Re: [PATCH] page_alloc: Fix freeing non-compound pages

[peterz]

On Tue, Sep 22, 2020 at 03:00:17PM +0100, Matthew Wilcox (Oracle) wrote:
> Here is a very rare race which leaks memory:
> 
> Page P0 is allocated to the page cache.
> Page P1 is free.
> 
> Thread A		Thread B		Thread C
> find_get_entry():
> xas_load() returns P0
> 						Removes P0 from page cache
> 						Frees P0
> 						P0 merged with its buddy P1
> 			alloc_pages(GFP_KERNEL, 1) returns P0
> 			P0 has refcount 1
> page_cache_get_speculative(P0)
> P0 has refcount 2
> 			__free_pages(P0)
> 			P0 has refcount 1
> put_page(P0)
> P1 is not freed
> 
> Fix this by freeing all the pages in __free_pages() that won't be freed
> by the call to put_page().  It's usually not a good idea to split a page,
> but this is a very unlikely scenario.
> 
> Fixes: e286781d5f2e ("mm: speculative page references")
> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> ---
>  mm/page_alloc.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index fab5e97dc9ca..5db74797db39 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -4943,10 +4943,19 @@ static inline void free_the_page(struct page *page, unsigned int order)
>  		__free_pages_ok(page, order);
>  }
>  
> +/*
> + * If we free a non-compound allocation, another thread may have a
> + * speculative reference to the first page.  It has no way of knowing
> + * about the rest of the allocation, so we have to free all but the
> + * first page here.
> + */
>  void __free_pages(struct page *page, unsigned int order)
>  {
>  	if (put_page_testzero(page))
>  		free_the_page(page, order);
> +	else if (!PageHead(page))
> +		while (order-- > 0)
> +			free_the_page(page + (1 << order), order);
>  }
>  EXPORT_SYMBOL(__free_pages);

So the obvious question I have here is why not teach put_page() to free
the whole thing?