[PATCH] Use correct page frame number to grab page zone lock

[PATCH] Use correct page frame number to grab page zone lock

[Nathan Fontenot]

There are situations in which the 'pfn' variable used to calculate the
page zone can be invalid.  One case is in the hotplug memory remove path
where we are trying to remove memory the last valid page frame for the
system.  Passing an invalid page frame number to the pfn_to_page() macro
will cause the system to oops.

Additionally I think this also solves an issue where we may be grabbing
the wrong zone->lock.  This would occur when 'pfn' refers to a page
frame beyond what we are examining and lies in a different zone.

Using the starting page frame number passed in appears to resolve
both of these issues.  This patch also removes the redundant
initialization of 'pfn', it is set explicitly and then again in the for loop.


Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
---

Index: linux-2.6/mm/page_isolation.c
===================================================================
--- linux-2.6.orig/mm/page_isolation.c	2008-10-13 12:00:46.000000000 -0500
+++ linux-2.6/mm/page_isolation.c	2008-10-29 11:18:39.000000000 -0500
@@ -119,7 +119,6 @@
 	struct zone *zone;
 	int ret;
 
-	pfn = start_pfn;
 	/*
 	 * Note: pageblock_nr_page != MAX_ORDER. Then, chunks of free page
 	 * is not aligned to pageblock_nr_pages.
@@ -133,7 +132,7 @@
 	if (pfn < end_pfn)
 		return -EBUSY;
 	/* Check all pages are free or Marked as ISOLATED */
-	zone = page_zone(pfn_to_page(pfn));
+	zone = page_zone(pfn_to_page(start_pfn));
 	spin_lock_irqsave(&zone->lock, flags);
 	ret = __test_page_isolated_in_pageblock(start_pfn, end_pfn);
 	spin_unlock_irqrestore(&zone->lock, flags);

Re: [PATCH] Use correct page frame number to grab page zone lock

[Nathan Fontenot]

Disregard.  It appears that Gerald Schaefer submitted a patch to fix
this already.

http://lkml.org/lkml/2008/10/29/142

-Nathan

Nathan Fontenot wrote:
> There are situations in which the 'pfn' variable used to calculate the
> page zone can be invalid.  One case is in the hotplug memory remove path
> where we are trying to remove memory the last valid page frame for the
> system.  Passing an invalid page frame number to the pfn_to_page() macro
> will cause the system to oops.
> 
> Additionally I think this also solves an issue where we may be grabbing
> the wrong zone->lock.  This would occur when 'pfn' refers to a page
> frame beyond what we are examining and lies in a different zone.
> 
> Using the starting page frame number passed in appears to resolve
> both of these issues.  This patch also removes the redundant
> initialization of 'pfn', it is set explicitly and then again in the for 
> loop.
> 
> 
> Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
> ---
> 
> Index: linux-2.6/mm/page_isolation.c
> ===================================================================
> --- linux-2.6.orig/mm/page_isolation.c    2008-10-13 12:00:46.000000000 
> -0500
> +++ linux-2.6/mm/page_isolation.c    2008-10-29 11:18:39.000000000 -0500
> @@ -119,7 +119,6 @@
>     struct zone *zone;
>     int ret;
> 
> -    pfn = start_pfn;
>     /*
>      * Note: pageblock_nr_page != MAX_ORDER. Then, chunks of free page
>      * is not aligned to pageblock_nr_pages.
> @@ -133,7 +132,7 @@
>     if (pfn < end_pfn)
>         return -EBUSY;
>     /* Check all pages are free or Marked as ISOLATED */
> -    zone = page_zone(pfn_to_page(pfn));
> +    zone = page_zone(pfn_to_page(start_pfn));
>     spin_lock_irqsave(&zone->lock, flags);
>     ret = __test_page_isolated_in_pageblock(start_pfn, end_pfn);
>     spin_unlock_irqrestore(&zone->lock, flags);
> -- 
> 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/