[PATCH] mm/memblock.c: Correctly check whether to trim a block

[PATCH] mm/memblock.c: Correctly check whether to trim a block

[Laura Abbott]

Currently in __memblock_remove, the check to trim the top of
a block off only checks if the requested base is less than the
memblock end. If the end of the requested region is equal to
the start of a memblock, this will incorrectly try to remove
the block, possibly causing an integer underflow:

   ---------------------------------------
   |                    |                |
   |                    |                |
  base              end = rgn->base    rend

An additional check is needed to see if the end of the requested
region is greater than the memblock region:

   ----------------------
   |                     |
   |                     |
  rgn->base    base     rend      end
                |                  |
                |                  |
                --------------------

Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
---
 mm/memblock.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/mm/memblock.c b/mm/memblock.c
index 5338237..e174ee0 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -459,7 +459,7 @@ static long __init_memblock __memblock_remove(struct memblock_type *type,
 		}
 
 		/* And check if we need to trim the top of a block */
-		if (base < rend)
+		if (base < rend && end > rend)
 			rgn->size -= rend - base;
 
 	}
-- 
1.7.8.3

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm/memblock.c: Correctly check whether to trim a block

[Benjamin Herrenschmidt]

On Wed, 2012-03-28 at 19:25 -0700, Laura Abbott wrote:
> Currently in __memblock_remove, the check to trim the top of
> a block off only checks if the requested base is less than the
> memblock end. If the end of the requested region is equal to
> the start of a memblock, this will incorrectly try to remove
> the block, possibly causing an integer underflow:
> 
>    ---------------------------------------
>    |                    |                |
>    |                    |                |
>   base              end = rgn->base    rend
> 
> An additional check is needed to see if the end of the requested
> region is greater than the memblock region:

__memblock_remove() open coded logic is gone now, re-implemented
in term of memblock_isolate_range()... though I suppose your
patch might have value in -stable...

Cheers,
Ben.


>    ----------------------
>    |                     |
>    |                     |
>   rgn->base    base     rend      end
>                 |                  |
>                 |                  |
>                 --------------------
> 
> Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
> ---
>  mm/memblock.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 5338237..e174ee0 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -459,7 +459,7 @@ static long __init_memblock __memblock_remove(struct memblock_type *type,
>  		}
>  
>  		/* And check if we need to trim the top of a block */
> -		if (base < rend)
> +		if (base < rend && end > rend)
>  			rgn->size -= rend - base;
>  
>  	}


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>