Add a new callback 'ptep_prepare_range' to allow the architecture
code to optimize the modification of multiple page table entries.
The background for the callback is an instruction found on s390.
The IPTE-range instruction can be used to invalidate up to 256 ptes
with a single IPI, including the flush of the TLB entries associated
to the address range.
This has similarities to the arch_[enter|leave]_lazy_mmu_mode, but for
a more specific situation. ptep_prepare_range is called for the update
of a block of ptes.
ptep_prepare_range is called optimistically, the callback may choose
to do nothing. In this case the individual single pte operation and
the arch_[enter|leave]_lazy_mmu_mode mechanics need to deal with the
invalidation and the associated TLB flush.
Signed-off-by: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
---
include/asm-generic/pgtable.h | 4 ++++
mm/memory.c | 2 ++
mm/mprotect.c | 1 +
mm/mremap.c | 1 +
4 files changed, 8 insertions(+)
diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h
index 9401f48..b29f360 100644
--- a/include/asm-generic/pgtable.h
+++ b/include/asm-generic/pgtable.h
@@ -192,6 +192,10 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addres
}
#endif
+#ifndef ptep_prepare_range
+#define ptep_prepare_range(mm, start, end, ptep, full) do {} while (0)
+#endif
+
#ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
static inline void pmdp_set_wrprotect(struct mm_struct *mm,
diff --git a/mm/memory.c b/mm/memory.c
index 07493e3..eeecb92 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -934,6 +934,7 @@ again:
orig_src_pte = src_pte;
orig_dst_pte = dst_pte;
arch_enter_lazy_mmu_mode();
+ ptep_prepare_range(src_mm, addr, end, src_pte, 0);
do {
/*
@@ -1114,6 +1115,7 @@ again:
start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
pte = start_pte;
arch_enter_lazy_mmu_mode();
+ ptep_prepare_range(mm, addr, end, pte, tlb->fullmm);
do {
pte_t ptent = *pte;
if (pte_none(ptent)) {
diff --git a/mm/mprotect.c b/mm/mprotect.c
index b650c54..3fa15b5 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -74,6 +74,7 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
return 0;
arch_enter_lazy_mmu_mode();
+ ptep_prepare_range(mm, addr, end, pte, 0);
do {
oldpte = *pte;
if (pte_present(oldpte)) {
diff --git a/mm/mremap.c b/mm/mremap.c
index 3fa0a467..5f4d0af 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -135,6 +135,7 @@ static void move_ptes(struct vm_area_struct *vma, pmd_t *old_pmd,
if (new_ptl != old_ptl)
spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
arch_enter_lazy_mmu_mode();
+ ptep_prepare_range(mm, old_addr, old_end, old_pte, 0);
for (; old_addr < old_end; old_pte++, old_addr += PAGE_SIZE,
new_pte++, new_addr += PAGE_SIZE) {
--
2.6.6
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>