Re: [PATCH] sched: mark parent and real_parent as __rcu

[Ingo Molnar <mingo@elte.hu>]

* Kees Cook <keescook@chromium.org> wrote:

> On Wed, Dec 14, 2011 at 11:08 PM, Ingo Molnar <mingo@elte.hu> wrote:
> > * Kees Cook <keescook@chromium.org> wrote:
> >
> >> The parent and real_parent pointers should be considered __rcu, since
> >> they should be held under either tasklist_lock or rcu_read_lock.
> >>
> >> Signed-off-by: Kees Cook <keescook@chromium.org>
> >> ---
> >>  include/linux/sched.h |    4 ++--
> >>  1 files changed, 2 insertions(+), 2 deletions(-)
> >
> > Did you get some warning that alerted you to this problem? If
> > yes then please include it in the changelog, so that people
> > hitting the message can easily find it.
> >
> > Same consideration applies the second patch as well.
> 
> No, this was via code inspection related to reviews of the 
> Yama LSM's use of these structures and comparing the RCU and 
> tasklist_lock behavior/requirements seen in the rest of the 
> kernel. Most attempts at using sparse after this change were 
> not very successful due to the high volume of unrelated noise 
> currently seen with "make C=2", but I was able to confirm that 
> the apparmor and tomoyo patches were flagged by sparse, at 
> least. When this __rcu marking is in place, if there is a 
> place that might need rcu_dereference, it would show up as 
> "warning: dereference of noderef expression".

ok - i'll apply the first patch, please send the second one 
against -tip, it does not apply anymore. (scheduler got moved to 
kernel/sched/)

Thanks,

	Ingo