On Mon, Sep 10, 2012 at 03:58:45PM -0700, Paul E. McKenney wrote:

quoted

Anyway, should I resend the patchset with this thing fixed?

Please feel free to do so.  ;-)

							Thanx, Paul

So here it is.

On the menu for the updated branch:

- fixed stable tags (use # 3.3+)
- not sure what to do with "alpha: Fix preemption handling in idle loop",
  tag as stable or not?
- added tested-by Michael Cree [off-list ref] for alpha patches
- propagate rename to rcu_user_*_after_irq() in the commit changelog
- remove trailing #endif in "rcu: New rcu_user_enter_after_irq() and rcu_user_exit_after_irq() APIs"
  and early rcu_user_hooks_switch() header as well (reported by WU Fenguang).
- add rcu_user_hooks_switch() header in "rcu: Switch task's syscall hooks on context switch"

I believe this all should also fix what WU Fenguang reported:

	[rcu:rcu:next 54/82] include/linux/rcupdate.h:202:20: error: redefinition of 'rcu_user_exit'
	[rcu:rcu:next 54/82] include/linux/rcupdate.h:401:19: error: redefinition of 'rcu_read_lock_bh_held'
	[rcu:rcu:idle 2/27] include/linux/rcupdate.h:969:2: error: #endif without #if

I'm not reposting the patches because few things have actually changed, except a few neats and some stuff
that have been moved from a patch to another to unbreak bisection. But the end result is about the same.
You can check the difference with a git diff between our branches.

If you want to, though, I can repost the patches.

This is pullable from:

git://github.com/fweisbec/linux-dynticks.git
	rcu/idle-for-v3.7

Thanks.