mm, rmap: handle anon_vma_prepare() common case inline

anon_vma_prepare() is mostly a large "if (unlikely(...))" block, as the
expected common case is that an anon_vma already exists.  We could turn
the condition around and return 0, but it also makes sense to do it
inline and avoid a call for the common case.

Bloat-o-meter naturally shows that inlining the check has some code size
costs:

add/remove: 1/1 grow/shrink: 4/0 up/down: 475/-373 (102)
function                                     old     new   delta
__anon_vma_prepare                             -     359    +359
handle_mm_fault                             2744    2796     +52
hugetlb_cow                                 1146    1170     +24
hugetlb_fault                               2123    2145     +22
wp_page_copy                                1469    1487     +18
anon_vma_prepare                             373       -    -373

Checking the asm however confirms that the hot paths now avoid a call,
which is moved away.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20161116074005.22768-1-vbabka@suse.czSigned-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/rmap.h

@@ -137,11 +137,19 @@ static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
  * anon_vma helper functions.
  */
 void anon_vma_init(void);	/* create anon_vma_cachep */
-int  anon_vma_prepare(struct vm_area_struct *);
+int  __anon_vma_prepare(struct vm_area_struct *);
 void unlink_anon_vmas(struct vm_area_struct *);
 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
 
+static inline int anon_vma_prepare(struct vm_area_struct *vma)
+{
+	if (likely(vma->anon_vma))
+		return 0;
+
+	return __anon_vma_prepare(vma);
+}
+
 static inline void anon_vma_merge(struct vm_area_struct *vma,
 				  struct vm_area_struct *next)
 {

mm/rmap.c

@@ -141,14 +141,15 @@ static void anon_vma_chain_link(struct vm_area_struct *vma,
 }
 
 /**
- * anon_vma_prepare - attach an anon_vma to a memory region
+ * __anon_vma_prepare - attach an anon_vma to a memory region
  * @vma: the memory region in question
  *
  * This makes sure the memory mapping described by 'vma' has
  * an 'anon_vma' attached to it, so that we can associate the
  * anonymous pages mapped into it with that anon_vma.
  *
- * The common case will be that we already have one, but if
+ * The common case will be that we already have one, which
+ * is handled inline by anon_vma_prepare(). But if
  * not we either need to find an adjacent mapping that we
  * can re-use the anon_vma from (very common when the only
  * reason for splitting a vma has been mprotect()), or we
@@ -167,15 +168,13 @@ static void anon_vma_chain_link(struct vm_area_struct *vma,
  *
  * This must be called with the mmap_sem held for reading.
  */
-int anon_vma_prepare(struct vm_area_struct *vma)
+int __anon_vma_prepare(struct vm_area_struct *vma)
 {
-	struct anon_vma *anon_vma = vma->anon_vma;
+	struct mm_struct *mm = vma->vm_mm;
+	struct anon_vma *anon_vma, *allocated;
 	struct anon_vma_chain *avc;
 
 	might_sleep();
-	if (unlikely(!anon_vma)) {
-		struct mm_struct *mm = vma->vm_mm;
-		struct anon_vma *allocated;
 
 	avc = anon_vma_chain_alloc(GFP_KERNEL);
 	if (!avc)
@@ -208,7 +207,7 @@ int anon_vma_prepare(struct vm_area_struct *vma)
 		put_anon_vma(allocated);
 	if (unlikely(avc))
 		anon_vma_chain_free(avc);
-	}
+
 	return 0;
 
  out_enomem_free_avc: