Do not leave invalid pages in the object after the short read for a

network file systems (not only NFS proper). Short reads cause pages
other then the requested one, which were not filled by read response,
to stay invalid.

Change the vm_page_readahead_finish() interface to not take the error
code, but instead to make a decision to free or to (de)activate the
page only by its validity. As result, not requested invalid pages are
freed even if the read RPC indicated success.

Noted and reviewed by:	alc
MFC after:	1 week

sys/fs/nfsclient/nfs_clbio.c

@@ -218,14 +218,18 @@ ncl_getpages(struct vop_getpages_args *ap)
 			    ("nfs_getpages: page %p is dirty", m));
 		} else {
 			/*
-			 * Read operation was short.  If no error occured
-			 * we may have hit a zero-fill section.   We simply
-			 * leave valid set to 0.
+			 * Read operation was short.  If no error
+			 * occured we may have hit a zero-fill
+			 * section.  We leave valid set to 0, and page
+			 * is freed by vm_page_readahead_finish() if
+			 * its index is not equal to requested, or
+			 * page is zeroed and set valid by
+			 * vm_pager_get_pages() for requested page.
 			 */
 			;
 		}
 		if (i != ap->a_reqpage)
-			vm_page_readahead_finish(m, error);
+			vm_page_readahead_finish(m);
 	}
 	VM_OBJECT_UNLOCK(object);
 	return (0);

sys/fs/nwfs/nwfs_io.c

@@ -459,7 +459,7 @@ nwfs_getpages(ap)
 		}
 
 		if (i != ap->a_reqpage)
-			vm_page_readahead_finish(m, error);
+			vm_page_readahead_finish(m);
 	}
 	VM_OBJECT_UNLOCK(object);
 	return 0;

sys/fs/smbfs/smbfs_io.c

@@ -523,7 +523,7 @@ smbfs_getpages(ap)
 		}
 
 		if (i != reqpage)
-			vm_page_readahead_finish(m, error);
+			vm_page_readahead_finish(m);
 	}
 	VM_OBJECT_UNLOCK(object);
 	return 0;

sys/nfsclient/nfs_bio.c

@@ -212,14 +212,18 @@ nfs_getpages(struct vop_getpages_args *ap)
 			    ("nfs_getpages: page %p is dirty", m));
 		} else {
 			/*
-			 * Read operation was short.  If no error occured
-			 * we may have hit a zero-fill section.   We simply
-			 * leave valid set to 0.
+			 * Read operation was short.  If no error
+			 * occured we may have hit a zero-fill
+			 * section.  We leave valid set to 0, and page
+			 * is freed by vm_page_readahead_finish() if
+			 * its index is not equal to requested, or
+			 * page is zeroed and set valid by
+			 * vm_pager_get_pages() for requested page.
 			 */
 			;
 		}
 		if (i != ap->a_reqpage)
-			vm_page_readahead_finish(m, error);
+			vm_page_readahead_finish(m);
 	}
 	VM_OBJECT_UNLOCK(object);
 	return (0);

sys/vm/vm_page.c

@@ -699,10 +699,10 @@ vm_page_free_zero(vm_page_t m)
  * array which is not the request page.
  */
 void
-vm_page_readahead_finish(vm_page_t m, int error)
+vm_page_readahead_finish(vm_page_t m)
 {
 
-	if (error == 0) {
+	if (m->valid != 0) {
 		/*
 		 * Since the page is not the requested page, whether
 		 * it should be activated or deactivated is not
@@ -721,6 +721,12 @@ vm_page_readahead_finish(vm_page_t m, int error)
 		}
 		vm_page_wakeup(m);
 	} else {
+		/*
+		 * Free the completely invalid page.  Such page state
+		 * occurs due to the short read operation which did
+		 * not covered our page at all, or in case when a read
+		 * error happens.
+		 */
 		vm_page_lock(m);
 		vm_page_free(m);
 		vm_page_unlock(m);

sys/vm/vm_page.h

@@ -388,7 +388,7 @@ vm_page_t vm_page_next(vm_page_t m);
 int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *);
 vm_page_t vm_page_prev(vm_page_t m);
 void vm_page_putfake(vm_page_t m);
-void vm_page_readahead_finish(vm_page_t m, int error);
+void vm_page_readahead_finish(vm_page_t m);
 void vm_page_reference(vm_page_t m);
 void vm_page_remove (vm_page_t);
 void vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);

sys/vm/vnode_pager.c

@@ -987,7 +987,7 @@ vnode_pager_generic_getpages(vp, m, bytecount, reqpage)
 		}
 		
 		if (i != reqpage)
-			vm_page_readahead_finish(mt, error);
+			vm_page_readahead_finish(mt);
 	}
 	VM_OBJECT_UNLOCK(object);
 	if (error) {