[PATCH] 2.4.19-pre3 NFS close-to-open fix part 2 (VFS change)

[Trond Myklebust]

Hi,

  The following patch fixes the one remaining hole in the
NFS close-to-open checking. It is due to the VFS assuming that it
doesn't have to revalidate the dentry when one does open("."), or
open(".."). This means that the NFS layer is never notified that it
needs to check the data cache integrity for this case.

Al has said he plans to plug this hole in 2.5.x as part of his unionfs
changes, however we also need a fix in 2.4.x. Other networked
filesystems (Alan has mentioned OpenGFS) are also having the same
problem...
Unless Al has a different suggestion (Al?), the following patch
inserts the necessary lines into link_path_walk().

Cheers,
  Trond

diff -u --recursive --new-file linux-2.4.19-fix_put/fs/namei.c linux-2.4.19-cto/fs/namei.c
--- linux-2.4.19-fix_put/fs/namei.c	Tue Mar 12 14:58:54 2002
+++ linux-2.4.19-cto/fs/namei.c	Tue Mar 12 15:35:02 2002
@@ -457,7 +457,7 @@
 	while (*name=='/')
 		name++;
 	if (!*name)
-		goto return_base;
+		goto return_reval;
 
 	inode = nd->dentry->d_inode;
 	if (current->link_count)
@@ -576,7 +576,7 @@
 				inode = nd->dentry->d_inode;
 				/* fallthrough */
 			case 1:
-				goto return_base;
+				goto return_reval;
 		}
 		if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
 			err = nd->dentry->d_op->d_hash(nd->dentry, &this);
@@ -627,6 +627,19 @@
 			nd->last_type = LAST_DOT;
 		else if (this.len == 2 && this.name[1] == '.')
 			nd->last_type = LAST_DOTDOT;
+return_reval:
+		/*
+		 * We bypassed the ordinary revalidation routines.
+		 * Check the cached dentry for staleness.
+		 */
+		dentry = nd->dentry;
+		if (dentry && dentry->d_op && dentry->d_op->d_revalidate) {
+			err = -ESTALE;
+			if (!dentry->d_op->d_revalidate(dentry, 0)) {
+				d_invalidate(dentry);
+				break;
+			}
+		}
 return_base:
 		return 0;
 out_dput: