Re: /hurd/tmpfs problem

[Neal H Walfield]

> Maybe somewhat happens in tmfs/{dir,node}.c or in libdiskfs. Any
> suggestions where to set breakpoints and which information to look
> at?

The problem that you are seeing is likely solved by the following patch:

--- tmpfs.c     Sun May  6 02:50:06 2001
+++ tmpfs.c.new Sun Jan  6 19:24:24 2002
@@ -265,7 +265,6 @@ main (int argc, char **argv)
   /* Propagate permissions, owner, etc. from underlying node to
      the root directory of the new (empty) filesystem.  */
   err = io_stat (realnode, &st);
-  mach_port_deallocate (mach_task_self (), realnode);
   if (err)
     {
       error (0, err, "cannot stat underlying node");


What happens in that the last reference is dropped to the underlying
node and the parent translator just orphans the active translator.

Thomas had an idea one day when he was driving down a highway that
would fix the protocol to not require that the translator itself hold
a reference to the underlying node, however, he has not yet shared it.


Once you conquer this problem, tmpfs should basically work -- in so
far as, tmpfs is correct, however, bugs in GNU Mach will prevent you
from actually using it effectively.