Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm
Sender: cygwin-developers-owner AT sourceware DOT cygnus DOT com
Delivered-To: mailing list cygwin-developers AT sourceware DOT cygnus DOT com
Message-ID: <00F8D6E8AB0DD3118F1A006008186C965368@server1.axonet.com.au>
From: Andrew Dalgleish <andrewd AT axonet DOT com DOT au>
To: cygwin-developers AT sourceware DOT cygnus DOT com
Subject: patch for bug in mount_info::del_item()
Date: Sat, 24 Jul 1999 02:09:17 +1000
X-Priority: 3
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.0.1458.49)
Content-Type: text/plain

mount_info::del_item() leaves holes in the mount_info::mount[] table.

This causes "umount --remove-all-mounts" to go into an endless loop.

You can also get other strange symptoms if you call umount followed by
mount.
This depends on your mount table, and what directories you umount and
mount.

To reproduce the bug, create a mount table like this:
(note that the posix root is not the last line)
Device              Directory           Type         Flags
c:\foobar           /                   user         textmode
c:\foo              /foo                user         textmode
then type "umount --remove-all-mounts".


The mount_info::posix_sorted[] and mount_info::native_sorted[] tables
correctly ignore the holes, 
but there are many accesses to mount_info::mount[] which do not use the
XXX_sorted[] tables
and do not check for the holes.


This patch is relative to winsup-src-19990718.tar.gz.

Sat Jul 24 01:35:38 1999  Andrew Dalgleish <andrewd AT axonet DOT com DOT au>
	* path.cc (mount_info::del_item): Don't leave holes in the mount
table.

--- winsup/path.cc.orig Thu Jul 15 12:49:24 1999
+++ winsup/path.cc      Sat Jul 24 01:35:38 1999
@@ -1673,12 +1674,16 @@
            || strcmp (mount[i].native_path, pathtmp) == 0)) &&
          ((mount[i].flags & MOUNT_SYSTEM) == (flags & MOUNT_SYSTEM)))
        {
-         /* Delete by emptying mount point in question, then sorting
-            the mount table, which will put the empty one to the end.
-            Inefficient but simple.  */
-         mount[i].init ("", "", 0);
-         sort ();
+        /* Move all the remaining mounts down one in the table. */
+        for (; i < nmounts - 1; i++)
+          mount[i] = mount[i+1];
+
+        /* Remember we have one less entry */
          nmounts--;
+
+        /* Sort the remaining entries. */
+         sort ();
+
          return 0;
        }
     }



Regards,
Andrew Dalgleish