JFFS: rmdir wierdness

JFFS: rmdir wierdness

[David Vrabel]

Hi,

# mkdir a
# ls
a
# rmdir a
# ls
# mkdir a
mkdir: a: File exists

Also

# mkdir b
# touch b/a
# rm b/a
# rmdir b
rmdir: b: Directory not empty

rmdir() thinks `b' still has children (I think).

This was done with the latest CVS with kernel 2.2.18.

David Vrabel.



To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: JFFS: rmdir wierdness

[David Woodhouse]

dvrabel@arcom.co.uk said:
> # mkdir a
> # ls
> a
> # rmdir a
> # ls
> # mkdir a
> mkdir: a: File exists
> 
> Also
> 
> # mkdir b
> # touch b/a
> # rm b/a
> # rmdir b
> rmdir: b: Directory not empty
> 
> rmdir() thinks `b' still has children (I think).

More problems with the deletion code since I did the use counts.
Hmmm. This was sort of why I didn't want to commit it immediately, till I'd 
had time to do some sanity checks on it.

Try this:

Index: intrep.c
===================================================================
RCS file: /home/cvs/mtd/fs/jffs/intrep.c,v
retrieving revision 1.84
diff -u -r1.84 intrep.c
--- intrep.c	2000/12/18 13:00:23	1.84
+++ intrep.c	2000/12/18 13:02:30
@@ -1387,7 +1387,7 @@
 	D3(printk("jffs_find_child()\n"));
 
 	for (f = dir->children; f; f = f->sibling_next) {
-		if (f->name
+		if (!f->deleted && f->name
 		    && !strncmp(f->name, name, len)
 		    && f->name[len] == '\0') {
 			break;


--
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: JFFS: rmdir wierdness

[David Vrabel]

Hi,

Are these problems anything to do with jffs_find_child() not ignoring
deleted files?

David Vrabel



To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: JFFS: rmdir wierdness

[David Vrabel]

Hi,

> # mkdir a
> # ls
> a
> # rmdir a
> # ls
> # mkdir a
> mkdir: a: File exists

This one has been fixed.

> # mkdir b
> # touch b/a
> # rm b/a
> # rmdir b
> rmdir: b: Directory not empty

This one is still broken.

In jffs_remove():
	if (S_ISDIR(type)) {
		/* ... */
		if (del_f->children) {
			result = -ENOTEMPTY;
			goto jffs_remove_end;
		}
	}

As I understand it del_f->children is the head of a list of the
directory's children.  If this is the case then won't you have to go
through the list checking that all the children have been deleted?  Find
a patch that does just this (it works too).

David Vrabel.

Index: inode-v22.c
===================================================================
RCS file: /home/cvs/mtd/fs/jffs/inode-v22.c,v
retrieving revision 1.55
diff -u -r1.55 inode-v22.c
--- inode-v22.c 2000/12/18 12:11:08     1.55
+++ inode-v22.c 2000/12/18 14:05:50
@@ -960,14 +960,20 @@
        }
 
        if (S_ISDIR(type)) {
+               struct jffs_file *child=del_f->children;
+
                result = -ENOTDIR;
                if (!S_ISDIR(del_f->mode)) { /* Is this really needed
for 2.2? */
                        D(printk("jffs_remove(): S_ISDIR but isn't\n"));
                        goto jffs_remove_end;
                }
-               if (del_f->children) {
-                       result = -ENOTEMPTY;
-                       goto jffs_remove_end;
+
+               while(child) {
+                       if( !child->deleted ) {
+                               result = -ENOTEMPTY;
+                               goto jffs_remove_end;
+                       }
+                       child = child->sibling_next;
                }
        }
        else if (S_ISDIR(del_f->mode)) {
Index: inode-v23.c
===================================================================
RCS file: /home/cvs/mtd/fs/jffs/inode-v23.c,v
retrieving revision 1.52
diff -u -r1.52 inode-v23.c
--- inode-v23.c 2000/12/18 12:11:08     1.52
+++ inode-v23.c 2000/12/18 14:05:56
@@ -956,11 +956,15 @@
        }
 
        if (S_ISDIR(type)) {
-               if (del_f->children) {
-                       result = -ENOTEMPTY;
-                       goto jffs_remove_end;
+               struct jffs_file *child=del_f->children;
+               while(child) {
+                       if( !child->deleted ) {
+                               result = -ENOTEMPTY;
+                               goto jffs_remove_end;
+                       }
+                       child = child->sibling_next;
                }
-       }
+       }
        else if (S_ISDIR(del_f->mode)) {
                D(printk("jffs_remove(): node is a directory "
                         "but it shouldn't be.\n"));



To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org