Re: [patch] mm, oom: remove task_lock protecting comm printing

[Vladimir Davydov <vdavydov@parallels.com>]

On Wed, Sep 23, 2015 at 06:50:22PM +0900, Sergey Senozhatsky wrote:
> On (09/23/15 11:43), Michal Hocko wrote:
> [..]
> > > > the previous name was already null terminated,
> > > 
> > > Yeah, but if the old name is shorter than the new one, set_task_comm()
> > > overwrites the terminating null of the old name before writing the new
> > > terminating null, so there is a short time window during which tsk->comm
> > > might be not null-terminated, no?
> > 
> > Not really:
> >         case PR_SET_NAME:
> >                 comm[sizeof(me->comm) - 1] = 0;
> >                 if (strncpy_from_user(comm, (char __user *)arg2,
> >                                       sizeof(me->comm) - 1) < 0)
> >                         return -EFAULT;
> > 
> > So it first writes the terminating 0 and only then starts copying.

It writes 0 to a temporary buffer, not to tsk->comm, so I don't think
it's related. However, reading tsk->comm w/o locking must be safe
anyway, because tsk->comm[TASK_COMM_LEN-1] is always 0 (inherited from
init_task) and it never gets overwritten, because __set_task_comm() uses
strlcpy().

> 
> right.
> 
> hm, shouldn't set_task_comm()->__set_task_comm() do the same?

I don't think so - see above.

Thanks,
Vladimir

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>