Re: [PATCH 2/2] ptrace: Convert ptrace_attach() to use lock guards

[Oleg Nesterov]

On 11/02, Peter Zijlstra wrote:
>
> Created as testing for the conditional guard infrastructure.

This patch scares me ;) I need to get used to guard/etc.
But looks correct.

Reviewed-by: Oleg Nesterov <oleg@xxxxxxxxxx>






> +			/* SEIZE doesn't trap tracee on attach */
> +			if (!seize)
> +				send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);

This is offtopic, but with or without this patch it is a bit
ugly to drop ->siglock and take it again right after that.
We can do (later) a minor cleanup on top of this change.


--- x/kernel/ptrace.c	2023-11-02 16:03:37.646838530 +0100
+++ x/kernel/ptrace.c	2023-11-02 16:05:52.171052506 +0100
@@ -386,10 +386,14 @@
 	return 0;
 }
 
-static inline void ptrace_set_stopped(struct task_struct *task)
+static inline void ptrace_set_stopped(struct task_struct *task, bool seize)
 {
 	guard(spinlock)(&task->sighand->siglock);
 
+	/* SEIZE doesn't trap tracee on attach */
+	if (!seize)
+		send_signal_locked(SIGSTOP, SEND_SIG_PRIV, task, PIDTYPE_PID);
+
 	/*
 	 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
 	 * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
@@ -470,11 +474,6 @@
 			task->ptrace = flags;
 
 			ptrace_link(task, current);
-
-			/* SEIZE doesn't trap tracee on attach */
-			if (!seize)
-				send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
-
 			ptrace_set_stopped(task);
 		}
 	}