Re: [PATCH 0/6] Optimize the cpu hotplug locking -v2

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

On Thu, Oct 10, 2013 at 08:50:32PM +0200, Peter Zijlstra wrote:
> On Thu, Oct 10, 2013 at 02:43:27PM -0400, Steven Rostedt wrote:
> > On Thu, 10 Oct 2013 11:10:35 -0700
> > Linus Torvalds <torvalds@linux-foundation.org> wrote:
> >      .. now we can free all the percpu data and kill the CPU ..
> > > 
> > > without any locking anywhere - not stop-machine, not anything. If
> > > somebody is doing a "for_each_cpu()" (under just a regular
> > > rcu_read_lock()) and they see the bit set while it's going down, who
> > > cares? The CPU is still there, the data is accessible..
> > 
> > The problem is that rcu_read_lock() requires preemption disabled unless
> > you are using the preemptable rcu tree version. There's always
> > srcu_read_lock() but that's not so free. It's basically the same as
> > what Peter is doing.
> 
> No, srcu is actually more expensive in the fast path. Although possibly
> we could make SRCU more complex ;-)

Indeed.  Especially if we wanted to get rid of the memory barriers in
srcu_read_lock() and srcu_read_unlock() and still allow SRCU to be used
from idle and offline CPUs!  ;-)

							Thanx, Paul

> > There's places in the kernel that does for_each_cpu() that I'm sure you
> > don't want to disable preemption for. Especially when you start having
> > 4096 CPU machines!
> 
> :-)

But shouldn't you only have to disable over each separate pass through
the loop rather across the entire loop?  "I guarantee that the CPU I
just handed you will remain online through the body of the loop" rather
than "I guarantee that no CPU is going anywhere through the entire loop."

							Thanx, Paul