[PATCH] rmap 22 flush_dcache_mmap_lock

From: Hugh Dickins <hugh@veritas.com>

arm and parisc __flush_dcache_page have been scanning the i_mmap(_shared) list
without locking or disabling preemption.  That may be even more unsafe now
it's a prio tree instead of a list.

It looks like we cannot use i_shared_lock for this protection: most uses of
flush_dcache_page are okay, and only one would need lock ordering fixed
(get_user_pages holds page_table_lock across flush_dcache_page); but there's a
few (e.g.  in net and ntfs) which look as if they're using it in I/O
completion - and it would be restrictive to disallow it there.

So, on arm and parisc only, define flush_dcache_mmap_lock(mapping) as
spin_lock_irq(&(mapping)->tree_lock); on i386 (and other arches left to the
next patch) define it away to nothing; and use where needed.

While updating locking hierarchy in filemap.c, remove two layers of the fossil
record from add_to_page_cache comment: no longer used for swap.

I believe all the #includes will work out, but have only built i386.  I can
see several things about this patch which might cause revulsion: the name
flush_dcache_mmap_lock?  the reuse of the page radix_tree's tree_lock for this
different purpose?  spin_lock_irqsave instead?  can't we somehow get
i_shared_lock to handle the problem?

9 files changed, 36 insertions, 3 deletions

diff --git a/arch/arm/mm/fault-armv.c b/arch/arm/mm/fault-armv.c
index 5ff9c9f8323f..81485b5b4de4 100644
--- a/arch/arm/mm/fault-armv.c
+++ b/arch/arm/mm/fault-armv.c
@@ -94,6 +94,8 @@ void __flush_dcache_page(struct page *page)
 	 * and invalidate any user data.
 	 */
 	pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
+
+	flush_dcache_mmap_lock(mapping);
 	while ((mpnt = vma_prio_tree_next(mpnt, &mapping->i_mmap,
 					&iter, pgoff, pgoff)) != NULL) {
 		/*
@@ -106,6 +108,7 @@ void __flush_dcache_page(struct page *page)
 		offset = (pgoff - mpnt->vm_pgoff) << PAGE_SHIFT;
 		flush_cache_page(mpnt, mpnt->vm_start + offset);
 	}
+	flush_dcache_mmap_unlock(mapping);
 }
 
 static void
@@ -129,6 +132,7 @@ make_coherent(struct vm_area_struct *vma, unsigned long addr, struct page *page,
 	 * space, then we need to handle them specially to maintain
 	 * cache coherency.
 	 */
+	flush_dcache_mmap_lock(mapping);
 	while ((mpnt = vma_prio_tree_next(mpnt, &mapping->i_mmap,
 					&iter, pgoff, pgoff)) != NULL) {
 		/*
@@ -143,6 +147,7 @@ make_coherent(struct vm_area_struct *vma, unsigned long addr, struct page *page,
 		offset = (pgoff - mpnt->vm_pgoff) << PAGE_SHIFT;
 		aliases += adjust_pte(mpnt, mpnt->vm_start + offset);
 	}
+	flush_dcache_mmap_unlock(mapping);
 	if (aliases)
 		adjust_pte(vma, addr);
 	else
diff --git a/arch/parisc/kernel/cache.c b/arch/parisc/kernel/cache.c
index 8f0bd9e72482..9fc061440122 100644
--- a/arch/parisc/kernel/cache.c
+++ b/arch/parisc/kernel/cache.c
@@ -249,6 +249,7 @@ void __flush_dcache_page(struct page *page)
 	 * declared as MAP_PRIVATE or MAP_SHARED), so we only need
 	 * to flush one address here for them all to become coherent */
 
+	flush_dcache_mmap_lock(mapping);
 	while ((mpnt = vma_prio_tree_next(mpnt, &mapping->i_mmap,
 					&iter, pgoff, pgoff)) != NULL) {
 		offset = (pgoff - mpnt->vm_pgoff) << PAGE_SHIFT;
@@ -266,8 +267,9 @@ void __flush_dcache_page(struct page *page)
 
 		__flush_cache_page(mpnt, addr);
 
-		return;
+		break;
 	}
+	flush_dcache_mmap_unlock(mapping);
 }
 EXPORT_SYMBOL(__flush_dcache_page);
 
diff --git a/include/asm-arm/cacheflush.h b/include/asm-arm/cacheflush.h
index 0e053301c09f..224c8d729775 100644
--- a/include/asm-arm/cacheflush.h
+++ b/include/asm-arm/cacheflush.h
@@ -303,6 +303,11 @@ static inline void flush_dcache_page(struct page *page)
 		__flush_dcache_page(page);
 }
 
+#define flush_dcache_mmap_lock(mapping) \
+	spin_lock_irq(&(mapping)->tree_lock)
+#define flush_dcache_mmap_unlock(mapping) \
+	spin_unlock_irq(&(mapping)->tree_lock)
+
 #define flush_icache_user_range(vma,page,addr,len) \
 	flush_dcache_page(page)
 
diff --git a/include/asm-i386/cacheflush.h b/include/asm-i386/cacheflush.h
index 9f78028b23e4..183361ebe0bc 100644
--- a/include/asm-i386/cacheflush.h
+++ b/include/asm-i386/cacheflush.h
@@ -10,6 +10,8 @@
 #define flush_cache_range(vma, start, end)	do { } while (0)
 #define flush_cache_page(vma, vmaddr)		do { } while (0)
 #define flush_dcache_page(page)			do { } while (0)
+#define flush_dcache_mmap_lock(mapping)		do { } while (0)
+#define flush_dcache_mmap_unlock(mapping)	do { } while (0)
 #define flush_icache_range(start, end)		do { } while (0)
 #define flush_icache_page(vma,pg)		do { } while (0)
 #define flush_icache_user_range(vma,pg,adr,len)	do { } while (0)
diff --git a/include/asm-parisc/cacheflush.h b/include/asm-parisc/cacheflush.h
index 84ba7dbe6493..a5700c29f18a 100644
--- a/include/asm-parisc/cacheflush.h
+++ b/include/asm-parisc/cacheflush.h
@@ -78,6 +78,11 @@ static inline void flush_dcache_page(struct page *page)
 	}
 }
 
+#define flush_dcache_mmap_lock(mapping) \
+	spin_lock_irq(&(mapping)->tree_lock)
+#define flush_dcache_mmap_unlock(mapping) \
+	spin_unlock_irq(&(mapping)->tree_lock)
+
 #define flush_icache_page(vma,page)	do { flush_kernel_dcache_page(page_address(page)); flush_kernel_icache_page(page_address(page)); } while (0)
 
 #define flush_icache_range(s,e)		do { flush_kernel_dcache_range_asm(s,e); flush_kernel_icache_range_asm(s,e); } while (0)
diff --git a/kernel/fork.c b/kernel/fork.c
index 3eb6ca91d29a..ef85a909e171 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -332,7 +332,9 @@ static inline int dup_mmap(struct mm_struct * mm, struct mm_struct * oldmm)
       
 			/* insert tmp into the share list, just after mpnt */
 			spin_lock(&file->f_mapping->i_mmap_lock);
+			flush_dcache_mmap_lock(mapping);
 			vma_prio_tree_add(tmp, mpnt);
+			flush_dcache_mmap_unlock(mapping);
 			spin_unlock(&file->f_mapping->i_mmap_lock);
 		}
 
diff --git a/mm/filemap.c b/mm/filemap.c
index 804d96e88fd7..6e59faaa6afb 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -65,7 +65,9 @@
  *    ->i_mmap_lock		(truncate->unmap_mapping_range)
  *
  *  ->mmap_sem
- *    ->i_mmap_lock		(various places)
+ *    ->i_mmap_lock
+ *      ->page_table_lock	(various places, mainly in mmap.c)
+ *        ->mapping->tree_lock	(arch-dependent flush_dcache_mmap_lock)
  *
  *  ->mmap_sem
  *    ->lock_page		(access_process_vm)
diff --git a/mm/fremap.c b/mm/fremap.c
index c624d3d6970f..eb056db90eb4 100644
--- a/mm/fremap.c
+++ b/mm/fremap.c
@@ -202,11 +202,13 @@ asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size,
 		    !(vma->vm_flags & VM_NONLINEAR)) {
 			mapping = vma->vm_file->f_mapping;
 			spin_lock(&mapping->i_mmap_lock);
+			flush_dcache_mmap_lock(mapping);
 			vma->vm_flags |= VM_NONLINEAR;
 			vma_prio_tree_remove(vma, &mapping->i_mmap);
 			vma_prio_tree_init(vma);
 			list_add_tail(&vma->shared.vm_set.list,
 					&mapping->i_mmap_nonlinear);
+			flush_dcache_mmap_unlock(mapping);
 			spin_unlock(&mapping->i_mmap_lock);
 		}
 
diff --git a/mm/mmap.c b/mm/mmap.c
index baa5d6c1600c..ad64fb036791 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -25,6 +25,7 @@
 
 #include <asm/uaccess.h>
 #include <asm/pgalloc.h>
+#include <asm/cacheflush.h>
 #include <asm/tlb.h>
 
 /*
@@ -74,10 +75,12 @@ static inline void __remove_shared_vm_struct(struct vm_area_struct *vma,
 	if (vma->vm_flags & VM_SHARED)
 		mapping->i_mmap_writable--;
 
+	flush_dcache_mmap_lock(mapping);
 	if (unlikely(vma->vm_flags & VM_NONLINEAR))
 		list_del_init(&vma->shared.vm_set.list);
 	else
 		vma_prio_tree_remove(vma, &mapping->i_mmap);
+	flush_dcache_mmap_unlock(mapping);
 }
 
 /*
@@ -266,11 +269,13 @@ static inline void __vma_link_file(struct vm_area_struct *vma)
 		if (vma->vm_flags & VM_SHARED)
 			mapping->i_mmap_writable++;
 
+		flush_dcache_mmap_lock(mapping);
 		if (unlikely(vma->vm_flags & VM_NONLINEAR))
 			list_add_tail(&vma->shared.vm_set.list,
 					&mapping->i_mmap_nonlinear);
 		else
 			vma_prio_tree_insert(vma, &mapping->i_mmap);
+		flush_dcache_mmap_unlock(mapping);
 	}
 }
 
@@ -350,14 +355,17 @@ void vma_adjust(struct vm_area_struct *vma, unsigned long start,
 	}
 	spin_lock(&mm->page_table_lock);
 
-	if (root)
+	if (root) {
+		flush_dcache_mmap_lock(mapping);
 		vma_prio_tree_remove(vma, root);
+	}
 	vma->vm_start = start;
 	vma->vm_end = end;
 	vma->vm_pgoff = pgoff;
 	if (root) {
 		vma_prio_tree_init(vma);
 		vma_prio_tree_insert(vma, root);
+		flush_dcache_mmap_unlock(mapping);
 	}
 
 	if (next) {