK) == 0) {
+                       pmd_t old_pmd =*pmd;
                        set_pmd(pmd, __pmd(phys | prot_sect_kernel));
-               else
+                       /*
+                        * Check for previous table entries created during
+                        * boot (__create_page_tables) and flush them.
+                        */
+                       if (!pmd_none(old_pmd))
+                               flush_tlb_all();
+               } else {
                        alloc_init_pte(pmd, addr, next, __phys_to_pfn(phys));
+               }
                phys += next - addr;
        } while (pmd++, addr = next, addr != end);
 }