[PATCH] ATTR_MODE breakage

(originally caught by Paul Larson <plars@austin.ibm.com>)

	Patch fixes a couple of idiotic bugs in notify_change() (my
fault).  Unary operations have higher priority than binary ones, so
if (!valid & ATTR_MODE) doesn't do the right thing.  I plead temporary
braindamage...

fs/attr.c

@@ -134,7 +134,7 @@ int notify_change(struct dentry * dentry, struct iattr * attr)
 		attr->ia_mtime = now;
 	if (ia_valid & ATTR_KILL_SUID) {
 		if (mode & S_ISUID) {
-			if (!ia_valid & ATTR_MODE) {
+			if (!(ia_valid & ATTR_MODE)) {
 				ia_valid = attr->ia_valid |= ATTR_MODE;
 				attr->ia_mode = inode->i_mode;
 			}
@@ -143,7 +143,7 @@ int notify_change(struct dentry * dentry, struct iattr * attr)
 	}
 	if (ia_valid & ATTR_KILL_SGID) {
 		if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
-			if (!ia_valid & ATTR_MODE) {
+			if (!(ia_valid & ATTR_MODE)) {
 				ia_valid = attr->ia_valid |= ATTR_MODE;
 				attr->ia_mode = inode->i_mode;
 			}