ptrace: mv send-SIGSTOP from do_fork() to ptrace_init_task()

If the new child is traced, do_fork() adds the pending SIGSTOP.
It assumes that either it is traced because of auto-attach or the
tracer attached later, in both cases sigaddset/set_thread_flag is
correct even if SIGSTOP is already pending.

Now that we have PTRACE_SEIZE this is no longer right in the latter
case. If the tracer does PTRACE_SEIZE after copy_process() makes the
child visible the queued SIGSTOP is wrong.

We could check PT_SEIZED bit and change ptrace_attach() to set both
PT_PTRACED and PT_SEIZED bits simultaneously but see the next patch,
we need to know whether this child was auto-attached or not anyway.

So this patch simply moves this code to ptrace_init_task(), this
way we can never race with ptrace_attach().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>

1 files changed, 3 insertions, 0 deletions

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index fd8669fc339f..9b5d2c901d06 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -227,6 +227,9 @@ static inline void ptrace_init_task(struct task_struct *child, bool ptrace)
         if (unlikely(ptrace) && current->ptrace) {
                 child->ptrace = current->ptrace;
                 __ptrace_link(child, current->parent);
+
+                sigaddset(&child->pending.signal, SIGSTOP);
+                set_tsk_thread_flag(child, TIF_SIGPENDING);
         }
 }