Date: Fri, 7 Mar 2008 11:53:40 +0000
From: Mel Gorman <mel@csn.ul.ie>
To: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Paul Jackson <pj@sgi.com>, akpm@linux-foundation.org,
       linux-kernel@vger.kernel.org, linux-mm@kvack.org, rientjes@google.com,
       nacc@us.ibm.com, kamezawa.hiroyu@jp.fujitsu.com, clameter@sgi.com,
       Eric Whitney <eric.whitney@hp.com>
Subject: Re: [PATCH] 2.6.25-rc3-mm1 - Mempolicy - update stale documentation and comments
Message-ID: <20080307115339.GD26229@csn.ul.ie>
References: <20071109143226.23540.12907.sendpatchset@skynet.skynet.ie> <20071109143406.23540.41284.sendpatchset@skynet.skynet.ie> <20080228230140.321581a4.pj@sgi.com> <1204662057.5338.104.camel@localhost>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-15
Content-Disposition: inline
In-Reply-To: <1204662057.5338.104.camel@localhost>
User-Agent: Mutt/1.5.13 (2006-08-11)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 4892
Lines: 108

On (04/03/08 15:20), Lee Schermerhorn didst pronounce:
> Was Re: [PATCH 5/6] Filter based on a nodemask as well as a gfp_mask
> 
> On Thu, 2008-02-28 at 23:01 -0600, Paul Jackson wrote:
> > Mel wrote:
> > > A positive benefit of
> > > this is that allocations using MPOL_BIND now use the local-node-ordered
> > > zonelist instead of a custom node-id-ordered zonelist.
> > 
> > Could you update the now obsolete documentation (perhaps just delete
> > the no longer correct remark):
> > 
> > Documentation/vm/numa_memory_policy.txt:
> > 
> >         MPOL_BIND:  This mode specifies that memory must come from the
> >         set of nodes specified by the policy.
> > 
> >             The memory policy APIs do not specify an order in which the nodes
> >             will be searched.  However, unlike "local allocation", the Bind
> >             policy does not consider the distance between the nodes.  Rather,
> >             allocations will fallback to the nodes specified by the policy in
> >             order of numeric node id.  Like everything in Linux, this is subject
> >             to change.
> > 
> 
> How's this:
> 
> PATCH Mempolicy:  update documentation and comments
> 
> Address stale comments and numa_memory_policy.txt discussion
> based on Mel Gorman's changes to page allocation zonelist handling.
> 
> Specifically:  mpol_free() and mpol_copy() no longer need to deal
> with a custom zonelist for MPOL_BIND policy, and MPOL_BIND now
> allocates memory from the nearest node with available memory in the
> specified nodemask.
> 
> In "fixing" the mpol_free() and mpol_copy() comments in mempolicy.h,
> I wanted to replace them with something, rather than just deleting
> them.  So, I described the reference counting.  Not directly related
> to the zonelist changes, but useful, IMO.
> 
> Signed-off-by:  Lee Schermerhorn <lee.schermerhorn@hp.com>
> 

Looks great.

Acked-by: Mel Gorman <mel@csn.ul.ie>

>  Documentation/vm/numa_memory_policy.txt |   11 +++--------
>  include/linux/mempolicy.h               |    8 +++++---
>  2 files changed, 8 insertions(+), 11 deletions(-)
> 
> Index: linux-2.6.25-rc3-mm1/Documentation/vm/numa_memory_policy.txt
> ===================================================================
> --- linux-2.6.25-rc3-mm1.orig/Documentation/vm/numa_memory_policy.txt	2008-01-24 17:58:37.000000000 -0500
> +++ linux-2.6.25-rc3-mm1/Documentation/vm/numa_memory_policy.txt	2008-03-04 14:44:51.000000000 -0500
> @@ -182,14 +182,9 @@ Components of Memory Policies
>  	    The Default mode does not use the optional set of nodes.
>  
>  	MPOL_BIND:  This mode specifies that memory must come from the
> -	set of nodes specified by the policy.
> -
> -	    The memory policy APIs do not specify an order in which the nodes
> -	    will be searched.  However, unlike "local allocation", the Bind
> -	    policy does not consider the distance between the nodes.  Rather,
> -	    allocations will fallback to the nodes specified by the policy in
> -	    order of numeric node id.  Like everything in Linux, this is subject
> -	    to change.
> +	set of nodes specified by the policy.  Memory will be allocated from
> +	the node in the set with sufficient free memory that is closest to
> +	the node where the allocation takes place.
>  
>  	MPOL_PREFERRED:  This mode specifies that the allocation should be
>  	attempted from the single node specified in the policy.  If that
> Index: linux-2.6.25-rc3-mm1/include/linux/mempolicy.h
> ===================================================================
> --- linux-2.6.25-rc3-mm1.orig/include/linux/mempolicy.h	2008-03-04 14:19:06.000000000 -0500
> +++ linux-2.6.25-rc3-mm1/include/linux/mempolicy.h	2008-03-04 14:38:29.000000000 -0500
> @@ -54,11 +54,13 @@ struct mm_struct;
>   * mmap_sem.
>   *
>   * Freeing policy:
> - * When policy is MPOL_BIND v.zonelist is kmalloc'ed and must be kfree'd.
> - * All other policies don't have any external state. mpol_free() handles this.
> + * Mempolicy objects are reference counted.  A mempolicy will be freed when
> + * mpol_free() decrements the reference count to zero.
>   *
>   * Copying policy objects:
> - * For MPOL_BIND the zonelist must be always duplicated. mpol_clone() does this.
> + * mpol_copy() allocates a new mempolicy and copies the specified mempolicy
> + * to the new storage.  The reference count of the new object is initialized
> + * to 1, representing the caller of mpol_copy().
>   */
>  struct mempolicy {
>  	atomic_t refcnt;
> 
> 

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/