[PATCH v2 04/20] mm: VMA sequence count

[Laurent Dufour]

From: Peter Zijlstra <peterz at infradead.org>

Wrap the VMA modifications (vma_adjust/unmap_page_range) with sequence
counts such that we can easily test if a VMA is changed.

The unmap_page_range() one allows us to make assumptions about
page-tables; when we find the seqcount hasn't changed we can assume
page-tables are still valid.

The flip side is that we cannot distinguish between a vma_adjust() and
the unmap_page_range() -- where with the former we could have
re-checked the vma bounds against the address.

Signed-off-by: Peter Zijlstra (Intel) <peterz at infradead.org>

[Port to 4.12 kernel]
[Fix lock dependency between mapping->i_mmap_rwsem and vma->vm_sequence]
Signed-off-by: Laurent Dufour <ldufour at linux.vnet.ibm.com>
---
 include/linux/mm_types.h |  1 +
 mm/memory.c              |  2 ++
 mm/mmap.c                | 21 ++++++++++++++++++---
 3 files changed, 21 insertions(+), 3 deletions(-)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 3cadee0a3508..642aad26b32f 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -342,6 +342,7 @@ struct vm_area_struct {
 	struct mempolicy *vm_policy;	/* NUMA policy for the VMA */
 #endif
 	struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
+	seqcount_t vm_sequence;
 } __randomize_layout;
 
 struct core_thread {
diff --git a/mm/memory.c b/mm/memory.c
index fa598889eb0e..4a2736fe2ef6 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1408,6 +1408,7 @@ void unmap_page_range(struct mmu_gather *tlb,
 	unsigned long next;
 
 	BUG_ON(addr >= end);
+	write_seqcount_begin(&vma->vm_sequence);
 	tlb_start_vma(tlb, vma);
 	pgd = pgd_offset(vma->vm_mm, addr);
 	do {
@@ -1417,6 +1418,7 @@ void unmap_page_range(struct mmu_gather *tlb,
 		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
 	} while (pgd++, addr = next, addr != end);
 	tlb_end_vma(tlb, vma);
+	write_seqcount_end(&vma->vm_sequence);
 }
 
 
diff --git a/mm/mmap.c b/mm/mmap.c
index f19efcf75418..140b22136cb7 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -557,6 +557,8 @@ void __vma_link_rb(struct mm_struct *mm, struct vm_area_struct *vma,
 	else
 		mm->highest_vm_end = vm_end_gap(vma);
 
+	seqcount_init(&vma->vm_sequence);
+
 	/*
 	 * vma->vm_prev wasn't known when we followed the rbtree to find the
 	 * correct insertion point for that vma. As a result, we could not
@@ -798,6 +800,11 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 		}
 	}
 
+	write_seqcount_begin(&vma->vm_sequence);
+	if (next && next != vma)
+		write_seqcount_begin_nested(&next->vm_sequence,
+					    SINGLE_DEPTH_NESTING);
+
 	anon_vma = vma->anon_vma;
 	if (!anon_vma && adjust_next)
 		anon_vma = next->anon_vma;
@@ -902,6 +909,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 		mm->map_count--;
 		mpol_put(vma_policy(next));
 		kmem_cache_free(vm_area_cachep, next);
+		write_seqcount_end(&next->vm_sequence);
 		/*
 		 * In mprotect's case 6 (see comments on vma_merge),
 		 * we must remove another next too. It would clutter
@@ -931,11 +939,14 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 		if (remove_next == 2) {
 			remove_next = 1;
 			end = next->vm_end;
+			write_seqcount_end(&vma->vm_sequence);
 			goto again;
-		}
-		else if (next)
+		} else if (next) {
+			if (next != vma)
+				write_seqcount_begin_nested(&next->vm_sequence,
+							    SINGLE_DEPTH_NESTING);
 			vma_gap_update(next);
-		else {
+		} else {
 			/*
 			 * If remove_next == 2 we obviously can't
 			 * reach this path.
@@ -961,6 +972,10 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 	if (insert && file)
 		uprobe_mmap(insert);
 
+	if (next && next != vma)
+		write_seqcount_end(&next->vm_sequence);
+	write_seqcount_end(&vma->vm_sequence);
+
 	validate_mm(mm);
 
 	return 0;
-- 
2.7.4