Re: [PATCH 1/1] futex: check PF_KTHREAD rather than !p->mm to filter out kthreads

[Davidlohr Bueso <dave@stgolabs.net>]

On Mon, 2015-02-02 at 15:05 +0100, Oleg Nesterov wrote:
> attach_to_pi_owner() checks p->mm to prevent attaching to kthreads and
> this looks doubly wrong:
> 
> 1. It should actually check PF_KTHREAD, kthread can do use_mm().
> 
> 2. If this task is not kthread and it is actually the lock owner we can
>    wrongly return -EPERM instead of -ESRCH or retry-if-EAGAIN.
> 
>    And note that this wrong EPERM is the likely case unless the exiting
>    task is (auto)reaped quickly, we check ->mm before PF_EXITING.
> 
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
>  kernel/futex.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/kernel/futex.c b/kernel/futex.c
> index 63678b5..b101381 100644
> --- a/kernel/futex.c
> +++ b/kernel/futex.c
> @@ -900,7 +900,7 @@ static int attach_to_pi_owner(u32 uval, union futex_key *key,
>  	if (!p)
>  		return -ESRCH;
>  
> -	if (!p->mm) {
> +	if (unlikely(p->flags & PF_KTHREAD)) {
>  		put_task_struct(p);
>  		return -EPERM;
>  	}

Futexes aren't the only naughty checkers, a quick search shows that, at
least, the oom killer and proc have this same problem. Should we make
this generic and update accordingly? ie:

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 8db31ef..b0d37d6 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1991,6 +1991,11 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
 #define used_math() tsk_used_math(current)
 
+static inline bool task_is_kthread(struct task_struct *task)
+{
+	return task->flags & PF_KTHREAD;
+}
+
 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
  * __GFP_FS is also cleared as it implies __GFP_IO.
  */


Thanks,
Davidlohr