Avoid an exclusive acquisition of the object lock on the expected execution

path through the NFS clients' getpages functions.

Introduce vm_pager_free_nonreq().  This function can be used to eliminate
code that is duplicated in many getpages functions.  Also, in contrast to
the code that currently appears in those getpages functions,
vm_pager_free_nonreq() avoids acquiring an exclusive object lock in one
case.

Reviewed by:	kib
MFC after:	6 weeks
Sponsored by:	EMC / Isilon Storage Division

5 files changed, 49 insertions, 56 deletions

diff --git a/sys/fs/nfsclient/nfs_clbio.c b/sys/fs/nfsclient/nfs_clbio.c
index dc106ee..4c8c2dc 100644
--- a/sys/fs/nfsclient/nfs_clbio.c
+++ b/sys/fs/nfsclient/nfs_clbio.c
@@ -129,23 +129,20 @@ ncl_getpages(struct vop_getpages_args *ap)
 	npages = btoc(count);
 
 	/*
+	 * Since the caller has busied the requested page, that page's valid
+	 * field will not be changed by other threads.
+	 */
+	vm_page_assert_xbusied(pages[ap->a_reqpage]);
+
+	/*
 	 * If the requested page is partially valid, just return it and
 	 * allow the pager to zero-out the blanks.  Partially valid pages
 	 * can only occur at the file EOF.
 	 */
-	VM_OBJECT_WLOCK(object);
 	if (pages[ap->a_reqpage]->valid != 0) {
-		for (i = 0; i < npages; ++i) {
-			if (i != ap->a_reqpage) {
-				vm_page_lock(pages[i]);
-				vm_page_free(pages[i]);
-				vm_page_unlock(pages[i]);
-			}
-		}
-		VM_OBJECT_WUNLOCK(object);
-		return (0);
+		vm_pager_free_nonreq(object, pages, ap->a_reqpage, npages);
+		return (VM_PAGER_OK);
 	}
-	VM_OBJECT_WUNLOCK(object);
 
 	/*
 	 * We use only the kva address for the buffer, but this is extremely
@@ -175,15 +172,7 @@ ncl_getpages(struct vop_getpages_args *ap)
 
 	if (error && (uio.uio_resid == count)) {
 		ncl_printf("nfs_getpages: error %d\n", error);
-		VM_OBJECT_WLOCK(object);
-		for (i = 0; i < npages; ++i) {
-			if (i != ap->a_reqpage) {
-				vm_page_lock(pages[i]);
-				vm_page_free(pages[i]);
-				vm_page_unlock(pages[i]);
-			}
-		}
-		VM_OBJECT_WUNLOCK(object);
+		vm_pager_free_nonreq(object, pages, ap->a_reqpage, npages);
 		return (VM_PAGER_ERROR);
 	}
 
diff --git a/sys/nfsclient/nfs_bio.c b/sys/nfsclient/nfs_bio.c
index 1f07b46..1a5b156 100644
--- a/sys/nfsclient/nfs_bio.c
+++ b/sys/nfsclient/nfs_bio.c
@@ -123,23 +123,20 @@ nfs_getpages(struct vop_getpages_args *ap)
 	npages = btoc(count);
 
 	/*
+	 * Since the caller has busied the requested page, that page's valid
+	 * field will not be changed by other threads.
+	 */
+	vm_page_assert_xbusied(pages[ap->a_reqpage]);
+
+	/*
 	 * If the requested page is partially valid, just return it and
 	 * allow the pager to zero-out the blanks.  Partially valid pages
 	 * can only occur at the file EOF.
 	 */
-	VM_OBJECT_WLOCK(object);
 	if (pages[ap->a_reqpage]->valid != 0) {
-		for (i = 0; i < npages; ++i) {
-			if (i != ap->a_reqpage) {
-				vm_page_lock(pages[i]);
-				vm_page_free(pages[i]);
-				vm_page_unlock(pages[i]);
-			}
-		}
-		VM_OBJECT_WUNLOCK(object);
-		return (0);
+		vm_pager_free_nonreq(object, pages, ap->a_reqpage, npages);
+		return (VM_PAGER_OK);
 	}
-	VM_OBJECT_WUNLOCK(object);
 
 	/*
 	 * We use only the kva address for the buffer, but this is extremely
@@ -169,15 +166,7 @@ nfs_getpages(struct vop_getpages_args *ap)
 
 	if (error && (uio.uio_resid == count)) {
 		nfs_printf("nfs_getpages: error %d\n", error);
-		VM_OBJECT_WLOCK(object);
-		for (i = 0; i < npages; ++i) {
-			if (i != ap->a_reqpage) {
-				vm_page_lock(pages[i]);
-				vm_page_free(pages[i]);
-				vm_page_unlock(pages[i]);
-			}
-		}
-		VM_OBJECT_WUNLOCK(object);
+		vm_pager_free_nonreq(object, pages, ap->a_reqpage, npages);
 		return (VM_PAGER_ERROR);
 	}
 
diff --git a/sys/vm/vm_pager.c b/sys/vm/vm_pager.c
index c7d038b..65193c3 100644
--- a/sys/vm/vm_pager.c
+++ b/sys/vm/vm_pager.c
@@ -282,6 +282,33 @@ vm_pager_object_lookup(struct pagerlst *pg_list, void *handle)
 }
 
 /*
+ * Free the non-requested pages from the given array.
+ */
+void
+vm_pager_free_nonreq(vm_object_t object, vm_page_t ma[], int reqpage,
+    int npages)
+{
+	int i;
+	boolean_t object_locked;
+
+	VM_OBJECT_ASSERT_UNLOCKED(object);
+	object_locked = FALSE;
+	for (i = 0; i < npages; ++i) {
+		if (i != reqpage) {
+			if (!object_locked) {
+				VM_OBJECT_WLOCK(object);
+				object_locked = TRUE;
+			}
+			vm_page_lock(ma[i]);
+			vm_page_free(ma[i]);
+			vm_page_unlock(ma[i]);
+		}
+	}
+	if (object_locked)
+		VM_OBJECT_WUNLOCK(object);
+}
+
+/*
  * initialize a physical buffer
  */
 
diff --git a/sys/vm/vm_pager.h b/sys/vm/vm_pager.h
index 62265a2..e3d292d 100644
--- a/sys/vm/vm_pager.h
+++ b/sys/vm/vm_pager.h
@@ -106,6 +106,8 @@ static __inline int vm_pager_get_pages(vm_object_t, vm_page_t *, int, int);
 static __inline boolean_t vm_pager_has_page(vm_object_t, vm_pindex_t, int *, int *);
 void vm_pager_init(void);
 vm_object_t vm_pager_object_lookup(struct pagerlst *, void *);
+void vm_pager_free_nonreq(vm_object_t object, vm_page_t ma[], int reqpage,
+    int npages);
 
 /*
  *	vm_page_get_pages:
diff --git a/sys/vm/vnode_pager.c b/sys/vm/vnode_pager.c
index 072f197..1b0dc7e 100644
--- a/sys/vm/vnode_pager.c
+++ b/sys/vm/vnode_pager.c
@@ -722,14 +722,7 @@ vnode_pager_generic_getpages(struct vnode *vp, vm_page_t *m, int bytecount,
 		VM_OBJECT_WUNLOCK(object);
 		return (error);
 	} else if (error != 0) {
-		VM_OBJECT_WLOCK(object);
-		for (i = 0; i < count; i++)
-			if (i != reqpage) {
-				vm_page_lock(m[i]);
-				vm_page_free(m[i]);
-				vm_page_unlock(m[i]);
-			}
-		VM_OBJECT_WUNLOCK(object);
+		vm_pager_free_nonreq(object, m, reqpage, count);
 		return (VM_PAGER_ERROR);
 
 		/*
@@ -739,14 +732,7 @@ vnode_pager_generic_getpages(struct vnode *vp, vm_page_t *m, int bytecount,
 		 */
 	} else if ((PAGE_SIZE / bsize) > 1 &&
 	    (vp->v_mount->mnt_stat.f_type != nfs_mount_type)) {
-		VM_OBJECT_WLOCK(object);
-		for (i = 0; i < count; i++)
-			if (i != reqpage) {
-				vm_page_lock(m[i]);
-				vm_page_free(m[i]);
-				vm_page_unlock(m[i]);
-			}
-		VM_OBJECT_WUNLOCK(object);
+		vm_pager_free_nonreq(object, m, reqpage, count);
 		PCPU_INC(cnt.v_vnodein);
 		PCPU_INC(cnt.v_vnodepgsin);
 		return vnode_pager_input_smlfs(object, m[reqpage]);