Re: [PATCH] mm: Fix NULL ptr dereference in __count_immobile_pages

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Wed, 11 Jan 2012 09:48:02 +0100
Michal Hocko <mhocko@suse.cz> wrote:

> On Tue 10-01-12 13:31:08, David Rientjes wrote:
> > On Tue, 10 Jan 2012, Michal Hocko wrote:
> [...]
> > >  mm/page_alloc.c |   11 +++++++++++
> > >  1 files changed, 11 insertions(+), 0 deletions(-)
> > > 
> > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > > index 2b8ba3a..485be89 100644
> > > --- a/mm/page_alloc.c
> > > +++ b/mm/page_alloc.c
> > > @@ -5608,6 +5608,17 @@ __count_immobile_pages(struct zone *zone, struct page *page, int count)
> > >  bool is_pageblock_removable_nolock(struct page *page)
> > >  {
> > >  	struct zone *zone = page_zone(page);
> > > +	unsigned long pfn = page_to_pfn(page);
> > > +

Hmm, I don't like to use page_zone() when we know the page may not be initialized.
Shouldn't we add

	if (!node_online(page_to_nid(page))
		return false;
?

But...hmm. I think we should return 'true' here for removing a section with a hole
finally....(Now, false will be safe.)



> > > +	/*
> > > +	 * We have to be careful here because we are iterating over memory
> > > +	 * sections which are not zone aware so we might end up outside of
> > > +	 * the zone but still within the section.
> > > +	 */
> > > +	if (!zone || zone->zone_start_pfn > pfn ||
> > > +			zone->zone_start_pfn + zone->spanned_pages <= pfn)
> > > +		return false;
> > > +
> > >  	return __count_immobile_pages(zone, page, 0);
> > >  }
> > >  
> > 
> > This seems partially bogus, why would
> > 
> > 	page_zone(page)->zone_start_pfn > page_to_pfn(page) ||
> > 	page_zone(page)->zone_start_pfn + page_zone(page)->spanned_pages <= page_to_pfn(page)
> > 
> > ever be true?  That would certainly mean that the struct zone is corrupted 
> > and seems to be unnecessary to fix the problem you're addressing.
> 
> Not really. Consider the case when the node 0 is present. Uninitialized
> page would lead to node=0, zone=0 and then we have to check for the zone
> boundaries.
> 



> > I think this should be handled in is_mem_section_removable() on the pfn 
> > rather than using the struct page in is_pageblock_removable_nolock() and 
> > converting back and forth.  We should make sure that any page passed to 
> > is_pageblock_removable_nolock() is valid.
> 
> Yes, I do not like pfn->page->pfn dance as well and in fact I do not
> have a strong opinion which one is better. I just put it at the place
> where we care about zone to be more obvious. If others think that I
> should move the check one level higher I'll do that. I just think this
> is more obvious.
> 
Hmm, mem_section and pageblock is a different chunk...
And, IIUC, in some IBM machines, section may includes several zones.
Please taking care of that if you move this to is_mem_section_removable()...

Thanks,
-Kame








--
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>