[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_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 = inode->i_sb->s_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, int mode,
			      struct buffer_head **pbh, sector_t *submit_ptr)
{
	struct buffer_head *bh;
	struct inode *inode = NILFS_BTNC_I(btnc);
	struct page *page;
	int err;

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

	err = -EEXIST; /* internal code */
	page = bh->b_page;

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

	if (pblocknr == 0) {
		pblocknr = blocknr;
		if (inode->i_ino != NILFS_DAT_INO) {
			struct the_nilfs *nilfs = inode->i_sb->s_fs_info;

			/* blocknr is a virtual block number */
			err = nilfs_dat_translate(nilfs->ns_dat, blocknr,
						  &pblocknr);
			if (unlikely(err)) {
				brelse(bh);
				goto out_locked;
			}
		}
	}

	if (mode == READA) {
		if (pblocknr != *submit_ptr + 1 || !trylock_buffer(bh)) {
			err = -EBUSY; /* internal code */
			brelse(bh);
			goto out_locked;
		}
	} else { /* mode == READ */
		lock_buffer(bh);
	}
	if (buffer_uptodate(bh)) {
		unlock_buffer(bh);
		err = -EEXIST; /* internal code */
		goto found;
	}
	set_buffer_mapped(bh);
	bh->b_bdev = inode->i_sb->s_bdev;
	bh->b_blocknr = pblocknr; /* set block address for read */
	bh->b_end_io = end_buffer_read_sync;
	get_bh(bh);
	submit_bh(mode, bh);
	bh->b_blocknr = blocknr; /* set back to the given block address */
	*submit_ptr = pblocknr;
	err = 0;
found:
	*pbh = bh;

out_locked:
	unlock_page(page);
	page_cache_release(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);
		mark_buffer_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);
		mark_buffer_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
a60be987 RK	37	void nilfs_btnode_cache_clear(struct address_space *btnc)
	38	{
	39	invalidate_mapping_pages(btnc, 0, -1);
	40	truncate_inode_pages(btnc, 0);
	41	}
	42
d501d736 RK	43	struct buffer_head *
	44	nilfs_btnode_create_block(struct address_space *btnc, __u64 blocknr)
	45	{
	46	struct inode *inode = NILFS_BTNC_I(btnc);
	47	struct buffer_head *bh;
	48
	49	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	50	if (unlikely(!bh))
	51	return NULL;
	52
	53	if (unlikely(buffer_mapped(bh) \|\| buffer_uptodate(bh) \|\|
	54	buffer_dirty(bh))) {
	55	brelse(bh);
	56	BUG();
	57	}
	58	memset(bh->b_data, 0, 1 << inode->i_blkbits);
0ef28f9a	59	bh->b_bdev = inode->i_sb->s_bdev;
d501d736 RK	60	bh->b_blocknr = blocknr;
	61	set_buffer_mapped(bh);
	62	set_buffer_uptodate(bh);
	63
	64	unlock_page(bh->b_page);
	65	page_cache_release(bh->b_page);
	66	return bh;
	67	}
	68
a60be987	69	int nilfs_btnode_submit_block(struct address_space *btnc, __u64 blocknr,
26dfdd8e RK	70	sector_t pblocknr, int mode,
26dfdd8e RK	71	struct buffer_head *pbh, sector_t submit_ptr)
a60be987 RK	72	{
	73	struct buffer_head *bh;
	74	struct inode *inode = NILFS_BTNC_I(btnc);
f8e6cc01	75	struct page *page;
a60be987 RK	76	int err;
	77
	78	bh = nilfs_grab_buffer(inode, btnc, blocknr, 1 << BH_NILFS_Node);
	79	if (unlikely(!bh))
	80	return -ENOMEM;
	81
	82	err = -EEXIST; /* internal code */
f8e6cc01	83	page = bh->b_page;
a60be987 RK	84
	85	if (buffer_uptodate(bh) \|\| buffer_dirty(bh))
	86	goto found;
	87
	88	if (pblocknr == 0) {
	89	pblocknr = blocknr;
	90	if (inode->i_ino != NILFS_DAT_INO) {
0ef28f9a	91	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
a60be987 RK	92
a60be987 RK	93	/* blocknr is a virtual block number */
0ef28f9a RK	94	err = nilfs_dat_translate(nilfs->ns_dat, blocknr,
0ef28f9a RK	95	&pblocknr);
a60be987 RK	96	if (unlikely(err)) {
	97	brelse(bh);
	98	goto out_locked;
	99	}
	100	}
	101	}
26dfdd8e RK	102
	103	if (mode == READA) {
	104	if (pblocknr != *submit_ptr + 1 \|\| !trylock_buffer(bh)) {
	105	err = -EBUSY; /* internal code */
	106	brelse(bh);
	107	goto out_locked;
	108	}
	109	} else { /* mode == READ */
	110	lock_buffer(bh);
	111	}
a60be987 RK	112	if (buffer_uptodate(bh)) {
	113	unlock_buffer(bh);
	114	err = -EEXIST; /* internal code */
	115	goto found;
	116	}
	117	set_buffer_mapped(bh);
0ef28f9a	118	bh->b_bdev = inode->i_sb->s_bdev;
a60be987 RK	119	bh->b_blocknr = pblocknr; /* set block address for read */
	120	bh->b_end_io = end_buffer_read_sync;
	121	get_bh(bh);
26dfdd8e	122	submit_bh(mode, bh);
a60be987	123	bh->b_blocknr = blocknr; /* set back to the given block address */
26dfdd8e	124	*submit_ptr = pblocknr;
a60be987 RK	125	err = 0;
	126	found:
	127	*pbh = bh;
	128
	129	out_locked:
f8e6cc01 RK	130	unlock_page(page);
f8e6cc01 RK	131	page_cache_release(page);
a60be987 RK	132	return err;
	133	}
	134
a60be987 RK	135	/**
	136	* nilfs_btnode_delete - delete B-tree node buffer
	137	* @bh: buffer to be deleted
	138	*
	139	* nilfs_btnode_delete() invalidates the specified buffer and delete the page
	140	* including the buffer if the page gets unbusy.
	141	*/
	142	void nilfs_btnode_delete(struct buffer_head *bh)
	143	{
	144	struct address_space *mapping;
	145	struct page *page = bh->b_page;
	146	pgoff_t index = page_index(page);
	147	int still_dirty;
	148
	149	page_cache_get(page);
	150	lock_page(page);
	151	wait_on_page_writeback(page);
	152
	153	nilfs_forget_buffer(bh);
	154	still_dirty = PageDirty(page);
	155	mapping = page->mapping;
	156	unlock_page(page);
	157	page_cache_release(page);
	158
	159	if (!still_dirty && mapping)
	160	invalidate_inode_pages2_range(mapping, index, index);
	161	}
	162
	163	/**
	164	* nilfs_btnode_prepare_change_key
	165	* prepare to move contents of the block for old key to one of new key.
	166	* the old buffer will not be removed, but might be reused for new buffer.
	167	* it might return -ENOMEM because of memory allocation errors,
	168	* and might return -EIO because of disk read errors.
	169	*/
	170	int nilfs_btnode_prepare_change_key(struct address_space *btnc,
	171	struct nilfs_btnode_chkey_ctxt *ctxt)
	172	{
	173	struct buffer_head obh, nbh;
	174	struct inode *inode = NILFS_BTNC_I(btnc);
	175	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	176	int err;
	177
	178	if (oldkey == newkey)
	179	return 0;
	180
	181	obh = ctxt->bh;
	182	ctxt->newbh = NULL;
	183
	184	if (inode->i_blkbits == PAGE_CACHE_SHIFT) {
	185	lock_page(obh->b_page);
	186	/*
	187	* We cannot call radix_tree_preload for the kernels older
	188	* than 2.6.23, because it is not exported for modules.
	189	*/
b1f1b8ce	190	retry:
a60be987 RK	191	err = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	192	if (err)
	193	goto failed_unlock;
	194	/* BUG_ON(oldkey != obh->b_page->index); */
	195	if (unlikely(oldkey != obh->b_page->index))
	196	NILFS_PAGE_BUG(obh->b_page,
	197	"invalid oldkey %lld (newkey=%lld)",
	198	(unsigned long long)oldkey,
	199	(unsigned long long)newkey);
	200
a60be987 RK	201	spin_lock_irq(&btnc->tree_lock);
	202	err = radix_tree_insert(&btnc->page_tree, newkey, obh->b_page);
	203	spin_unlock_irq(&btnc->tree_lock);
	204	/*
	205	* Note: page->index will not change to newkey until
	206	* nilfs_btnode_commit_change_key() will be called.
	207	* To protect the page in intermediate state, the page lock
	208	* is held.
	209	*/
	210	radix_tree_preload_end();
	211	if (!err)
	212	return 0;
	213	else if (err != -EEXIST)
	214	goto failed_unlock;
	215
	216	err = invalidate_inode_pages2_range(btnc, newkey, newkey);
	217	if (!err)
	218	goto retry;
	219	/* fallback to copy mode */
	220	unlock_page(obh->b_page);
	221	}
	222
45f4910b RK	223	nbh = nilfs_btnode_create_block(btnc, newkey);
	224	if (!nbh)
	225	return -ENOMEM;
	226
	227	BUG_ON(nbh == obh);
	228	ctxt->newbh = nbh;
	229	return 0;
a60be987 RK	230
	231	failed_unlock:
	232	unlock_page(obh->b_page);
	233	return err;
	234	}
	235
	236	/**
	237	* nilfs_btnode_commit_change_key
	238	* commit the change_key operation prepared by prepare_change_key().
	239	*/
	240	void nilfs_btnode_commit_change_key(struct address_space *btnc,
	241	struct nilfs_btnode_chkey_ctxt *ctxt)
	242	{
	243	struct buffer_head obh = ctxt->bh, nbh = ctxt->newbh;
	244	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	245	struct page *opage;
	246
	247	if (oldkey == newkey)
	248	return;
	249
	250	if (nbh == NULL) { /* blocksize == pagesize */
	251	opage = obh->b_page;
	252	if (unlikely(oldkey != opage->index))
	253	NILFS_PAGE_BUG(opage,
	254	"invalid oldkey %lld (newkey=%lld)",
	255	(unsigned long long)oldkey,
	256	(unsigned long long)newkey);
5fc7b141	257	mark_buffer_dirty(obh);
a60be987 RK	258
	259	spin_lock_irq(&btnc->tree_lock);
	260	radix_tree_delete(&btnc->page_tree, oldkey);
	261	radix_tree_tag_set(&btnc->page_tree, newkey,
	262	PAGECACHE_TAG_DIRTY);
	263	spin_unlock_irq(&btnc->tree_lock);
	264
	265	opage->index = obh->b_blocknr = newkey;
	266	unlock_page(opage);
	267	} else {
	268	nilfs_copy_buffer(nbh, obh);
5fc7b141	269	mark_buffer_dirty(nbh);
a60be987 RK	270
	271	nbh->b_blocknr = newkey;
	272	ctxt->bh = nbh;
	273	nilfs_btnode_delete(obh); /* will decrement bh->b_count */
	274	}
	275	}
	276
	277	/**
	278	* nilfs_btnode_abort_change_key
	279	* abort the change_key operation prepared by prepare_change_key().
	280	*/
	281	void nilfs_btnode_abort_change_key(struct address_space *btnc,
	282	struct nilfs_btnode_chkey_ctxt *ctxt)
	283	{
	284	struct buffer_head *nbh = ctxt->newbh;
	285	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	286
	287	if (oldkey == newkey)
	288	return;
	289
	290	if (nbh == NULL) { /* blocksize == pagesize */
	291	spin_lock_irq(&btnc->tree_lock);
	292	radix_tree_delete(&btnc->page_tree, newkey);
	293	spin_unlock_irq(&btnc->tree_lock);
	294	unlock_page(ctxt->bh->b_page);
	295	} else
	296	brelse(nbh);
	297	}