MFp4: Several clean-ups and improvements over tmpfs:

 - Remove tmpfs_zone_xxx KPI, the uma(9) wrapper, since
   they does not bring any value now.
 - Use |= instead of = when applying VV_ROOT flag.
 - Remove tm_avariable_nodes list.  Use uma to hold the
   released nodes.
 - init/destory interlock mutex of node when init/fini
   instead of ctor/dtor.
 - Change memory computing using u_int to fix negative
   value in 2G mem machine.
 - Remove unnecessary bzero's
 - Rely uma logic to make file id allocation harder to
   guess.
 - Fix some unsigned/signed related things.  Make sure
   we respect -o size=xxxx
 - Use wire instead of hold a page.
 - Pass allocate_zero to obtain zeroed pages upon first
   use.

Submitted by:	Howard Su
Approved by:	re (tmpfs blanket, kensmith)

6 files changed, 106 insertions, 122 deletions

diff --git a/sys/fs/tmpfs/tmpfs.h b/sys/fs/tmpfs/tmpfs.h
index 2947b9b..61999e1 100644
--- a/sys/fs/tmpfs/tmpfs.h
+++ b/sys/fs/tmpfs/tmpfs.h
@@ -314,7 +314,6 @@ struct tmpfs_mount {
 	 * they will go into the available list, remaining alive until the
 	 * file system is unmounted. */
 	struct tmpfs_node_list	tm_nodes_used;
-	struct tmpfs_node_list	tm_nodes_avail;
 
 	/* All node lock to protect the node list and tmp_pages_used */
 	struct mtx allnode_lock;
@@ -470,10 +469,14 @@ TMPFS_PAGES_MAX(struct tmpfs_mount *tmp)
 }
 
 /* Returns the available space for the given file system. */
-#define TMPFS_META_SIZE(tmp) ((tmp)->tm_nodes_inuse * (sizeof(struct tmpfs_node) \
-				+ sizeof(struct dirent)))
-#define TMPFS_PAGES_AVAIL(tmp) (TMPFS_PAGES_MAX(tmp) - (tmp)->tm_pages_used - \
-			  	TMPFS_META_SIZE(tmp) / PAGE_SIZE - 1) 
+#define TMPFS_META_PAGES(tmp) ((tmp)->tm_nodes_inuse * (sizeof(struct tmpfs_node) \
+				+ sizeof(struct tmpfs_dirent))/PAGE_SIZE + 1)
+#define TMPFS_FILE_PAGES(tmp) ((tmp)->tm_pages_used)
+
+#define TMPFS_PAGES_AVAIL(tmp) (TMPFS_PAGES_MAX(tmp) > \
+			TMPFS_META_PAGES(tmp)+TMPFS_FILE_PAGES(tmp)? \
+			TMPFS_PAGES_MAX(tmp) - TMPFS_META_PAGES(tmp) \
+			- TMPFS_FILE_PAGES(tmp):0)
 
 #endif
 
diff --git a/sys/fs/tmpfs/tmpfs_subr.c b/sys/fs/tmpfs/tmpfs_subr.c
index a45febb..739747b 100644
--- a/sys/fs/tmpfs/tmpfs_subr.c
+++ b/sys/fs/tmpfs/tmpfs_subr.c
@@ -100,47 +100,23 @@ tmpfs_alloc_node(struct tmpfs_mount *tmp, enum vtype type,
 	MPASS(IFF(type == VLNK, target != NULL));
 	MPASS(IFF(type == VBLK || type == VCHR, rdev != VNOVAL));
 
-	nnode = NULL;
+	if (tmp->tm_nodes_inuse > tmp->tm_nodes_max)
+		return (ENOSPC);
 
-	TMPFS_LOCK(tmp);
-	if (LIST_EMPTY(&tmp->tm_nodes_avail)) {
-		MPASS(tmp->tm_nodes_last <= tmp->tm_nodes_max);
-		if (tmp->tm_nodes_last == tmp->tm_nodes_max) {
-			TMPFS_UNLOCK(tmp);
-			return ENOSPC;
-		}
-		TMPFS_UNLOCK(tmp);
-		nnode = (struct tmpfs_node *)tmpfs_zone_alloc(
-						tmp->tm_node_pool, M_WAITOK);
-		if (nnode == NULL)
-			return ENOSPC;
-		nnode->tn_id = tmp->tm_nodes_last++;
-		nnode->tn_gen = arc4random();
-	} else {
-		nnode = LIST_FIRST(&tmp->tm_nodes_avail);
-		LIST_REMOVE(nnode, tn_entries);
-		TMPFS_UNLOCK(tmp);
-		nnode->tn_gen++;
-	}
-	MPASS(nnode != NULL);
+	nnode = (struct tmpfs_node *)uma_zalloc_arg(
+				tmp->tm_node_pool, tmp, M_WAITOK);
+	if (nnode == NULL)
+		return (ENOSPC);
 
 	/* Generic initialization. */
 	nnode->tn_type = type;
-	nnode->tn_size = 0;
-	nnode->tn_status = 0;
-	nnode->tn_flags = 0;
-	nnode->tn_links = 0;
 	nanotime(&nnode->tn_atime);
 	nnode->tn_birthtime = nnode->tn_ctime = nnode->tn_mtime =
 	    nnode->tn_atime;
 	nnode->tn_uid = uid;
 	nnode->tn_gid = gid;
 	nnode->tn_mode = mode;
-	nnode->tn_lockf = NULL;
-	nnode->tn_vnode = NULL;
 	
-	nnode->tn_vpstate = 0;
-	mtx_init(&nnode->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
 	/* Type-specific initialization. */
 	switch (nnode->tn_type) {
 	case VBLK:
@@ -169,7 +145,7 @@ tmpfs_alloc_node(struct tmpfs_mount *tmp, enum vtype type,
 		    M_WAITOK, nnode->tn_size);
 		if (nnode->tn_link == NULL) {
 			nnode->tn_type = VNON;
-			tmpfs_free_node(tmp, nnode);
+			uma_zfree(tmp->tm_node_pool, nnode);
 			return ENOSPC;
 		}
 		memcpy(nnode->tn_link, target, nnode->tn_size);
@@ -216,9 +192,7 @@ tmpfs_alloc_node(struct tmpfs_mount *tmp, enum vtype type,
 void
 tmpfs_free_node(struct tmpfs_mount *tmp, struct tmpfs_node *node)
 {
-	ino_t id;
-	unsigned long gen;
-	size_t pages;
+	size_t pages = 0;
 
 	TMPFS_LOCK(tmp);
 	LIST_REMOVE(node, tn_entries);
@@ -240,38 +214,28 @@ tmpfs_free_node(struct tmpfs_mount *tmp, struct tmpfs_node *node)
 	case VFIFO:
 		/* FALLTHROUGH */
 	case VSOCK:
-		pages = 0;
 		break;
 
 	case VLNK:
 		tmpfs_str_zone_free(&tmp->tm_str_pool, node->tn_link,
 		    node->tn_size);
-		pages = 0;
 		break;
 
 	case VREG:
 		if (node->tn_reg.tn_aobj != NULL) {
 			vm_object_deallocate(node->tn_reg.tn_aobj);
-			node->tn_reg.tn_aobj = 0;
 		}
 		pages = node->tn_reg.tn_aobj_pages;
 		break;
 
 	default:
 		MPASS(0);
-		pages = 0; /* Shut up gcc when !DIAGNOSTIC. */
 		break;
 	}
 
-	id = node->tn_id;
-	gen = node->tn_gen;
-	memset(node, 0, sizeof(struct tmpfs_node));
-	node->tn_id = id;
-	node->tn_type = VNON;
-	node->tn_gen = gen;
+	uma_zfree(tmp->tm_node_pool, node);
 
 	TMPFS_LOCK(tmp);
-	LIST_INSERT_HEAD(&tmp->tm_nodes_avail, node, tn_entries);
 	tmp->tm_pages_used -= pages;
 	TMPFS_UNLOCK(tmp);
 }
@@ -293,14 +257,14 @@ tmpfs_alloc_dirent(struct tmpfs_mount *tmp, struct tmpfs_node *node,
 {
 	struct tmpfs_dirent *nde;
 
-	nde = (struct tmpfs_dirent *)tmpfs_zone_alloc(
+	nde = (struct tmpfs_dirent *)uma_zalloc(
 					tmp->tm_dirent_pool, M_WAITOK);
 	if (nde == NULL)
 		return ENOSPC;
 
 	nde->td_name = tmpfs_str_zone_alloc(&tmp->tm_str_pool, M_WAITOK, len);
 	if (nde->td_name == NULL) {
-		tmpfs_zone_free(tmp->tm_dirent_pool, nde);
+		uma_zfree(tmp->tm_dirent_pool, nde);
 		return ENOSPC;
 	}
 	nde->td_namelen = len;
@@ -339,7 +303,7 @@ tmpfs_free_dirent(struct tmpfs_mount *tmp, struct tmpfs_dirent *de,
 	}
 
 	tmpfs_str_zone_free(&tmp->tm_str_pool, de->td_name, de->td_namelen);
-	tmpfs_zone_free(tmp->tm_dirent_pool, de);
+	uma_zfree(tmp->tm_dirent_pool, de);
 }
 
 /* --------------------------------------------------------------------- */
@@ -576,7 +540,6 @@ tmpfs_dir_attach(struct vnode *vp, struct tmpfs_dirent *de)
 	dnode->tn_size += sizeof(struct tmpfs_dirent);
 	dnode->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \
 	    TMPFS_NODE_MODIFIED;
-	vnode_pager_setsize(vp, dnode->tn_size);
 }
 
 /* --------------------------------------------------------------------- */
@@ -602,8 +565,6 @@ tmpfs_dir_detach(struct vnode *vp, struct tmpfs_dirent *de)
 	dnode->tn_size -= sizeof(struct tmpfs_dirent);
 	dnode->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \
 	    TMPFS_NODE_MODIFIED;
-
-	vnode_pager_setsize(vp, dnode->tn_size);
 }
 
 /* --------------------------------------------------------------------- */
@@ -905,7 +866,7 @@ tmpfs_reg_resize(struct vnode *vp, off_t newsize)
 	node->tn_size = newsize;
 	vnode_pager_setsize(vp, newsize);
 	if (newsize < oldsize) {
-		int zerolen = MIN(round_page(newsize), node->tn_size) - newsize;
+		size_t zerolen = MIN(round_page(newsize), node->tn_size) - newsize;
 		struct vm_object *uobj = node->tn_reg.tn_aobj;
 		vm_page_t m;
 
diff --git a/sys/fs/tmpfs/tmpfs_uma.c b/sys/fs/tmpfs/tmpfs_uma.c
index 73066e1..6f9e4db 100644
--- a/sys/fs/tmpfs/tmpfs_uma.c
+++ b/sys/fs/tmpfs/tmpfs_uma.c
@@ -32,29 +32,17 @@ __FBSDID("$FreeBSD$");
 
 #include <fs/tmpfs/tmpfs.h>
 
-uma_zone_t
-tmpfs_zone_create(char *name, int size, int align, struct tmpfs_mount *tmp)
-{
-	uma_zone_t z;
-	z = uma_zcreate(name, size, NULL, NULL, NULL, NULL, align, M_WAITOK);
-	return z;
-}
-
-void
-tmpfs_zone_destroy(uma_zone_t zone)
-{
-	uma_zdestroy(zone);
-}
-
 void
-tmpfs_str_zone_create(struct tmpfs_str_zone *tsz, struct tmpfs_mount *tmp)
+tmpfs_str_zone_create(struct tmpfs_str_zone *tsz)
 {
 	int i, len;
 
 	len = TMPFS_STRZONE_STARTLEN;
 	for (i = 0; i < TMPFS_STRZONE_ZONECOUNT; ++i) {
-		tsz->tsz_zone[i] = tmpfs_zone_create(
-					"TMPFS str", len, UMA_ALIGN_PTR, tmp);
+		tsz->tsz_zone[i] = uma_zcreate(
+					"TMPFS str", len, 
+					NULL, NULL, NULL, NULL,
+					UMA_ALIGN_PTR, 0);
 		len <<= 1;
 	}
 }
@@ -66,7 +54,7 @@ tmpfs_str_zone_destroy(struct tmpfs_str_zone *tsz)
 
 	len = TMPFS_STRZONE_STARTLEN;
 	for (i = 0; i < TMPFS_STRZONE_ZONECOUNT; ++i) {
-		tmpfs_zone_destroy(tsz->tsz_zone[i]);
+		uma_zdestroy(tsz->tsz_zone[i]);
 		len <<= 1;
 	}
 }
diff --git a/sys/fs/tmpfs/tmpfs_uma.h b/sys/fs/tmpfs/tmpfs_uma.h
index a402a4f..ceaf2ab 100644
--- a/sys/fs/tmpfs/tmpfs_uma.h
+++ b/sys/fs/tmpfs/tmpfs_uma.h
@@ -44,26 +44,14 @@ uma_zone_t	tmpfs_zone_create(char *name, int size, int align,
 		    struct tmpfs_mount *m);
 void		tmpfs_zone_destroy(uma_zone_t zone);
 
-static __inline void* 
-tmpfs_zone_alloc(uma_zone_t zone, int flags)
-{
-	return uma_zalloc(zone, flags);
-}
-
-static __inline void
-tmpfs_zone_free(uma_zone_t zone, void *item)
-{
-	uma_zfree(zone, item); 
-} 
-
-void	tmpfs_str_zone_create(struct tmpfs_str_zone *, struct tmpfs_mount *); 
+void	tmpfs_str_zone_create(struct tmpfs_str_zone *); 
 void	tmpfs_str_zone_destroy(struct tmpfs_str_zone *);
 
 static __inline char*
 tmpfs_str_zone_alloc(struct tmpfs_str_zone *tsz, int flags, size_t len)
 {
 
-	int i, zlen;
+	size_t i, zlen;
 	char *ptr;
 
 	MPASS(len <= (TMPFS_STRZONE_STARTLEN << (TMPFS_STRZONE_ZONECOUNT-1)));
@@ -74,14 +62,14 @@ tmpfs_str_zone_alloc(struct tmpfs_str_zone *tsz, int flags, size_t len)
 		++i;
 		zlen  <<= 1;
 	}
-	ptr = (char *)tmpfs_zone_alloc(tsz->tsz_zone[i], flags);
+	ptr = (char *)uma_zalloc(tsz->tsz_zone[i], flags);
 	return  ptr;
 }
 
 static __inline void
 tmpfs_str_zone_free(struct tmpfs_str_zone *tsz, char *item, size_t len)
 {
-	int i, zlen;
+	size_t i, zlen;
 
 	MPASS(len <= (TMPFS_STRZONE_STARTLEN << (TMPFS_STRZONE_ZONECOUNT-1)));
 
@@ -91,7 +79,7 @@ tmpfs_str_zone_free(struct tmpfs_str_zone *tsz, char *item, size_t len)
 		++i;
 		zlen  <<= 1;
 	}
-	tmpfs_zone_free(tsz->tsz_zone[i], item);
+	uma_zfree(tsz->tsz_zone[i], item);
 }
 
 #endif
diff --git a/sys/fs/tmpfs/tmpfs_vfsops.c b/sys/fs/tmpfs/tmpfs_vfsops.c
index e5d3d6e..fe8f455 100644
--- a/sys/fs/tmpfs/tmpfs_vfsops.c
+++ b/sys/fs/tmpfs/tmpfs_vfsops.c
@@ -91,13 +91,13 @@ static const char *tmpfs_opts[] = {
 
 #define SWI_MAXMIB	3
 
-static int
+static u_int
 get_swpgtotal(void)
 {
 	struct xswdev xsd;
 	char *sname = "vm.swap_info";
 	int soid[SWI_MAXMIB], oid[2];
-	int unswdev, total, dmmax, nswapdev;
+	u_int unswdev, total, dmmax, nswapdev;
 	size_t mibi, len;
 
 	total = 0;
@@ -139,6 +139,60 @@ get_swpgtotal(void)
 }
 
 /* --------------------------------------------------------------------- */
+static int
+tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
+{
+	struct tmpfs_node *node = (struct tmpfs_node *)mem;
+	
+	if (node->tn_id == 0) {
+		/* if this node structure first time used */
+		struct tmpfs_mount *tmp = (struct tmpfs_mount *)arg;
+		TMPFS_LOCK(tmp);
+		node->tn_id = tmp->tm_nodes_last++;
+		TMPFS_UNLOCK(tmp);
+		node->tn_gen = arc4random();	
+	}
+	else {
+		node->tn_gen++;
+	}
+	
+	node->tn_size = 0;
+	node->tn_status = 0;
+	node->tn_flags = 0;
+	node->tn_links = 0;
+	node->tn_lockf = NULL;
+	node->tn_vnode = NULL;
+	node->tn_vpstate = 0;
+	node->tn_lookup_dirent = NULL;
+
+	return (0);
+}
+
+static void
+tmpfs_node_dtor(void *mem, int size, void *arg)
+{
+	struct tmpfs_node *node = (struct tmpfs_node *)mem;
+	node->tn_type = VNON;
+}
+
+static int 
+tmpfs_node_init(void *mem, int size, int flags)
+{
+	struct tmpfs_node *node = (struct tmpfs_node *)mem;
+	node->tn_id = 0;
+
+	mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
+	
+	return (0);
+}
+
+static void
+tmpfs_node_fini(void *mem, int size)
+{
+	struct tmpfs_node *node = (struct tmpfs_node *)mem;
+	
+	mtx_destroy(&node->tn_interlock);
+}
 
 static int
 tmpfs_mount(struct mount *mp, struct thread *l)
@@ -208,21 +262,23 @@ tmpfs_mount(struct mount *mp, struct thread *l)
 	tmp->tm_nodes_inuse = 0;
 	tmp->tm_maxfilesize = get_swpgtotal() * PAGE_SIZE;
 	LIST_INIT(&tmp->tm_nodes_used);
-	LIST_INIT(&tmp->tm_nodes_avail);
-
+	
 	tmp->tm_pages_max = pages;
 	tmp->tm_pages_used = 0;
-	tmp->tm_dirent_pool = tmpfs_zone_create(
+	tmp->tm_dirent_pool = uma_zcreate(
 					"TMPFS dirent",
 					sizeof(struct tmpfs_dirent),
+					NULL, NULL, NULL, NULL,
 					UMA_ALIGN_PTR,
-	    				tmp);
-	tmp->tm_node_pool = tmpfs_zone_create(
+	    				0);
+	tmp->tm_node_pool = uma_zcreate(
 					"TMPFS node",
 					sizeof(struct tmpfs_node),
+					tmpfs_node_ctor, tmpfs_node_dtor,
+					tmpfs_node_init, tmpfs_node_fini,
 					UMA_ALIGN_PTR,
-					tmp);
-	tmpfs_str_zone_create(&tmp->tm_str_pool, tmp);
+					0);
+	tmpfs_str_zone_create(&tmp->tm_str_pool);
 
 	/* Allocate the root node. */
 	error = tmpfs_alloc_node(tmp, VDIR, args.ta_root_uid,
@@ -231,8 +287,8 @@ tmpfs_mount(struct mount *mp, struct thread *l)
 
 	if (error != 0 || root == NULL) {
 	    tmpfs_str_zone_destroy(&tmp->tm_str_pool);
-	    tmpfs_zone_destroy(tmp->tm_node_pool);
-	    tmpfs_zone_destroy(tmp->tm_dirent_pool);
+	    uma_zdestroy(tmp->tm_node_pool);
+	    uma_zdestroy(tmp->tm_dirent_pool);
 	    free(tmp, M_TMPFSMNT);
 	    return error;
 	}
@@ -300,18 +356,9 @@ tmpfs_unmount(struct mount *mp, int mntflags, struct thread *l)
 		tmpfs_free_node(tmp, node);
 		node = next;
 	}
-	node = LIST_FIRST(&tmp->tm_nodes_avail);
-	while (node != NULL) {
-		struct tmpfs_node *next;
-
-		next = LIST_NEXT(node, tn_entries);
-		LIST_REMOVE(node, tn_entries);
-		tmpfs_zone_free(tmp->tm_node_pool, node);
-		node = next;
-	}
 
-	tmpfs_zone_destroy(tmp->tm_dirent_pool);
-	tmpfs_zone_destroy(tmp->tm_node_pool);
+	uma_zdestroy(tmp->tm_dirent_pool);
+	uma_zdestroy(tmp->tm_node_pool);
 	tmpfs_str_zone_destroy(&tmp->tm_str_pool);
 
 	mtx_destroy(&tmp->allnode_lock);
@@ -336,7 +383,7 @@ tmpfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
 	error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, vpp, td);
 
 	if (!error)
-		(*vpp)->v_vflag = VV_ROOT;
+		(*vpp)->v_vflag |= VV_ROOT;
 
 	return error;
 }
diff --git a/sys/fs/tmpfs/tmpfs_vnops.c b/sys/fs/tmpfs/tmpfs_vnops.c
index d7aea01..3527261 100644
--- a/sys/fs/tmpfs/tmpfs_vnops.c
+++ b/sys/fs/tmpfs/tmpfs_vnops.c
@@ -501,13 +501,10 @@ tmpfs_uio_xfer(struct tmpfs_mount *tmp, struct tmpfs_node *node,
 		idx = OFF_TO_IDX(uio->uio_offset);
 		d = uio->uio_offset - IDX_TO_OFF(idx);
 		len = MIN(len, (PAGE_SIZE - d));
-		m = vm_page_grab(uobj, idx, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
+		m = vm_page_grab(uobj, idx, VM_ALLOC_WIRED | VM_ALLOC_ZERO | 
+				VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
 		if (uio->uio_rw == UIO_READ && m->valid != VM_PAGE_BITS_ALL)
-			if (vm_pager_get_pages(uobj, &m, 1, 0) != VM_PAGER_OK)
-				vm_page_zero_invalid(m, TRUE);
-		vm_page_lock_queues();
-		vm_page_hold(m);
-		vm_page_unlock_queues();
+			vm_page_zero_invalid(m, TRUE);
 		VM_OBJECT_UNLOCK(uobj);
 		sched_pin();
 		sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
@@ -521,7 +518,7 @@ tmpfs_uio_xfer(struct tmpfs_mount *tmp, struct tmpfs_node *node,
 			vm_page_zero_invalid(m, TRUE);
 			vm_page_dirty(m);
 		}
-		vm_page_unhold(m);
+		vm_page_unwire(m, 0);
 		vm_page_activate(m);
 		vm_page_wakeup(m);
 		vm_page_unlock_queues();