[PATCH v3 4/5] arm64: mm: support additional contiguous kernel mapping region sizes
Ard Biesheuvel
ard.biesheuvel at linaro.org
Wed Oct 12 04:23:44 PDT 2016
Extend the basic support for kernel mappings using contiguous regions
by adding support for contiguous PUDs (4k granule only), either as a
discrete level or folded into the PGDs. In the same way, handle folded
PMDs so that contiguous PMDs (for 16k and 64k granule kernels) will
work as expected for 2 levels of translation as well.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel at linaro.org>
---
arch/arm64/include/asm/pgtable-hwdef.h | 6 +++
arch/arm64/mm/mmu.c | 40 +++++++++++++++++++-
2 files changed, 44 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable-hwdef.h b/arch/arm64/include/asm/pgtable-hwdef.h
index eb0c2bd90de9..4192072af932 100644
--- a/arch/arm64/include/asm/pgtable-hwdef.h
+++ b/arch/arm64/include/asm/pgtable-hwdef.h
@@ -93,12 +93,15 @@
#ifdef CONFIG_ARM64_64K_PAGES
#define CONT_PTE_SHIFT 5
#define CONT_PMD_SHIFT 5
+#define CONT_PUD_SHIFT 0
#elif defined(CONFIG_ARM64_16K_PAGES)
#define CONT_PTE_SHIFT 7
#define CONT_PMD_SHIFT 5
+#define CONT_PUD_SHIFT 0
#else
#define CONT_PTE_SHIFT 4
#define CONT_PMD_SHIFT 4
+#define CONT_PUD_SHIFT 4
#endif
#define CONT_PTES (1 << CONT_PTE_SHIFT)
@@ -107,6 +110,9 @@
#define CONT_PMDS (1 << CONT_PMD_SHIFT)
#define CONT_PMD_SIZE (CONT_PMDS * PMD_SIZE)
#define CONT_PMD_MASK (~(CONT_PMD_SIZE - 1))
+#define CONT_PUDS (1 << CONT_PUD_SHIFT)
+#define CONT_PUD_SIZE (CONT_PUDS * PUD_SIZE)
+#define CONT_PUD_MASK (~(CONT_PUD_SIZE - 1))
/* the the numerical offset of the PTE within a range of CONT_PTES */
#define CONT_RANGE_OFFSET(addr) (((addr)>>PAGE_SHIFT)&(CONT_PTES-1))
diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 40be4979102d..0e0eca45b54a 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -233,6 +233,7 @@ static void alloc_init_pud(pgd_t *pgd, unsigned long addr, unsigned long end,
phys_addr_t (*pgtable_alloc)(void),
bool page_mappings_only)
{
+ pgprot_t __prot = prot;
pud_t *pud;
unsigned long next;
@@ -254,7 +255,19 @@ static void alloc_init_pud(pgd_t *pgd, unsigned long addr, unsigned long end,
* For 4K granule only, attempt to put down a 1GB block
*/
if (use_1G_block(addr, next, phys) && !page_mappings_only) {
- pud_set_huge(pud, phys, prot);
+ /*
+ * Set the contiguous bit for the subsequent group of
+ * PUDs if its size and alignment are appropriate.
+ */
+ if (((addr | phys) & ~CONT_PUD_MASK) == 0) {
+ if (end - addr >= CONT_PUD_SIZE)
+ __prot = __pgprot(pgprot_val(prot) |
+ PTE_CONT);
+ else
+ __prot = prot;
+ }
+
+ pud_set_huge(pud, phys, __prot);
/*
* After the PUD entry has been populated once, we
@@ -284,6 +297,7 @@ static void __create_pgd_mapping(pgd_t *pgdir, phys_addr_t phys,
{
unsigned long addr, length, end, next;
pgd_t *pgd = pgd_offset_raw(pgdir, virt);
+ pgprot_t __prot = prot;
/*
* If the virtual and physical address don't have the same offset
@@ -299,7 +313,29 @@ static void __create_pgd_mapping(pgd_t *pgdir, phys_addr_t phys,
end = addr + length;
do {
next = pgd_addr_end(addr, end);
- alloc_init_pud(pgd, addr, next, phys, prot, pgtable_alloc,
+
+ /*
+ * If any intermediate levels are folded into the PGDs, we
+ * need to deal with the contiguous attributes here, since
+ * the contiguity can only be observed at this level.
+ */
+ if (PGDIR_SHIFT == PMD_SHIFT && !page_mappings_only &&
+ ((addr | phys) & ~CONT_PMD_MASK) == 0) {
+ if (end - addr >= CONT_PMD_SIZE)
+ __prot = __pgprot(pgprot_val(prot) |
+ PTE_CONT);
+ else
+ __prot = prot;
+ } else if (PGDIR_SHIFT == PUD_SHIFT && CONT_PUD_SHIFT > 0 &&
+ !page_mappings_only &&
+ ((addr | phys) & ~CONT_PUD_MASK) == 0) {
+ if (end - addr >= CONT_PUD_SIZE)
+ __prot = __pgprot(pgprot_val(prot) |
+ PTE_CONT);
+ else
+ __prot = prot;
+ }
+ alloc_init_pud(pgd, addr, next, phys, __prot, pgtable_alloc,
page_mappings_only);
phys += next - addr;
} while (pgd++, addr = next, addr != end);
--
2.7.4
More information about the linux-arm-kernel
mailing list