ARM: Only initialize max_mapnr when using flatmem

Commit Message

Russell King - ARM Linux Jan. 8, 2013, 4:46 p.m. UTC

On Mon, Oct 22, 2012 at 03:51:54PM -0400, Michael Spang wrote:
> The expression used to initialize max_mapnr is not appropriate for
> sparsemem, since mem_map is NULL and pfn_to_page() only works on valid
> pages. This patch leaves max_mapnr uninitialized for sparsemem.
> 
> For flatmem, the expression simplifies to copying max_pfn.
> 
> Signed-off-by: Michael Spang <[e-mail]>
> ---
>  arch/arm/mm/init.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> I'm not sure we need this variable it all - it looks like there are no
> users.

Indeed, I've just confirmed that. I suggest we just get rid of it instead,
like this:

 arch/arm/mm/init.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

Comments

Russell King - ARM Linux April 3, 2013, 4:37 p.m. UTC | #1

On Tue, Jan 08, 2013 at 04:46:40PM +0000, Russell King - ARM Linux wrote:
> On Mon, Oct 22, 2012 at 03:51:54PM -0400, Michael Spang wrote:
> > The expression used to initialize max_mapnr is not appropriate for
> > sparsemem, since mem_map is NULL and pfn_to_page() only works on valid
> > pages. This patch leaves max_mapnr uninitialized for sparsemem.
> > 
> > For flatmem, the expression simplifies to copying max_pfn.
> > 
> > Signed-off-by: Michael Spang <[e-mail]>
> > ---
> >  arch/arm/mm/init.c |    4 +++-
> >  1 files changed, 3 insertions(+), 1 deletions(-)
> > 
> > I'm not sure we need this variable it all - it looks like there are no
> > users.
> 
> Indeed, I've just confirmed that. I suggest we just get rid of it instead,
> like this:
> 
>  arch/arm/mm/init.c |    6 +++++-
>  1 files changed, 5 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
> index ad722f1..bd93bdc 100644
> --- a/arch/arm/mm/init.c
> +++ b/arch/arm/mm/init.c
> @@ -600,7 +600,11 @@ void __init mem_init(void)
>  	extern u32 itcm_end;
>  #endif
>  
> -	max_mapnr   = pfn_to_page(max_pfn + PHYS_PFN_OFFSET) - mem_map;
> +	/*
> +	 * We used to initialize max_mapnr here to be the same as max_pfn,
> +	 * but as nothing uses it (we don't use the generic pfn_valid())
> +	 * this has become unnecessary.
> +	 */
>  
>  	/* this will put all unused low memory onto the freelists */
>  	free_unused_memmap(&meminfo);
> 

I notice there wasn't a response to this, and the patch is in the patch
system.  I'm going to discard that patch now, and assume that this issue
is dead.  Thanks.

Patch

diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
index ad722f1..bd93bdc 100644
--- a/arch/arm/mm/init.c
+++ b/arch/arm/mm/init.c
@@ -600,7 +600,11 @@  void __init mem_init(void)
 	extern u32 itcm_end;
 #endif
 
-	max_mapnr   = pfn_to_page(max_pfn + PHYS_PFN_OFFSET) - mem_map;
+	/*
+	 * We used to initialize max_mapnr here to be the same as max_pfn,
+	 * but as nothing uses it (we don't use the generic pfn_valid())
+	 * this has become unnecessary.
+	 */
 
 	/* this will put all unused low memory onto the freelists */
 	free_unused_memmap(&meminfo);