[linux.git] / fs / nilfs2 / btnode.c

/*
 * btnode.c - NILFS B-tree node cache
 *
 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 *
 * This file was originally written by Seiji Kihara <[email protected]>
 * and fully revised by Ryusuke Konishi <[email protected]> for
 * stabilization and simplification.
 *
 */

#include <linux/types.h>
#include <linux/buffer_head.h>
#include <linux/mm.h>
#include <linux/backing-dev.h>
#include <linux/gfp.h>
#include "nilfs.h"
#include "mdt.h"
#include "dat.h"
#include "page.h"
#include "btnode.h"


void nilfs_btnode_cache_init_once(struct address_space *btnc)
{
	memset(btnc, 0, sizeof(*btnc));
	INIT_RADIX_TREE(&btnc->page_tree, GFP_ATOMIC);
	spin_lock_init(&btnc->tree_lock);
	INIT_LIST_HEAD(&btnc->private_list);
	spin_lock_init(&btnc->private_lock);

	spin_lock_init(&btnc->i_mmap_lock);
	INIT_RAW_PRIO_TREE_ROOT(&btnc->i_mmap);
	INIT_LIST_HEAD(&btnc->i_mmap_nonlinear);
}

static const struct address_space_operations def_btnode_aops = {
	.sync_page		= block_sync_page,
};

void nilfs_btnode_cache_init(struct address_space *btnc,
			     struct backing_dev_info *bdi)
{
	btnc->host = NULL;  /* can safely set to host inode ? */
	btnc->flags = 0;
	mapping_set_gfp_mask(btnc, GFP_NOFS);
	btnc->assoc_mapping = NULL;
	btnc->backing_dev_info = bdi;
	btnc->a_ops = &def_btnode_aops;
}

void nilfs_btnode_cache_clear(struct address_space *btnc)
{
	invalidate_mapping_pages(btnc, 0, -1);
	truncate_inode_pages(btnc, 0);
}

struct buffer_head *
nilfs_btnode_create_block(struct address_space *btnc, __u64 blocknr)
{
	struct inode *inode = NILFS_BTNC_I(btnc);
	struct buffer_head *bh;

	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	if (unlikely(!bh))
		return NULL;

	if (unlikely(buffer_mapped(bh) || buffer_uptodate(bh) ||
		     buffer_dirty(bh))) {
		brelse(bh);
		BUG();
	}
	memset(bh->b_data, 0, 1 << inode->i_blkbits);
	bh->b_bdev = NILFS_I_NILFS(inode)->ns_bdev;
	bh->b_blocknr = blocknr;
	set_buffer_mapped(bh);
	set_buffer_uptodate(bh);

	unlock_page(bh->b_page);
	page_cache_release(bh->b_page);
	return bh;
}

int nilfs_btnode_submit_block(struct address_space *btnc, __u64 blocknr,
			      sector_t pblocknr, struct buffer_head **pbh)
{
	struct buffer_head *bh;
	struct inode *inode = NILFS_BTNC_I(btnc);
	int err;

	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	if (unlikely(!bh))
		return -ENOMEM;

	err = -EEXIST; /* internal code */

	if (buffer_uptodate(bh) || buffer_dirty(bh))
		goto found;

	if (pblocknr == 0) {
		pblocknr = blocknr;
		if (inode->i_ino != NILFS_DAT_INO) {
			struct inode *dat =
				nilfs_dat_inode(NILFS_I_NILFS(inode));

			/* blocknr is a virtual block number */
			err = nilfs_dat_translate(dat, blocknr, &pblocknr);
			if (unlikely(err)) {
				brelse(bh);
				goto out_locked;
			}
		}
	}
	lock_buffer(bh);
	if (buffer_uptodate(bh)) {
		unlock_buffer(bh);
		err = -EEXIST; /* internal code */
		goto found;
	}
	set_buffer_mapped(bh);
	bh->b_bdev = NILFS_I_NILFS(inode)->ns_bdev;
	bh->b_blocknr = pblocknr; /* set block address for read */
	bh->b_end_io = end_buffer_read_sync;
	get_bh(bh);
	submit_bh(READ, bh);
	bh->b_blocknr = blocknr; /* set back to the given block address */
	err = 0;
found:
	*pbh = bh;

out_locked:
	unlock_page(bh->b_page);
	page_cache_release(bh->b_page);
	return err;
}

/**
 * nilfs_btnode_delete - delete B-tree node buffer
 * @bh: buffer to be deleted
 *
 * nilfs_btnode_delete() invalidates the specified buffer and delete the page
 * including the buffer if the page gets unbusy.
 */
void nilfs_btnode_delete(struct buffer_head *bh)
{
	struct address_space *mapping;
	struct page *page = bh->b_page;
	pgoff_t index = page_index(page);
	int still_dirty;

	page_cache_get(page);
	lock_page(page);
	wait_on_page_writeback(page);

	nilfs_forget_buffer(bh);
	still_dirty = PageDirty(page);
	mapping = page->mapping;
	unlock_page(page);
	page_cache_release(page);

	if (!still_dirty && mapping)
		invalidate_inode_pages2_range(mapping, index, index);
}

/**
 * nilfs_btnode_prepare_change_key
 *  prepare to move contents of the block for old key to one of new key.
 *  the old buffer will not be removed, but might be reused for new buffer.
 *  it might return -ENOMEM because of memory allocation errors,
 *  and might return -EIO because of disk read errors.
 */
int nilfs_btnode_prepare_change_key(struct address_space *btnc,
				    struct nilfs_btnode_chkey_ctxt *ctxt)
{
	struct buffer_head *obh, *nbh;
	struct inode *inode = NILFS_BTNC_I(btnc);
	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	int err;

	if (oldkey == newkey)
		return 0;

	obh = ctxt->bh;
	ctxt->newbh = NULL;

	if (inode->i_blkbits == PAGE_CACHE_SHIFT) {
		lock_page(obh->b_page);
		/*
		 * We cannot call radix_tree_preload for the kernels older
		 * than 2.6.23, because it is not exported for modules.
		 */
retry:
		err = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
		if (err)
			goto failed_unlock;
		/* BUG_ON(oldkey != obh->b_page->index); */
		if (unlikely(oldkey != obh->b_page->index))
			NILFS_PAGE_BUG(obh->b_page,
				       "invalid oldkey %lld (newkey=%lld)",
				       (unsigned long long)oldkey,
				       (unsigned long long)newkey);

		spin_lock_irq(&btnc->tree_lock);
		err = radix_tree_insert(&btnc->page_tree, newkey, obh->b_page);
		spin_unlock_irq(&btnc->tree_lock);
		/*
		 * Note: page->index will not change to newkey until
		 * nilfs_btnode_commit_change_key() will be called.
		 * To protect the page in intermediate state, the page lock
		 * is held.
		 */
		radix_tree_preload_end();
		if (!err)
			return 0;
		else if (err != -EEXIST)
			goto failed_unlock;

		err = invalidate_inode_pages2_range(btnc, newkey, newkey);
		if (!err)
			goto retry;
		/* fallback to copy mode */
		unlock_page(obh->b_page);
	}

	nbh = nilfs_btnode_create_block(btnc, newkey);
	if (!nbh)
		return -ENOMEM;

	BUG_ON(nbh == obh);
	ctxt->newbh = nbh;
	return 0;

 failed_unlock:
	unlock_page(obh->b_page);
	return err;
}

/**
 * nilfs_btnode_commit_change_key
 *  commit the change_key operation prepared by prepare_change_key().
 */
void nilfs_btnode_commit_change_key(struct address_space *btnc,
				    struct nilfs_btnode_chkey_ctxt *ctxt)
{
	struct buffer_head *obh = ctxt->bh, *nbh = ctxt->newbh;
	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	struct page *opage;

	if (oldkey == newkey)
		return;

	if (nbh == NULL) {	/* blocksize == pagesize */
		opage = obh->b_page;
		if (unlikely(oldkey != opage->index))
			NILFS_PAGE_BUG(opage,
				       "invalid oldkey %lld (newkey=%lld)",
				       (unsigned long long)oldkey,
				       (unsigned long long)newkey);
		nilfs_btnode_mark_dirty(obh);

		spin_lock_irq(&btnc->tree_lock);
		radix_tree_delete(&btnc->page_tree, oldkey);
		radix_tree_tag_set(&btnc->page_tree, newkey,
				   PAGECACHE_TAG_DIRTY);
		spin_unlock_irq(&btnc->tree_lock);

		opage->index = obh->b_blocknr = newkey;
		unlock_page(opage);
	} else {
		nilfs_copy_buffer(nbh, obh);
		nilfs_btnode_mark_dirty(nbh);

		nbh->b_blocknr = newkey;
		ctxt->bh = nbh;
		nilfs_btnode_delete(obh); /* will decrement bh->b_count */
	}
}

/**
 * nilfs_btnode_abort_change_key
 *  abort the change_key operation prepared by prepare_change_key().
 */
void nilfs_btnode_abort_change_key(struct address_space *btnc,
				   struct nilfs_btnode_chkey_ctxt *ctxt)
{
	struct buffer_head *nbh = ctxt->newbh;
	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;

	if (oldkey == newkey)
		return;

	if (nbh == NULL) {	/* blocksize == pagesize */
		spin_lock_irq(&btnc->tree_lock);
		radix_tree_delete(&btnc->page_tree, newkey);
		spin_unlock_irq(&btnc->tree_lock);
		unlock_page(ctxt->bh->b_page);
	} else
		brelse(nbh);
}
Commit	Line	Data
a60be987 RK	1	/*
	2	* btnode.c - NILFS B-tree node cache
	3	*
	4	* Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License as published by
	8	* the Free Software Foundation; either version 2 of the License, or
	9	* (at your option) any later version.
	10	*
	11	* This program is distributed in the hope that it will be useful,
	12	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	13	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	14	* GNU General Public License for more details.
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* along with this program; if not, write to the Free Software
	18	* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
	19	*
	20	* This file was originally written by Seiji Kihara <[email protected]>
	21	* and fully revised by Ryusuke Konishi <[email protected]> for
	22	* stabilization and simplification.
	23	*
	24	*/
	25
	26	#include <linux/types.h>
	27	#include <linux/buffer_head.h>
	28	#include <linux/mm.h>
	29	#include <linux/backing-dev.h>
5a0e3ad6	30	#include <linux/gfp.h>
a60be987 RK	31	#include "nilfs.h"
	32	#include "mdt.h"
	33	#include "dat.h"
	34	#include "page.h"
	35	#include "btnode.h"
	36
	37
	38	void nilfs_btnode_cache_init_once(struct address_space *btnc)
	39	{
1f28fcd9	40	memset(btnc, 0, sizeof(*btnc));
a60be987 RK	41	INIT_RADIX_TREE(&btnc->page_tree, GFP_ATOMIC);
	42	spin_lock_init(&btnc->tree_lock);
	43	INIT_LIST_HEAD(&btnc->private_list);
	44	spin_lock_init(&btnc->private_lock);
	45
	46	spin_lock_init(&btnc->i_mmap_lock);
	47	INIT_RAW_PRIO_TREE_ROOT(&btnc->i_mmap);
	48	INIT_LIST_HEAD(&btnc->i_mmap_nonlinear);
	49	}
	50
7f09410b	51	static const struct address_space_operations def_btnode_aops = {
fa032744 RK	52	.sync_page = block_sync_page,
fa032744 RK	53	};
a60be987	54
a53b4751 RK	55	void nilfs_btnode_cache_init(struct address_space *btnc,
a53b4751 RK	56	struct backing_dev_info *bdi)
a60be987 RK	57	{
	58	btnc->host = NULL; /* can safely set to host inode ? */
	59	btnc->flags = 0;
	60	mapping_set_gfp_mask(btnc, GFP_NOFS);
	61	btnc->assoc_mapping = NULL;
a53b4751	62	btnc->backing_dev_info = bdi;
a60be987 RK	63	btnc->a_ops = &def_btnode_aops;
	64	}
	65
	66	void nilfs_btnode_cache_clear(struct address_space *btnc)
	67	{
	68	invalidate_mapping_pages(btnc, 0, -1);
	69	truncate_inode_pages(btnc, 0);
	70	}
	71
d501d736 RK	72	struct buffer_head *
	73	nilfs_btnode_create_block(struct address_space *btnc, __u64 blocknr)
	74	{
	75	struct inode *inode = NILFS_BTNC_I(btnc);
	76	struct buffer_head *bh;
	77
	78	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	79	if (unlikely(!bh))
	80	return NULL;
	81
	82	if (unlikely(buffer_mapped(bh) \|\| buffer_uptodate(bh) \|\|
	83	buffer_dirty(bh))) {
	84	brelse(bh);
	85	BUG();
	86	}
	87	memset(bh->b_data, 0, 1 << inode->i_blkbits);
	88	bh->b_bdev = NILFS_I_NILFS(inode)->ns_bdev;
	89	bh->b_blocknr = blocknr;
	90	set_buffer_mapped(bh);
	91	set_buffer_uptodate(bh);
	92
	93	unlock_page(bh->b_page);
	94	page_cache_release(bh->b_page);
	95	return bh;
	96	}
	97
a60be987	98	int nilfs_btnode_submit_block(struct address_space *btnc, __u64 blocknr,
75f65edf	99	sector_t pblocknr, struct buffer_head **pbh)
a60be987 RK	100	{
	101	struct buffer_head *bh;
	102	struct inode *inode = NILFS_BTNC_I(btnc);
	103	int err;
	104
	105	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	106	if (unlikely(!bh))
	107	return -ENOMEM;
	108
	109	err = -EEXIST; /* internal code */
a60be987 RK	110
	111	if (buffer_uptodate(bh) \|\| buffer_dirty(bh))
	112	goto found;
	113
	114	if (pblocknr == 0) {
	115	pblocknr = blocknr;
	116	if (inode->i_ino != NILFS_DAT_INO) {
	117	struct inode *dat =
	118	nilfs_dat_inode(NILFS_I_NILFS(inode));
	119
	120	/* blocknr is a virtual block number */
	121	err = nilfs_dat_translate(dat, blocknr, &pblocknr);
	122	if (unlikely(err)) {
	123	brelse(bh);
	124	goto out_locked;
	125	}
	126	}
	127	}
	128	lock_buffer(bh);
	129	if (buffer_uptodate(bh)) {
	130	unlock_buffer(bh);
	131	err = -EEXIST; /* internal code */
	132	goto found;
	133	}
	134	set_buffer_mapped(bh);
	135	bh->b_bdev = NILFS_I_NILFS(inode)->ns_bdev;
	136	bh->b_blocknr = pblocknr; /* set block address for read */
	137	bh->b_end_io = end_buffer_read_sync;
	138	get_bh(bh);
	139	submit_bh(READ, bh);
	140	bh->b_blocknr = blocknr; /* set back to the given block address */
	141	err = 0;
	142	found:
	143	*pbh = bh;
	144
	145	out_locked:
	146	unlock_page(bh->b_page);
	147	page_cache_release(bh->b_page);
	148	return err;
	149	}
	150
a60be987 RK	151	/**
	152	* nilfs_btnode_delete - delete B-tree node buffer
	153	* @bh: buffer to be deleted
	154	*
	155	* nilfs_btnode_delete() invalidates the specified buffer and delete the page
	156	* including the buffer if the page gets unbusy.
	157	*/
	158	void nilfs_btnode_delete(struct buffer_head *bh)
	159	{
	160	struct address_space *mapping;
	161	struct page *page = bh->b_page;
	162	pgoff_t index = page_index(page);
	163	int still_dirty;
	164
	165	page_cache_get(page);
	166	lock_page(page);
	167	wait_on_page_writeback(page);
	168
	169	nilfs_forget_buffer(bh);
	170	still_dirty = PageDirty(page);
	171	mapping = page->mapping;
	172	unlock_page(page);
	173	page_cache_release(page);
	174
	175	if (!still_dirty && mapping)
	176	invalidate_inode_pages2_range(mapping, index, index);
	177	}
	178
	179	/**
	180	* nilfs_btnode_prepare_change_key
	181	* prepare to move contents of the block for old key to one of new key.
	182	* the old buffer will not be removed, but might be reused for new buffer.
	183	* it might return -ENOMEM because of memory allocation errors,
	184	* and might return -EIO because of disk read errors.
	185	*/
	186	int nilfs_btnode_prepare_change_key(struct address_space *btnc,
	187	struct nilfs_btnode_chkey_ctxt *ctxt)
	188	{
	189	struct buffer_head obh, nbh;
	190	struct inode *inode = NILFS_BTNC_I(btnc);
	191	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	192	int err;
	193
	194	if (oldkey == newkey)
	195	return 0;
	196
	197	obh = ctxt->bh;
	198	ctxt->newbh = NULL;
	199
	200	if (inode->i_blkbits == PAGE_CACHE_SHIFT) {
	201	lock_page(obh->b_page);
	202	/*
	203	* We cannot call radix_tree_preload for the kernels older
	204	* than 2.6.23, because it is not exported for modules.
	205	*/
b1f1b8ce	206	retry:
a60be987 RK	207	err = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	208	if (err)
	209	goto failed_unlock;
	210	/* BUG_ON(oldkey != obh->b_page->index); */
	211	if (unlikely(oldkey != obh->b_page->index))
	212	NILFS_PAGE_BUG(obh->b_page,
	213	"invalid oldkey %lld (newkey=%lld)",
	214	(unsigned long long)oldkey,
	215	(unsigned long long)newkey);
	216
a60be987 RK	217	spin_lock_irq(&btnc->tree_lock);
	218	err = radix_tree_insert(&btnc->page_tree, newkey, obh->b_page);
	219	spin_unlock_irq(&btnc->tree_lock);
	220	/*
	221	* Note: page->index will not change to newkey until
	222	* nilfs_btnode_commit_change_key() will be called.
	223	* To protect the page in intermediate state, the page lock
	224	* is held.
	225	*/
	226	radix_tree_preload_end();
	227	if (!err)
	228	return 0;
	229	else if (err != -EEXIST)
	230	goto failed_unlock;
	231
	232	err = invalidate_inode_pages2_range(btnc, newkey, newkey);
	233	if (!err)
	234	goto retry;
	235	/* fallback to copy mode */
	236	unlock_page(obh->b_page);
	237	}
	238
45f4910b RK	239	nbh = nilfs_btnode_create_block(btnc, newkey);
	240	if (!nbh)
	241	return -ENOMEM;
	242
	243	BUG_ON(nbh == obh);
	244	ctxt->newbh = nbh;
	245	return 0;
a60be987 RK	246
	247	failed_unlock:
	248	unlock_page(obh->b_page);
	249	return err;
	250	}
	251
	252	/**
	253	* nilfs_btnode_commit_change_key
	254	* commit the change_key operation prepared by prepare_change_key().
	255	*/
	256	void nilfs_btnode_commit_change_key(struct address_space *btnc,
	257	struct nilfs_btnode_chkey_ctxt *ctxt)
	258	{
	259	struct buffer_head obh = ctxt->bh, nbh = ctxt->newbh;
	260	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	261	struct page *opage;
	262
	263	if (oldkey == newkey)
	264	return;
	265
	266	if (nbh == NULL) { /* blocksize == pagesize */
	267	opage = obh->b_page;
	268	if (unlikely(oldkey != opage->index))
	269	NILFS_PAGE_BUG(opage,
	270	"invalid oldkey %lld (newkey=%lld)",
	271	(unsigned long long)oldkey,
	272	(unsigned long long)newkey);
b1e19e56	273	nilfs_btnode_mark_dirty(obh);
a60be987 RK	274
	275	spin_lock_irq(&btnc->tree_lock);
	276	radix_tree_delete(&btnc->page_tree, oldkey);
	277	radix_tree_tag_set(&btnc->page_tree, newkey,
	278	PAGECACHE_TAG_DIRTY);
	279	spin_unlock_irq(&btnc->tree_lock);
	280
	281	opage->index = obh->b_blocknr = newkey;
	282	unlock_page(opage);
	283	} else {
	284	nilfs_copy_buffer(nbh, obh);
	285	nilfs_btnode_mark_dirty(nbh);
	286
	287	nbh->b_blocknr = newkey;
	288	ctxt->bh = nbh;
	289	nilfs_btnode_delete(obh); /* will decrement bh->b_count */
	290	}
	291	}
	292
	293	/**
	294	* nilfs_btnode_abort_change_key
	295	* abort the change_key operation prepared by prepare_change_key().
	296	*/
	297	void nilfs_btnode_abort_change_key(struct address_space *btnc,
	298	struct nilfs_btnode_chkey_ctxt *ctxt)
	299	{
	300	struct buffer_head *nbh = ctxt->newbh;
	301	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	302
	303	if (oldkey == newkey)
	304	return;
	305
	306	if (nbh == NULL) { /* blocksize == pagesize */
	307	spin_lock_irq(&btnc->tree_lock);
	308	radix_tree_delete(&btnc->page_tree, newkey);
	309	spin_unlock_irq(&btnc->tree_lock);
	310	unlock_page(ctxt->bh->b_page);
	311	} else
	312	brelse(nbh);
	313	}