blob: 77615aabc7e20722a2c1cc43814da387a2bd650a [file] [log] [blame]
Ryusuke Konishia3d93f72009-04-06 19:01:40 -07001/*
Ryusuke Konishi047180f2009-04-06 19:01:45 -07002 * gcinode.c - dummy inodes to buffer blocks for garbage collection
Ryusuke Konishia3d93f72009-04-06 19:01:40 -07003 *
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 * Written by Seiji Kihara <kihara@osrg.net>, Amagai Yoshiji <amagai@osrg.net>,
21 * and Ryusuke Konishi <ryusuke@osrg.net>.
22 * Revised by Ryusuke Konishi <ryusuke@osrg.net>.
23 *
24 */
Ryusuke Konishi047180f2009-04-06 19:01:45 -070025/*
26 * This file adds the cache of on-disk blocks to be moved in garbage
27 * collection. The disk blocks are held with dummy inodes (called
28 * gcinodes), and this file provides lookup function of the dummy
29 * inodes and their buffer read function.
30 *
31 * Since NILFS2 keeps up multiple checkpoints/snapshots accross GC, it
32 * has to treat blocks that belong to a same file but have different
33 * checkpoint numbers. To avoid interference among generations, dummy
34 * inodes are managed separatly from actual inodes, and their lookup
35 * function (nilfs_gc_iget) is designed to be specified with a
36 * checkpoint number argument as well as an inode number.
37 *
38 * Buffers and pages held by the dummy inodes will be released each
39 * time after they are copied to a new log. Dirty blocks made on the
40 * current generation and the blocks to be moved by GC never overlap
41 * because the dirty blocks make a new generation; they rather must be
42 * written individually.
43 */
Ryusuke Konishia3d93f72009-04-06 19:01:40 -070044
45#include <linux/buffer_head.h>
46#include <linux/mpage.h>
47#include <linux/hash.h>
48#include <linux/swap.h>
49#include "nilfs.h"
50#include "page.h"
51#include "mdt.h"
52#include "dat.h"
53#include "ifile.h"
54
55static struct address_space_operations def_gcinode_aops = {};
56/* XXX need def_gcinode_iops/fops? */
57
58/*
59 * nilfs_gccache_submit_read_data() - add data buffer and submit read request
60 * @inode - gc inode
61 * @blkoff - dummy offset treated as the key for the page cache
62 * @pbn - physical block number of the block
63 * @vbn - virtual block number of the block, 0 for non-virtual block
64 * @out_bh - indirect pointer to a buffer_head struct to receive the results
65 *
66 * Description: nilfs_gccache_submit_read_data() registers the data buffer
67 * specified by @pbn to the GC pagecache with the key @blkoff.
68 * This function sets @vbn (@pbn if @vbn is zero) in b_blocknr of the buffer.
69 *
70 * Return Value: On success, 0 is returned. On Error, one of the following
71 * negative error code is returned.
72 *
73 * %-EIO - I/O error.
74 *
75 * %-ENOMEM - Insufficient amount of memory available.
76 *
77 * %-ENOENT - The block specified with @pbn does not exist.
78 */
79int nilfs_gccache_submit_read_data(struct inode *inode, sector_t blkoff,
80 sector_t pbn, __u64 vbn,
81 struct buffer_head **out_bh)
82{
83 struct buffer_head *bh;
84 int err;
85
86 bh = nilfs_grab_buffer(inode, inode->i_mapping, blkoff, 0);
87 if (unlikely(!bh))
88 return -ENOMEM;
89
90 if (buffer_uptodate(bh))
91 goto out;
92
93 if (pbn == 0) {
94 struct inode *dat_inode = NILFS_I_NILFS(inode)->ns_dat;
95 /* use original dat, not gc dat. */
96 err = nilfs_dat_translate(dat_inode, vbn, &pbn);
97 if (unlikely(err)) { /* -EIO, -ENOMEM, -ENOENT */
98 brelse(bh);
99 goto failed;
100 }
101 }
102
103 lock_buffer(bh);
104 if (buffer_uptodate(bh)) {
105 unlock_buffer(bh);
106 goto out;
107 }
108
109 if (!buffer_mapped(bh)) {
110 bh->b_bdev = NILFS_I_NILFS(inode)->ns_bdev;
111 set_buffer_mapped(bh);
112 }
113 bh->b_blocknr = pbn;
114 bh->b_end_io = end_buffer_read_sync;
115 get_bh(bh);
116 submit_bh(READ, bh);
117 if (vbn)
118 bh->b_blocknr = vbn;
119 out:
120 err = 0;
121 *out_bh = bh;
122
123 failed:
124 unlock_page(bh->b_page);
125 page_cache_release(bh->b_page);
126 return err;
127}
128
129/*
130 * nilfs_gccache_submit_read_node() - add node buffer and submit read request
131 * @inode - gc inode
132 * @pbn - physical block number for the block
133 * @vbn - virtual block number for the block
134 * @out_bh - indirect pointer to a buffer_head struct to receive the results
135 *
136 * Description: nilfs_gccache_submit_read_node() registers the node buffer
137 * specified by @vbn to the GC pagecache. @pbn can be supplied by the
138 * caller to avoid translation of the disk block address.
139 *
140 * Return Value: On success, 0 is returned. On Error, one of the following
141 * negative error code is returned.
142 *
143 * %-EIO - I/O error.
144 *
145 * %-ENOMEM - Insufficient amount of memory available.
146 */
147int nilfs_gccache_submit_read_node(struct inode *inode, sector_t pbn,
148 __u64 vbn, struct buffer_head **out_bh)
149{
150 int ret = nilfs_btnode_submit_block(&NILFS_I(inode)->i_btnode_cache,
151 vbn ? : pbn, pbn, out_bh, 0);
152 if (ret == -EEXIST) /* internal code (cache hit) */
153 ret = 0;
154 return ret;
155}
156
157int nilfs_gccache_wait_and_mark_dirty(struct buffer_head *bh)
158{
159 wait_on_buffer(bh);
160 if (!buffer_uptodate(bh))
161 return -EIO;
162 if (buffer_dirty(bh))
163 return -EEXIST;
164
165 if (buffer_nilfs_node(bh))
166 nilfs_btnode_mark_dirty(bh);
167 else
168 nilfs_mdt_mark_buffer_dirty(bh);
169 return 0;
170}
171
172/*
173 * nilfs_init_gccache() - allocate and initialize gc_inode hash table
174 * @nilfs - the_nilfs
175 *
176 * Return Value: On success, 0.
177 * On error, a negative error code is returned.
178 */
179int nilfs_init_gccache(struct the_nilfs *nilfs)
180{
181 int loop;
182
183 BUG_ON(nilfs->ns_gc_inodes_h);
184
185 INIT_LIST_HEAD(&nilfs->ns_gc_inodes);
186
187 nilfs->ns_gc_inodes_h =
188 kmalloc(sizeof(struct hlist_head) * NILFS_GCINODE_HASH_SIZE,
189 GFP_NOFS);
190 if (nilfs->ns_gc_inodes_h == NULL)
191 return -ENOMEM;
192
193 for (loop = 0; loop < NILFS_GCINODE_HASH_SIZE; loop++)
194 INIT_HLIST_HEAD(&nilfs->ns_gc_inodes_h[loop]);
195 return 0;
196}
197
198/*
199 * nilfs_destroy_gccache() - free gc_inode hash table
200 * @nilfs - the nilfs
201 */
202void nilfs_destroy_gccache(struct the_nilfs *nilfs)
203{
204 if (nilfs->ns_gc_inodes_h) {
205 nilfs_remove_all_gcinode(nilfs);
206 kfree(nilfs->ns_gc_inodes_h);
207 nilfs->ns_gc_inodes_h = NULL;
208 }
209}
210
211static struct inode *alloc_gcinode(struct the_nilfs *nilfs, ino_t ino,
212 __u64 cno)
213{
214 struct inode *inode = nilfs_mdt_new_common(nilfs, NULL, ino, GFP_NOFS);
215 struct nilfs_inode_info *ii;
216
217 if (!inode)
218 return NULL;
219
220 inode->i_op = NULL;
221 inode->i_fop = NULL;
222 inode->i_mapping->a_ops = &def_gcinode_aops;
223
224 ii = NILFS_I(inode);
225 ii->i_cno = cno;
226 ii->i_flags = 0;
227 ii->i_state = 1 << NILFS_I_GCINODE;
228 ii->i_bh = NULL;
229 ii->i_dtime = 0;
230 nilfs_bmap_init_gc(ii->i_bmap);
231
232 return inode;
233}
234
235static unsigned long ihash(ino_t ino, __u64 cno)
236{
237 return hash_long((unsigned long)((ino << 2) + cno),
238 NILFS_GCINODE_HASH_BITS);
239}
240
241/*
242 * nilfs_gc_iget() - find or create gc inode with specified (ino,cno)
243 */
244struct inode *nilfs_gc_iget(struct the_nilfs *nilfs, ino_t ino, __u64 cno)
245{
246 struct hlist_head *head = nilfs->ns_gc_inodes_h + ihash(ino, cno);
247 struct hlist_node *node;
248 struct inode *inode;
249
250 hlist_for_each_entry(inode, node, head, i_hash) {
251 if (inode->i_ino == ino && NILFS_I(inode)->i_cno == cno)
252 return inode;
253 }
254
255 inode = alloc_gcinode(nilfs, ino, cno);
256 if (likely(inode)) {
257 hlist_add_head(&inode->i_hash, head);
258 list_add(&NILFS_I(inode)->i_dirty, &nilfs->ns_gc_inodes);
259 }
260 return inode;
261}
262
263/*
264 * nilfs_clear_gcinode() - clear and free a gc inode
265 */
266void nilfs_clear_gcinode(struct inode *inode)
267{
268 nilfs_mdt_clear(inode);
269 nilfs_mdt_destroy(inode);
270}
271
272/*
273 * nilfs_remove_all_gcinode() - remove all inodes from the_nilfs
274 */
275void nilfs_remove_all_gcinode(struct the_nilfs *nilfs)
276{
277 struct hlist_head *head = nilfs->ns_gc_inodes_h;
278 struct hlist_node *node, *n;
279 struct inode *inode;
280 int loop;
281
282 for (loop = 0; loop < NILFS_GCINODE_HASH_SIZE; loop++, head++) {
283 hlist_for_each_entry_safe(inode, node, n, head, i_hash) {
284 hlist_del_init(&inode->i_hash);
285 list_del_init(&NILFS_I(inode)->i_dirty);
286 nilfs_clear_gcinode(inode); /* might sleep */
287 }
288 }
289}