From mboxrd@z Thu Jan  1 00:00:00 1970
From: Martin Jambor <jambormartin@gmail.com>
Subject: pagevec_lookup_tag() and pagevec_release() substitutes and an OOM
Date: Wed, 8 Feb 2006 01:29:44 +0100
Message-ID: <9615ac9b0602071629y5054dc07w8d1f18e070b0b01f@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7BIT
Return-path: <linux-fsdevel-owner@vger.kernel.org>
Received: from xproxy.gmail.com ([66.249.82.207]:25308 "EHLO xproxy.gmail.com")
	by vger.kernel.org with ESMTP id S1030314AbWBHA3r convert rfc822-to-8bit
	(ORCPT <rfc822;linux-fsdevel@vger.kernel.org>);
	Tue, 7 Feb 2006 19:29:47 -0500
Received: by xproxy.gmail.com with SMTP id s14so1052896wxc
        for <linux-fsdevel@vger.kernel.org>; Tue, 07 Feb 2006 16:29:44 -0800 (PST)
To: Linux FS Development List <linux-fsdevel@vger.kernel.org>
Content-Disposition: inline
Sender: linux-fsdevel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

Hi all,

I am implementing aops->writepages of a filesystem and for various
reasons I cannot use mpage_writepages. On the other hand, I would like
to do be able to get a vector of pages and process it page by page (in
my way) just like mpage_writepages gets pages.

However, pagevec functions are not exported (I don't understand why)
and so I cannot use them directly by my module but rather I had to do
a bit of cutting and pasting from 2.6.14 kernel source :-(

So my approach was to use a plain array of pointers to pages and
instead of calling

unsigned pagevec_lookup_tag(struct pagevec *pvec, struct address_space *mapping,
		pgoff_t *index, int tag, unsigned nr_pages)

I use my own function which is basically copied find_get_pages_tag()
(which also isn't exported):

static unsigned my_find_get_pages_tag(struct address_space *mapping,
				   pgoff_t *index, int tag, unsigned int nr_pages,
                                   struct page **pages)
{
	unsigned int i;
	unsigned int ret;

	read_lock_irq(&mapping->tree_lock);
	ret = radix_tree_gang_lookup_tag(&mapping->page_tree,
				(void **)pages, *index, nr_pages, tag);
	for (i = 0; i < ret; i++)
		page_cache_get(pages[i]);
	if (ret)
		*index = pages[ret - 1]->index + 1;
	read_unlock_irq(&mapping->tree_lock);
	return ret;
}

instead of pagevec_release() I use a much simpler function:

static void release_pagearray(struct page **pages, unsigned int nr_pages)
{
	unsigned i;
	
	for (i = 0; i < nr_pages; i++)
		page_cache_release(pages[i]);
}

Everything seems to be working well except that I get an OOM. The free
and vmstat utilities show that "cache" is the only thing that is
getting unreasonably bigger. There are many reasons why I believe the
page cache pages are still somehow pinned down by my code and cannot
be freed when I am done with them. I must admit that I find it quite
difficult to uderstand the pagevec source and if someone knew what I
am doing wrong, I would be very grateful.

(Yes, I have carefully checked several times very carefully that I
set&clear writeback as indicated in Documentation/filesystmes/Locking
and how it is done by mpage_writepages.)

Thank you very much in advance for any comment.

Martin