Re: [v4][PATCH 2/6] mm: swap: make 'struct page' and swp_entry_tvariants of swapcache_free().

[Dave Hansen]

On 06/02/2013 11:13 PM, Minchan Kim wrote:
> I lost from this description.
> 
> Old behavior
> 
> delete_from_swap_cache
>         swap_address_space
>         __delete_from_swap_cache
>                 swap_address_space
> 
> 
> New behavior
> 
> delete_from_swap_cache
>         __delete_from_swap_cache
>                 swap_address_space
>                 
> So you removes a swap_address_space, not adding a extra call.
> Am I missing something?

I think I got the page->swp_entry_t lookup confused with the
page->swap_address_space lookup when I was writing the description.  The
bit that you missed is that I _added_ a page_mapping() call, which calls
swap_address_space() internally:

Old behavior:

delete_from_swap_cache
        swap_address_space
        __delete_from_swap_cache
                swap_address_space

New behavior:

delete_from_swap_cache
	page_mapping
		swap_address_space
        __delete_from_swap_cache
                swap_address_space

--

New description (last paragraph changed).  Andrew, I'll resend the
series since there are a few of these cleanups.

From: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>

swapcache_free() takes two arguments:

	void swapcache_free(swp_entry_t entry, struct page *page)

Most of its callers (5/7) are from error handling paths haven't even
instantiated a page, so they pass page=NULL.  Both of the callers
that call in with a 'struct page' create and pass in a temporary
swp_entry_t.

Now that we are deferring clearing page_private() until after
swapcache_free() has been called, we can just create a variant
that takes a 'struct page' and does the temporary variable in
the helper.

That leaves all the other callers doing

	swapcache_free(entry, NULL)

so create another helper for them that makes it clear that they
need only pass in a swp_entry_t.

One downside here is that delete_from_swap_cache() now calls
swap_address_space() via page_mapping() instead of calling
swap_address_space() directly.  In doing so, it removes one more
case of the swap cache code being special-cased, which is a good
thing in my book.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/