Re: [PATCH] ptrace: restore smp_rmb() in __ptrace_may_access()

[Oleg Nesterov <oleg@redhat.com>]

On 05/29, Jann Horn wrote:
>
> (I have no clue whatsoever what the relevant tree for this is, but I
> guess Oleg is the relevant maintainer?)

we usually route ptrace changes via -mm tree, plus I lost my account on korg.

> --- a/kernel/ptrace.c
> +++ b/kernel/ptrace.c
> @@ -324,6 +324,16 @@ static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
>  	return -EPERM;
>  ok:
>  	rcu_read_unlock();
> +	/*
> +	 * If a task drops privileges and becomes nondumpable (through a syscall
> +	 * like setresuid()) while we are trying to access it, we must ensure
> +	 * that the dumpability is read after the credentials; otherwise,
> +	 * we may be able to attach to a task that we shouldn't be able to
> +	 * attach to (as if the task had dropped privileges without becoming
> +	 * nondumpable).
> +	 * Pairs with a write barrier in commit_creds().
> +	 */
> +	smp_rmb();

(I am wondering if smp_acquire__after_ctrl_dep() could be used instead, just to
 make this code look more confusing)

>  	mm = task->mm;

while at it, could you also change this into mm = READ_ONCE(task->mm) ?

Oleg.