--- //depot/projects/smpng/sys/fs/fifofs/fifo_vnops.c	2008/03/13 21:35:00
+++ //depot/user/jhb/lock/fs/fifofs/fifo_vnops.c	2008/10/22 22:21:18
@@ -180,10 +180,14 @@
 	struct file *fp = ap->a_fp;
 	struct socket *rso, *wso;
 	int error;
+	int ltype;
 
-	ASSERT_VOP_ELOCKED(vp, "fifo_open");
+	ASSERT_VOP_LOCKED(vp, "fifo_open");
 	if (fp == NULL)
 		return (EINVAL);
+	ltype = VOP_ISLOCKED(vp);
+	if (vp->v_fifoinfo == NULL && ltype != LK_EXCLUSIVE)
+		vn_lock(vp, LK_UPGRADE | LK_RETRY);
 	if ((fip = vp->v_fifoinfo) == NULL) {
 		MALLOC(fip, struct fifoinfo *, sizeof(*fip), M_VNODE, M_WAITOK);
 		error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, cred, td);
@@ -257,7 +261,7 @@
 			VOP_UNLOCK(vp, 0);
 			error = msleep(&fip->fi_readers, &fifo_mtx,
 			    PDROP | PCATCH | PSOCK, "fifoor", 0);
-			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+			vn_lock(vp, ltype | LK_RETRY);
 			if (error) {
 				fip->fi_readers--;
 				if (fip->fi_readers == 0) {
@@ -277,7 +281,7 @@
 			VOP_UNLOCK(vp, 0);
 			error = msleep(&fip->fi_writers, &fifo_mtx,
 			    PDROP | PCATCH | PSOCK, "fifoow", 0);
-			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+			vn_lock(vp, ltype | LK_RETRY);
 			if (error) {
 				fip->fi_writers--;
 				if (fip->fi_writers == 0) {
--- //depot/projects/smpng/sys/kern/kern_descrip.c	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/kern/kern_descrip.c	2008/10/23 21:14:12
@@ -1241,7 +1241,7 @@
 	if (vp != NULL) {
 		int vfslocked;
 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+		vn_lock(vp, LK_SHARED | LK_RETRY);
 		error = VOP_PATHCONF(vp, uap->name, td->td_retval);
 		VOP_UNLOCK(vp, 0);
 		VFS_UNLOCK_GIANT(vfslocked);
--- //depot/projects/smpng/sys/kern/vfs_lookup.c	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/kern/vfs_lookup.c	2008/10/23 03:52:37
@@ -331,12 +331,48 @@
 static int
 compute_cn_lkflags(struct mount *mp, int lkflags)
 {
+
 	if (mp == NULL || 
 	    ((lkflags & LK_SHARED) && !(mp->mnt_kern_flag & MNTK_LOOKUP_SHARED))) {
 		lkflags &= ~LK_SHARED;
 		lkflags |= LK_EXCLUSIVE;
 	}
-	return lkflags;
+	return (lkflags);
+}
+
+static __inline int
+needs_exclusive_leaf(struct mount *mp, int flags)
+{
+
+	/*
+	 * Intermediate nodes can use shared locks, we only need to
+	 * force an exclusive lock for leaf nodes.
+	 */
+	if ((flags & (ISLASTCN | LOCKLEAF)) != (ISLASTCN | LOCKLEAF))
+		return (0);
+
+	/* Always use exclusive locks if LOCKSHARED isn't set. */
+	if (!(flags & LOCKSHARED))
+		return (1);
+
+	/*
+	 * For lookups during open(), if the mount point supports
+	 * extended shared operations, then use a shared lock for the
+	 * leaf node, otherwise use an exclusive lock.
+	 */
+	if (flags & ISOPEN) {
+		if (mp != NULL &&
+		    (mp->mnt_kern_flag & MNTK_EXTENDED_SHARED))
+			return (0);
+		else
+			return (1);
+	}
+
+	/*
+	 * Lookup requests outside of open() that specify LOCKSHARED
+	 * only need a shared lock on the leaf vnode.
+	 */
+	return (1);
 }
 
 /*
@@ -596,8 +632,7 @@
 	 * If we're looking up the last component and we need an exclusive
 	 * lock, adjust our lkflags.
 	 */
-	if ((cnp->cn_flags & (ISLASTCN|LOCKSHARED|LOCKLEAF)) ==
-	    (ISLASTCN|LOCKLEAF))
+	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags))
 		cnp->cn_lkflags = LK_EXCLUSIVE;
 #ifdef NAMEI_DIAGNOSTIC
 	vprint("lookup in", dp);
@@ -797,8 +832,8 @@
 	 * Because of lookup_shared we may have the vnode shared locked, but
 	 * the caller may want it to be exclusively locked.
 	 */
-	if ((cnp->cn_flags & (ISLASTCN | LOCKSHARED | LOCKLEAF)) ==
-	    (ISLASTCN | LOCKLEAF) && VOP_ISLOCKED(dp) != LK_EXCLUSIVE) {
+	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags) &&
+	    VOP_ISLOCKED(dp) != LK_EXCLUSIVE) {
 		vn_lock(dp, LK_UPGRADE | LK_RETRY);
 	}
 	if (vfslocked && dvfslocked)
--- //depot/projects/smpng/sys/kern/vfs_subr.c	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/kern/vfs_subr.c	2008/10/24 18:10:20
@@ -3899,6 +3899,22 @@
 }
 
 void
+vop_setattr_pre(void *ap)
+{
+#ifdef DEBUG_VFS_LOCKS
+	struct vop_setattr_args *a = ap;
+
+	/*
+	 * VOP_SETATTR() normally needs an exclusive lock except for
+	 * the special case of vfs_mark_atime() when a shared lock is
+	 * sufficient.
+	 */
+	if (!(a->a_vap->va_flags & VA_MARK_ATIME))
+		ASSERT_VOP_ELOCKED(a->a_vp, "VOP_SETATTR");
+#endif
+}
+
+void
 vop_setattr_post(void *ap, int rc)
 {
 	struct vop_setattr_args *a = ap;
--- //depot/projects/smpng/sys/kern/vfs_syscalls.c	2008/09/17 20:27:47
+++ //depot/user/jhb/lock/kern/vfs_syscalls.c	2008/10/23 21:14:12
@@ -286,8 +286,8 @@
 	int error;
 	struct nameidata nd;
 
-	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
-	    pathseg, path, td);
+	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
+	    AUDITVNODE1, pathseg, path, td);
 	error = namei(&nd);
 	if (error)
 		return (error);
@@ -368,7 +368,7 @@
 		return (error);
 	vp = fp->f_vnode;
 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(vp, LK_SHARED | LK_RETRY);
 #ifdef AUDIT
 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
 #endif
@@ -736,7 +736,7 @@
 	VREF(vp);
 	fdrop(fp, td);
 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(vp, LK_SHARED | LK_RETRY);
 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
 	error = change_dir(vp, td);
 	while (!error && (mp = vp->v_mountedhere) != NULL) {
@@ -800,8 +800,8 @@
 	struct vnode *vp;
 	int vfslocked;
 
-	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1 | MPSAFE,
-	    pathseg, path, td);
+	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
+	    MPSAFE, pathseg, path, td);
 	if ((error = namei(&nd)) != 0)
 		return (error);
 	vfslocked = NDHASGIANT(&nd);
@@ -886,8 +886,8 @@
 	error = priv_check(td, PRIV_VFS_CHROOT);
 	if (error)
 		return (error);
-	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
-	    UIO_USERSPACE, uap->path, td);
+	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
+	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
 	error = namei(&nd);
 	if (error)
 		goto error;
@@ -1938,7 +1938,7 @@
 		offset += fp->f_offset;
 		break;
 	case L_XTND:
-		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+		vn_lock(vp, LK_SHARED | LK_RETRY);
 		error = VOP_GETATTR(vp, &vattr, cred);
 		VOP_UNLOCK(vp, 0);
 		if (error)
@@ -2123,8 +2124,8 @@
 		td->td_ucred = tmpcred;
 	} else
 		cred = tmpcred = td->td_ucred;
-	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
-	    pathseg, path, fd, td);
+	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
+	    AUDITVNODE1, pathseg, path, fd, td);
 	if ((error = namei(&nd)) != 0)
 		goto out1;
 	vfslocked = NDHASGIANT(&nd);
@@ -2490,8 +2491,8 @@
 	struct nameidata nd;
 	int error, vfslocked;
 
-	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
-	    pathseg, path, td);
+	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
+	    AUDITVNODE1, pathseg, path, td);
 	if ((error = namei(&nd)) != 0)
 		return (error);
 	vfslocked = NDHASGIANT(&nd);
@@ -2566,8 +2567,8 @@
 	struct nameidata nd;
 	int vfslocked;
 
-	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
-	    pathseg, path, fd, td);
+	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
+	    AUDITVNODE1, pathseg, path, fd, td);
 
 	if ((error = namei(&nd)) != 0)
 		return (error);
@@ -2732,7 +2733,7 @@
 		return (error);
 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 #ifdef AUDIT
-	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
 	VOP_UNLOCK(fp->f_vnode, 0);
 #endif
@@ -2894,7 +2895,7 @@
 		return (error);
 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 #ifdef AUDIT
-	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
 	VOP_UNLOCK(fp->f_vnode, 0);
 #endif
@@ -3072,7 +3073,7 @@
 		return (error);
 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 #ifdef AUDIT
-	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
 	VOP_UNLOCK(fp->f_vnode, 0);
 #endif
@@ -3309,7 +3310,7 @@
 		return (error);
 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 #ifdef AUDIT
-	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
 	VOP_UNLOCK(fp->f_vnode, 0);
 #endif
--- //depot/projects/smpng/sys/kern/vfs_vnops.c	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/kern/vfs_vnops.c	2008/10/23 16:11:06
@@ -187,6 +187,8 @@
 		ndp->ni_cnd.cn_flags = ISOPEN |
 		    ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) |
 		    LOCKLEAF | MPSAFE | AUDITVNODE1;
+		if (!(fmode & FWRITE))
+			ndp->ni_cnd.cn_flags |= LOCKSHARED;
 		if ((error = namei(ndp)) != 0)
 			return (error);
 		if (!mpsafe)
@@ -239,7 +241,7 @@
 	if (fmode & FWRITE)
 		vp->v_writecount++;
 	*flagp = fmode;
-	ASSERT_VOP_ELOCKED(vp, "vn_open_cred");
+	ASSERT_VOP_LOCKED(vp, "vn_open_cred");
 	if (!mpsafe)
 		VFS_UNLOCK_GIANT(vfslocked);
 	return (0);
@@ -284,12 +286,18 @@
 	struct thread *td;
 {
 	struct mount *mp;
-	int error;
+	int error, lock_flags;
+
+	if (!(flags & FWRITE) && vp->v_mount != NULL &&
+	    vp->v_mount->mnt_kern_flag & MNTK_EXTENDED_SHARED)
+		lock_flags = LK_SHARED;
+	else
+		lock_flags = LK_EXCLUSIVE;
 
 	VFS_ASSERT_GIANT(vp->v_mount);
 
 	vn_start_write(vp, &mp, V_WAIT);
-	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(vp, lock_flags | LK_RETRY);
 	if (flags & FWRITE) {
 		VNASSERT(vp->v_writecount > 0, vp, 
 		    ("vn_close: negative writecount"));
@@ -672,7 +680,7 @@
 	int error;
 
 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(vp, LK_SHARED | LK_RETRY);
 	error = vn_stat(vp, sb, active_cred, fp->f_cred, td);
 	VOP_UNLOCK(vp, 0);
 	VFS_UNLOCK_GIANT(vfslocked);
--- //depot/projects/smpng/sys/kern/vnode_if.src	2008/09/17 20:27:47
+++ //depot/user/jhb/lock/kern/vnode_if.src	2008/10/24 17:58:56
@@ -133,7 +133,7 @@
 };
 
 
-%% close	vp	E E E
+%% close	vp	L L L
 
 vop_close {
 	IN struct vnode *vp;
@@ -162,7 +162,8 @@
 };
 
 
-%% setattr	vp	E E E
+%% setattr	vp	L L L
+%! setattr	pre	vop_setattr_pre
 %! setattr	post	vop_setattr_post
 
 vop_setattr {
--- //depot/projects/smpng/sys/security/audit/audit_arg.c	2008/09/17 20:27:47
+++ //depot/user/jhb/lock/security/audit/audit_arg.c	2008/10/23 20:58:05
@@ -631,7 +631,7 @@
 		 */
 		vp = fp->f_vnode;
 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+		vn_lock(vp, LK_SHARED | LK_RETRY);
 		audit_arg_vnode(vp, ARG_VNODE1);
 		VOP_UNLOCK(vp, 0);
 		VFS_UNLOCK_GIANT(vfslocked);
@@ -849,7 +849,7 @@
 
 	vp = fp->f_vnode;
 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
-	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+	vn_lock(vp, LK_SHARED | LK_RETRY);
 	audit_arg_vnode(vp, ARG_VNODE1);
 	VOP_UNLOCK(vp, 0);
 	VFS_UNLOCK_GIANT(vfslocked);
--- //depot/projects/smpng/sys/sys/mount.h	2008/09/24 21:56:18
+++ //depot/user/jhb/lock/sys/mount.h	2008/10/23 03:06:27
@@ -295,10 +295,7 @@
 #define	MNT_BYFSID	0x08000000	/* specify filesystem by ID. */
 #define MNT_CMDFLAGS   (MNT_UPDATE	| MNT_DELEXPORT	| MNT_RELOAD	| \
 			MNT_FORCE	| MNT_SNAPSHOT	| MNT_BYFSID)
-/*
- * Still available.
- */
-#define	MNT_SPARE_0x00000010	0x00000010
+
 /*
  * Internal filesystem control flags stored in mnt_kern_flag.
  *
@@ -319,6 +316,7 @@
 #define MNTK_ASYNC	0x00000002	/* filtered async flag */
 #define MNTK_SOFTDEP	0x00000004	/* async disabled by softdep */
 #define MNTK_NOINSMNTQ	0x00000008	/* insmntque is not allowed */
+#define MNTK_EXTENDED_SHARED	0x00000010 /* Allow shared locking for more ops */
 #define MNTK_UNMOUNT	0x01000000	/* unmount in progress */
 #define	MNTK_MWAIT	0x02000000	/* waiting for unmount to finish */
 #define	MNTK_SUSPEND	0x08000000	/* request write suspension */
--- //depot/projects/smpng/sys/sys/vnode.h	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/sys/vnode.h	2008/10/24 18:16:32
@@ -676,6 +676,7 @@
 void	vop_rename_pre(void *a);
 void	vop_rmdir_post(void *a, int rc);
 void	vop_setattr_post(void *a, int rc);
+void	vop_setattr_pre(void *a);
 void	vop_strategy_pre(void *a);
 void	vop_symlink_post(void *a, int rc);
 void	vop_unlock_post(void *a, int rc);
--- //depot/projects/smpng/sys/ufs/ffs/ffs_vfsops.c	2008/10/21 22:11:05
+++ //depot/user/jhb/lock/ufs/ffs/ffs_vfsops.c	2008/10/23 20:06:26
@@ -883,6 +883,9 @@
 	 */
 	MNT_ILOCK(mp);
 	mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED;
+#ifndef NO_UFS_EXTENDED_SHARED
+	mp->mnt_kern_flag |= MNTK_EXTENDED_SHARED;
+#endif
 	MNT_IUNLOCK(mp);
 #ifdef UFS_EXTATTR
 #ifdef UFS_EXTATTR_AUTOSTART
@@ -1423,10 +1426,9 @@
 		return (error);
 	}
 	/*
-	 * FFS supports recursive and shared locking.
+	 * FFS supports recursive locking.
 	 */
 	VN_LOCK_AREC(vp);
-	VN_LOCK_ASHARE(vp);
 	vp->v_data = ip;
 	vp->v_bufobj.bo_bsize = fs->fs_bsize;
 	ip->i_vnode = vp;
@@ -1499,6 +1501,10 @@
 	/*
 	 * Finish inode initialization.
 	 */
+	if (vp->v_type != VFIFO) {
+		/* FFS supports shared locking for all files except fifos. */
+		VN_LOCK_ASHARE(vp);
+	}
 
 	/*
 	 * Set up a generation number for this inode if it does not
--- //depot/projects/smpng/sys/ufs/ufs/ufs_vnops.c	2008/09/24 21:56:18
+++ //depot/user/jhb/lock/ufs/ufs/ufs_vnops.c	2008/10/24 14:22:31
@@ -488,7 +488,9 @@
 	 * checks get handled by other levels (e.g., ffs_update()).
 	 */
 	if (vap->va_vaflags & VA_MARK_ATIME) {
+		VI_LOCK(vp);
 		ip->i_flag |= IN_ACCESS;
+		VI_UNLOCK(vp);
 		return (0);
 	}
 	if (vap->va_flags != VNOVAL) {
--- //depot/projects/smpng/sys/vm/vm_mmap.c	2008/09/24 21:56:18
+++ //depot/user/jhb/lock/vm/vm_mmap.c	2008/10/23 19:03:09
@@ -1164,7 +1164,7 @@
 	mp = vp->v_mount;
 	cred = td->td_ucred;
 	vfslocked = VFS_LOCK_GIANT(mp);
-	if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) {
+	if ((error = vget(vp, LK_SHARED, td)) != 0) {
 		VFS_UNLOCK_GIANT(vfslocked);
 		return (error);
 	}
@@ -1181,7 +1181,7 @@
 		if (obj->handle != vp) {
 			vput(vp);
 			vp = (struct vnode*)obj->handle;
-			vget(vp, LK_EXCLUSIVE, td);
+			vget(vp, LK_SHARED, td);
 		}
 		type = OBJT_VNODE;
 		handle = vp;