mm: always lock the root (oldest) anon_vma
Rik van Riel [Tue, 10 Aug 2010 00:18:40 +0000 (17:18 -0700)]
Always (and only) lock the root (oldest) anon_vma whenever we do something
in an anon_vma.  The recently introduced anon_vma scalability is due to
the rmap code scanning only the VMAs that need to be scanned.  Many common
operations still took the anon_vma lock on the root anon_vma, so always
taking that lock is not expected to introduce any scalability issues.

However, always taking the same lock does mean we only need to take one
lock, which means rmap_walk on pages from any anon_vma in the vma is
excluded from occurring during an munmap, expand_stack or other operation
that needs to exclude rmap_walk and similar functions.

Also add the proper locking to vma_adjust.

Signed-off-by: Rik van Riel <riel@redhat.com>
Tested-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/rmap.h
mm/ksm.c
mm/migrate.c
mm/mmap.c

index 41fa6dd..af43cb9 100644 (file)
@@ -104,24 +104,24 @@ static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
 {
        struct anon_vma *anon_vma = vma->anon_vma;
        if (anon_vma)
-               spin_lock(&anon_vma->lock);
+               spin_lock(&anon_vma->root->lock);
 }
 
 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 {
        struct anon_vma *anon_vma = vma->anon_vma;
        if (anon_vma)
-               spin_unlock(&anon_vma->lock);
+               spin_unlock(&anon_vma->root->lock);
 }
 
 static inline void anon_vma_lock(struct anon_vma *anon_vma)
 {
-       spin_lock(&anon_vma->lock);
+       spin_lock(&anon_vma->root->lock);
 }
 
 static inline void anon_vma_unlock(struct anon_vma *anon_vma)
 {
-       spin_unlock(&anon_vma->lock);
+       spin_unlock(&anon_vma->root->lock);
 }
 
 /*
index eb9f680..da6037c 100644 (file)
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -325,7 +325,7 @@ static void drop_anon_vma(struct rmap_item *rmap_item)
 {
        struct anon_vma *anon_vma = rmap_item->anon_vma;
 
-       if (atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->lock)) {
+       if (atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->root->lock)) {
                int empty = list_empty(&anon_vma->head);
                anon_vma_unlock(anon_vma);
                if (empty)
index 1855f86..5208fa1 100644 (file)
@@ -682,7 +682,7 @@ skip_unmap:
 rcu_unlock:
 
        /* Drop an anon_vma reference if we took one */
-       if (anon_vma && atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->lock)) {
+       if (anon_vma && atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->root->lock)) {
                int empty = list_empty(&anon_vma->head);
                anon_vma_unlock(anon_vma);
                if (empty)
index f5db18d..fb89360 100644 (file)
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -506,6 +506,7 @@ int vma_adjust(struct vm_area_struct *vma, unsigned long start,
        struct vm_area_struct *importer = NULL;
        struct address_space *mapping = NULL;
        struct prio_tree_root *root = NULL;
+       struct anon_vma *anon_vma = NULL;
        struct file *file = vma->vm_file;
        long adjust_next = 0;
        int remove_next = 0;
@@ -578,6 +579,17 @@ again:                     remove_next = 1 + (end > next->vm_end);
                }
        }
 
+       /*
+        * When changing only vma->vm_end, we don't really need anon_vma
+        * lock. This is a fairly rare case by itself, but the anon_vma
+        * lock may be shared between many sibling processes.  Skipping
+        * the lock for brk adjustments makes a difference sometimes.
+        */
+       if (vma->anon_vma && (insert || importer || start != vma->vm_start)) {
+               anon_vma = vma->anon_vma;
+               anon_vma_lock(anon_vma);
+       }
+
        if (root) {
                flush_dcache_mmap_lock(mapping);
                vma_prio_tree_remove(vma, root);
@@ -617,6 +629,8 @@ again:                      remove_next = 1 + (end > next->vm_end);
                __insert_vm_struct(mm, insert);
        }
 
+       if (anon_vma)
+               anon_vma_unlock(anon_vma);
        if (mapping)
                spin_unlock(&mapping->i_mmap_lock);
 
@@ -2470,23 +2484,23 @@ static DEFINE_MUTEX(mm_all_locks_mutex);
 
 static void vm_lock_anon_vma(struct mm_struct *mm, struct anon_vma *anon_vma)
 {
-       if (!test_bit(0, (unsigned long *) &anon_vma->head.next)) {
+       if (!test_bit(0, (unsigned long *) &anon_vma->root->head.next)) {
                /*
                 * The LSB of head.next can't change from under us
                 * because we hold the mm_all_locks_mutex.
                 */
-               spin_lock_nest_lock(&anon_vma->lock, &mm->mmap_sem);
+               spin_lock_nest_lock(&anon_vma->root->lock, &mm->mmap_sem);
                /*
                 * We can safely modify head.next after taking the
-                * anon_vma->lock. If some other vma in this mm shares
+                * anon_vma->root->lock. If some other vma in this mm shares
                 * the same anon_vma we won't take it again.
                 *
                 * No need of atomic instructions here, head.next
                 * can't change from under us thanks to the
-                * anon_vma->lock.
+                * anon_vma->root->lock.
                 */
                if (__test_and_set_bit(0, (unsigned long *)
-                                      &anon_vma->head.next))
+                                      &anon_vma->root->head.next))
                        BUG();
        }
 }
@@ -2577,7 +2591,7 @@ out_unlock:
 
 static void vm_unlock_anon_vma(struct anon_vma *anon_vma)
 {
-       if (test_bit(0, (unsigned long *) &anon_vma->head.next)) {
+       if (test_bit(0, (unsigned long *) &anon_vma->root->head.next)) {
                /*
                 * The LSB of head.next can't change to 0 from under
                 * us because we hold the mm_all_locks_mutex.
@@ -2588,10 +2602,10 @@ static void vm_unlock_anon_vma(struct anon_vma *anon_vma)
                 *
                 * No need of atomic instructions here, head.next
                 * can't change from under us until we release the
-                * anon_vma->lock.
+                * anon_vma->root->lock.
                 */
                if (!__test_and_clear_bit(0, (unsigned long *)
-                                         &anon_vma->head.next))
+                                         &anon_vma->root->head.next))
                        BUG();
                anon_vma_unlock(anon_vma);
        }