Re: [PATCH] fix PTRACE_ATTACH race with real parent's wait calls

[Roland McGrath]

I've thought of one additional race associated with this.  If the real
parent bogusly tried a ptrace (rather than a wait) in that same interval,
it could pass the ptrace_check_attach test and then a moment later not
actually be that task's ->parent any more, which might break all sorts of
stuff and hit BUGs or who knows what.

This further patch should rule out that bogus ptrace call.

Thanks,
Roland


Index: linux-2.6/kernel/ptrace.c
===================================================================
RCS file: /home/roland/redhat/bkcvs/linux-2.5/kernel/ptrace.c,v
retrieving revision 1.36
diff -u -b -B -p -r1.36 ptrace.c
--- linux-2.6/kernel/ptrace.c 23 Sep 2004 23:35:16 -0000 1.36
+++ linux-2.6/kernel/ptrace.c 25 Sep 2004 01:35:02 -0000
@@ -82,7 +82,8 @@ int ptrace_check_attach(struct task_stru
 	 */
 	read_lock(&tasklist_lock);
 	if ((child->ptrace & PT_PTRACED) && child->parent == current &&
-	    child->signal != NULL) {
+	    (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
+	    && child->signal != NULL) {
 		ret = 0;
 		spin_lock_irq(&child->sighand->siglock);
 		if (child->state == TASK_STOPPED) {

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/