Bugzilla Bug 119538

due to new mutex system submount compiles not more.

here is my fix, tested and working:

--- a/subfs-0.9/subfs.c 2004-02-26 23:54:58.000000000 +0100
+++ b/subfs-0.9/subfs.c 2006-01-19 14:12:26.000000000 +0100
@@ -221,14 +221,14 @@
        struct vfsmount *child;

        /* This is ugly, but prevents a lockup during mount. */
-       up(&dir->i_sem);
+       mutex_lock(&dir->i_mutex);
        if (down_interruptible(&sfs_mnt->sem)) {
-               down(&dir->i_sem);/*put the dir sem back down if interrupted*/
+               mutex_unlock(&dir->i_mutex);
                return ERR_PTR(-ERESTARTSYS);
        }
        child = get_child_mount(sfs_mnt);
        up(&sfs_mnt->sem);
-       down(&dir->i_sem);
+       mutex_unlock(&dir->i_mutex);
        if (IS_ERR(child))
                return (void *) child;
        subfs_send_signal();

please apply.
full story link here:
http://www.patriziobassi.it/blog/?p=8

Created an attachment (id=77525) [details]
Submount compile fix

and file attachment too.

big warning

compilation works, but there are some deadlock that take process (like a file
manager or simple ls) to D status.

i've no idea how to solve those, keeping the mutexes.

fixed!

the problem was in a typo in the patch which changed lock/unlock sequence.

now works.

i've edited the patch so you can apply without any kernel version tests, done
in the patch.

please apply to portage and mark as fixed

Created an attachment (id=82851) [details]
patch to fix 2.6.16

full story and download link here:
http://www.patriziobassi.it/blog/?p=24

fixed and sent upstream, thanks!