[Unionfs] Can't mount NFS-exported union anymore in unionfs-1.4

[hooanon05 at yahoo.co.jp]

Wilhelm Meier:
> I tested it with two client doing a normal nfs-mount and performing a find 
> through the mount: works fine with and without the mount-option nolock.

What I wanted to point out is nfsroot may not use lockd, and I was
afraid when two nfs clinets use the single nfs export as thier root
filesystem, file lock will not work correctly.


> Did it now (debug=18) with the diskless-client mounting the nfs as root-fs. I 
> attached the log. Most messages correspond to the tftp-transfer of the kernel 
> (which comes also from the unionfs via tftp and which is obviously o.k.). 
> Seach for "authenticated mount request" and you get the start of the 
> nfs-sequence.

Thanx.

(in your log)
Nov 30 10:31:40 gs unionfs: 255, fh{i22, g0, hr0, hi0, hpi0}

This shows an error which means the file handle was not encoded or
decoded correctly.
Will you apply this patch which only prints debug messages, and test
both of normal-mount and root-mount cases.


Junjiro Okajima


Index: super.c
===================================================================
RCS file: /home/cvs/unionfs/unionfs/super.c,v
retrieving revision 1.101
diff -u -p -r1.101 super.c
--- super.c	4 Nov 2006 22:27:51 -0000	1.101
+++ super.c	30 Nov 2006 13:19:58 -0000
@@ -712,7 +712,8 @@ static int unionfs_encode_fh(struct dent
 	ino_t h_ino, hp_ino;
 	static int warn;
 
-	print_entry("dentry %p", dentry);
+	//print_entry("dentry %p, max %d, conn %d", dentry, *max_len, connectable);
+	printk("dentry %p, max %d, conn %d", dentry, *max_len, connectable);
 	BUG_ON(is_anon(dentry) || !dentry->d_inode
 	       || is_anon(dentry->d_parent));
 
@@ -732,6 +733,7 @@ static int unionfs_encode_fh(struct dent
 
 	len = *max_len;
 	type = export_op_default.encode_fh(dentry, fh, max_len, connectable);
+	printk("type %d, max %d\n", type, *max_len);
 	if (type == 255 || *max_len > FhHead || len < FhTail) {
 		type = 255;
 		goto out;