Re: [PATCH] local_t : Documentation

[Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>]

* Andrew Morton (akpm@osdl.org) wrote:
> On Mon, 8 Jan 2007 22:14:46 -0500
> Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> wrote:
> 
> > +* How to use local atomic operations
> > +
> > +#include <linux/percpu.h>
> > +#include <asm/local.h>
> > +
> > +static DEFINE_PER_CPU(local_t, counters) = LOCAL_INIT(0);
> > +
> > +
> > +* Counting
> > +
> > +In preemptible context, use get_cpu_var() and put_cpu_var() around local atomic
> > +operations : it makes sure that preemption is disabled around write access to
> > +the per cpu variable. For instance :
> > +
> > +	local_inc(&get_cpu_var(counters));
> > +	put_cpu_var(counters);
> 
> Confused.  The whole point behind local_t is that we can do
> atomic-wrt-interrupts inc and dec on them.
> 
> Consequently, as atomic-wrt-interrupts means atomic-wrt-preemption, there
> is no need to do a preempt_disable() around local_inc() and local_dec().
> 

Hi Andrew,

Not exactly : the increment operation is atomic, but not the selection of the
local variable. local_inc(&__get_cpu_var()) implies the following sequence 
of operations :

1 - Get the variable copy corresponding to the currently running CPU.
2 - atomically increment the variable.

It would be wrong to be scheduled on another CPU between 1 and 2, because the
atomic increment should only be done by the CPU "owner" of the local variable,
as the local atomic increment is not atomic wrt other CPUs.

Mathieu

-- 
OpenPGP public key:              http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint:     8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68