On Wed, Sep 10, 2014 at 02:26:54AM -0500, Chuck Ebbert wrote:

On Tue,  9 Sep 2014 10:42:27 +0100
Aaron Tomlin [off-list ref] wrote:

quoted

+void task_stack_end_magic(struct task_struct *tsk)
+{
+	unsigned long *stackend;
+
+	stackend = end_of_stack(tsk);
+	*stackend = STACK_END_MAGIC;	/* for overflow detection */
+}
+

For clarity this should probably be called set_task_stack_end_magic().

Agreed.

And has this been tested on parisc and metag, which use STACK_GROWSUP ?
I can't see how end_of_stack() as it's defined now could work on those archs.

AFAIU, dup_task_struct() has always done this explicitly.
I see no reason why init_task requires special attention.

-- 
Aaron Tomlin