[PATCH] handling of failed copy_thread() in do_fork()

[Alexander Viro]

	If copy_thread() fails (it can't happen on x86, but on other
architectures, e.g. on sparc, it's possible) do_fork() doesn't do
full cleanup - it forgets to do exit_mm(). Obvious fix follows.
Please, apply.
								Al

diff -urN S4-pre7/kernel/fork.c S4-pre7-fork/kernel/fork.c
--- S4-pre7/kernel/fork.c	Wed Apr 25 20:43:12 2001
+++ S4-pre7-fork/kernel/fork.c	Wed Apr 25 22:12:39 2001
@@ -652,7 +652,7 @@
 		goto bad_fork_cleanup_sighand;
 	retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
 	if (retval)
-		goto bad_fork_cleanup_sighand;
+		goto bad_fork_cleanup_mm;
 	p->semundo = NULL;
 	
 	/* Our parent execution domain becomes current domain
@@ -708,6 +708,8 @@
 		down(&sem);
 	return retval;
 
+bad_fork_cleanup_mm:
+	exit_mm(p);
 bad_fork_cleanup_sighand:
 	exit_sighand(p);
 bad_fork_cleanup_fs: