Re: [RCU PATCH 06/14] net: sched: fw use RCU

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Fri, Mar 14, 2014 at 11:50:05AM -0700, Paul E. McKenney wrote:
> On Fri, Mar 14, 2014 at 08:38:10AM -0700, Paul E. McKenney wrote:
> > On Fri, Mar 14, 2014 at 06:46:38AM -0700, Eric Dumazet wrote:
> > > On Fri, 2014-03-14 at 06:28 -0700, Paul E. McKenney wrote:
> > > 
> > > > Now I just need to figure out why it -causes- a BUILD_BUG_ON for the
> > > > cris and m68k architectures in Fengguang Wu's testing...
> > > 
> > > Right, they don't necessarily align pointers ;)
> > > 
> > > alignof(void *) ?
> > 
> > It complained about the size rather than the alignment, but it would
> > certainly make sense to constrain the rcu_head structure's alignment,
> > now that you mention it.  I will do that and see if it helps.  ;-)
> > 
> > Thank you for the tip!
> 
> And you were quite right...  When the alignment is bad, it complains
> about the size.  And although I can specify that struct rcu_head be
> aligned, but __attribute__(packed) overrides this when specified on an
> enclosing structure.  As does the -fpack-struct gcc command-line option.
> 
> Other thoughts on getting around this problem?
> 
> Hmmm...  Maybe if cris and m68k guaranteed to keep functions out of the
> bottom four pages of memory...

Or I could find some other large area of memory that never contains
functions, for example, the beginning of data or bss (if we still have
those).  Then check for a range rather than a cutoff.

							Thanx, Paul