lkml.org 
[lkml]   [2012]   [Feb]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH RFC tip/core/rcu 04/41] rcu: Add diagnostic for misaligned rcu_head structures
On Thu, Feb 02, 2012 at 08:22:35AM -0800, Paul E. McKenney wrote:
> On Wed, Feb 01, 2012 at 05:00:26PM -0800, Josh Triplett wrote:
> > On Wed, Feb 01, 2012 at 11:41:22AM -0800, Paul E. McKenney wrote:
> > > The push for energy efficiency will require that RCU tag rcu_head
> > > structures to indicate whether or not their invocation is time critical.
> > > This tagging is best carried out in the bottom bits of the ->next
> > > pointers in the rcu_head structures. This tagging requires that the
> > > rcu_head structures be properly aligned, so this commit adds the required
> > > diagnostics.
> >
> > > --- a/kernel/rcutree.c
> > > +++ b/kernel/rcutree.c
> > > @@ -1707,6 +1707,7 @@ __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
> > > unsigned long flags;
> > > struct rcu_data *rdp;
> > >
> > > + WARN_ON_ONCE((unsigned long)head & 0x3); /* Misaligned rcu_head! */
> >
> > I wonder if this should go out in one of the wrapper macros, so that the
> > _ONCE means once per call site?
>
> Certainly past experience introducing lockdep-RCU would argue that way. ;-)
>
> However, my testing thus far (famous last words!) has found no
> misalignments. So I expect that these will be caught as they are
> inserted, so one at a time should be fine.

Fair enough. Doesn't seem worth introducing a new wrapper macro for,
then, since it doesn't look like an appropriate one currently exists.

- Josh Triplett


\
 
 \ /
  Last update: 2012-02-02 21:13    [W:0.841 / U:0.072 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site