[mirror_ubuntu-artful-kernel.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.
 *
 * Originally written by Seiji Kihara.
 * Fully revised by Ryusuke Konishi 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, BIT(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, i_blocksize(inode));
	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);
	put_page(bh->b_page);
	return bh;
}

int nilfs_btnode_submit_block(struct address_space *btnc, __u64 blocknr,
			      sector_t pblocknr, int mode, int mode_flags,
			      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, BIT(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_flags & REQ_RAHEAD) {
		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, mode_flags, 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);
	put_page(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;

	get_page(page);
	lock_page(page);
	wait_on_page_writeback(page);

	nilfs_forget_buffer(bh);
	still_dirty = PageDirty(page);
	mapping = page->mapping;
	unlock_page(page);
	put_page(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_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	*
4b420ab4 RK	16	* Originally written by Seiji Kihara.
4b420ab4 RK	17	* Fully revised by Ryusuke Konishi for stabilization and simplification.
a60be987 RK	18	*
	19	*/
	20
	21	#include <linux/types.h>
	22	#include <linux/buffer_head.h>
	23	#include <linux/mm.h>
	24	#include <linux/backing-dev.h>
5a0e3ad6	25	#include <linux/gfp.h>
a60be987 RK	26	#include "nilfs.h"
	27	#include "mdt.h"
	28	#include "dat.h"
	29	#include "page.h"
	30	#include "btnode.h"
	31
a60be987 RK	32	void nilfs_btnode_cache_clear(struct address_space *btnc)
	33	{
	34	invalidate_mapping_pages(btnc, 0, -1);
	35	truncate_inode_pages(btnc, 0);
	36	}
	37
d501d736 RK	38	struct buffer_head *
	39	nilfs_btnode_create_block(struct address_space *btnc, __u64 blocknr)
	40	{
	41	struct inode *inode = NILFS_BTNC_I(btnc);
	42	struct buffer_head *bh;
	43
4ce5c342	44	bh = nilfs_grab_buffer(inode, btnc, blocknr, BIT(BH_NILFS_Node));
d501d736 RK	45	if (unlikely(!bh))
	46	return NULL;
	47
	48	if (unlikely(buffer_mapped(bh) \|\| buffer_uptodate(bh) \|\|
	49	buffer_dirty(bh))) {
	50	brelse(bh);
	51	BUG();
	52	}
93407472	53	memset(bh->b_data, 0, i_blocksize(inode));
0ef28f9a	54	bh->b_bdev = inode->i_sb->s_bdev;
d501d736 RK	55	bh->b_blocknr = blocknr;
	56	set_buffer_mapped(bh);
	57	set_buffer_uptodate(bh);
	58
	59	unlock_page(bh->b_page);
09cbfeaf	60	put_page(bh->b_page);
d501d736 RK	61	return bh;
	62	}
	63
a60be987	64	int nilfs_btnode_submit_block(struct address_space *btnc, __u64 blocknr,
2a222ca9	65	sector_t pblocknr, int mode, int mode_flags,
26dfdd8e	66	struct buffer_head *pbh, sector_t submit_ptr)
a60be987 RK	67	{
	68	struct buffer_head *bh;
	69	struct inode *inode = NILFS_BTNC_I(btnc);
f8e6cc01	70	struct page *page;
a60be987 RK	71	int err;
a60be987 RK	72
4ce5c342	73	bh = nilfs_grab_buffer(inode, btnc, blocknr, BIT(BH_NILFS_Node));
a60be987 RK	74	if (unlikely(!bh))
	75	return -ENOMEM;
	76
	77	err = -EEXIST; /* internal code */
f8e6cc01	78	page = bh->b_page;
a60be987 RK	79
	80	if (buffer_uptodate(bh) \|\| buffer_dirty(bh))
	81	goto found;
	82
	83	if (pblocknr == 0) {
	84	pblocknr = blocknr;
	85	if (inode->i_ino != NILFS_DAT_INO) {
0ef28f9a	86	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
a60be987 RK	87
a60be987 RK	88	/* blocknr is a virtual block number */
0ef28f9a RK	89	err = nilfs_dat_translate(nilfs->ns_dat, blocknr,
0ef28f9a RK	90	&pblocknr);
a60be987 RK	91	if (unlikely(err)) {
	92	brelse(bh);
	93	goto out_locked;
	94	}
	95	}
	96	}
26dfdd8e	97
2a222ca9	98	if (mode_flags & REQ_RAHEAD) {
26dfdd8e RK	99	if (pblocknr != *submit_ptr + 1 \|\| !trylock_buffer(bh)) {
	100	err = -EBUSY; /* internal code */
	101	brelse(bh);
	102	goto out_locked;
	103	}
	104	} else { /* mode == READ */
	105	lock_buffer(bh);
	106	}
a60be987 RK	107	if (buffer_uptodate(bh)) {
	108	unlock_buffer(bh);
	109	err = -EEXIST; /* internal code */
	110	goto found;
	111	}
	112	set_buffer_mapped(bh);
0ef28f9a	113	bh->b_bdev = inode->i_sb->s_bdev;
a60be987 RK	114	bh->b_blocknr = pblocknr; /* set block address for read */
	115	bh->b_end_io = end_buffer_read_sync;
	116	get_bh(bh);
2a222ca9	117	submit_bh(mode, mode_flags, bh);
a60be987	118	bh->b_blocknr = blocknr; /* set back to the given block address */
26dfdd8e	119	*submit_ptr = pblocknr;
a60be987 RK	120	err = 0;
	121	found:
	122	*pbh = bh;
	123
	124	out_locked:
f8e6cc01	125	unlock_page(page);
09cbfeaf	126	put_page(page);
a60be987 RK	127	return err;
	128	}
	129
a60be987 RK	130	/**
	131	* nilfs_btnode_delete - delete B-tree node buffer
	132	* @bh: buffer to be deleted
	133	*
	134	* nilfs_btnode_delete() invalidates the specified buffer and delete the page
	135	* including the buffer if the page gets unbusy.
	136	*/
	137	void nilfs_btnode_delete(struct buffer_head *bh)
	138	{
	139	struct address_space *mapping;
	140	struct page *page = bh->b_page;
	141	pgoff_t index = page_index(page);
	142	int still_dirty;
	143
09cbfeaf	144	get_page(page);
a60be987 RK	145	lock_page(page);
	146	wait_on_page_writeback(page);
	147
	148	nilfs_forget_buffer(bh);
	149	still_dirty = PageDirty(page);
	150	mapping = page->mapping;
	151	unlock_page(page);
09cbfeaf	152	put_page(page);
a60be987 RK	153
	154	if (!still_dirty && mapping)
	155	invalidate_inode_pages2_range(mapping, index, index);
	156	}
	157
	158	/**
	159	* nilfs_btnode_prepare_change_key
	160	* prepare to move contents of the block for old key to one of new key.
	161	* the old buffer will not be removed, but might be reused for new buffer.
	162	* it might return -ENOMEM because of memory allocation errors,
	163	* and might return -EIO because of disk read errors.
	164	*/
	165	int nilfs_btnode_prepare_change_key(struct address_space *btnc,
	166	struct nilfs_btnode_chkey_ctxt *ctxt)
	167	{
	168	struct buffer_head obh, nbh;
	169	struct inode *inode = NILFS_BTNC_I(btnc);
	170	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	171	int err;
	172
	173	if (oldkey == newkey)
	174	return 0;
	175
	176	obh = ctxt->bh;
	177	ctxt->newbh = NULL;
	178
09cbfeaf	179	if (inode->i_blkbits == PAGE_SHIFT) {
a60be987 RK	180	lock_page(obh->b_page);
	181	/*
	182	* We cannot call radix_tree_preload for the kernels older
	183	* than 2.6.23, because it is not exported for modules.
	184	*/
b1f1b8ce	185	retry:
a60be987 RK	186	err = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	187	if (err)
	188	goto failed_unlock;
	189	/* BUG_ON(oldkey != obh->b_page->index); */
	190	if (unlikely(oldkey != obh->b_page->index))
	191	NILFS_PAGE_BUG(obh->b_page,
	192	"invalid oldkey %lld (newkey=%lld)",
	193	(unsigned long long)oldkey,
	194	(unsigned long long)newkey);
	195
a60be987 RK	196	spin_lock_irq(&btnc->tree_lock);
	197	err = radix_tree_insert(&btnc->page_tree, newkey, obh->b_page);
	198	spin_unlock_irq(&btnc->tree_lock);
	199	/*
	200	* Note: page->index will not change to newkey until
	201	* nilfs_btnode_commit_change_key() will be called.
	202	* To protect the page in intermediate state, the page lock
	203	* is held.
	204	*/
	205	radix_tree_preload_end();
	206	if (!err)
	207	return 0;
	208	else if (err != -EEXIST)
	209	goto failed_unlock;
	210
	211	err = invalidate_inode_pages2_range(btnc, newkey, newkey);
	212	if (!err)
	213	goto retry;
	214	/* fallback to copy mode */
	215	unlock_page(obh->b_page);
	216	}
	217
45f4910b RK	218	nbh = nilfs_btnode_create_block(btnc, newkey);
	219	if (!nbh)
	220	return -ENOMEM;
	221
	222	BUG_ON(nbh == obh);
	223	ctxt->newbh = nbh;
	224	return 0;
a60be987 RK	225
	226	failed_unlock:
	227	unlock_page(obh->b_page);
	228	return err;
	229	}
	230
	231	/**
	232	* nilfs_btnode_commit_change_key
	233	* commit the change_key operation prepared by prepare_change_key().
	234	*/
	235	void nilfs_btnode_commit_change_key(struct address_space *btnc,
	236	struct nilfs_btnode_chkey_ctxt *ctxt)
	237	{
	238	struct buffer_head obh = ctxt->bh, nbh = ctxt->newbh;
	239	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	240	struct page *opage;
	241
	242	if (oldkey == newkey)
	243	return;
	244
	245	if (nbh == NULL) { /* blocksize == pagesize */
	246	opage = obh->b_page;
	247	if (unlikely(oldkey != opage->index))
	248	NILFS_PAGE_BUG(opage,
	249	"invalid oldkey %lld (newkey=%lld)",
	250	(unsigned long long)oldkey,
	251	(unsigned long long)newkey);
5fc7b141	252	mark_buffer_dirty(obh);
a60be987 RK	253
	254	spin_lock_irq(&btnc->tree_lock);
	255	radix_tree_delete(&btnc->page_tree, oldkey);
	256	radix_tree_tag_set(&btnc->page_tree, newkey,
	257	PAGECACHE_TAG_DIRTY);
	258	spin_unlock_irq(&btnc->tree_lock);
	259
	260	opage->index = obh->b_blocknr = newkey;
	261	unlock_page(opage);
	262	} else {
	263	nilfs_copy_buffer(nbh, obh);
5fc7b141	264	mark_buffer_dirty(nbh);
a60be987 RK	265
	266	nbh->b_blocknr = newkey;
	267	ctxt->bh = nbh;
	268	nilfs_btnode_delete(obh); /* will decrement bh->b_count */
	269	}
	270	}
	271
	272	/**
	273	* nilfs_btnode_abort_change_key
	274	* abort the change_key operation prepared by prepare_change_key().
	275	*/
	276	void nilfs_btnode_abort_change_key(struct address_space *btnc,
	277	struct nilfs_btnode_chkey_ctxt *ctxt)
	278	{
	279	struct buffer_head *nbh = ctxt->newbh;
	280	__u64 oldkey = ctxt->oldkey, newkey = ctxt->newkey;
	281
	282	if (oldkey == newkey)
	283	return;
	284
	285	if (nbh == NULL) { /* blocksize == pagesize */
	286	spin_lock_irq(&btnc->tree_lock);
	287	radix_tree_delete(&btnc->page_tree, newkey);
	288	spin_unlock_irq(&btnc->tree_lock);
	289	unlock_page(ctxt->bh->b_page);
	290	} else
	291	brelse(nbh);
	292	}